NSC DP8496

December 1991
DP8496/DP8497
SCSI-2 Disk Data Controller
General Description
Features
The DP8496/7 is a highly integrated, high-performance
CMOS SCSI disk data controller. It is designed for use inside intelligent hard disk drives that utilize the Small Computer System Interface (SCSI) standard. It can also be used
in ESDI, SMD and ST506 bridging controller applications.
The DP8496/7 includes most of the data path functions
needed to implement a complete hard disk controller. It includes a full featured SCSI Bus Controller, Buffer Memory
Interface with pipelined pointers, fast Disk Data Controller,
and a Processor Interface.
With the addition of National Semiconductor read-channel
chips such as a PLL Synchronizer, and Encoder/Decoder, a
pulse detector, and a head amplifier, complete data-path
electronics of a SCSI drive can be implemented. A microcontroller, such as National’s HPCTM , may be used to manage the SCSI commands and the drive specific control signals. The high level of intelligence implemented on the
DP8496/7 means lower overhead for the disk-drive’s embedded microcontroller, making possible a high-performance design employing only one micro-controller.
The DP8496 provides on-chip single-ended transceivers for
driving the SCSI bus. The DP8497 provides all control signals necessary for direct interfacing with differential transceivers recommended for Fast SCSI option of SCSI-2.
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
High disk data rates:
Ð DP8496/7-33 33 Mbit/sec
Ð DP8496/7-50 50 Mbit/sec
Synchronous SCSI-2 transfer rates up to 10 MByte/sec
with offset up to 16 (Fast option)
Asynchronous SCSI transfer rates up to 5 MByte/sec
Support for Fast Page Mode and Static Column Decode type DRAMs
Attains sustained buffer bandwidths above 9 MByte/sec
with byte-wide memory configuration
Word-wide buffer memory port allows sustained bandwidths of 17 MByte/sec
Buffer memory up to 4 MBytes DRAM or 1 MByte
SRAM
On-chip DMA with buffered pointer addresses
Multi-phase type SCSI commands
Parity error checking on SCSI, buffer memory, and all
internal data paths
Programmable format and sectoring modes including
soft, pseudo-hard, and hard
32, 48 or 56-bit computer generated ECC with on-chip
correction
On-chip single-ended transceivers on DP8496. DP8497
interfaces directly with differential transceivers for Fast
SCSI
Available in 100-pin PQFP package
System Diagram
TL/F/11212 – 1
TRI-STATEÉ is a registered trademark of National Semiconductor Corporation.
HPCTM is a trademark of National Semiconductor Corporation.
C1995 National Semiconductor Corporation
TL/F/11212
RRD-B30M105/Printed in U. S. A.
DP8496/DP8497 SCSI-2 Disk Data Controller
PRELIMINARY
Table of Contents
7.0 A.C. SPECIFICATIONS (Continued)
1.0 OVERVIEW
7.4 Ready Pin
2.0 PIN DESCRIPTION
7.5 Mode Pin (Mot.)
3.0 REGISTER LIST
7.6 Register Read (Mot.)
3.1 Reset Summary
7.7 Register Write (Mot.)
3.2 Initialization Registers
7.8 Interrupts
4.0 FUNCTIONAL DESCRIPTION
7.9 Buffer Memory Static Read
4.1 Processor Interface
7.10 Buffer Memory Static Write
4.1.1 Access by Different Type Processors
7.11 Buffer Memory Dynamic RAM Read/Write
4.2 Buffer Memory Interface
7.12 Buffer Memory DRAM Refresh
4.2.1 Static RAM
4.2.2 Dynamic RAM
4.2.3 Data Transfer Timing
4.2.4 Pointers
4.2.5 Buffer Management
4.2.6 Processor Access to Buffer Memory
4.2.7 16-Bit Wide Access
4.2.8 Sequential Accesses of Buffer Memory
4.2.9 Parity
4.2.10 Register Description
7.13 Disk Read Data Timing
7.14 Disk Write Data Timing
7.15 Disk Address Mark
7.16 SCSI Reset
7.17 SCSI Arbitration
7.18 SCSI Selection as Init (w/o Arb.)
7.19 SCSI Selection as Init (with Arb.)
7.20 SCSI Reselection as Init
7.21 SCSI Selection (Targ)
4.3 Disk Data Controller
4.3.1
4.3.2
4.3.3
4.3.4
7.22 SCSI Reselection (Targ)
Command Operation
Disk Command and Control Registers
Format Registers
Disk Operations
7.23 SCSI Lost Arbitration
7.24 SCSI Abort (Re)Selection
7.25 SCSI Async Info In (Init)
7.26 SCSI Async Info In (Targ)
4.4 SCSI Interface
7.27 SCSI Async Info Out (Init)
4.4.1 Simple SCSI Operations
4.4.2 SCSI Registers
4.4.3 SCSI Operations
7.28 SCSI Async Info Out (Targ)
7.29 SCSI Sync Info In (Init)
7.30 SCSI Sync Info In (Targ)
4.5 Timer
7.31 SCSI Sync Info Out (Init)
4.5.1 Timer Register Descriptions
5.0 APPLICATION INFORMATION: BUFFER MEMORY
INTERFACING IN A FAST SCSI IMPLEMENTATION
7.32 SCSI Sync Info Out (Targ)
6.0 D.C. SPECIFICATIONS
7.34 SCSI Bus Free (Targ)
7.33 SCSI Bus Free (Init)
7.35 Differential Transceiver Direction Control
7.0 A.C. SPECIFICATIONS
7.36 Direction Control for Target and Initiator
7.1 Clock Timing
7.2 Register Read (Intel)
8.0 A.C TEST CONDITIONS
7.3 Register Write (Intel)
9.0 PHYSICAL DIMENSIONS
2
1.0 Overview
Buffer RAM is needed for all disk or SCSI data transfers.
This RAM is connected to the Buffer Memory Interface section. The DP8496/7 assumes exclusive access to this buffer
RAM. This enables the chip to utilize the full bandwidth of
the RAM to streamline any combination of disk, SCSI or
processor transfers. All transfer of data is done with on-chip
DMA. Address pointers may be pipelined which will allow
different groups of data to be placed in non-consecutive
locations in buffer memory.
The Disk Data Controller section transfers NRZ data to the
serial disk data path. Sector size, gaps, synch bytes, etc. are
programmable. It can work with hard or soft sectored drives.
ESDI soft sectored (pseudo-hard) AMF/AME handshaking
is programmable. Fixed 32/48/56-bit ECC polynomial hardware automatically generates and checks error correction
fields. Correction calculation is done on chip which will relieve the processor of the time and code space overhead of
this function.
The SCSI Bus Controller section saves the user board area
by integrating the 48 mA open drain drivers. The controller
was designed to minimize the number of interrupts generated due to phase changes, parity errors and the like. Groups
of often used phase sequences can be invoked with a single command.
An internal timer is also available on the DP8496/7 which
may be used to accurately control the execution of certain
commands for the disk or SCSI sections.
The DP8496/7 contains four major sections. Each section is
listed below along with the major functions performed within
that section.
1. Processor Interface
Processor Interface for chip control
2. Buffer Memory Interface
SRAM/DRAM control timing
Memory access arbitration
Memory access prioritization
3. Disk Data Controller
Serializer/Deserializer (SERDES)
Read/Write/Format Control
CRC/ECC generation/checking/correcting
4. SCSI Bus Controller
SCSI Data Transfer Control
SCSI Bus ControlÐphase changes
Parity generation/checking
On-chip bus transceivers
The Processor Interface section allows the drive’s processor access to all programmable features of the chip. This
interface is used to initiate and control any function or operation on both disk and SCSI data. All DP8496/7 registers
are accessed through this section.
Block Diagram
TL/F/11212 – 2
3
2.0 Pin Description
**Extra Ground Pins Only on DP8496
TL/F/11212 – 3
Connection Diagram
100-Pin QFP
Signals shown in bold are unique to the DP8497
TL/F/11212 – 4
Order Number DP8496VF or DP8497VF
See NS Package Number VF100B
4
2.0 Pin Description (Continued)
Symbol
Pin
Type
Function
DISK DATA CONTROLLER PINS
RGATE
20
O
READ GATE: This active high output is asserted while the Disk Data Controller is reading data
from the disk. It commands an external data separator to acquire lock and enables the RDATA
pin.
RCLK
21
I
READ CLOCK: This input is the disk data rate clock. When RGATE is deasserted low, this pin
will receive the crystal or servo derived clock. When RGATE is asserted high, this pin will
receive the recovered NRZ clock from the decoder. Each rising edge of the clock at this input
is used to strobe RDATA into the Disk Data Controller. The AC timing characteristics should
not be violated, even during the time of transition between clock sources.
RDATA
22
I
READ DATA: This active high input accepts NRZ disk data from the data synchronizer/
decoder.
WGATE
23
O
WRITE GATE: This active high output is asserted while writing data to the disk.
WCLK
24
O
WRITE CLOCK: This output is derived from RCLK and is used to clock NRZ WDATA out from
the DP8496/7. The rising edge indicates vaild WDATA.
WDATA
25
O
WRITE DATA: This active high output is the NRZ data to be written to the disk. It is
synchronized to WCLK. It is held deasserted any time WGATE is deasserted.
INDEX
26
I
INDEX: This active high input from the disk drive signifies the start of a track. Disk Data
Controller commands may be synchronized to it.
AMF/SEC
27
I
ADDRESS MARK FOUND/SECTOR: This active high input denotes the start of a sector in
hard and pseudo-hard sectored drives. Soft sectored drives use AMF to denote the start of
header and data fields.
AME
28
O
ADDRESS MARK ENABLE: This active high output forces the read-channel encoder to
generate an Address Mark. It is also used to enable Address Mark detection for pseudo-hard
sectored drives.
PROCESSOR INTERFACE PINS
CRST
37
I
40–47
I/O
ADSi
29
I
ADDRESS STROBE IN: This active high input latches the address from PADB0 – 7 on the
falling edge. The latched address is used to select internal registers.
RDi or DS
30
I
The function of this pin is determined by the processor mode initialized by the RDY/MODE pin.
READ STROBE (IN): NSC/Intel mode: This active low input combined with a low level on CS
will assert data from the addressed register onto the PADB0 – 7 bus.
DATA STROBE: Zilog/Motorola Mode: This active low input combined with a low level on CS
will either assert data from the addressed register onto the PADB0 – 7 bus, or it will write data
present on the PADB0 – 7 bus into the register. The data direction is determined by the W/R
pin.
WRi or
R/W
31
I
The function of this pin is determined by the processor mode initialized by the RDY/MODE pin.
WRITE STROBE (IN)ÐNSC/Intel Mode: This active low input combined with a low level on CS
will write data present on the PADB0 – 7 bus into the addressed register.
READ/WRITEÐZilog/Motorola Mode: This pin determines the direction of data transfer on
the PADB0–7 bus while the CS and DS pins are both asserted. High e Read, Low e Write.
CS
32
I
CHIP SELECT: This active low input must be asserted to access any of the registers.
PADB0:7
RESET: This active low Schmitt input will reset the DP8496/7 immediately without regard to
data transfers which may be in progress. Registers affected are listed in Table 3.1.
PROCESSOR/ADDRESS/DATA BUS 0 – 7: These eight active high, bi-directional lines
transfer information between the DP8496/7 and the processor.
5
2.0 Pin Description (Continued)
Symbol
Pin
Type
Function
PROCESSOR INTERFACE PINS (Continued)
RDY or
MODE
33
I/O
The function of this pin is determined by the state of the CRST pin.
READY (CRST Pin Deasserted): When low, this output indicates that the Buffer Memory Data
register access must be extended by the processor. This pin can be connected to the
processor’s wait-state input. Refer to Section 4.2.8 for a more complete description of this option.
This pin is only active while in the NSC/Intel mode. This pin will be driven low internally while in
the Zilog/Motorola mode.
MODE (CRST Pin Asserted): This input will determine the processor access mode of the
DP8496/7. If this pin is left floating or pulled or driven high while CRST is asserted, the
National/Intel mode will be enabled. If this pin is pulled or driven low while CRST is asserted, the
Zilog/Motorola mode will be enabled.
While CRST is asserted, an internal pull-up resister is active. See the DC specifications for
characteristics.
DINT
34
O
DISK INTERRUPT: This active low output will be asserted on any Disk Data Controller condition
enabled by the Disk Interrupt Enable register. If the CI (Combine Interrupts) bit in the Setup 3
register is set to ‘‘1’’, this pin will become asserted if either a Disk Data Controller or a SCSI Bus
Controller interrupt occurs.
SINT
35
O
SCSI INTERRUPT: This active low output will be asserted on any SCSI Bus Controller condition
enabled by the SCSI Interrupt Enable register.
BCLK
36
I
BUS CLOCK: This input is used by the Buffer Memory Interface for access timing and DMA
arbitration. It is also used by the SCSI Bus Controller for timing.
SCSI BUS CONTROLLER PINS (All SCSI Signals are Active High on the DP8497)
SDB0:7,
SDBP
57 – 59
61 – 63
65 – 67
I/O
SCSI DATA, PARITY BUS: These nine active low, open drain, bi-directional lines should be
connected directly to the SCSI data bus.
BSY
70
I/O
BUSY: Same as above. Should be connected to SCSI control bus.
SEL
75
I/O
SELECT: Same as above.
C/D
76
I/O
COMMAND/DATA: Same as above.
I/O
78
I/O
INPUT/OUTPUT: Same as above.
REQ
77
I/O
REQUEST: Same as above.
ACK
71
I/O
ACKNOWLEDGE: Same as above.
ATN
69
I/O
ATTENTION: Same as above.
MSG
73
I/O
MESSAGE: Same as above.
RST
72
I/O
SCSI RESET: If CRST is asserted during power-up, the RST pin will not produce a glitch. See
AC Specifications for input de-glitch description.
6
2.0 Pin Description (Continued)
Symbol
Pin
Type
Function
SCSI BUS TRANSCEIVER DIRECTION CONTROL PINS (DP8497 Only)
DSDB0:7,
DSDBP
48–56
O
DIRECTION CONTROL FOR SCSI DATA, PARITY BUS: Each of these signals connects to
the appropriate enable pin of the differential transceivers used for a Fast SCSI
implementation and controls the direction of its corresponding SCSI data bus signal. A high
level on these pins enables the driver while a low level enables the receiver.
DTARG
79
O
DIRECTION CONTROL FOR TARGET SIGNALS: Same as above. Controls the direction of
C/D, I/O, REQ and MSG signals.
DINIT
80
O
DIRECTION CONTROL FOR INITIATOR SIGNALS: Same as above. Controls the direction
of ACK and ATN signals.
DBSY
81
O
DIRECTION CONTROL FOR BSY: Same as above. Controls the direction of BSY signal.
DSEL
82
O
DIRECTION CONTROL FOR SEL: Same as above. Controls the direction of SEL signal.
DRST
83
O
DIRECTION CONTROL FOR RST: Same as above. Controls the direction of RST signal.
BUFFER MEMORY INTERFACE PINS
DB1 – 0:7,
DB1 – P
84–92
I/O
DATA BUS 1: These nine active high, bi-directional lines transfer data between the
DP8496/7 and the least significant 8-bits plus parity of buffer memory.
ADB2 – 0:7,
ADB2 – P
93–97
99–100
1– 2
I/O
ADDRESS DATA BUS 2: These nine active high, bi-directional lines transfer data between
the DP8496/7 and the most significant 8-bits plus parity of buffer memory if word mode is
enabled. These pins also contain address information for static RAM. See Tables 4.2 and
4.3.
AB3 – 0:7
4–11
O
ADDRESS BUS 3: These eight active high outputs represent address information for both
SRAM and DRAM. See Tables 4.2 to 4.5.
AB4 – 0:2
12–14
O
ADDRESS BUS 4: These three active high outputs represent address information for both
SRAM and DRAM. See Tables 4.2 to 4.5.
RDo or
CAS
18
O
The function of this pin is determined by the use of SRAM or DRAM.
READ STROBE (OUT): SRAM.
COLUMN ADDRESS STROBE: DRAM.
WRo
19
O
WRITE STROBE (OUT): Always.
ADSo or
RAS
17
O
The function of this pin is determined by the use of SRAM or DRAM.
ADDRESS STROBE (OUT): SRAM. In word mode, controls address latch for ADB2 bus.
High level on this signal should be used by an external latch to accept address inputsÐ
latching them on the falling edge.
ROW ADDRESS STROBE: DRAM.
7
2.0 Pin Description (Continued)
Pin
Symbol
Function
1
2
ADB2 – 7
ADB2 – P
Buffer Memory
Address/Data Bus 2
3
VCC2
a 5 VDC
4
5
6
7
8
9
10
11
AB3 – 0
AB3 – 1
AB3 – 2
AB3 – 3
AB3 – 4
AB3 – 5
AB3 – 6
AB3 – 7
Buffer Memory
Address Bus 3
12
13
14
AB4 – 0
AB4 – 1
AB4 – 2
Buffer Memory
Address Bus 4
15
16
GND1
GND5
GND
GND
17
18
19
ADSo/RAS
RDo/CAS
WRo
Buffer Memory
Strobes
20
21
22
23
24
25
26
27
RGATE
RCLK
RDATA
WGATE
WCLK
WDATA
INDEX
AMF/SEC
28
AME
Disk Read Gate
Disk Read Clock
Disk Read Data
Disk Write Gate
Disk Write Clock
Disk Write Data
Disk Index Input
Disk Address Mark Found or
Sector Input
Address Mark Enable
29
30
31
32
33
34
35
36
37
ADSi
RDi/DS
WRi/R/W
CS
RDY/MODE
DINT
SINT
BCLK
CRST
Processor Port
Strobes
38
39
VCC1
VCC3
a 5 VDC
a 5 VDC
40
41
42
43
44
45
46
47
PADB-0
PADB-1
PADB-2
PADB-3
PADB-4
PADB-5
PADB-6
PADB-7
Processor Address/
Data Bus
48
49
50
51
DSDB0
DSDB1
DSDB2
DSDB3
Transceiver Direction
Control for DP8497
(N/C on DP8496)
Ready/Mode
Disk Interrupt
SCSI Interrupt
Bus Clock
Chip Reset
8
Pin
Symbol
52
53
54
55
56
DSDB4
DSDB5
DSDB6
DSDB7
DSDBP(DP8497)
GND9 (DP8496)
Direction Control (Continued)
Function
57
58
59
SDB-0
SDB-1
SDB-2
SCSI Data Bus
60
GND2
GND
61
62
63
SDB-3
SDB-4
SDB-5
SCSI Data Bus
64
GND7
GND
65
66
67
SDB-6
SDB-7
SDB-P
SCSI Data Bus
68
GND3
GND
69
70
71
72
73
ATN
BSY
ACK
RST
MSG
SCSI Attention
SCSI Busy
SCSI Acknowledge
SCSI Reset
SCSI Message
74
GND6
GND
75
76
77
78
SEL
C/D
REQ
I/O
SCSI Select
SCSI Cmd/Data
SCSI Request
SCSI Input/Output
79
80
81
82
83
GND8 (DP8496)
DTARG (’97)
DINIT
DBSY
DSEL
DRST
Additional GND on DP8496
Transceiver Direction
Control on DP8497 only
(N/C on DP8496)
84
85
86
87
88
89
90
91
92
DB1 – 0
DB1 – 1
DB1 – 2
DB1 – 3
DB1 – 4
DB1 – 5
DB1 – 6
DB1 – 7
DB1 – P
Buffer Memory Data
Bus 1
93
94
95
96
97
ADB2 – 0
ADB2 – 1
ADB2 – 2
ADB2 – 3
ADB2 – 4
Buffer Memory
Address/Data Bus 2
98
GND4
GND
99
100
ADB2 – 5
ADB2 – 6
Buffer Memory
Address/Data Bus 2
GND on DP8496 Only
SCSI Parity
3.0 Register List
DISK DATA CONTROLLER REGISTERS
Addr.
00 – 06
07
08 – 0F
10
11
12
13
14
15
16
17
18
19
1A – 1F
Name
Label
SCSI REGISTERS
R/W Pg.
Addr.
Name
Label
R/W
Pg.
ECC Shift Registers 0–6
ESRx
R
29
ECC Control
EC
R/W 28
Reserved
Disk Command
DCMD
W
23
Disk Control
DCTL R/W 25
Disk Interrupt
DINT R/W 27
Disk Interrupt Enable
DINTE R/W 27
Disk Status
DSTAT
R
26
Sector Number
SN
R/W 28
Header Byte Count/Interlock HBC R/W 28
Sector Count
SC
R/W 28
ECC Byte Count (Low)
ECCL
R
29
ECC Byte Count (High)
ECCH
R
29
Reserved
40
41
42
43
44
45
46
47
48
49
4A
4B
4C
4D
4E
4F
50 – 5F
SCSI Command
SCSI Data
SCSI Control
SCSI Operation
Synchronous Transfer
Identify
Destination ID
Source ID
SCSI Status
SCSI Interrupt
SCSI Interrupt Enable
SCSI Block Count
SCSI Block Size (MSB)
SCSI Block Size (LSB)
Differential SCSI 1
Differential SCSI 2
Reserved
SCMD
SDAT
SCTL
SOP
SYNC
IDENT
DID
SID
SSTAT
SINT
SINTE
SBC
SBSH
SBSL
SDIF1
SDIF2
W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R
R
R/W
R/W
R/W
R/W
R/W
R/W
39
53
53
53
54
55
56
56
56
57
60
60
60
60
60
60
DISK FORMAT REGISTERS
Addr.
20
21
22
23
24
25
26
27
28
29
2A
2B
2C
2D
2E
2F
Name
Post Sector/Index Count
Header Preamble Count
Preamble Pattern
Header Synch Ý1, Ý2 Counts
Header Synch Ý1 Pattern
Header Synch Ý2 Pattern
Header Byte 0, 1 Control
Header Byte 0 Pattern
Header Byte 1 Pattern
Header Byte 2, 3 Control
Header Byte 2 Pattern
Header Byte 3 Pattern
Header Byte 4, 5 Control
Header Byte 4 Pattern
Header Byte 5 Pattern
Header ECC & Postamble
Count
30
Postamble Pattern
31
Data Preamble Count
32
Data Synch Ý1, Ý2 Count
33
Data Synch Ý1 Pattern
34
Data Synch Ý2 Pattern
35
Sector Byte Count (Low)
36
Sector Byte Count (High)
37
Data Format Pattern
38
Data ECC & Postamble
Count
39
Gap 3 Byte Count
3A
Gap 3 & PSIG Pattern
3B – 3F Reserved
Label R/W Pg.
PSIG
HPC
PREP
HSC
HSP1
HSP2
HC01
HP0
HP1
HC23
HP2
HP3
HC45
HP4
HP5
HECC
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
29
29
30
30
30
30
30
31
31
31
31
31
31
31
31
31
POSTP
DPC
DSC
DSP1
DSP2
SBCL
SBCH
DFP
DECC
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
31
31
31
32
32
32
32
32
32
SETUP AND TIMER REGISTERS
Addr.
60
61
62
63
64
Name
Label
R/W
Pg.
Setup 1
Setup 2
Setup 3
Timer Prescale
Timer Count
SUP1
SUP2
SUP3
TPRE
TCNT
R/W
R/W
R/W
R/W
R/W
10
11
12
64
64
BUFFER MEMORY REGISTERS
GAPC R/W 32
GAPP R/W 32
9
Addr.
Name
Label
R/W
Pg.
65
66
67
68
69
6A
6B
6C
6D
6E
6F – 7F
Disk Pointer (MSB)
Disk Pointer
Disk Pointer (LSB)
SCSI Pointer (MSB)
SCSI Pointer
SCSI Pointer (LSB)
Processor Pointer (MSB)
Processor Pointer
Processor Pointer (LSB)
Buffer Memory Data
Reserved
DP2
DP1
DP0
SP2
SP1
SP0
PP2
PP1
PP0
BMD
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
R/W
19
19
19
19
19
19
19
19
19
19
3.0 Register List (Continued)
3.1 RESET SUMMARY
Setup 1 (SUP1)
7
6
There are four different types of resets that may be issued.
Each reset has its own function and purpose. These are
summarized in Table 3.1.
CLK1
CLK0
60h
5
4
3
2
1
SWS
DFP
DD1
DD0
RSEL
R/W
0
RWID
The contents of this register will be invalid after asserting
the CRST pin. It must be initialized before proper chip operation.
3.2 INITIALIZATION REGISTERS
The three basic Setup Registers, that must be initialized before proper chip operation can begin, are described in this
section. All other registers are described in their respective
sections in the Functional Description chapter; i.e., in the
Disk Data Controller, SCSI Bus Controller, Processor Interface, Buffer Memory Interface, and Timer sections.
Setup 1 register is typically written only once after a reset to
establish the physical path between the DP8496/7 and its
buffer memory.
Setup 2 register contains control bits which may be modified
by the processor for power-up tests, error recovery or normal operations.
Setup 3 register is typically written only once after a reset to
establish pin configuration.
CLK(1:0): BCLK Frequency
The Bus Clock (BCLK) is used for many functions within the
DP8496/7. It is used in the SCSI Bus Controller, Buffer
Memory Interface, and Timer sections. Therefore, the
choice of Bus Clock frequency is very important. Table 3.2
lists the frequency range allowable for each bit combination
of this field to guarantee timing specifications within the
SCSI standard.
The minimum BCLK frequency allowed is 10 MHz. Choosing
a clock frequency nearest the upper limit in a range results
in the optimal and fastest execution fo SCSI operations.
TABLE 3.1. Summary of Resets
Reset Type
Purpose
Assert RST Bit in
Disk Control
Register
Reset Disk
Data Controller
Clears and terminates current Disk command and any pipelined command.
Clears Disk Status register.
Clears bits 0–5 and 7 of the Disk Interrupt register.
Deasserts RGATE, WGATE, and AME pins.
Result
Issue SCSI
Reset Command
Reset SCSI
Bus Controller
Clears all SCSI Bus signals.
Clears and terminates current SCSI command and any pipelined command.
Clears the following registers: SCSI Data, SCSI Control, SCSI Operation except the PEP
(Parity Error to Processor) bit, Identify, Destination ID except the ID field, SCSI Status
except the DBR (Data Buffer Ready) bit.
Generates an ‘‘Operation Complete’’ interrupt.
SCSI Bus Reset
Reset All SCSI
Devices on
SCSI Bus
Clears all SCSI Bus signals except the RST signal.
Clears and terminates current command and any pipelined command.
Generates a ‘‘SCSI Bus Reset’’ interrupt.
Assert CRST
Pin
Power-Up Reset
or Reset All
Hardware
Sets RST bit in Disk Control register (see RST bit in Disk Control register result above).
Same result as issuing a SCSI Reset command (see above) except does NOT generate an
interrupt.
Clears the SCSI Operation register.
Clears the SCSI Interrupt and SCSI Interrupt Enable registers.
The contents of the Setup 1 register becomes invalid.
The contents of the Setup 2 register becomes invalid except the SPP (SCSI Parity Polarity)
bit which is set to a ‘‘1’’ and the SPE bit which is reset to ‘‘0’’.
Resets Timer Prescale and Timer Count registers.
Clears Buffer Memory Data register. (This register will not reflect the contents of buffer
memory until the Processor Pointer registers are initialized.)
The RDY/MODE pin functions as an input.
10
3.0 Register List (Continued)
RWID: RAM Data Path Width
TABLE 3.2. BCLK Frequency Range
BCLK Freq.
Code
CLK1
CLK0
0
1
0
1
0
0
1
1
The value of this bit is used by the SDDC to determine
whether to provide addresses for a byte-wide memory configuration or word-wide configurationÐwhen disk or SCSI
data accesses are made. The state of this bit has no effect
on the processor address pointer, which always increments
by one.
1 Byte Mode. 8-bit wide buffer memory is being used. All
address pointers will increment by one.
0 Word Mode. 16-bit wide buffer memory is being used.
Disk and SCSI address pointers will increment by two.
BCLK Range
10.00 s BCLK s 15.00 MHz
15.00 s BCLK s 17.50 MHz
17.50 s BCLK s 20.00 MHz
20.00 s BCLK s 25.00 MHz
Note: Though not in binary order, this table is correct.
SWS: SRAM Wait State
This bit allows the user to extend SRAM wait states so that
slower SRAMs may be used.
0: Normal wait states. Single byte burst takes 5 cycles, 4
word bursts take 12 cycles, and 6 byte bursts take 17
BCLK cycles.
1: Extended wait states. Single byte burst is extended to 6
cycles; the 4 word burst transfer is extended to 16 cycles; and the 6 byte burst transfer is extended to 23 cycles.
Setup 2 (SUP2)
7
6
AI
2
1
BPE
SPP
SPE
SID2
SID1
R/W
0
SID0
BPP: Buffer Parity Polarity
1 Even parity is checked or generated at the buffer memory port if enabled by the BPE bit.
0 Odd parity is checked or generated at the buffer memory port if enabled by the BPE bit.
BPE: Buffer Parity Enable
1 Enables the checking or generation of parity at the buffer memory port for various types of transfers:
When Writing Parity is generated for the data exiting at
to Buffer:
the buffer port; except in the case of
SCSI to Buffer transfer, when, the parity
is generated only if SCSI Parity is not enabled (SPE e 0). If SPE e 1, then the
parity accompanying SCSI Bus Data is
checked upon exiting the chip at buffer
port.
When Reading Parity is checked at the port where the
from Buffer:
data exits the chip; except in the case of
Buffer to SCSI transfer with the SCSI
Parity disabled: here the parity is
checked on the incoming data at the
buffer port. Also, in the case of processor reads of buffer memory, the parity is
checked on the incoming data at the
buffer port.
0 Parity is not generated or checked at the buffer memory
port. Parity may still be employed at the SCSI port by
using the SPE bit of this register.
TABLE 3.3. Organization of DRAM
0
1
0
1
3
AI: Auto Increment
1 The Processor Pointer will increment after each access, read or write, of the Buffer Memory Data register.
0 No incrementing will occur. The Processor Pointer will
remain unchanged unless it is loaded with a new value.
DD(1:0): DRAM Size
If DRAM is used, the DP8496/7 needs to know the organization, e.g., 64k x n, 256k x n, etc. This field controls the row
and column address so that the least significant address
bits are always on the column address. This will guarantee
proper refresh timing during Buffer Memory transfers. The
proper bit settings are shown in Table 3.3.
If SRAM is used, these bits are ‘‘don’t care’’.
0
0
1
1
4
The contents of this register will be invalid after assertion of
the CRST pin. This register must be initialized before proper
chip operation.
DFP: Disable Fast Page
1: Fast Page DRAM access mode will be disabled. This
way normal DRAMs can be used. If SRAMs are being
used, then SRAM ‘‘Fast mode’’ will also be disabled;
thus SRAM transfers will occur in single byte/word
bursts.
0: Fast Page DRAM access mode will be enabled.
DRAM Depth
RSZ1
RSZ0
BPP
61h
5
Organization of DRAM
64k x n, (64k x 1, 64k x 4 . . . )
256k x n, (256k x 1, 256k x 4 . . . )
1M x n, (1M x 1, 1M x 4 . . . )
4M x n, (4M x 1, 4M x 4 . . . )
RSEL: DRAM/SRAM Select
The DRAM/SRAM switch. This bit changes the memory access timing to control static or dynamic RAM memory. This
bit must be set before any accesses are attempted to buffer
memory. If DRAM is selected, the proper refresh timing to
initiate CAS before RAS automatic refreshing is provided
without any further intervention through the Processor Interface.
1 DRAM
0 SRAM
11
3.0 Register List (Continued)
REV(3:0): Revision Number
SPP: SCSI Parity Polarity
Set to ‘‘1’’ when CRST is asserted or a SCSI Reset command is issued. This is to prevent the parity bit from driving
the bus immediately after a reset while in Manual Mode.
(Manual Mode described in Section 4.4.3.)
1 Even parity is checked or generated on the SCSI bus if
enabled by the SPE bit.
0 Odd parity is checked or generated on the SCSI bus if
enabled by the SPE bit.
These four bits reflect the functional version number of the
chip. If a modification is made to the DP8496/7 that may
require software or hardware modifications in a system, this
value will be modified as well. The Revision Number corresponding to this data sheet is 3h.
The Revision Number will not be affected by a write to this
register.
4.0 Functional Description
SPE: SCSI Parity Enable
Set to ‘‘0’’ when CRST is asserted or a SCSI Reset command is issued. Other than in Arbitration phase, parity is
always provided when writing to the SCSI bus; parity is
checked if buffer parity is being used (BPE e 1), or generated if buffer parity is not being used (BPE e 0). Parity verification while reading the SCSI bus is dependent upon this bit
and some other variables as desribed below.
1 SCSI Parity is verified on read data conditional to A/M
bit of the SCSI Operation Register (43h) and HE bit of
the Synchronous Transfer Register (44h)Ðas specified
in Table 3.4.
0 SCSI Parity is never verified.
4.1 PROCESSOR INTERFACE
The processor port of the DP8496/7 can interface with
equal ease with NSC/Intel-type and Motorola/Zilog-type
processors. An 8-bit multiplexed Address/Data port is provided, and through it the processor can randomly access all
SDDC registers. The chip also provides two interrupt lines to
allow separate disk and SCSI interrupts.
To the processor, the DP8496/7 appears as a slave peripheral at all times and can be accessed with programmed I/O
in memory mapped or I/O mapped systems. All control registers are eight bits wide. Some functions are programmed
through more than one register. For example, address
pointers are accessed through three different registers because the address may be up to 22 bits wide. The buffer
memory data is accessed through one register called the
Buffer Memory Data register (6Eh). There are two interrupt
registers in the DP8496/7, the SCSI Interrupt register (49h)
and the Disk Interrupt register (12h). Each of these interrupt
registers has an associated mask and status register.
TABLE 3.4. SCSI Parity
Auto/
Man.
Handshake
Enable
When is parity checked during
a read from the SCSI bus?
1
X
Target: Parity checked when
ACK asserted.
Initiator: Parity checked when
REQ asserted.
(Re)Select: Parity Checked while
SCSI ID matches, SEL true, BSY
false.
0
1
0
TABLE 4.1. DP8496/DP8497 Address Map
Target: Parity checked when
ACK asserted.
Initiator: Parity checked when
REQ asserted.
0
Parity not checked.
CI
x
4
62h
3
2
1
x
x
REV3
REV2
REV1
Disk Data Controller
Disk Format Registers
SCSI Bus Controller
Buffer Memory, Setup & Timer
4.1.1 Access by Different Type Processors
At the time of chip reset, the user can configure the
DP8496/7 to work with NSC/Intel-type or Motorola/Zilogtype processors. While the CRST pin is asserted, the RDY/
MODE pin functions as a MODE input. This input will determine the processor access mode of the DP8496/7. If this
pin is left floating or pulled or driven high while CRST is
asserted, the National/Intel mode will be enabled. If this pin
is pulled or driven low while CRST is asserted, the Zilog/
Motorola mode will be enabled.
If NSC/Intel-type mode is selected, then pins RDi/DS and
WRi/R/W act as Read Strobe and Write Strobe and they
are used to read and write data from/to the addressed register. If the Motorola/Zilog-type mode is selected, then the
same pins act as Data Strobe and Read/Write; where Data
Strobe is used to strobe data when reading or writing, while
the direction of data is determined by the Read/Write signal. In either mode, the address is strobed into the SDDC
with the ADSi signal.
SID(2:0): SCSI ID
The unique binary address which identifies the SCSI device
is loaded here. This number is usually obtained by the processor after a Chip Reset from a PROM or from switches or
jumpers. This is not to be confused with the LUN (Logical
Unit Number) of which each SCSI device can have eight.
This field will be internally translated to the one-of-eight
form required for the SCSI bus.
5
DP8496/7 Section
00 – 1F
20 – 3F
40 – 5F
60 – 7F
All the registers in the DP8496/7 have unique addresses
and most are readable and writable. They can be accessed
in single instructions with a local processor. The DP8496/7
never becomes a bus master on the processor bus.
Refer to Table 4.10 in Section 4.2.9 for more detail on parity.
Setup 3 (SUP3)
7
6
Address
R/W
0
REV0
Asserting the CRST pin will not effect this register. This register must be initialized before proper interrupt operation.
CI: Combine Interrupts
1 Enabled Disk and SCSI interrupts appear on the DINT
pin. The SINT pin is never asserted.
0 Enabled Disk interrupts appear on the DINT pin. Enabled SCSI interrupts appear on the SINT pin.
12
4.0 Functional Description (Continued)
The Processor Interface bus timing can be asynchronous to
any clock on the DP8496/7.
TABLE 4.3. SRAM Address Pins: 16-Bit Interface
ADB2 (MUXed)
4.2 BUFFER MEMORY INTERFACE
All data transfers to and from the disk and to and from the
SCSI bus go through buffer memory. Buffer memory is attached directly to the Buffer Memory Interface. Access by
the DP8496/7 to buffer memory is performed via the internal DMA controller. All access to the buffer memory is prioritized and arbitrated through the DP8496/7. Thus, concurrent disk, SCSI, and processor accesses to the buffer memory are allowed. In normal operation no other bus masters
access the buffer memory.
The user has the option of configuring the memory in a bytewide or word-wide arrangement. If byte wide SRAM or any
type of DRAM is used, no components other than the memory chips are necessary on the buffer memory interface.
The only external component that may be needed is an 8-bit
address demultiplexing latch if word-wide SRAM is used.
Appropriate buffer memory configuration must be specified
in the setup registers, Setup 1 and Setup 2, before proper
chip operation can begin. Setup 1 register is used to configure the port timing to static or dynamic modes, program the
optional wait states, specify the DRAM depth and data bus
width, and to enable fast page mode DRAM control. Setup 2
register is used to enable and configure buffer memory parity.
4.2.2 Dynamic RAM
Since dynamic RAMs contain their own address latches, no
external latches are needed in byte- or word-wide mode.
Multiplexed row and column addresses are issued on the 11
physical address lines of AB3 and AB4, with the least significant address bits being issued on the CAS strobe. A full
4 MB address range is available in byte wide mode which
can be apportioned on 64k, 256k, 1M or 4M by n-bit boundaries. A full 2M address range is available in word wide
mode which can be apportioned on 64k, 256k, or 1M by
n-bit boundaries.
The specified pins that are used in byte mode are described
in Table 4.4. The specific pins that are used in word mode
are described in Table 4.5. The redundant signals on AB4
may be ignored.
TABLE 4.4. DRAM Address Pins: 8-Bit Interface
Addr.
AB4
Strobe
Depth
2 1 0
Pins 7
6
5
4
3
2
1
0
7
6
5
AB3
4 3
2
1
0
8
0
64k
RAS
CAS
21 19 17 15 14 13 12 11 10 9
10 9 8 7 6 5 4 3 2 1
256k
RAS
CAS
21 19 17 15 14 13 12 11 10 9 16
10 9 8 7 6 5 4 3 2 1 0
1M
RAS
CAS
21 19 17 15 14 13 12 11 10 18 16
10 9 8 7 6 5 4 3 2 1 0
4M
RAS
CAS
21 19 17 15 14 13 12 11 20 18 16
10 9 8 7 6 5 4 3 2 1 0
TABLE 4.2. SRAM Address Pins: 8-Bit Interface
AB4
AB3
Addr 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
Bit
4.2.1 Static RAM
In SRAM mode there are 19 address bits available to support up to 512 kBytes in byte mode, or 1 MBytes in word
mode. There is a large degree of flexibility of SRAM configurations that may be interfaced to the Buffer Memory Interface. The specific pins that are used in byte mode are described in Table 4.2. The address bits 19, 20, and 21 still
exist in all pointer registers but they are not output to the
interface.
ADB2
AB4
Pins 7 6 5 4 3 2 1 0 2 1 0 7 6 5 4 3 2 1 0
AB3
TABLE 4.5. DRAM Address Pins: 16-Bit Interface
2 1 0 7 6 5 4 3 2 1 0
Addr.
AB4
Strobe
Depth
2 1 0
Addr 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Bit
SRAM used word-wide can transfer a little less than twice
as fast as byte-wide given equal clock rates. The address
increments by two in word mode making address bit 1 of the
pointers the least significant address bit presented on the
physical interface. The processor will still access the RAM
as individual upper and lower bytes of a single word by using
address bit 0 in the Processor Pointer. A0 will determine
which byte is present in the Buffer Memory Data registerÐ
low byte or high byte.
The specific pins that are used in word mode are described
in Table 4.3. The ADSo pin is used to strobe an external 8bit latch, capturing A12–19 so the ADB2 bus can be used
for bi-directional data during the remainder of the memory
cycle.
7
6
5
AB3
4 3
1
0
64k
RAS
CAS
X 20 18 16 15 14 13 12 11 10 9
11 10 9 8 7 6 5 4 3 2 1
256k
RAS
CAS
X 20 18 16 15 14 13 12 11 10 17
11 10 9 8 7 6 5 4 3 2 1
1M
RAS
CAS
X 20 18 16 15 14 13 12 11 19 17
11 10 9 8 7 6 5 4 3 2 1
2M
RAS
CAS
X 20 18 16 15 14 13 12 21 19 17
11 10 9 8 7 6 5 4 3 2 1
Note: The address signal x will always be equal to VDD.
13
2
4.0 Functional Description (Continued)
The on-chip arbitration circuitry uses the prioritization
scheme shown in Table 4.7 to smoothly manage the data
flow and insure that the FIFOs don’t overflow or underflow.
The maximum capacitive loading on the Buffer Memory Interface should be taken into account when designing the
buffer memory interface. The address/data pins are designed to directly drive up to 12 DRAM chips with board
traces of reasonable length. By using 256k x 1 chips, 256k
can easily be addressed in 8-bit mode. By using higher density DRAMs, like 1M x 4 and 1M x 1, the full addressing
range of 4M in byte-wide mode, and 2M in word-wide mode,
can be utilized with chip counts up to or less than 12.
No refresh address counters are provided on-chip. This necessitates the use of DRAMs with on-board counters and
the ability to do a ‘‘CAS before RAS’’ type refresh.
TABLE 4.7. Buffer Memory Transfer Priority
4.2.3 Data Transfer Timing
Two basic transfer modes are possible, normal mode and
the fast page mode. If the fast page mode is disabled, by
setting the DFP bit to 1 in Setup 1 register (60h), then a
single memory transfer will occur every 5 BCLK cycles for
DRAM and every 5 or 6 cycles for SRAM, depending on the
SWS bit of the SUP1 register.
If the fast page mode is selected, by setting the DFP bit to 0,
then the DP8496/7 will transfer up to 6 bytes (or 4 words in
word-mode) of data at a time in a ‘‘burst’’, as long as only
the least significant 8 bits of address change. In the case of
DRAMs then, the DP8496/7 only has to change the address and toggle the CAS line for each byte or word of the
burst transferred after the first one. The number of bytes or
words transferred in a ‘‘burst’’ are shown in Table 4.6 and
depend on number of bytes or words present in the on-chip
FIFOs, the address pointers’ relation to the page boundary,
and refresh, etc. Operation in this mode yields highest attainable buffer memory bandwidths, but it requires the use
of fast page mode type DRAMs.
DRAM
Transfer
Period
SRAM
Transfer
Period
6 Bytes
12*BCLK Period
17*BCLK Pd (SWS e 0)
23*BCLK Pd (SWS e 1)
4 Bytes
1 Byte
9*BCLK Period
5*BCLK Period
4 Words
9*BCLK Period
3 Words
2 Words
1 Word
8*BCLK Period
6*BCLK Period
5*BCLK Period
Type of Transfer
1
2
3
4
5
6
7
DRAM Refresh
Disk Data Burst
SCSI Data Burst
Processor Access
Disk DataÐSingle Byte/Word
SCSI DataÐSingle Byte/Word
Idle
The DP8496/7 arbiter does not waste any cycles between
consecutive transfers, even if those transfers used different
pointers. Pointers are seamlessly switched so that all memory transfers use consecutive BCLK cycles.
Keeping in mind the above priority scheme, some design
trade-offs must be made between the available bandwidth
of the buffer memory port and rates of disk, SCSI and processor accesses that are to be supported. A sound design
would assure that enough bandwidth is left over after accommodating simultaneous full-speed disk and SCSI transfers so that occasional processor access can be serviced
within reasonable time. Section 4.2.6 describes how processor accesses of buffer memory are handled.
Detailed timing specifications for the buffer memory port
can be found in Chapter 6 of this document. For a given bus
clock frequency, calculations for gross available bandwidth
can be carried out by using Table 4.6. For example, with bus
clock frequency set at 20 MHz, and with 6 bytes of fastpage mode transfers taking 12 cycles, bandwidth of
10 MByte/sec is calculated for a byte-wide operation. However, after this number is adjusted for the DRAM refresh
cycle overhead, maximum bandwidth is reduced to
9.6 MByte/sec. There are other practical factors that would
further reduce this number to the attainable, sustained
bandwidth for a particular transfer. For the case of BCLK e
20 MHz, this value turns out to be 9.3 MHz.
As a calculation based on the timing specifications in Chapter 6 would show, 100 ns DRAMs would work with the
DP8496/7 operating with bus clock of 20 MHz. By using
faster BCLK and DRAMs the user can achieve sustained
byte-wide bandwidths of above 11 MByte/sec.
TABLE 4.6. Fast Page Mode Transfer Periods
Number of
Transfers
Rank
12*BCLK Pd (SWS e 0)
16*BCLK Pd (SWS e 1)
TABLE 4.8. DRAM Bandwidth Capabilities
Note: SWS is bit 5 in SUP1 register (60h)
The DP8496/7 issues a DRAM refresh cycle every 128 bys
clock periods. Of these, 5 bus clock periods are used by the
refresh cycle to complete, so 123 bus clock periods are left
for actual data transfers. Thus, the minimum specified bus
clock rate of 10 MHz will guarantee a refresh cycle time of
less than 12.8 ms per row.
The DP8496/7 automatically manages concurrent buffer
accesses by disk, SCSI and processor, while also executing
DRAM refresh cycles within the requisite time. Small FIFO’s
have been integrated into the DP8496/7 to allow for buffer
memory latency due to contention. The Disk Data Controller
has a 14 byte (or 7 word) FIFO and the SCSI Bus Controller
has a 32 byte (or 16 word) FIFO. The word-wide FIFOs are
used if the 16-bit mode is used for buffer memory.
Attainable Bandwidth
Desired
DRAM
Fastest Usable
BCLK
Byte-Wide
Word-Wide
100 ns
18 MHz
8.4 MHz
15.3 MHz
100 ns
20 MHz
9.3 MHz
17.0 MHz
80 ns
22 MHz
10.2 MHz
18.7 MHz
80 ns
24 MHz
11.2 MHz
20.4 MHz
The maximum BCLK frequency is specified in Chapter 7.
For maximum performance, however, word-wide memory
architecture can be employed; and then, with 100 ns
DRAMs and 20 MHz BCLK, sustained bandwidth of
17 MB/s can be attained.
14
4.0 Functional Description (Continued)
The processor is allowed read/write access to the Holding
registers. It is only allowed to read the Active registers.
4.2.4 Pointers
Three 22-bit pointers represent the three possible addresses into the buffer memory. These addresses are used by the
Disk Data Controller, SCSI Bus Controller, and the Processor Interface. The Disk and SCSI pointers are double buffered in what are called the Active and Holding register sets.
Both of these Active pointers automatically increment on
every transfer to buffer memory. The Processor pointer is
only single buffered. It will increment automatically if enabled in the Setup 2 register.
Normally the processor will only access the Holding registers. The Active registers are really the block counters and
address pointers actually used to transfer the data. Thus
only in unusual situations will the Active registers need to be
read. The A/H (Active/Holding) bit in the SCSI Operation
register determines which register set the processor will access.
The Processor Pointer does not have a Holding register.
The Processor Pointer Active register will be accessed independent of the A/H bit.
The Disk, SCSI, and Processor Pointer registers contents
have no relationship to one another, thus one pointer is not
automatically kept a certain distance from the other.
Instead, the monitoring by the processor of block and sector
transfers is facilitated through the use of interrupts. [Polling
could also be used by reading the SCSI Interrupt (49h) and
Disk Interrupt (12h) registers.]
Interrupts are programmable at three different levels (see
Figure 4.2 ):
1. At the end of each block transferred.
2. After a certain number of blocks have been transferred.
This is called a ‘‘group’’.
3. After all programmed blocks have been transferred, or
‘‘command completion’’.
(See the descriptions of DINTE (13h) and SINTE (4Ah) registers in Sections 4.3 and 4.4, respectively)
This provides a point of synchronization for the processor to
safely monitor the progress of both disk and SCSI transfers
and thus prevent overflow or underflow of the buffer memory. Overflow and underflow are analogous to pointers overtaking one another. The processor is responsible for preventing this.
4.2.5 Buffer Management
The processor will normally initialize the Holding registers
with values before an operation is started. When the operation begins the DP8496/7 will copy the contents of the
Holding registers into the Active registers (which are actually counters). This double buffering allows the processor to
load new values into the Holding registers while the previous operation is still continuing. This feature allows for versatile management of contiguous or non-contiguous buffers.
TL/F/11212 – 5
FIGURE 4.1. Holding and Active Registers
TL/F/11212 – 6
FIGURE 4.2. Flexible Interrupt Levels
15
4.0 Functional Description (Continued)
Byte wide data written to the Buffer Memory Data register
by the processor replaces the low byte or high byte depending on the least significant bit in the Processor Pointer 0
register. After the other byte has been obtained by reading
the addressed word location, the 16-bit word is written back
to buffer memory. This ensures that only one byte is effected with each write.
If the AI (Auto Increment) mode is enabled in the Setup 2
register, the Processor Pointer will increment by one after
each read of the Buffer Memory Data register by the processor or after each transfer from the Buffer Memory Data register to the buffer memory. To complete a 16-bit processor
read the processor must read the Buffer Memory Data register again. When writing, the processor must write to the
Buffer Memory Data register again to change the other byte
of the memory word. After this second byte is altered, the
16-bit word is again written back to buffer memory. This
sequence completes a processor modification of a 16-bit
buffer memory location.
If Auto Increment mode is enabled and the processor writes
to the Buffer Memory Data register twiceÐloading the low
byte and the high byteÐ before the DP8496/7 initiates a
buffer memory read, no buffer memory read will occur. This
new 16-bit value will then be written to buffer memory twice.
The data is transferred twice to buffer memory to increment
the processor pointer to the next word.
4.2.6 Processor Access to Buffer Memory
The processor gains access to the buffer memory by loading the Processor Pointer 0–2 registers with the address of
the byte in buffer memory that needs to be accessed. The
Processor Pointer 0 register should be loaded last. The data
is then read from or written to the Buffer Memory Data register.
The Buffer Memory Data register can be read or written to
just like any other register on the DP8496/7. The same AC
specifications apply to this register. However, the data read
from the Buffer Memory Data register may not be valid immediately and the data written to the Buffer Memory Data
register may not be transferred to the buffer memory immediately.
The Buffer Memory Data register is simply another DMA
channel trying to access the buffer memory. Since the arbitration priority of the Processor Interface is not the highest,
there may be a delay before the contents of the Buffer
Memory Data register are actually transferred to or from the
buffer memory. The amount of the delay depends on how
much extra bandwidth was allowed by the system designer
in the buffer memory data path.
The Processor Pointer registers should be loaded by the
processor from most-significant byte to least-significant
byte. Immediately after the Processor Pointer 0 register is
written to, the DP8496/7 will initiate an arbitration for a read
from buffer memory at the location pointed to by the Processor Pointer registers. Once granted, the buffer memory
location is read and the value is placed in the Buffer Memory Data register.
In the case of a write operation, the Buffer Memory Data
register may be written to immediately after loading the
Processor Pointer 0 register. If the processor writes to the
buffer memory data register before the DP8496/7 completes a buffer memory read, no buffer memory read will
occur.
To determine when the transfer has completed, one of the
three techniques described in Section 4.2.8 must be used.
Same techniques should be used if sequential reads or
writes are issued.
4.2.8 Sequential Accesses of Buffer Memory
To make sequential reads and writes by the processor more
convenient, the Processor Pointer registers may be programmed to automatically increment after each access to
the Buffer Memory Data register. The AI (Auto Increment)
bit in the Setup 2 register determines whether the Processor
Pointer will automatically increment. The Processor Pointer
will increment after each read or write of the Buffer Memory
Data register, then the DP8496/7 will read the next RAM
location from buffer memory. Sequential writes are accomplished in the same way. Mixed reads and writes are also
allowed.
There are three methods to insure safe transfer of data
to/from buffer memory by the processor: Using the RDY
pin, software polling, or using the Interrupt Error.
The method chosen will largely be based upon the processor facilities available. If the processor has a Ready, DTACK
or some other asynchronous access control pin, the RDY
output pin of the DP8496/7 can be used to automatically
control access to Buffer Memory Data register. If a processor is chosen without such a facility, the software polling
operation is the next best choice.
The Disk Interrupt Enable register is used to select between
the different processor access techniques. Only bit 6 of this
register is used for this purpose. The other bits of this register will be described in the Disk Data Controller description
in Section 4.3.2. A summary of the RDY pin function and the
buffer memory interrupt function is shown in Table 4.9.
4.2.7 16-Bit Wide Access
If the 16-bit buffer memory mode is enabled in the Setup 1
register, byte wide processor access to DP8496/7 does not
change. The processor should load the Processor Pointer
registers from most-significant byte to the least-significant
byte. Immediately after the Processor Pointer 0 register is
written to, the DP8496/7 will initiate an arbitration for a read
from buffer memory at the location pointed to by the Processor Pointer registers.
This will result in the entire 16-bit word being read from
buffer memory. However, the Buffer Memory Data register
will only allow access to the 8 bits that were requested
based on the least significant bit in the Processor Pointer 0
register. The least significant bit in the Processor Pointer 0
register determines whether the low byte (0) or high byte (1)
can be accessed by the processor. By using an even processor pointer address and setting AI (Auto Increment) to
‘‘1’’, the low byte and high bytes of each word will be accessed in order.
Disk Interrupt Enable
7
6
5
TI
16
BMD
I/S
13h
4
3
2
1
GCI
SCI
HNC
HC
R/W
0
CCI
4.0 Functional Description (Continued)
time between each access to the Buffer Memory register. If
an interrupt occurs, simply clear the interrupt, wait an additional amount of time, and try the access again.
Interrupt after Indicates that the previous write has not yet
a write:
been transferred to the buffer memory. The
previous data is still retained and the DMA
process is unaffected by the new write. The
data just written will be ignored.
Interrupt after Indicates that the Buffer Memory Data regisa Read:
ter has not yet been bloaded with the data
requested. The data just read is invalid.
Ready Pin
The RDY pin may be connected to the processor’s memory
access control pin (or ‘‘wait state’’ pin). After WR or RD
goes low, the processor will be ‘‘wait-stated’’ until the data
from buffer memory becomes available. This is the easiest
method of access to buffer memory since the job of regulating the speed of the processor access is done entirely in
hardware by the DP8496/7. The only requirement is that the
processor used with the DP8496/7 have some type of asynchronous access line.
When the processor reads the Buffer Memory Data register,
the RDY pin will become deasserted if the data is not yet
available. It will become asserted when the data is valid. For
a write operation, the RDY pin will become deasserted if a
previous write has not yet been transferred to buffer memory.
To enable this mode, the BMD (Buffer Memory Data) bit in
the Disk Interrupt Enable register must be set to zero. The
RDY pin will be always deasserted while in the Zilog/Motorola mode.
If non-contiguous address locations are to be accessed, the
processor must poll the E/R (Error/Ready) bit in the Disk
Interrupt register for a ‘‘1’’ before the Processor Pointer registers may be modified.
TABLE 4.9. RDY Pin and Interrupt Summary
TI
E/R
12h
5
4
3
2
I/S
GCI
SCI
HNC
1
Disabled
Buffer Memory Error
0
Enabled
Buffer Memory Ready
Function of E/R Bit in
Disk Interrupt Register
4.2.9 Parity
Buffer memory parity is generated when writing to buffer
memory and verified while reading from buffer memory only
if enabled by the BPE (Buffer Parity Enable) bit in the Setup
2 register.
In general, parity (if enabled) is always checked at the point
where the data exits the chip. Therefore, parity errors generated within the DP8496/7 are checked as well as external
parity errors.
When the processor accesses data from buffer memory,
parity information cannot be determined through the Buffer
Memory Data register. If the Setup 2 register is configured
to check buffer memory parity, a parity error detected while
reading the Buffer Memory Data register will be reported in
the SCSI Operation register (43h). The SCSI Operation register may be polled after a block of data has been read from
the Buffer Memory Data register. Once a parity error is detected, it is cleared only by reading the SCSI Operation register.
R/W
0
1
RDY Pin
If non-contiguous address locations are to be accessed, the
processor must verify that the Processor Pointer registers
can be modified. To do this, the BMD bit in the Disk Interrupt
Enable register must be set to ‘‘0’’. This enables the software polling mode. While in this mode, the Processor must
poll the E/R bit in the Disk Interrupt register. When this bit is
‘‘1’’, the Processor Pointer registers may be modified. The
Interrupt Error mode may then be re-enabled by setting the
BMD bit in the Disk Interrupt Enable register back to ‘‘1’’.
Software Polling
If no asynchronous access pin is available, or you choose
not to use it, software polling may be used. With the BMD
(Buffer Memory Data) bit in the Disk Interrupt Enable register set to ‘‘0’’, not only is the Ready pin enabled, but the
R/E (Ready/Error) bit in the Disk Interrupt register will operate as ‘‘Buffer Memory Data Ready’’ and may be polled. A
‘‘1’’ in the E/R bit indicates that the Buffer Memory Data
register is ready for either reading or writing, a ‘‘0’’ means
that it is not ready.
A ‘‘1’’ in the E/R bit also indicates that the Processor Pointer registers may be modified by the Processor.
Disk Interrupt
7
6
BMD Bit in Disk
Interrupt Enable
Register
CCI
Interrupt Error
By setting the BMD (Buffer Memory Data) bit in the Disk
Interrupt Enable register to a ‘‘1’’, the RDY pin is disabled.
Also, the E/R bit in Disk Interrupt register will operate as a
true interrupt. This ‘‘Error’’ interrupt indicates that the Buffer
Memory Data register has been accessed before the data is
valid. The processor should simply wait a certain amount of
17
4.0 Functional Description (Continued)
TABLE 4.10. Operation of Parity Generation and Checking
Control Bits (SUP2)
Action on Port
Disk
Buffer
SCSI
Error
Report
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Ð
Check
Check
Ð
Ð
Generate
Check
Check
Inv. a Check
Inv. a Check
Ð
Check
Check
Inv. a Check
Inv. a Check
Ð
Ð
Invert
Invert
Ð
Generate
Invert
Ð
Generate
Check
Ð
Check
Check
Check
Check
Generate
Ð
Ð
Ð
Ð
Check
Check
Check
Check
Ð
Ð
Ð
Ð
Ð
Ð
Ð
1
1
1
1
Ð
2
2
2
2
2
2
2
2
Ð
Ð
3
3
Ð
4
Direction of Transfer
SPEN
DBPEN
SPP
DBPP
0
1
1
1
1
x
0
0
0
0
1
1
1
1
x
x
x
x
x
x
1
x
x
x
x
0
1
1
1
1
1
1
1
1
0
1
1
1
1
1
x
0
1
0
1
x
0
1
0
1
0
1
0
1
x
x
x
x
x
x
x
0
1
1
0
x
0
1
1
0
0
1
1
0
x
x
0
1
x
x
SCSI to Buffer
SCSI to Buffer
SCSI to Buffer
SCSI to Buffer
SCSI to Buffer
Buffer to SCSI
Buffer to SCSI
Buffer to SCSI
Buffer to SCSI
Buffer to SCSI
Buffer to SCSI
Buffer to SCSI
Buffer to SCSI
Buffer to SCSI
Not between SCSI & Buffer
Disk to Buffer
Buffer to Disk
Buffer to Disk
Processor to Buffer
Buffer to Processor
Error Reporting:
1. SCSI Bus Error in CIC field of SINT register (49h) and SCSI Parity Error in SSTAT register (48h).
2. Buffer Parity Error in CIC field of SINT register and Buffer Memory Error in SSTAT register.
3. Error in DINT register (12h) and Disk Parity error in EC field of DSTAT register (14h).
4. PEP bit in the SOP register (43h).
18
4.0 Functional Description (Continued)
4.2.10 Register Description
Disk PointerÐMSB (DP2)
7
6
5
x
65h
4
x
3
2
1
R/W
0
3
2
1
R/W
0
Address
in the active registers is incremented each time an access
to buffer memory is made by the disk controller. The active
register set may be read only by the processor by using the
same register address, but by first setting the A/H bit in the
SCSI Operation register.
Processor PointerÐMSB (PP2) 6Bh
7
6
5
4
3
2
1
R/W
0
3
2
1
R/W
0
The 22-bit address value programmed in these registers is
used as pointer into the buffer memory for all data transfers
between the disk port and buffer. The DP8496/7 uses these
registers as ‘‘holding registers’’ in that the contents of these
registers are loaded into a set of ‘‘active registers’’ upon
start of operation. Thus the holding registers may be accessed even while an operation is continuing. The value in
the active registers is incremented each time an access to
buffer memory is made by the disk controller. The active
register set may be read only by the processor by using the
same register address, but by first setting the A/H bit in the
SCSI Operation register.
Processor PointerÐLSB (PP0)
6Dh
7
6
5
4
3
2
1
R/W
0
SPSI PointerÐMSB (SP2)
7
6
5
Buffer Memory Data (BMD)
7
6
5
Disk Pointer (DP1)
7
6
66h
5
4
x
Address
Disk PointerÐLSB (DP0)
7
6
5
67h
4
3
2
1
x
Processor Pointer (PP1)
7
6
5
R/W
0
Address
6Ch
4
Address
Address
x
68h
4
x
SCSI Pointer (SP1)
7
6
3
2
1
Address
The 22-bit address value programmed in these registers is
used as pointer into the buffer memory for all data transfers
between processor and buffer. The value in these registers
is incremented automatically each time an access to buffer
memory is completedÐif the AI bit is Setup 2 register is set.
The user should write into the PP0 register lastÐas that
initiates a read operation. In case of a write operation, the
BMD register should next be written.
R/W
0
Address
4
3
2
1
R/W
0
3
2
1
R/W
0
6Ah
4
2
1
R/W
0
This register provides the processor a window into the buffer memory for reading or writing at the address pointed to
by the PP2, PP1, and PP0 registers. While the writing of
pointer LSB into the PP0 register initiates a read operation,
the act of writing to this register initiates a write operationÐ
terminating the read operation in effect.
For 16-bit wide memory operations, the processor has to
read or write this register twice to access both the low-order
and high-order bytes. Bit 0 and PP0 register is used to determine if the byte in BMD register is low-order (PP0:0 e 0) or
high-order (PP0:0 e 1). However, the processor does not
have to wait for one byte to be transferred before accessing
the other byte of the same word since the DP8496/7 will
read or write from the memory the full word.
Address
SCSI PointerÐLSB (SP0)
7
6
5
3
Data
69h
5
6Eh
4
Address
The 22-bit address value programmed in these registers is
used as pointer into the buffer memory for all data transfers
between the SCSI port and buffer. The DP8496/7 uses
these registers as ‘‘holding registers’’ in that the contents of
these registers are loaded into a set of ‘‘active registers’’
upon start of operation. Thus the holding registers may be
accessed even while an operation is continuing. The value
19
4.0 Functional Description (Continued)
Additional features include a Sector or Index pulse interrupt
option. This makes processor monitoring of disk rotational
speed easier and also location of sectors based on time
since last index pulse for soft sectored drives or count of
sector interrupts for hard sectored drives. Also programmable control over AME and AMF/Sector pins supports pseudo-hard sectored (ESDI soft sectored) as well as hard and
soft sectored drives.
All registers are buffered such that individual bits will not
change while RD strobe is active except where specified.
4.3 DISK DATA CONTROLLER
The Disk Data Controller is concerned only with the data
aspects of the disk drive. Control signals for head movement, head selection, track zero detection, etc. are left to
the processor or through an external I/O port. Because of
this, the DP8496/7 can easily be used in a wide variety of
SCSI disk systems.
Track format is specified through the internal pattern and
count registers loaded by the host processor at initialization.
A unique format can be developed for each application taking into account speed tolerances, PLL lock-on time and
write splice length, etc.
An additional field before the Header and Data Preambles
(Header Gap and Data Gap) controls RGATE vs. WGATE
assertion times. This assures that the PLL never sees a
write splice between the header and data field. Byte count
programmability of these fields ensures that minimum track
area for the write splice.
The data field length (sector size) can range from 16 kBytes
to 64 kBytes with a two byte resolution.
4.3.1 Command Operation
User issues disk commands by writing to the Disk Command register (DCMD, 10h) after format registers and other
disk control registers have been specified. Most disk commands follow the flowcharts shown in Figures 4.3 , 4.4 , and
4.5 . The command will begin after the starting condition
specified by the SCT (Start Control) bit in the Disk Command register and the DCOZ (Drive Command On Zero) bit
in the Timer Prescale register. Each command is really broken into two sub-commands: a Header Segment command
and a Data Segment command. If a single sector command
is issued (determined by the MSO (Multi-Sector Operation)
bit in the Drive Control register) the command will terminate
at the end of the specified sector. Multi-sector commands
will update the Sector Number register and the Sector
Count register after processing each sector and will continue until the Sector Count register reaches zero.
Finally, both the ID and Data portions of the sector can be
protected with CRC or ECC ranging in length from 16 bits
(CRC) to 32, 48 or 56 bits (ECC). All the polynomials are
fixed computer generated codes except the 16-bit CRC
which is a CCITT standard. Very little processor overhead is
needed for disk data error correction due to on-chip logic.
TL/F/11212 – 10
FIGURE 4.3. Disk Command Flowchart
20
4.0 Functional Description (Continued)
TL/F/11212 – 11
Note: HFASM e Header Failed Although Sector Matched as enabled in the Header Byte Control registers.
FIGURE 4.4. Header Segment Flowchart
21
4.0 Functional Description (Continued)
TL/F/11212 – 12
FIGURE 4.5. Data Segment Flowchart
22
4.0 Functional Description (Continued)
TABLE 4.12. Disk Command List
4.3.2 Disk Command and Control Registers
Disk Command(DCMD)
7
6
5
MSO
SCT
X
10h
4
3
2
1
Disk Comand
Opcode (Hex)
W Only
0
Disk Command Opcode
Writing to this register will initiate a disk command. Normally, other disk format and control registers are configured
before writing to this register, as described in Figure 4.6 in
the section describing Disck Operations (4.3.4). After all
other registers are loaded, writing to this register will initiate
the operation.
Disk commands may be pipelined. Just before a new command begins, the Disk Command register is downloaded
internally. A new command may then be written to the Disk
Command register. See Section 4.3.4 for additional comments on pipelining.
MSO: Multi-Sector Operation
1 Multi-sector operation the Sector Number register and
the Sector Count register.
0 Single sector operation.
SCT: Start Control
The Disk Command will not start until the conditions specified in Table 4.11 are satisfied. Note, however, that the
DCOZ bit in the TIMER PRESCALE register overrides the
SCT bit. When the DCOZ bit is set the command will start
when the time counts down to zeroÐregardless of the state
of SCT bit.
1
0
Drive Type
Starting Condition
Soft
Immediate
Pseudo-Hard
After AMF
Hard
After Index or Sector
Soft
After Index
Pseudo-Hard
Hard
0
No Operation
4
5
6
7
Ignore Header/Ignore Data
Ignore Header/Verify Data
Ignore Header/Write Data
Ignore Header/Read Data
8
9
A
B
Compare Header/Ignore Data(*)
Compare Header/Verify Data
Compare Header/Write Data
Compare Header/Read Data
0C
0D
Write Header/Ignore Data
Write Header/Write Data
10
11
12
13
Format Type A
Format Type B
Format Type C
Format Type D
14
15
Read Header/Ignore Data
Read Header/Read Data
18
19
Write Unformatted
Read Unformatted
1C
Start Correction Cycle
*The Compare Header/Ignore Data command transfers all the headers to
buffer memory.
TABLE 4.11. Start Control
SCT
Operation
Ignore Header/Ignore Data
This command will optionally check the Header CRC/ECC
or Data CRC/ECC fields, or both without comparing the
header or reading the data. The CRC/ECC verification is
enabled in the ECC Control register. This command will
search for a Header Synch field, count through the other
header fields, and optionally check the Header CRC/ECC. It
will then search for the Data Synch field, count through the
data field, and optionally check the Data CRC/ECC.
After AMF after First Index
After Index
Ignore Header/Verify Data
This command will search for a Header Synch field, count
through the other header fields, and optionally check the
Header CRC/ECC. It will then perform a byte for byte comparison with the data field and buffer memory and optionally
check the Data CRC/ECC.
Disk Command Opcode
This field contains the opcode of the actual command that
the Disk Data Controller will perform.
Table 4.12 shows a list of valid command opcodes. No other opcodes may be used.
Ignore Header/Write Data
This command will search for a Header Synch field, count
through the other header fields, and optionally check the
Header CRC/ECC. It will then write data from the buffer
memory to the data field.
No Operation
NOP corresponds to a sector operation. When the operation begins (depending upon the specified starting condition) the disk sequencer will blindly count through the Post
Index Gap, Data CRC/ECC, and Data Postamble fields and
then terminate with a Disk Operation Complete interrupt.
RGATE and WGATE are never asserted but signals associated with the specified starting condition must be present in
addition to RCLK. In soft sectored mode, the NOP will begin
immediately if the SCT (Start Control) bit is set to one.
Multi-sector NOPs are permitted in which case the Sector
Count active register will be decremented and the Sector
Number register incremented for each sector operation.
Ignore Header/Read Data
This command will search for a Header Synch field, count
through the other header fields, and optionally check the
Header CRC/ECC. It will then read the data field, transferring it to the buffer memory, and optionally check the Data
CRC/ECC.
23
4.0 Functional Description (Continued)
The ECC/CRC for the ID Segment and the Data Segment
will be automatically generated during the format operation.
This is true for all four types of the Format command (A, B,
C, and D).
Compare Header/Ignore Data
This command is unique. As the Header Segments read
from the disk are being scanned for a comparison to the
Header Byte Control and Pattern registers, the header bytes
are all transferred to buffer memory. The header bytes from
each sector read will be placed in the buffer memory after
the previous sector’s header bytes. If the number of header
bytes from a sector is odd, an extra dummy byte will be
transferred to buffer memory at the end of each sector.
Only the Header Bytes are transferred. No Synch or
CRC/ECC data will be transferred.
This command is normally used in the single sector mode
(MSO e 0). This command will terminate under the same
conditions as any other single sector command: a header
match is found, an error occurs, or 2 index pulses are received without a header match.
Format Type B
Internal header bytes, data from buffer memory
All header fields are generated from the internal Disk Data
Controller registers. The data field of the Data Segment is
obtained from buffer memory. This allows formatting with
real data in the data field. This mode is usually used for an
interleave of one. The multi-sector operation capability of
the DP8496/7 increments the Sector Number register which
is written in each sector.
Format Type C
Header from buffer memory, internal data pattern.
The header bytes in the Header Segment are obtained from
buffer memory. All data fields are from the internal registers.
The header bytes should be arranged contiguously in buffer
memory. If the number of header bytes in each sector is
odd, an extra dummy byte must be appended to the end of
each sector’s header in buffer memory. This approach is
ideal for sector interleaving of greater than one and offers
the minimum of processor intervention during the formatting
process.
Compare Header/Verify Data
Headers are compared as defined in the Header Byte Control and Pattern registers. Header CRC/ECC is optionally
checked. When a matching header is found, a byte for byte
comparison between the data field and buffer memory will
be performed. Data CRC/ECC will be optionally checked.
Buffer memory parity is not checked during this operation.
Compare Header/Write Data
This is the ‘‘normal’’ write data command. Headers are
compared as defined in the Header Byte Control and Pattern registers. Header CRC/ECC is optionally checked.
When a matching header is found, data will be written from
the buffer memory to the data field.
Format Type D
Header from buffer memory, data from buffer memory.
The header bytes in the Header Segment and the data field
in the Data Segment are obtained from buffer memory. The
buffer memory should be arranged with the header bytes for
one sector followed by its data field. This should be repeated for each sector to be formatted. For the case where
word-wide memory configuration is employed, if the number
of header bytes in each sector is odd, an extra dummy byte
must be appended to the end of each sector’s header in
buffer memory.
Compare Header/Read Data
This is the ‘‘normal’’ read data command. Headers are compared as defined in the Header Byte Control and Pattern
registers. Header CRC/ECC is optionally checked. When a
matching header is found, the data field will be read from
the disk and transferred to buffer memory. Data CRC/ECC
will be optionally checked.
Read Header/Ignore Data
This is a ‘‘normal’’ read header command. After the first
Header Sync is found, the Header Bytes are transferred to
buffer memory. If the number of Header Bytes read is odd,
an extra dummy byte will be transferred to buffer memory
after each sector. Header CRC/ECC is optionally checked.
The data field CRC/ECC may also be optionally checked.
Write Header/Ignore Data
This operation will work properly only with hard sectored
drives. The Post Sector/Index Gap will be counted through
or written (based on the WPSIG (Write Post/Sector Index
Gap) bit in the Disk Control register). The Header Gap,
Header Preamble, Synch fields, Header fields, CRC/ECC,
and Header Postamble will all be written. The data field
CRC/ECC will be optionally checked.
Write Header/Write Data
This operation will work properly only with hard sectored
drives. All Header fields and Data fields will be written to the
disk if they are enabled EXCEPT Gap3 . Gap3 will not be
written with this operation.
Read Header/Read Data
Header and data bytes are transferred to buffer memory.
After the first Header Synch is found, the Header Bytes are
transferred to buffer memory. If the number of Header Bytes
read is odd, an extra dummy byte will be transferred to buffer memory after each sector. Header CRC/ECC is optionally checked.
Format Type A
Internal header bytes, internal data pattern.
This is the most simple type of format. All fields in the header and data segments are generated from the internal Disk
Data Controller registers. This mode is usually used for an
interleave of one. The multi-sector operation capability of
the DP8496/7 increments the Sector Number register which
is written in each sector.
Write Unformatted
This command will write a Data Segment only (although the
Data Segment registers may be programmed to appear as a
Header Segment). No header fields will be read, counted
through, or written. After the specified starting condition, the
Post Sector/Index Gap (if enabled), Data Preamble, Data
Synch, Data Field, CRC/ECC (if enabled) and Data Postamble are written to the disk.
24
4.0 Functional Description (Continued)
1
This command may be used for many special applications
such as:
1. Writing data fields that are split by servo information.
2. Writing to non-hard sectored Header Segments.
3. Headerless sectoring.
This command could be useful when used with the LOI
(Load On Index) bit in the Timer Prescaler register. Write
Unformatted can be ‘‘dropped’’ accurately on the track using the time from Index as a count.
0
HSS: Hard or Soft Sectored
Works in conjunction with the DT (Drive Type) bit to control
the AME and the AMF/Sector pins. In general, this bit
should be set for hard or pseudo-hard sectored drives and
cleared for soft sectored drives. Refer to Table 4.17 for the
specific use of this bit.
Read Unformatted
This command will read a Data Segment only (although the
Data Segment registers may be programmed to appear as a
Header Segment). No header fields will be read, counted
through, or written. After the specified starting condition, a
Data Synch will be searched for. After Data Synch is found,
the data field will be read and transferred to buffer memory.
Data CRC/ECC will be optionally checked.
This command could be useful when used with the LOI
(Load On Index) bit in the Timer Prescale register. Read
Unformatted can be ‘‘dropped’’ accurately on the track using the time from Index as a count. It is also useful for reading split data fields due to embedded servo.
DT: Drive Type
Works in conjunction with the HSS (Hard/Soft Sectored) bit
to control the AME and the AMF/Sector pins. In general,
this bit should be set for pseudo-hard sectored drives. Refer
to Table 4.17 for the specific use of this bit.
WPSIG: Write Post Sector/Index Gap
This bit is only used in Write Header or Format operations.
1 WGATE is asserted during the Post Sector/Index Gap
when a header is written.
0 WGATE is not asserted during the Post Sector/Index
Gap.
Start Correction Cycle
This command will start the internal ECC correction cycle.
When the correction cycle has completed, A Disk Completion Interrupt will be generated. If the correction cycle was
successful, no error will be indicated. If unsuccessful, a Disk
Complete Interrupt will occur with an error indication. The
Disk Status register will report the Correction Failed condition. This operation, like all other operations, may be started
any time another disk operation is not in progress. Refer to
Section 4.3.4 for a discussion of the procedure for correcting the data.
Disk Control (DCTL)
7
6
5
IR
I/S
HSS
11h
4
3
2
1
DT
WPSIG
WGF
PLL
An interrupt will be generated as determined by Table
4.17. In general this will generate an interrupt at each
Index pulse for soft sectored drives, each Index or AMF
signal for pseudo-hard sectored drives, or at each Index or Sector pulse for hard sectored drives.
An interrupt will be generated by each Index pulse.
WGF: Write Gate Format
This bit will allow a single revolution format operation with
an RLL code. In some applications, if WGATE pulses low for
2 bit times, an external ENDEC will generate a valid preamble pattern. This bit will only effect the operation of WGATE
during the Data Gap field.
1 WGATE is deasserted for 2 bit times at the beginning of
the Data Gap field of each sector written.
0 WGATE remains asserted throughout the Data Gap
field.
R/W
0
PLL: PLL Recovery Time
This bit is only used in soft sectored mode. It is used to
control the amount of PLL relock time given to an external
synchronizer. When used with the National’s DP8459 or
DP8469 the two byte choice gives the shortest necessary
preamble lengths and takes advantage of the zero phase
start of these PLLs. When using other PLLs, the six byte
choice should be used. See Section 4.3.4.
1 RGATE is deasserted for six byte times following a
synch mis-compare in Soft Sector mode.
0 RGATE is deasserted for two byte times following a
synch mis-compare in Soft Sector mode.
RES
IR: Interlock Required
1 The Interlock register (Header Byte Counter register)
must be written to after the Disk Data Controller has
completed a header operation and before the beginning of the Data Postamble field. If the Interlock register
is not written to in time, the command will terminate
with the LI (Late Interlock) bit in the Disk Status register
set to ‘‘1’’. This allows the safe updating of header formatting registers during a Format operation. Normally
used with the H/NC (Header Complete or Next Command) bit in the Disk Interrupt Enable register set to ‘‘1’’
to enable Header Complete interrupts. See Section
4.3.4 for a description of this technique.
0 Normal operation. No interlock function.
RES: Reset
1 Reset Active: Setting this bit initiates a Disk Data Controller reset operation. Any disk command which may
be in progress will be terminated after transferring up to
four additional bytes. See Table 3.1 for a summary of
registers affected by this reset.
This bit will also be set by asserting the CRST pin.
Once this bit is set (even by CRST), the only method to
clear it is by writing a zero to this bit.
0 Normal Operation: Clearing this bit terminates the reset
condition. The processor must wait a minimum of 17
RCLK cycles after reset has begun before clearing reset.
I/S: Index/Sector/AMF Interrupt
This bit controls the generation of an interrupt at each index,
sector, or AMF. This interrupt may be masked by the I/S
(Index/Sector) bit in the Disk Interrupt Enable register.
25
4.0 Functional Description (Continued)
Disk Status (DSTAT)
7
6
5
LI
DL
SOR
14h
4
3
EC
2
1
011 Data Synch Failed: If a sector or index pulse occurs
while the DP8496/7 is waiting to byte align on the first
data synch field (Synch Ý1, or Synch Ý2 if Synch Ý1
is disabled). This error will occur at the sector or index
pulse.
This error code is also set if the DP8496/7 byte aligns
to the first synch byte of the data field but does not
match to subsequent bytes (more bytes of Synch Ý1
or Synch Ý2). This error will occur at the synch byte
that does not match.
100 Correction Failed: Only possible after a Start Correction Cycle command. This indicates that the error cannot be corrected by the DP8496/7.
101 Data Field Error:: Occurs when a CRC/ECC error is
detected during an Ignore, Verify, or Read Data Field
operation if the Data CRC/ECC verification is enabled
in the ECC Control register.
110 Read Header Fault: Occurs only when a CRC/ECC
error is detected during a Read Header operation if
Header CRC/ECC verification is enabled by the ECC
Control register. CRC/ECC faults during an Ignore or
Compare Header operations are reported in the
Status Code field as a Header Fault. This error occurs
at the end of the header field.
111 Disk Parity Error: Occurs if a parity error is detected
in the buffer memory during any disk write operation if
buffer parity is enabled in the Setup 2 register. The
data will be written to the disk normally as if there
were no error. This status register will be updated immediately. The current command will terminate after
the current sector is completely written to the disk.
R Only
0
SC
These codes allow processor monitoring of the progress of
disk commands and a check on completion. Some of the
information here is redundant to information provided in the
Disk Interrupt register.
LI: Late Interlock
Will only occur if the IR (Interlock Required) bit in the Disk
Control register is set to ‘‘1’’. The processor has failed to
write to the Interlock register (Header Byte Count register)
before the end of the data field of the present sector. The
current command will terminate at the end of the data field.
This status register will be updated immediately.
This error will only occur due to system design errors.
This error could occur in combination with the Data Field
Error, Sector Overrun Error, or the Data Lost Error.
DL: Data Lost
This error will occur during a disk transfer if the FIFO overflows or underflows as data is transferred to or from buffer
memory. The data transferred to or from the disk during this
condition is invalid. The status register will be updated immediately. The current command will terminate after the
current sector is completely written to the disk. The command will terminate immediately for a non-write operation.
This error will only occur due to system design errors. For
example, when not enough memory bandwidth has been
allocated for the disk data rate. Since disk transfers have
the highest priority, this should be a rare occurrence indeed!
This error could occur in combination with the Sector Overrun Error, Late Interlock Error, Data Field Error, Header
Fault Error, or the Disk Parity Error.
SC: Status Codes
00 Idle: The Disk Data Controller is not executing a command, or has started executing a command but is waiting for the starting condition.
01 Header In Progress: The sequencer in the Disk Data
Controller is in the Header section of the flowchart in
Figure 4.4 . The disk sequencer is searching for a
match on a compare header operation, reading a header in a read header operation, or writing a header in a
write header or format operation.
If a ‘‘Sector Not Found’’ or ‘‘Header Failed Although
Sector Matched’’ error occurs, this Status Code will remain frozen until the next Disk Command or Reset.
10 Header Fault: While the Disk Data Controller is searching for the requested header during any header operation, all headers read will be scanned for CRC/ECC
errors if enabled. This Header Fault code (10) will be
set if a CRC/ECC error is detected in any header field
encountered. However, if the header being sought is
found and has no CRC/ECC error, the Header Fault
code is changed to Data Operation In Progress (11). If
the header being sought is not found and a Header
Fault occurred, the Header Fault code (10) will remain
frozen until the next Disk Command or Reset. This
code could provide useful diagnostic information if a
Sector Not Found error occurs.
SOR: Sector Overrun
This error will occur when a sector or index pulse is detected while reading or writing in the middle of a sector. Specifically, for soft sectored drives, if RGATE is active, the first
header synch byte has already been found, and a sector or
index pulse is received, this error will be set. For hard sectored drives, the same conditions apply except no synch
bytes need to be found.
Also, if WGATE is active, this error will occur when a sector
or index pulse is received, except in the case of a soft sectored Format command while GAP 3 is being written.
This error could occur in combination with the Data Lost
Error, Late Interlock Error, Data Field Error, Header Fault
Error, or the Disk Parity Error.
EC: Error Code
000 No Encoded Error.
001 Header Failed Although Sector Matched: At least
one of the header bytes marked with the SM (Sector
Matched) bit in the corresponding Header Control register(s) matched correctly, but other header bytes
were in error. This error will occur at the end of the
header field who’s sector number matched.
010 Sector Not Found: When the desired header cannot
be found before two consecutive index pulses in any
non-write Header operation. This error will occur at the
second index pulse.
26
4.0 Functional Description (Continued)
11 Data Operation In Progress: The sequencer in the
Disk Data Controller is in the Data section of the flowchart in Figure 4.5 . In compare header operation, the
header has been found. In a read header operation, a
header has been read. In a write header operation a
header has been written. This state is entered at the
end of the Data Synch field which coincides with Header Complete Interrupt if enabled. This is at the point at
which the Sector Byte Count registers are down-loaded.
OR
Correction Cycle In Progress: Set when the Start
Correction Cycle command is written to the Disk Command register and remains until the Disk Complete Interrupt.
Disk Interrupt (DINT)
7
6
5
TI
E/R
I/S
12h
4
3
2
GCI
SCI
H/NC
GCI: Group Complete Interrupt
This indicates that the Sector Count register can be reloaded. This occurs at the end of the data field of the last sector
in a multi-sector command. This interrupt will only be set if
the Sector Count register has been written to since the last
down-load of the Sector Count register.
SCI: Sector Complete Interrupt
This interrupt is set after the last byte of a sector has been
read and the CRC/ECC was good. This interrupt will not be
set at the end of the last sector of a multi-sector operation.
H/NC: Header Complete or Next Command
This is a dual function interrupt defined by HC bit in the Disk
Interrupt Enable register.
When HC bit in the Disk Interrupt Enable register is ‘‘1’’
(Header Complete), an interrupt will be generated at the end
of the ID Segment. This is also where the Sector Byte Count
registers are down-loaded. This will indicate when a new
sector byte count or other format register can be loaded.
This feature could be used to alter the length of sectors on
the same track. See Section 4.3.4.
When HC in the Disk Interrupt Enable register is ‘‘0’’ (Next
Command), an interrupt will be generated, as soon as a
pending disk command is down-loaded. This will indicate
when a new command may be written to the Disk Command
register for disk command pipelining. See Section 4.3.4.
R/W
0
1
CCI
This register indicates interrupts that have occurred. There
are two types of interrupts, checkpoint and completion.
Checkpoint interrupts (bits 2–7) are separated into separate
bits and may be asserted simultaneously. These interrupts
may be cleared only by writing a ‘’1’’ into the corresponding
bit location. These bits may be cleared individually or simultaneously.
Completion interrupts are encoded into bits 0–1. Only one
completion interrupt may be read at time. Completion interrupts may be cleared only by writing the same two bit pattern back to the Disk Interrupt register. A completion interrupt may be cleared simultaneously with other checkpoint
interrupts.
Interrupts will be reported in the Disk Interrupt register always, independent of the Disk Interrupt Enable register.
If enabled interrupts remain after writing to the Disk Interrupt
register, the DINT pin will deassert momentarily to retrigger
an edge sensitive interrupt input to the processor.
This register is latched to prevent the contents from changing while reading the register.
CCI: Command Completion Interrupt
One of these three codes will be obtained upon completing
a disk command. They are cleared by writing the exact code
pattern back to this register. For instance, a ‘‘11’’ will not
clear a ‘‘01’’ coded interrupt.
00 No Interrupt.
01 No Error: Disk command complete. Set upon completion of any buffer memory activity and reads or writes to
the drive. Also set upon the successful completion of a
correcton cycle.
10 Error: Any error which aborts a disk command other
than a Verify command. The specific error condition is
coded in the Disk Status register.
11 Verify Data Error: If a data mis-compare occurs during
a Verify Data command, this code will result. If this error
occurs during a multi-sector operation, the rest of multisector operation will not complete. Other errors may
occur at the same time as this error and will be reported
appropriately in the Disk Status register.
TI: Timer Interrupt
This bit is set when the Timer Count register has reached
zero. The Timer description can be found in Section 4.5.
E/R: Buffer Memory Data Error/Ready
If the BMD (Buffer Memory Data) bit in the Disk Interrupt
Enable register is set to ‘‘0’’, the E/R bit indicates that the
Buffer Memory Data register is ready to be read or written.
Also, the Processor Pointer registers may be modified.
If the BMD bit in the Disk Interrupt Enable register is set to
‘‘1’’, this bit indicates that an interrupt error has occurred.
The processor has attempted to read or write to the Buffer
Memory Data register too quickly. The read or write was
unsuccessful. See Section 4.2.8.
Disk Interrupt Enable (DINTE)
13h
7
6
5
4
3
TI
BMD
I/S
GCI
SCI
2
1
H/NC
HC
R/W
0
CCI
TI, I/S, GCI, SCI, H/NC
Setting these bits to a ‘‘1’’ will individually enable the corresponding interrupt as described in the Disk Interrupt register. When set to a zero, the interrupt functions are still active
in the Disk Interrupt register, but will not assert the interrupt
pin. Interrupts should be cleared from the Disk Interrupt register whether or not they are enabled in the Disk Interrupt
Enable register.
I/S: Index or Sector
This interrupt is generated according to Table 4.17 if the I/S
(Index/Sector) bit in the Disk Control register is set to one. If
the I/S it is set to zero, this interrupt will be generated at
each Index pulse.
27
4.0 Functional Description (Continued)
BMD: Buffer Memory Data
1
Sector Count (SC)
7
6
Enables the Buffer Memory Data register Error interrupt
function. Interrupt will occur when the processor has
read or written the Buffer Memory Data register too
quickly.
Enables Buffer Memory Data Ready function (no interrupt). The processor can poll the Disk Interrupt register
to see if enough time has passed to read or write to the
Buffer Memory Data register.
0
15h
3
2
1
R/W
0
Pattern
The contents of this register will replace any header byte
where the SN (Sector Number Substitution) bit is set to ‘‘1’’
in the corresponding Header Byte Control register. This register will be incremented after each sector operation is completed during a multi-sector command. This register should
be initialized with the desired starting sector number for a
multi-sector command.
This allows sequential logical sectors to be accessed during
multi-sector commands without processor intervention.
Header Byte Count/Interlock (HBC) 16h
7
6
5
4
3
x
x
x
x
x
2
1
3
2
1
R/W
0
Sets the number of sectors to operate on in a multi-sector
operation.
Sector Count is really two registers called Active and Holding. The holding register is down-loaded by the Disk Data
Controller to the active register at the same time a new
command is down-loaded from the Disk Command register
(just before the new command begins).
The Sector Count register is also down-loaded following the
last sector operation which decrements the active Sector
Count to zero. This down-load will only occur if a new value
has been written to the Sector Count register after a previous down-load and before the zero detection. If this download occurs, the current disk command will continue with
this new Sector Count.
The processor may access either active or holding register
through this address as controlled by the A/H (Active/Holding) bit in the SCSI Operation register. Normally,
the processor should only access the holding register.
CCI: Command Complete Interrupt
1
Enables all disk Command Complete Interrupts (CCI
field in the Disk Interrupt register).
0
Disables all disk Command Complete Interrupts.
4
4
Count
HC: Header Complete
1
Defines the H/NC interrupt as ‘‘Header Complete’’.
0
Defines the H/NC interrupt as ‘‘Next Command’’.
Sector Number (SN)
7
6
5
17h
5
ECC Control (EC)
7
6
PRE
DE
07h
5
HE
4
3
2
1
R/W
0
Span
PRE: Preset Control
0 A zeros preset is used for all CRC/ECC calculations.
1 A ones preset is used for all CRC/ECC calculations.
This is the normal setting for most applications.
DE: Data Error Check Enable
This bit controls the verification of the CRC/ECC in the Data
Segment during a Read, Ignore, or Verify Data operation.
1 Data field CRC/ECC is verified.
0 Data field CRC/ECC is not verified.
R/W
0
Count
Count is the number of header bytes in the Header Segment. The Count field must be loaded by the processor only
for a Format C or Format D command. This count will allow
the Disk Data Controller to request the proper number of
header bytes from the buffer memory. Allowable values for
count are 3 to 6. An even number of header bytes will always be transferred to and from buffer memory even though
only the specified count will be used in the header.
This register is also used in interlock mode to signal completion to the Disk Data Controller if registers were updated
by the processor between sectors. If the processor fails to
write to this register before the end of the sector when the
interlock mode is enabled, a Late Interlock error will be generated and the disk operation will be aborted.
HE: Header Error Check Enable
This bit controls the verification of the CRC/ECC in the
Header Segment during an Ignore, Compare or Read Header operation.
1 Header field CRC/ECC is verified.
0 Header field CRC/ECC is not verified.
SPAN
Span bits 0 – 4 are set for the longest burst error that is to be
corrected. Allowable values are 3 to 28, depending on polynomial. Values under 3 will default to 3. Refer to Table 4.18
in Section 4.3.4 for recommended values for SPAN. This
value is only used during the Start Correction Cycle command.
28
4.0 Functional Description (Continued)
TABLE 4.13. ECC Byte Ordering
When Read
Syndrom Bytes (Most Significant to Least Significant)
(Individual Bits Are in Reverse Order)
Most Significant Byte
Least Significant Byte
CRC/ECC
Selected
No
Correction Cycle
After Successful
Correction Cycle
16-Bit
ESR6
ESR0
32-Bit
ESR6
ESR5
ESR1
ESR0
48-Bit
ESR6
ESR5
ESR4
ESR2
ESR1
ESR0
56-Bit
ESR6
ESR5
ESR4
ESR3
ESR2
ESR1
ESR0
Correction Mask (Read in Order)
(Individual Bits Are in Correct Order)
3rd
4th
5th
1st
2nd
32-Bit
ESR1
ESR5
ESR6
48-Bit
ESR1
ESR2
ESR4
ESR5
56-Bit
ESR1
ESR2
ESR3
ESR4
ESR5
Note: ESRx e ECC Shift Register x.
ECC Shift Register (ESRx)
7
6
5
00–06h
4
3
2
1
4.14 and Table 4.15 to relate each pattern and control register to the actual disk format.
These registers should be loaded during the initialization
process for the particular format chosen. Reset does not
effect these registers.
Be sure to observe the restrictions on count fields being set
to zero as specified in the individual register descriptions.
There are some count fields which may not be zero. All
fields listed in Table 4.14 and 4.15 are programmable.
RGATE and WGATE are asserted during certain fields. This
is shown in Figure 4.9 and Figure 4.10 at the end of the
section on Disk Operations (4.3.4).
R Only
0
Pattern
The shift register will yield different types of information depending on when it is read. If read after a successful disk
data read or write operation, they contain zeros.
Reading the shift register after an CRC/ECC error produces
the syndrome bytes for that particular error. This syndrome
may be stored and compared with syndromes from multiple
reads to verify a hard error condition.
An error correction command may be issued to attempt to
correct the data read from the disk. When the ECC Shift
Register is read after a successful correction cycle it contains a correction mask which can be used to correct the
error contained in Buffer Memory. See Section 4.3 for a
discussion of the procedure for error correction.
ECC Byte CountÐLow (ECCL)
18h
7
6
5
4
3
2
1
R Only
0
2
1
R Only
0
Post Sector/Index Count (PSIG) 20h
7
6
5
4
3
1
R/W
0
Count
Sets the number of bytes in the Post Sector/Index field. The
pattern used is from the Gap 3 Pattern register. WGATE will
be asserted during this field according to the WPSIG (Write
Post Sector/Index Gap) bit in the Disk Control register. This
field can be used to skip sectored or wedge servo fields.
RGATE is never asserted during this field for hard and pseudo-hard sectored drives. The Post Sector/Index field is not
used if the DCOZ (Drive Command On Zero) bit is set in the
Timer Prescale register).
For soft-sectored drives, PSIG is only used before the first
sector. For other sectors, PSIG is assumed to have a length
of one. Range is 1 – FF.
Byte Count (Low)
ECC Byte CountÐHigh (ECCH) 19h
7
6
5
4
3
2
Byte Count (High)
These two registers contain an offset byte count after a
correction cycle, which when added to the address of the
start of the offending sector, identifies the location of the
error. These registers only contain valid information after
the end of a correction cycle.
Header Preamble Count (HPC)
21h
7
6
5
4
3
4.3.3 Format Registers
The disk format is defined by using the format pattern and
control registers. Generally these registers are set up in
pairs: a pattern register and a control register. In each pair,
the pattern register is loaded with an appropriate 8-bit pattern that will be written to the disk during a Format or Write
operation, or will be used during a Read or Compare operation for byte alignment or a comparison in locating a sector.
The control register will generally determine how many
times the pattern is repeated, or the count. Refer to Table
Header Gap
2
1
R/W
0
Header Preamble
Header Gap
Sets the number of bytes in the Header Gap field. The pattern used is from the Preamble Pattern register. Used to
separate the time between RGATE and WGATE assertions.
WGATE is always asserted during this field for a write operation. RGATE is never asserted during this field for hard and
pseudo-hard sectored drives. Range is 1 – 7.
29
4.0 Functional Description (Continued)
TABLE 4.14. Register Addresses for Format of ID Segment (Hex)
Pattern
PSIG
Hdr
Gap
Hdr
Prmbl
Snyc1
Snyc2
Hdr0
Hdr1
Hdr2
Hdr3
Hdr4
Hdr5
CRC/
ECC
Hdr
Post
3A
22
22
24
25
27
28
2A
2B
2D
2E
auto
30
Control
20
21
21
23
23
26
26
29
29
2C
2C
2F
2F
Range
1 – FF
1–7
0–1F
0–7
0–1F
0–1
0–1
0–1
0–1
0–1
0–1
0–7
3 – 1F
TABLE 4.15. Register Addresses for Format of Data Segment (Hex)
Data
Gap
Data
Prmbl
Sync1
Sync2
Data Field
CRC/
ECC
Data
Post
Gap3
Pattern
22
22
33
34
37
auto
30
3A
Control
31
31
32
32
35, 36
38
38
39
Range
1–7
0 –1F
0–7
0–1F
10 – FFFE (Must Be Even)
0–7
1 – 1F
1 – FF
Header Preamble
Sets the number of bytes in the Header Preamble for the
PLL to lock to. The pattern used is from the Preamble Pattern register. Range is 0–1F.
Preamble Pattern (PREP)
7
6
5
22h
4
3
2
1
Header Byte 0,1 Control (HC01) 26h
7
6
5
4
3
NC0
Pattern used for the Header Gap, Header Preamble, Data
Gap, and Data Preamble fields. This pattern must be all
zeros for a soft-sectored drive.
Synch Ý1
2
1
Sets the number of bytes in the Synch Ý1 and Synch Ý2
fields. The range of Synch Ý1 is 0–7. The range of Synch
Ý2 is 0 – 1F. Both Synch Ý1 and Synch Ý2 cannot be zero.
At least one must be non-zero.
For soft sectored drives, AME is generated while writing the
Synch Ý1 field to the disk. AMF is expected while reading
Synch Ý1 from the disk.
2
1
R/W
0
1
SM1
SN1
HB1
SN: Sector Number Substitution
1 The contents of the Sector Number register are substituted for this Header Byte pattern during a Write Header operation and compared during a Compare Header
operation. This is normally used in multi-sector commands.
0 The contents of the Header Byte Pattern register is written to the disk for a Write Header operation and compared during a Compare Header operation.
Pattern used for the Synch Ý1 field. Can be used as an
Address Mark in soft sectored formats.
2
NC1
SM: Header Failed Although Sector Matched Enable
1 The Header Failed Although Sector Matched function is
enabled for this header byte. While searching for a
matching Header Segment, if this field matches while
other header bytes don’t, an error is reported in the
Disk Status register. Typically, this would be the sector
number field. See Section 4.3.4 for additional information.
0 The Header Failed Although Sector Matched function is
not enabled for this header byte.
Pattern
Header Synch Ý2 Pattern (HSP2) 25h
7
6
5
4
3
HB0
R/W
0
NC: Not Compared
1 This header byte will always match true for a Compare
Header operation.
0 This header byte will be compared normally without
modification of the outcome.
R/W
0
Synch Ý2
Header Synch Ý1 Pattern (HSP1) 24h
7
6
5
4
3
SN0
1
This register does not perform any pattern repetition, nor
does it define a field size. It is provided to control the function of each corresponding header byte. There is one Header Byte Control register for each pair of the six Header Byte
Pattern registers. The upper four bits correspond to one pattern register. The lower four bits correspond to another pattern register.
R/W
0
Pattern
Header Sync Ý1, Ý2 Count (HSC) 23h
7
6
5
4
3
SM0
2
R/W
0
Pattern
Pattern used for the Synch Ý2 field.
30
4.0 Functional Description (Continued)
HB: Header Byte Active
Header Byte 5 Pattern (HP5)
7
6
5
1
This header byte contains valid data and will be used in
the header operation.
These seven locations (00–06h) contain the CRC/ECC shift
register contents. The register ordering is determined by the
CRC/ECC polynomial selected and whether a correction cycle has taken place. This ordering is shown in Table 4.13.
0 This header byte is not included in the header byte field
and will not be used in the header operation. The SN,
SM, and NC bits for this header byte must also be set to
zero. Only the upper 4 bits or the lower 4 bits may be
set to all zero for each Header Byte Control register,
not both . This implies that the minimum number of
Header Bytes is three, and the maximum is six.
Header Byte 0 Pattern (HP0)
7
6
5
27h
4
3
2
1
2
1
R/W
0
2
1
R/W
0
Pattern used for Header Byte 5.
Header ECC and Postamble Count (HECC) 2Fh
7
6
5
4
3
CRC/ECC
Postamble Count
CRC/ECC
000: No CRC/ECC
010: 16-Bit CRC
100: 32-Bit ECC
110: 48-Bit ECC
111: 56-Bit ECC
R/W
0
Postamble Count
Number of bytes in the Header Postamble. Range is 3 – 1F.
Pattern used for Header Byte 0.
28h
4
3
Pattern
Pattern
Header Byte 1 Pattern (HP1)
7
6
5
2Eh
4
3
2
1
Postamble Pattern (POSTP)
7
6
5
R/W
0
30h
4
3
2
1
R/W
0
Pattern
Pattern
Pattern used for Header Postamble field and Data Postamble field.
Pattern used for Header Byte 1.
Header Byte 2,3 Control (HC23) 29h
7
6
5
4
3
NC2
SM2
SN2
HB2
NC3
2
1
SM3
SN3
R/W
0
Data Preamble Count (DPC)
7
6
5
HB3
2Ah
4
3
2
1
R/W
0
3
2
1
R/W
0
2
1
R/W
0
SM5
SN5
Pattern used for Header Byte 2.
2Bh
4
Pattern used for Header Byte 3.
NC4
SM4
SN4
HB4
NC5
Data Synch Ý1, Ý2 Count (DSC) 32h
7
6
5
4
3
Synch Ý1
HB5
2Dh
4
3
2
1
R/W
0
Data Preamble
2
1
R/W
0
Synch Ý2
Sets the number of bytes in the Synch Ý1 and Synch Ý2
fields. The range of Synch Ý1 is 0 – 7. The range of Synch
Ý2 is 0 –1F. Both Synch Ý1 and Synch Ý2 cannot be zero.
At least one must be non-zero.
For soft sectored drives, AME is generated while writing the
Synch Ý1 field to the disk. AMF is expected while reading
Synch Ý1 from the disk.
Same as Header Byte 0,1 Control but for Header Bytes 4
and 5.
Header Byte 4 Pattern (HP4)
7
6
5
1
Data Preamble
Sets the number of bytes in the Data Preamble for the PLL
to lock to. The pattern used is from the Preamble Pattern
register. Range is 0 – 1F.
Pattern
Header Byte 4,5 Control (HC45) 2Ch
7
6
5
4
3
2
Data Gap
Sets the number of bytes in the Data Gap field. The pattern
used is from the Preamble Pattern register. Used to separate the time between RGATE and WGATE assertions.
WGATE is always asserted during this field for a write operation. RGATE is never asserted during this field for hard and
pseudo-hard sectored drives. Range is 1 – 7.
Pattern
Header Byte 3 Pattern (HP3)
7
6
5
3
Data Gap
Same as Header Byte 0,1 Control but for Header Bytes 2
and 3.
Header Byte 2 Pattern (HP2)
7
6
5
31h
4
R/W
0
Pattern
Pattern used for Header Byte 4.
31
4.0 Functional Description (Continued)
Data Synch Ý1 Pattern (DSP1)
33h
7
6
5
4
3
2
1
R/W
0
Gap 3 Byte Count (GAPC)
7
6
5
39h
4
Pattern
3
2
1
R/W
0
Count
Pattern used for the Synch Ý1 field. Can be used as an
Address Mark in soft sectored formats.
Data Synch Ý2 Pattern (DSP2)
34h
7
6
5
4
3
2
1
R/W
0
2
1
R/W
0
Number of bytes in Gap 3. In hard sectored mode, this
count must be programmed small enough to time-out before
the next Index or Sector pulse.
During a soft-sectored Format operation, Gap 3 functions
normally except after the last sector. After the last sector,
the Gap 3 pattern is repeated until the Index pulse is received. The Gap 3 Byte Count is ignored after the last sector.
Pattern
Pattern used for the Synch Ý2 field.
Sector Byte CountÐLow (SBCL) 35h
7
6
5
4
3
Gap 3 Pattern/PSIG (GAPP)
7
6
5
Count
3Ah
4
3
2
1
R/W
0
Pattern
Sector Byte CountÐHigh (SBCH) 36h
7
6
5
4
3
2
1
R/W
0
Pattern used in Gap 3 field and Post Sector/Index Gap field.
Count
These two registers set the number of data bytes in a sector. The range of the combined value for the Byte Count is
10h – FFFEh. This must be an even number.
Data Format Pattern (DFP)
7
6
5
37h
4
3
2
1
R/W
0
Pattern
Pattern used for the data field during a Format Type A or C.
For Format Types B and C, the data field will be transferred
from buffer memory.
Data ECC and Postamble Count (DECC) 38h
7
6
5
4
3
CRC/ECC
2
1
R/W
0
Postamble Count
CRC/ECC
000: No CRC/ECC
010: 16-Bit CRC
100: 32-Bit ECC
110: 48-Bit ECC
111: 56-Bit ECC
Postamble Count
Number of bytes in the Data Postamble. Range is 1–1F.
32
4.0 Functional Description (Continued)
The flowchart in Figure 4.6 describes how to implement simple commands such as reading data from the disk, writing
data to the disk, and formatting the disk. This flowchart has
been kept quite simple. More advanced techniques may be
used to streamline performance, such as pipelining commands. Some of these techniques are described later.
4.3.4 Disk Operations
The DP8496/7 has been designed to offer a great deal of
flexibility in many areas. The disk formats that may be used
are quite varied. The procedures that the processor may
use to access information on the disk are varied as well.
TL/F/11212 – 13
FIGURE 4.6. Software Flowchart for Simple Disk Operations
33
4.0 Functional Description (Continued)
interrupt occurs, a new disk command may be written to the
Disk command register. This new disk command will be
downloaded and executed after the current disk command
has completed all of its operations. If the current disk command is multi-sectored, the new disk command will not be
downloaded until all the sector operations have completed.
The Sector Count register will also be downloaded at the
same time a disk command is downloaded. If a new value is
not written to this register, the value from the previous
download will be used. A new Disk Pointer will also be
downloaded at the same time as the new disk commandÐ
but only if the Disk Pointer has been written to since the last
disk command download.
Disk commands may be pipelined multiple times if desired.
Simply wait for the ‘‘Next Command’’ interrupt between
each update of the Disk Command register.
Extending the Current Disk Command
Disk commands with the MSO (Multi-Sector Operation) bit
set to ‘‘1’’ in the Disk Command register may be extended
to transfer more sectors than originally programmed. If a
new value is written to the Sector Count register while a disk
command is executing, the new value will be downloaded
from the holding register to the active register when the
active Sector Count reaches zero. A new Disk Pointer will
also be downloaded at the same time if a new value has
been written to the Disk Pointer registers.
A new, pipelined Sector Count may be written to the Sector
Count register any time before the active Sector Count
reaches zero. The Sector Count register should not be updated again (after the first pipelined value) until the previous
pipelined value has been downloaded. The ‘‘Group Complete’’ interrupt indicates when a pipelined Sector Count
and Disk Pointer is downloaded. After this interrupt is received, the Sector Count register and Disk Pointer registers
may be updated again.
A disk command may be extended multiple times if desired.
Simply wait for the ‘‘Group Complete’’ interrupt between
each update of the Sector Count register.
Table 4.16 summarizes these interrupts.
Special Formats
Some applications require tracks that are formatted with
sectors of varying lengths, or sectors with different synch
bytes, or other unique format variations. This requires modifying the header format registers ‘‘on the fly’’ during a format operation. It is important to modify these registers only
while they are not being used by the Disk Data Controller.
The processor can use the Interlock mode and the Header
Complete interrupt to allow modification of the disk format
registers for the ID Segment (addresses 21h, 23h – 2Fh) or
the Sector Byte Count registers (35h, 36h) during the format
operation. The Header Complete interrupt is generated after
the Sector Byte Count registers have been downloaded at
the beginning of the data field. The processor has until the
Data Postamble field to make any necessary changes to the
desired registers and then write to the Interlock (Header
Byte Count) register, telling the DP8496/7 that changes are
complete. If Interlock is not received in time, a Late Interlock
interrupt is generated.
Pipelining Disk Commands
Disk commands may be pipelined. If a new disk command is
written to the Disk Command register while an operation is
executing, the new disk command will be buffered and
downloaded internally when the current command has completed.
To assist in determining when a new disk command may be
written to the Disk Command register for pipelining, an interrupt may be used. If the H/NC (Header Complete/Next
Command) bit in the Disk Interrupt Enable register is set to
‘‘0’’, the ‘‘Next Command’’ interrupt is enabled. When this
TABLE 4.16. Disk Buffer Management Interrupts
Command
Type
Multi-Sector
(MSO e 1)
Sector
Count
Reached
Zero?
New Disk
Command
Been
Loaded?
New
Sector
Count
Been
Loaded?
Interrupt
Type
Comments
No
X
X
SCI
Interim sector completed of a multi-sector operation
Yes
No
Yes
GCI
The last sector of the previous Sector Count has been
transferred to/from the disk. Command is extended with new
Sector Count.
Yes
Yes
X
CCI
Previous disk command has completed. New pipelined
command is starting.
Yes
No
No
CCI
Previous disk command has completed. No new command.
Single
Sector
X
Yes
X
CCI
Single sector command has completed. New pipelined command
is starting.
(MSO e 0)
X
No
X
CCI
Single sector command has completed.
Note: MSO (Multi-Sector Operation) bit is in the Disk Command register. SCI e Sector Complete Interrupt, GCI e Group Complete Interrupt, CCI e Command
Complete Interrupt.
34
4.0 Functional Description (Continued)
The altered format registers will apply to the next sector to
be operated on, regardless of whether the drive command
is a multi-sector type or not.
is expected for each Synch Ý1 byte read. In MFM systems
this is commonly generated by a missing clock from an A1h
byte.
A new drive command may also be loaded along with updating the Format registers. This would facilitate a Read Header/Read Data command (to locate track position) followed
by a Compare Header/Read Data command (multi-sector)
to read the rest of the track. See the previous sub-section
for a description of how commands are pipelined.
Changing disk format registers other than the ones listed
above can have critical timing restrictions. In general, if the
field to be changed for the next sector has been, or is being
written to disk for the current sector, it is safe to change the
count and pattern registers for that field.
The ‘‘Header Complete’’ interrupt is enabled by setting the
H/NC (Header Complete/Next Command) bit in the Disk
Interrupt Enable register to a ‘‘1’’.
In the pseudo-hard sectored mode AME is used to perform
two different functions. During a format, AME is asserted
during the Header Gap field. This, along with the fact that
WGATE was already asserted instructs the drive to write an
Address Mark (the Header Gap Count should be loaded
with a 3 according to the ANSI draft proposed ESDI specification). During a read or write data operation AME is asserted after the specified starting condition for a soft sectored
drive (immediately, wait for index, wait for timer, etc.). The
Disk Data Controller will then wait for AMF to be asserted by
the drive before executing the command. AME will be asserted again before each sector operation, and the Disk
Data Controller will wait for the AMF before continuing.
Table 4.17 shows the proper state for the HSS and DT bits
(Hard/Soft Sectored and Drive Type) in the Disk Control
register based on the drive used.
Single Sector Formatting
Format operations normally start with an index pulse and
end with the next index pulse, thus formatting one track.
Individual sectors can be formatted for hard sectored drives
only, via the Write Header/Write Data command. A Compare Header/Read Data may be issued before the Write
Header/Write Data command to locate the sector preceding the sector to be formatted.
Disk Data Error Handling
The DP8496/7 uses a fixed 32, 48 or 56-bit ECC polynomial
for detection and correction of errors. Correction is performed entirely on-chip with one of the fixed polynomials.
The processor needs only to EX-OR the correction mask
bytes contained on-chip after a successful correction cycle
with the bytes in error in buffer memory.
It is not recommended to use the maximum correction span
listed in Table 4.18. The chance for a mis-correction is too
great. The recommended correction span produces the
best trade-off between correction span and probability of
mis-correction. The correction span is set in the ECC Control register.
AMF/AME Operation
Address mark found (input) and address mark enable (output) pins are usually used with soft sectored drives to synchronize the actual disk with DP8496/7 data operations. On
true soft sectored drives (ST506/412 type), AME is generated for each byte of Synch Ý1 generated during a format or
write operation. On a compare or read operation, AMF
TABLE 4.17. Drive Types
Drive Type
Soft Sectored
ST506 Type
Format
Operation
HSS
DT
PSIG
Written
AME Asserted
during . . .
AMF/Sector
Pulse Expected
Format
0
0
First Sector
of Track
Hdr. Synch Ý1,
Data Synch Ý1
X
Write
0
0
No
Data Synch Ý1
Hdr. Synch Ý1
Read
Pseudo Hard
Sectored
(ESDI Soft
Sectored)
Using Gap Type
Address Mark
Hard Sectored
Not Using Gap
Type Address
Mark
0
Index
Hdr. Synch Ý1,
0
No
Never
Hdr. Gap
X
Data Synch Ý1
Format
0
1
First Sector
of Track
Write
1
1
No
ESDI Handshake
AMF for ESDI
Handshake
Read
1
1
No
ESDI Handshake
AMF for ESDI
Handshake
Format
1
0
All Sectors
Never
Sector
Sector
Format
1
1
All Sectors
Post Sector/Index
Gap
Write
1
0
No
Never
Sector Pulse
Read
1
0
No
Never
Sector Pulse
Note: I/S (Index/Sector) bit is in Disk Control register. If the I/S bit is set to ‘‘0’’, an interrupt is generated at each Index pulse only.
35
If I/S* Set to ‘‘1’’,
Int. Generated:
Index
Index or AMF
Index or Sector
4.0 Functional Description (Continued)
The proper procedure to perform error correction is listed
below:
TABLE 4.18. ECC Correction Span
Polynomial
Maximum
Recommended
16-Bit CRC
32-Bit ECC
48-Bit ECC
56-Bit ECC
None
11 Bits
15 Bits
22 Bits
None
5 Bits
11 Bits
17 Bits
1. Clear Disk Interrupt register.
2. Set the Sector Byte Count registers to the sum of the
original sector length plus the length of the ECC polynomial (4, 6, or 7 bytes).
3. Issue the Start Correction Cycle command.
4. After the command has finished (interrupt generated),
check the Disk Interrupt register. If an error is indicated,
the Correction Failed bit in the Disk Status register will
also be set. The ECC error is not correctable by the
DP8496/7.
5. If no error is reported, the error is correctable. The address of the first byte in buffer memory that must be
corrected is given by the formula below:
[Address of Start of Sector] a [Contents of ECC Byte
Count Register] b 1
If the contents of the ECC Byte Count register is greater
than or equal to the sector length, the error is in the ECC
itself. The data in buffer memory is already correct and
should not be modified.
6. Use the Processor Pointer and Buffer Memory Data register to read the invalid bytes from buffer memory. Use
the data in the ECC Shift Registers specified in Table
4.13. EX-OR the first ECC Shift register byte with the first
buffer memory byte in error. EX-OR the second ECC
Shift register byte with the second buffer memory byte in
error, etc.
7. Write the corrected data back to buffer memory with the
Processor Pointer and the Buffer Memory Data register.
CRC may be used for either Header or Data fields, or both.
This is set in the Data or Header ECC count registers. The
CRC-CCITT polynomial used by the DP8496/7 is given below:
P(x) e x16 a x12 a x5 a 1
The 32-bit code is a public domain, computer generated
code. This is listed below:
P(x) e x32 a x28 a x26 a x19 a x17
a x10 a x6 a x2 a 1
The 48 and 56-bit polynomials must be licensed by National
Semiconductor to users of the DP8496/7 for exclusive use
in products containing the DP8496/7. The codes generated
on disks by these polynomials may be distributed freely, as
when used in floppy or removable hard disk media. There is
no charge for this license.
Operation during Correction
The DP8496/7 can be set to correct an error any time after
an error has been detected and before another command
has been issued. A correction cycle will not take place unless a Start Correction Cycle command has been issued.
By the time an error is reported by the DP8496/7, the data
read from the disk will have been transferred to buffer memory. To correct an error, a Start Correction command must
be issued by the processor. During the execution of this
command, the buffer memory will not be accessed by the
Disk Data Controller, which leaves the buffer memory free
for other non-disk operations. At the end of the correction
command, the processor must use the information provided
by the DP8496/7 to manually correct the data residing in
buffer memory.
The time it takes the Start Correction command to complete
is determined by the error’s location in the sector. The nearer to the start of the sector, the longer the DP8496/7 takes
to locate the error. This time can be determined using the
formula shown in Figure 4.7 . It should be noted that this is
the internal correction time only. More time is required for
the processor to perform additional operations.
Header Diagnostic Operations
The Header Failed Although Sector Matched function is enabled independently for each header byte by the SM (Header Failed Although Sector Matched) bit in the Header Control registers. If a Compare Header/Ignore Data operation is
performed and the total header did not compare but any
byte with the SM bit enabled did compare correctly, an ‘‘Error’’ completion interrupt will be generated. The ‘‘Header
Failed Although Sector Matched’’ status will be reported in
the Disk Status register. The processor could then read the
header bytes from buffer memory.
To find the last header bytes read from the disk in the buffer
memory, the following steps may be taken:
1. Set the A/H (Active/Holding) bit in the SCSI Operation
register to ‘‘1’’.
2. The first byte of the failing header information is located
in buffer memory at the address pointed to by:
[Contents of Disk Pointer Register] b [Number of Bytes
in Each Header]
3. Set the A/H bit in the SCSI Operation register back to
‘‘0’’.
TL/F/11212–14
Correction time &
LbX
f
L e Length of data and ECC fields (bits)
X e Distance from start of sector to first bit in error (bits)
f e Read clock frequency (Hz)
FIGURE 4.7. Time for ECC Correction
36
4.0 Functional Description (Continued)
TL/F/11212 – 15
FIGURE 4.8. RGATE Operation during Search for Synch
frequency while searching for a preamble and address
mark. This algorithm is shown in Figure 4.8 . While reading a
preamble (or what the DP8496/7 thinks is a preamble), the
first non-zero bit clears a bit counter. If the data read for this
bit and the next 7 bits do not match the proper Synch register, RGATE is deasserted. If a Synch match is made, the
read operation continues to its conclusion.
If RGATE is deasserted, it will remain deasserted for either
two or six byte times. This time is determined by the PLL
(PLL Recovery Time) bit in the Disk Control register.
RGATE Timing for Soft Sectors
Ideally RGATE should transition from deasserted to asserted only while the drive’s head is positioned over a preamble
field. Proper programming of the gap lengths in the Format
registers strive for this condition.
However, for soft sectored drives, it is impossible to avoid
asserting RGATE over non-preamble fields as shown in Figure 4.9 .
For soft sectored drives, the DP8496/7, employs a RGATE
algorithm to allow an external PLL to re-lock to the proper
37
4.0 Functional Description (Continued)
RGATE during Header Segment
(Hard and Pseudo-Hard Sectored)
RGATE during Header Segment (Soft Sectored)
TL/F/11212 – 17
TL/F/11212–16
WGATE during Header Segment (Hard Sectored)
WGATE during Header Segment
(Soft and Pseudo-Hard Sectored)
TL/F/11212–18
TL/F/11212 – 19
Note 1: While searching for the Header Preamble, RGATE will transition from asserted to deasserted following the algorithm shown in Figure 4.8.
Note 2: RGATE will become deasserted one byte time after the end of CRC/ECC field.
Note *: PSIG* field for soft sectored drives only appears before the first sector. For other sectors, PSIG* will have a field length of one byte.
FIGURE 4.9. RGATE and WGATE Timing during Header Segment
RGATE during Data Segment (Soft Sectored)
RGATE during Data Segment
(Hard and Pseudo-Hard Sectored)
TL/F/11212 – 21
TL/F/11212–20
WGATE during Data Segment (Hard Sectored)
WGATE during Data Segment
(Soft and Pseudo-Hard Sectored)
TL/F/11212–22
TL/F/11212 – 23
Note 1: RGATE will become deasserted one byte time after the end of CRC/ECC field. Also, if an additional sector is to be read, RGATE will become re-asserted
after a Header Gap time plus two bytes after this point.
FIGURE 4.10. RGATE and WGATE Timing during Data Segment
38
4.0 Functional Description (Continued)
4.4 SCSI INTERFACE
4.4.1 Simple SCSI Operations
Primary among the objectives for the SCSI interface is to
simplify the microprocessor’s interaction with the
DP8496/7. The most complex portion of SCSI related code
is usually dedicated to interrupt handling, whether vectored
or polled. In the DP8496/7, the multi-phase commands and
combination commands help to reduce the number of interrupts and to pre-define the reason for those interrupts which
do occur. An interrupt mask register enables the software to
control interrupt flow. Also, a physically separate pin dedicated for SCSI interrupts (SINT) can shorten the processor
response time. The chip is also capable of operating in either the target mode, for which it is optimized, or the initiator
mode.
The SCSI Port can be controlled by the processor by utilizing the DP8496/7 in one of two different modes: Automatic
or Manual. Automatic mode enables the SCSI Command
register and the state machines which make up the SCSI
protocol engine to interpret and perform multi-phase or
combination commands.
Manual mode disables the SCSI Command register and
SCSI protocol engine. The processor must then manipulate
both the SCSI control and data bus through programmed
I/O. The SCSI Control register and the SCSI Data register
are the windows into the SCSI bus which allow this manipulation.
The DP8496/7 may be switched between the Automatic
and Manual modes of operation. There are a few restrictions, however, and these are detailed in the SCSI Operation register description in Section 4.4.2.
The DP8496 provides 48 mA, single-ended transceivers on
all SCSI bus signal pins; so the bus lines can be driven
directly. The DP8497 version is designed to interface with
differential transceivers external to the chip for Fast SCSI
applications. Therefore, on the DP8497 all SCSI outputs are
the standard, 2 mA type. Fourteen additional pins on the
DP8497 provide direction control for the external differential
transceivers. In the manual mode of operation, the user
controls the transceiver direction by programming two registers, SDIF1 and SDIF2 (4Eh and 4Fh). Also note that on the
DP8497 all SCSI signals are active high.
The flowchart in Figure 4.11 describes how to implement
simple SCSI operations such as reading data from the disk
and writing data to the disk. This flowchart has been kept
quite simple. The next section describes the various registers that the user must program in order to configure and
control the SCSI Controller’s operation. More advanced
SCSI operations are then described in Section 4.4.3.
4.4.2 SCSI Registers
SCSI Command (SCMD)
7
6
5
40h
4
3
2
1
W Only
0
SCSI Command Opcode
While in Automatic mode, writing an opcode to this register
will initiate a SCSI command. Normally other SCSI registers
are configured before writing to this register. After the proper registers are loded, writing to this register will initiate the
operation. Invalid opcodes should NEVER be issued. After
any invalid opcode is written to this register, the DP8496/7
should be reset with the CRST pin.
In general SCSI commands cannot be pipelined. The combination commands eliminate the need for most pipelining.
New commands should be written to the SCSI Command
register only after the previous command has completed its
operation as indicated by its completion interrupt. See Section 4.4.3 for a description of certain commands that may be
pipelined.
SCSI commands which transfer information blocks may be
extended, however. If a new value is written to the SCSI
Block Count register, the command will continue by downloading the new holding register value to the active SCSI
Block Count register when the current active contents
reaches zero. This is described in Section 4.4.3.
39
4.0 Functional Description (Continued)
TL/F/11212 – 24
FIGURE 4.11. Software Flowchart for Simple SCSI Operation
40
4.0 Functional Description (Continued)
The valid SCSI Commands are listed in Table 4.19. Flowcharts of the multi-phase commands are shown in Figures
4.12 to 4.19 at the end of this description of the SCMD
register.
Clear Pending Command (Opcode e 00)
This command will clear the SCSI Command register of any
outstanding or pending commands. This may be used in the
following situations.
1. After a ‘‘Wait for Select’’ command is written to the SCSI
Command register, a Clear Pending Command may be
issued to prevent the ‘‘Wait for Select’’ command from
executing. See Section 4.4.3 for a description of the
proper sequence to clear the ‘‘Wait for Select’’.
2. If a new command is written to the SCSI Command register, and you want to prevent it from executing you may
issue the Clear Pending Command. The following condition may create this situation:
A. If a command is pipelined, the Clear Pending Command may be issued to clear the pipelined command
before it is executed. See 4.4.3 for pipelining restrictions.
B. If a command is pipelined, and the currently executing
command terminates with an error, the pipelined
command will not execute until the interrupt is
cleared. The Clear Pending Command may be issued
to clear the pipelined command before it is executed.
C. If an asynchronous event generates a completion interrupt (SCSI Bus Reset or (Re)Select) just before a
command is written to the SCSI Command register,
the command will be pipelined and will not execute
until the interrupt is cleared. The Clear Pending Command may be issued to clear the pipelined command
before it is executed.
3. Before modifying the T/I (Target/Initiator) bit in the SCSI
Operation register, the Clear Pending Command should
be issued. This will prevent an ‘‘Invalid Command’’ interrupt from being issued if the last command written to the
SCSI Command register is invalid for the new mode set
with the T/I bit. This operation is not required if you are
sure that the last command is not invalid for the new
mode.
4. The Clear Pending Command is used to clear the self
test mode. The Clear Pending Command should be issued after issuing the Disconnect Self Test command.
TABLE 4.19. SCSI Command Opcodes
OpCode
00
2B
32
30
10
08
3E
3F
3A
3B
0A
0B
0E
0F
3C
38
33
31
1F
1B
1C
18
1D
19
35
36
37
90
B2
8B
BF
BC
Command
Mode
Int?
Clear Pending Command
Reset
Disconnect
Message Accepted
(Re)Select
Abort (Re)Select
Receive Command
Receive Data
Receive Message
Receive Unspecified
Send Status
Send Data
Send Message
Send Unspecified
Transfer Info
Transfer Pad
Send Disconnect
Send End
Reselect-Receive Data
Reselect-Send Data
Reselect-Receive DataDisconnect
Reselect-Send DataDisconnect
Reselect-Receive Data-End
Reselect-Send Data-End
Wait for Sel.-Busy-End
Wait for Sel.-Receive
Command
Wait for Sel.-Receive
Command-Disconnect
(Re)Select Self Test
Disconnect Self Test
Send Data Loopback
Receive Data Loopback
Transfer Info Loopback
I/T
I/T
I/T
I
I/T
I/T
T
T
T
T
T
T
T
T
I
I
T
T
T
T
T
N
Y
N
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
T
Y
T
T
T
T
Y
Y
Y
Y
T
Y
I/T
I/T
T
T
I
Y
N
Y
Y
Y
Mode: I e Initiator, T e Target
INT e Interrupt generated at command termination? (Y/N)
41
4.0 Functional Description (Continued)
While in the Initiator mode, the ATN signal may be controlled during selection by the SWA (Select With Attention)
bit in the SCSI Operation register.
Reset (Opcode e 2B)
This command will terminate and clear the current command (if any) and will clear any pipelined command. Refer
to Table 3.1 for other registers that are affected by this command. This command may be issued at any time.
At the completion of this command, the ‘‘Operation Complete’’ interrupt will be generated. This interrupt may be buffered after another interrupt as described in the SCSI Interrupt register.
Abort (Re)Selection (Opcode e 08h)
This command is used to terminate a pending (Re)select
combination command. This command will function correctly only if the initial (re)selection has not yet begun, or while
the (re)selection is in progress.
If this command is issued before or during arbitration, the
SCSI bus is immediately released. This will result in an ‘‘Operation Complete’’ interrupt.
If issued after this point but before the desired SCSI device
has responded with the BSY signal, the SCSI Data Bus will
be released. If the BSY signal is not detected after a certain
delay, the SEL signal will be deasserted and the command
will terminate with an ‘‘Operation Complete’’ interrupt. If the
BSY signal is detected, a ‘‘Last Command Ignored’’ interrupt will be generated and the DP8496/7 will continue as if
the Abort (Re)Selection command had not been issued at
all (except for the ‘‘Last Command Ignored’’ interrupt). This
follows option 2 of paragraph 5.1.3.5 of the X3.131-1986
specification.
If this command is issued after the desired SCSI device has
already responded with the BSY signal, no interrupt will be
generated for this command. The DP8496/7 will continue
as if the Abort (Re)Selection command had not been issued
at all.
To verify the result of this command, the SCSI Status register should be read at least 20 BCLK periods after this command is issued. If the Status Code is less than 7 (0000 –
0111), the Abort (Re)Select command was issued in time.
Simply wait for the completion interrupt.
If the Status Code is 7 or greater (0111 – 1111), the Abort
(Re)Select command was issued too late. The Selection
has already completed, and there may be no completion
interrupt generated by the Abort (Re)Selection command.
However, a completion interrupt will be generated when the
original (Re)Select or (Re)Select combination command terminates.
Receive Command (Opcode e 3Eh)
Receive Data (Opcode e 3Fh)
Receive Message (Opcode e 3Ah)
Receive Unspecified (Opcode e 3Bh)
Send Status (Opcode e 0Ah)
Send Data (Opcode e 0Bh)
Send Message (Opcode e 0Eh)
Send Unspecified (Opcode e 0F)
The Send and Receive commands listed above are used to
transfer information on the SCSI bus. The information transferred will be to or from the buffer memory.
These commands are only valid if the DP8496/7 is configured as a Target in the Automatic mode. The DP8496/7
must already be connected (BSY asserted). The function of
these commands are all similar except for the I/O, C/D and
MSG lines. The state of these control lines are listed in
Table 4.20.
Disconnect (Opcode e 32h)
This command will immediately release the SCSI lines to
the Bus Free phase. This command must be issued only if
there are no pending commands (Status Code in SCSI
Status register e 0111).
This command will not generate an interrupt because it is
immediate, without the possibility of an error.
Message Accepted (Opcode e 30h)
Valid only as an initiator. This command should be used
after a Transfer Info or Transfer Pad command while in the
Message In phase. At the termination of the Transfer Info or
Transfer Pad command, the ACK signal will normally be
deasserted. However, if these commands are issued while
in the Message In phase, the ACK signal will remain asserted at the termination of the command. The Message Accepted command can be used to deassert the ACK signal in
this case.
If the message is to be accepted, issue the Message Accepted command which will deassert the ACK signal. If the
message is to be rejected, the ATN signal should be asserted before the Message Accepted command is issued. The
ATN signal may be asserted by setting the ATN bit in the
SCSI Control register.
(Re)Select (Opcode e 10h)
This command initiates a rather lengthy series of events
commencing with detecting the Bus Free phase. An Arbitration phase is executed next if enabled in the SCSI Operation
register. This phase asserts the SCSI ID (initialized in the
Setup 2 register) on the SCSI bus and detects if any device
of higher priority is arbitrating with it. The DP8496/7 will
cycle through this sequence until it wins arbitration. After
winning arbitration, the chip Selects the Target or Reselects
the Initiator whose ID is contained in the Destination ID register. It will wait for a response from that device indefinitely
with SEL and ID’s asserted and BSY deasserted. When the
selected device responds by asserting BSY, the command
will terminate with a completion interrupt.
None of the above sequences are protected by time-outs,
because it is much more efficient to bracket the entire selection process with a single, processor controlled timer.
The choice between Selection and Reselection is determined by the T/I (Target/Initiator) bit in the SCSI Operation
register. While in the Target mode, this command will execute a Reselection. While in the Initiator mode, a Selection
will be executed.
42
4.0 Functional Description (Continued)
The Receive Data command and the Send Data command
will automatically use synchronous transfers on the SCSI
bus if enabled in the Synchronous Transfer register.
Transfer Pad (Op-Code e 38h)
This command is identical to the Transfer Info command,
except the data transferred on the SCSI bus will not be read
or written to the buffer memory.
For Send type transfers, the last byte transferred to the Target in Automatic mode or a byte loaded into the SCSI Data
register is repeatedly sent. For Receive type transfers, the
DP8496/7 will blindly receive the bytes. The data will not be
transferred to buffer memory. Parity is checked if enabled in
the Setup 2 register.
The number of bytes transferred over the SCSI bus is determined by the product of the SCSI Block Size register and
the SCSI Block Count register.
This command may be useful if the Target requests more
information than the Initiator has to give it.
A SCSI bus parity error or the assertion of the ATN signal
will terminate these commands. The termination will occur
immediately or at the end of the current phase based on the
SE (Stop Enable) bit in the SCSI Operation register. Refer to
the SE bit description for other conditions which will terminate these commands.
The number of bytes transferred over the SCSI bus is determined by the product of the SCSI Block Size register and
the SCSI Block Count register.
After command termination, the SCSI bus control signals
will be left in the state used by the command.
TABLE 4.20. SCSI Control Signals
Command
Bus Phase
C/D
I/O
MSG
Receive Command
Receive Data
Receive Message
Receive Unspecified
Command
Data Out
Message Out
Undefined
1
0
1
0
0
0
0
0
0
0
1
1
Status
Data In
Message In
Undefined
1
0
1
0
1
1
1
1
0
0
1
1
Send Status
Send Data
Send Message
Send Unspecified
Send Disconnect (Opcode e 33h)
The DP8496/7 will change to the Message In phase and
transfer a Save Data Pointers message followed by a Disconnect message. The bus will then be released to Bus
Free phase. This command can be used to temporarily interrupt a data transfer and prepare the initiator for a subsequent reconnection.
If the ATN signal is asserted, this command will terminate
immediately with an ‘‘Uncompleted Command’’ interrupt.
Send End (Opcode e 31h)
While connected as a Target, the Send End command will
change to Status phase, send good status, change to Message In phase, and transfer a Command Complete message. The bus will then be released to Bus Free phase. This
command would be used to end a complete SCSI Command after the Data Phase.
If the ATN signal is asserted, this command will terminate
immediately with an ‘‘Uncompleted Command’’ interrupt.
Note: 1 e Asserted, 0 e Deasserted
Transfer Info (Opcode e 3Ch)
This command is used to transfer information over the SCSI
bus, similar to the Receive and Send commands. However,
this command is used only in the Initiator mode. The information transferred will be to or from the buffer memory depending on the state of the I/O signal.
The transfer type (Receive, Send, Command, Data, Message) is determined by the control signals present on the
SCSI bus listed in Table 4.20.
The correct transfer mode (synchronous or asynchronous)
must be set in the Synchronous Transfer register prior to
issuing this command. The transfer mode is not automatically selected by the bus phase. If a non-data phase is being
used, be sure to set for asynchronous transfers.
In any transfer type other than Mesasge In, the command
completes with ACK deasserted. In the case of a Mesage In
transfer (I/O e 1, C/D e 1, MSB e 1), ACK is left asserted
upon the last byte transferred. When a new Transfer Info or
Transfer Pad command is issued, the ACK signal will become deasserted. Since the ACK signal remains asserted
after the Message In phase, the target is prevented from
starting a synchronous data transfer before a new Transfer
Info command has been issued. This will prevent a FIFO
overflow. This handling of ACK will also allow the software
to accept or reject a message.
If the Target chooses to terminate the command, perhaps
by changing the phase lines, before the normal termination,
an ‘‘Uncompleted Command’’ interrupt will be generated.
The number of bytes transferred over the SCSI bus is determined by the product of the SCSI Block Size register and
the SCSI Block Count register.
Reselect/Receive Data (Opcode e 1Fh)
This combination command is identical to issuing the Reselect command (10h), followed by a single byte Message In
phase using the Identify register contents, followed by the
Receive Data Command (3Fh).
The Message In phase is unique. Instead of transferring
data from buffer memory, a single byte is transferred from
the identify register. The Identify register must be initialized
before this command is issued.
Time-out on Reselection can be monitored by checking the
Status Code in the SCSI Status register.
Reselect/Send Data (Opcode e 1Bh)
This combination command is identical to issuing the Reselect command (10h), followed by a single byte Message In
phase using the Identify register contents, followed by the
Send Data command (0Bh).
Reselect/Receive Data/Disconnect
(Opcode e 1Ch)
This combination command is identical to issuing the Reselect command (10h), followed by a single byte Message In
phase using the Identify register contents, followed by the
Receive Data command (3Fh), followed by the Send Disconnect command (33h).
43
4.0 Functional Description (Continued)
After the first byte of the Command Phase has been read
from the SCSI bus, the DP8496/7 will interpret the Group
Number in bits 5, 6 and 7. This determines the total number
of bytes in the SCSI Command Descriptor Block as shown
in Table 4.21. The correct number of bytes are then read
from the SCSI bus. The entire SCSI Command Descriptor
Block will be transferred to buffer memory.
All the normal interrupts will be generated during the data
transfer phase of the command (Block Transfer Complete,
Group Transfer Complete, etc.). In addition, a ‘‘Group Complete’’ interrupt will always be generated just before the Disconnect operation begins. This will allow the processor to
process the information read from the SCSI bus while the
disconnect operation is taking place.
Reselect/Send Data/Disconnect
(Opcode e 18h)
This combination command is identical to issuing the Reselect command (10h), followed by a single byte Message In
phase using the Identify register contents, followed by the
Send Data command (0Bh), followed by the Send Disconnect command (33h).
A ‘‘Group Complete’’ interrupt will always be generated just
before the Disconnect operation begins. This will allow the
processor to set up the SCSI Bus Controller for a new command while the disconnect operation is taking place.
TABLE 4.21. Group Numbers
Group Number
7 6 5
0
0
0
1
0
0
1
0
0
1
0
1
Ý of Bytes in
Command
6
10
10
12
Note: Any other group combinations are unknown length and will cause an
‘‘Uncompleted Command’’ interrupt.
A SCSI bus parity error or the assertion of the ATN signal
will terminate this command. The termination will occur immediately or at the end of the current phase based on the
SE (Stop Enable) bit in the SCSI Operation register. Also,
during a Message Out phase, if the ATN signal goes low,
the command will terminate based on the SE bit.
The command will terminate after the last Command Descriptor byte has been read. An ‘‘Operation Complete’’ interrupt will be generated. The SCSI Control signals will remain in the Command Phase.
The Clear Pending Command can be issued to terminate
this command if the DP8496/7 has not yet been selected.
If the DP8496/7 is reselected as an Initiator (before it has
been selected as a Target), the chip will respond properly
as an Initiator on the SCSI bus as if there were no active
command. The Wait-for-Select/Receive command will then
terminate with a ‘‘(Re)Selected’’ interrupt and the ‘‘Last
Command Ignored’’ bit set also.
Reselect/Receive Data/End (Opcode e 1Dh)
This combination command is identical to issuing the Reselect command (10h), followed by a single byte Message In
phase using the Identify register contents, followed by the
Receive Data command (3Fh), followed by the Send End
command (31h).
A ‘‘Group Complete’’ interrupt will always be generated just
before the Disconnect operation begins. This will allow the
processor to process the information read from the SCSI
bus while the disconnect operation is taking place.
Reselect/Send Data/End (Opcode e 19h)
This combination command is identical to issuing the Reselect command (10h), followed by a single byte Message In
phase using the Identify register contents, followed by the
Send Data command (0Bh), followed by the Send End command (31h).
A ‘‘Group Complete’’ interrupt will always be generated just
before the Disconnect operation begins. This will allow the
processor to set up the SCSI Bus Controller for a new command while the disconnect operation is taking place.
Wait-for-Select/Receive Command/Disconnect
(Op-Code e 37h)
This command is identical to the previous command, except
with the addition of an optional Disconnect function at the
end.
If the SCSI Command Descriptor Block that is read from the
SCSI bus is a read disk command, the DP8496/7 will enter
the ‘‘Message In’’ phase, send a ‘‘Disconnect’’ message
and terminate with an ‘‘Operation Complete’’ interrupt. A
read command is determined by the first byte of the Command Descriptor Block that was transferred during the Command Phase. If this byte is either a 08h, 28h, 48h, or A8h,
then this is interpreted as a read command.
If it is not a read command, the command will terminate
after the last Command Descriptor byte has been read. An
‘‘Uncompleted Command’’ interrupt will be generated. In
this case the SCSI Control signals may not remain in the
Command Phase.
Wait-for-Select/Receive Command
(Opcode e 36h)
This command should normally be issued any time there is
no other SCSI bus data transfers in progress. This will allow
new SCSI commands to be received by the DP8496/7 (configured as a target).
The DP8496/7 will wait until it recognizes itself being selected with BSY deasserted, SEL asserted, and its own ID present on the SCSI data bus. If the ATN signal was also asserted, a Message Out phase will be generated with an Identify
Message. The Identify byte will be loaded into the Identify
register, not buffer memory. If the ATN signal was not asserted the Command Phase will be entered immediately,
without a Message Out phase.
If a Message Out phase is executing and the ATN signal
remains asserted when the ACK signal is deasserted, the
command is terminated with an ‘‘Uncompleted Command’’
interrupt. This indicates that the Initiator is sending an extended or multi-byte message.
Wait-for-Select/Send Busy (Opcode e 35h)
If no command can be accepted by the Target, this command will turn away all Selections in a clean manner with
minimal processor overhead. This command operates the
44
4.0 Functional Description (Continued)
same as the Wait-for-Select/Receive command, except after receiving the Command Descriptor Block, the DP8496/7
will send a busy status, Command Complete message, and
will disconnect. It will then generate an ‘‘Operation Complete’’ interrupt.
The Command Descriptor Block read from the SCSI bus will
not be transferred to buffer memory.
This command will remain active until overwritten with another command. This command is unique in that it will re-enable itself after each successful ‘‘Operation Complete’’ interrupt. If the target needs to monitor selection attempts,
the desired registers (Identify register) must be read by the
processor before the next selection occurs.
Any other interrupt besides ‘‘Operation Complete’’ will
pause this command. The Clear Pending Command should
be issued before this interrupt is cleared to prevent the
Wait-for-Select/Send Busy command from re-enabling itself
again after the error is cleared.
Receive Data Loopback (Opcode e BFh)
Operates the same as the Receive Data command in the
Target mode. The processor should initialize the SCSI Data
register with a value before this command is issued. The
contents of the SCSI Data register plus parity will be asserted on the internal SCSI data bus. The SCSI parity bit is
calculated when a byte is written to the SCSI Data register
by the processor. This value will be read and transferred to
buffer memory. SCSI parity is checked if enabled. The asynchronous mode must be used.
Transfer Info Loopback (Opcode e BCh)
Operates the same as the Transfer Info command in the
Initiator mode. The I/O bit in the SCSI Control register
should be set high or low to indicate the direction of the
transfer. The asynchronous mode must be used.
If receiving data from the SCSI bus, the processor should
initialize the SCSI Data register with a value before this command is issued. The contents of the SCSI Data register plus
parity will be asserted on the internal SCSI data bus. The
SCSI parity bit is calculated when a byte is written to the
SCSI Data register by the processor. This value will be read
and transferred to buffer memory. SCSI parity is checked if
enabled.
If sending data to the SCSI bus, the data is read from buffer
memory and parity is checked if enabled. The data asserted
internally on the SCSI bus may be monitored by reading the
SCSI Data register.
Self Test Commands
The five self test commands operate exactly the same as
the corresponding non-self test commands. However, the
SCSI bus is held TRI-STATEÉ and the inputs are ignored.
The SCSI outputs are still active internal to the chip and the
SCSI inputs are emulated for the correct response. The other pins of the DP8496/7 (non-SCSI) operate normally.
Once the first self test command is issued, the DP8496/7
enters its self test mode. The Clear Pending Command
must be issued to clear the self test mode. Be sure to issue
the Disconnect Self Test command before leaving the self
test mode to ensure that all the internal SCSI bus signals
are deasserted.
Forcing Parity Errors in Self Test Mode
In order to force a SCSI bus parity error in either a Receive
Data Loopback or Transfer Info Loopback (with the I/O bit
set to ‘‘1’’ in the SCSI Control register), the following sequence may be used.
1. Set the SPE (SCSI Parity Enable) bit in the Setup 2 register to a ‘‘1’’.
2. Issue (Re)Select Self Test.
3. Write a byte into SCSI Data register.
4. Switch SCSI Parity Polarity (the SPP bit in the Setup 2
register)
5. Issue a receive type loopback command.
6. Should obtain parity error in the SCSI Status register.
(Re)Select Self Test (Opcode e 90h)
Operates the same as the (Re)Select command but with all
necessary input handshaking internally generated on chip.
This should be the first self test command issued.
Disconnect Self Test (Opcode e B2h)
Operates the same as the Disconnect command. This command should be issued before the self test mode is cleared.
Send Data Loopback (Opcode e 8Bh)
Operates the same as the Send Data command in the Target mode. The data is read from buffer memory and parity is
checked if enabled. The data asserted internally on the
SCSI bus may be monitored by reading the SCSI Data register. The asynchronous mode must be used.
Command Flowcharts
The following flowcharts describe in detail the operation of
the multiphase commands.
45
4.0 Functional Description (Continued)
TL/F/11212 – 25
*If this routine called as a subroutine, the Completion Interrupt is not generated here.
**After reaching this point, the ‘‘point of no return’’, the Abort (Re)Select opcode will not function.
FIGURE 4.12. (Re)Select Flowchart
46
4.0 Functional Description (Continued)
TL/F/11212 – 26
**Continue with whatever operation was occurring when Abort (Re)Select opcode was issued.
FIGURE 4.13. Abort (Re)Select Flowchart
47
4.0 Functional Description (Continued)
TL/F/11212 – 27
*If this routine called as a subroutine, the Completion Interrupt is not generated here.
FIGURE 4.14. Target Information Transfer Flowchart
48
4.0 Functional Description (Continued)
TL/F/11212 – 28
FIGURE 4.15. Initiator Information Transfer Flowchart
49
4.0 Functional Description (Continued)
TL/F/11212 – 29
*If this routine called as a subroutine, the Completion Interrupt is not generated here.
FIGURE 4.16. Send Disconnect Flowchart
TL/F/11212 – 30
*If this routine called as a subroutine, the Completion Interrupt is not generated here.
FIGURE 4.17. Send End Flowchart
50
4.0 Functional Description (Continued)
TL/F/11212 – 31
FIGURE 4.18. Reselect Combination Flowchart
51
4.0 Functional Description (Continued)
TL/F/11212 – 32
FIGURE 4.19. ‘‘Wait for Select’’ Type Flowchart
52
4.0 Functional Description (Continued)
SCSI Data (SDAT)
7
6
41h
5
4
3
2
1
R/W
0
SCSI Operation (SOP)
7
6
5
Data
T/I
This register allows access to the SCSI Data Bus by the
processor while in Manual Mode. See Section 4.4.3 for details about Manual Mode data transfers. These bits are asserted only if a ‘‘1’’ is written. On the single-ended SCSI bus
that is open-drain, writing a ‘‘0’’ turns off the driver and allows the bus to float.
This register should only be used while in the Manual Mode.
There are two exceptions:
1. Just before switching from Automatic to Manual Mode.
See Section 4.4.3.
2. Just before a Transfer Pad command. See the Transfer
Pad command description.
SCSI Control (SCTL)
7
6
5
BSY
SEL
42h
ATN
4
3
2
1
C/D
I/O
MSG
REQ
ACK
Man.
Dir.
Auto
Man.
BSY
SEL
ATN
I/O
I/O
I/O
R
R
R
R/W
R/W
R
I/O
I/O
O
R
R
R/W
R/W
R/W
R/W
C/D
I/O
MSG
O
O
O
R/W
R/W
R/W
R/W
R/W
R/W
I
I
I
R
R
R
R
R
R
REQ
ACK
O
I
R
R
R/W
R
I
O
R
R
R
R/W
SE
A/H
APE
PEP
R/W
0
RST
A/M: Automatic/Manual Mode
1 Enables the Automatic mode of operation, enabling the
SCSI sequencer. SCSI commands may be issued to the
SCSI Command register. The chip will respond properly
to (Re)Selections.
0 Manual Mode completely disables the SCSI sequencer
and clears any curent SCSI command. No SCSI commands may be issued in this mode, but it has no effect
on the Disk Data Controller. In this mode the processor
has complete control over the SCSI bus and any sequence, legal or illegal, can be performed.
Automatic to Manual Restrictions: Before switching from
Automatic mode to Manual mode, the T/I (Target/Initiator)
bit in this same register should be set to the desired value.
Also the HE (Handshake Enable) bit in the Synchronous
Transfer register should be programmed. If the DP8496/7 is
not currently connected to the SCSI bus, the HE bit should
be set to ‘‘0’’. This guarantees that all 18 SCSI signals will
remain glitchless during the A/M bit transition. If the HE bit
is set to ‘‘1’’, only the 9 SCSI Bus Control signals are guaranteed to remain glitchless.
Initiator
Auto
1
SWA: Select With Attention
1 When in the Initiator mode and Selecting, the ATN signal will be asserted as per SCSI specifications.
0 When in the Initiator mode and Selecting, the ATN line
will not be asserted.
When executing Transfer Info/Pad commands, ATN can be
controlled through the SCSI Control register. In addition,
when executing Transfer Info/Pad commands while executing any Information In phase, ATN will be asserted on a
SCSI bus parity error if enabled in the Setup 2 register.
TABLE 4.22. Bit Attributes in SCSI Control Register
Target
2
T/I: Target/Initiator
This bit must not be changed while an information transfer
phase is in progress. It can be changed after an unexpected
Selection or Reselection, but in these cases it must be
changed before a transfer command is issued. It also can
be changed while the SCSI bus is idle.
Caution: If the last command written to the SCSI Command
register is legal in the current mode only (Target or Initiator),
a Clear Pending Command should be issued before the (T/I
bit is modified. Even if the last command has completed and
the interrupt has been processed, the Clear Pending Command should be issued. If this is not done, an ‘‘Invalid Command’’ interrupt will be generated after the T/I bit is
changed.
1 Target Mode. All commands and modes relevant to the
Target functions are enabled. ((Re)Select command
becomes Reselect)
0 Initiator Mode. All commands and modes relevant to
the Initiator function are enabled. ((Re)Select command
becomes Select)
R/W
0
Dir.
A/M
3
This register must be loaded before SCSI Block Count and
SCSI Block Size registers.
The SCSI Control register is really an image of eight out of
the nine SCSI control lines. The direction and R/W attributes of these bits are determined by the Target or Initiator
mode and whether in Automatic or Manual operation. A
summary of bit attributes are given in Table 4.22. When they
are read, ATN, C/D, I/O, and MSG bits reflect the logical
OR of what is on the SCSI bus and what the chip is attempting to drive. These two values may not be the same under
some abnormal conditions. For example, on the DP8497,
incorrect use of the Differential SCSI register in manual
mode or an external circuit fault may cause such a conflict.
The SCSI RST signal does not appear in this register but is
driven by the DP8496/7 from the SCSI Operation register.
When RST is driven by another SCSI device it will generate
a ‘‘SCSI Reset’’ interrupt.
Note the ATN line is writable in Automatic mode when configured as an Initiator. However, the SWA (Select With Attention) bit in the SCSI Operation register should be used to
assert the ATN signal properly during the Selection phase.
The ATN bit is only intended to be written during the Transfer Info or Transfer Pad commands.
These bits are asserted only if a ‘‘1’’ is written. Since the
SCSI bus is open-drain, writing a ‘‘0’’ turns off the driver and
allows the bus to float.
Sig.
SWA
43h
4
53
4.0 Functional Description (Continued)
0
The Select command will not contain an Arbitration
Phase.
In Target mode, an Arbitration Phase will always be executed for Reselected and Reselect combination commands.
The APE bit is also used to clear an ‘‘Invalid Command’’
condition. Clearing an ‘‘Invalid Command’’ interrupt is a twostep process. First, the APE should be modified twice (set,
then reset, or reset, then set). Second, clear the interrupt by
writing the three-bit pattern for ‘‘Invalid Command’’ (100)
back to the SCSI Interrupt register.
The SCSI Data register should be programmed with the desired value to be driven on the SCSI bus. If the bus should
not be driven immediately, a zero should be written to the
SCSI Data register and the SPP (SCSI Parity Polarity) bit in
the Setup 2 register should be set to ‘‘1’’ for even parity to
prevent the SCSI parity bit being asserted on the SCSI bus.
If the DP8496/7 is currently connected, there should be no
SCSI bus activity during the A/M bit transition.
Immediately after switching to Manual mode a command
complete interrupt will be generated.
Manual to Automatic Restrictions: Before switching from
Manual mode to Automatic mode, a Clear Pending Command should be written to the SCSI Command register. The
DP8496/7 must be disconnected.
PEP: Parity Error to Processor
If the Setup 2 register is configured to check buffer memory
parity, a parity error detected while the processor reads the
Buffer Memory Data register will set this bit. This bit may be
read after a block of data has been read from the Buffer
Memory Data register. Once this bit is set, it can only be
cleared by reading this register. It will not be cleared by the
SCSI Reset command.
SE: Stop Enable
1 For the following conditions the current command will
be terminated immediately. There will be a ‘‘SCSI Bus
Error’’ interrupt for a SCSI parity error, or an ‘‘Uncompleted Command’’ with ‘‘Attention Detected’’ for the
ATN signal.
1. In Initiator or Target mode, a SCSI parity error is
detected while receiving information.
2. While in Target mode, the ATN signal is detected
going high during any SCSI transfer,
If in Target mode, during a Message Out phase, and the
ATN signal becomes deasserted before the leading
edge of the REQ signal for the last byte, the command
will terminate immediately with an ‘‘Uncompleted Command’’ interrupt.
0 Command will terminate immediately as desribed
above only for the following situations while in Target
mode:
1. SCSI parity error during Command Phase.
2. SCSI parity error during Message Out Phase.
3. ATN going high during Status Phase.
4. ATN going high during Message In Phase.
In all other situations the command will terminate at the
end of the current phase instead of immediately. The
reported interrupts will be the same as above.
If in target mode, during a Message Out phase, and the
ATN signal becomes deasserted before the leading
edge of the REQ signal for the last byte, the command
will terminate at the end of the Message Out phase with
an ‘‘Uncompleted Command’’ interrupt.
RST: SCSI Bus Reset
1 The SCSI RST signal will be held asserted as long as
this bit is set. This is valid in any mode; Initiator, Target,
Automatic or Manual. Be sure to keep this asserted for
the minimum time required by the SCSI specification
(25 ms).
0 Normal operating mode. The SCSI RST signal will not
be asserted by the DP8496/7.
Reading this bit will simply reflect the last value written to
this bit, not the current state of the SCSI RST signal. A
‘‘SCSI Reset’’ interrupt will be generated if the SCSI RST
signal is asserted. This interrupt is generated whether the
RST signal was asserted by the DP8496/7, or by any other
source.
Synchronous Transfer (SYNC)
44h
7
6
5
4
3
HE
Transfer
Period
2
1
R/W
0
Offset
Synchronous transfers are only legal in the Data Phase
while in Automatic mode.
Adequate memory bandwidth should be provided for full
speed disk and SCSI transfers. If the bandwidth is inadequate and the DP8496/7 is receiving SCSI data, overruns
may occur (this would be reported in the SCSI Status register). If the DP8496/7 is sending SCSI data, the data may not
be sent at full speed.
A/H: Active/Holding
1 The processor has access to the Active set of Disk and
SCSI pointers and block counts. Caution: the contents
of the Active registers may change as they are being
read if data transfers are currently taking place. See
Section 4.2.6 for more information.
0 The processor has access to the Holding set of registers. This is the normal mode of operation.
HE: Handshake Enable
This bit is valid only in the Manual mode and further defines
the way the SCSI Data register operates.
1 This mode is used to transfer single bytes over the
SCSI bus using the DBR (Data Buffer Ready) bit in the
SCSI Status register. The processor should poll the
DBR bit for the proper state depending on the data direction. The processor can then read or write to the
SCSI Data register. In this mode, the REQ and ACK
signals handshake properly for every byte. As an application, message and status phases could be implemented this way. HE should be set to ‘‘1’’ only after the
SCSI Operation register and SCSI Control register are
loaded correctly.
APE: Arbitration Phase Enable
This bit is only significant while in Automatic mode as an
Initiator. This bit should only be modified when there are no
active or pending commands and while disconnected.
1 The Select command will contain an Arbitration Phase
between detection of bus free and the Selection Phase.
54
4.0 Functional Description (Continued)
0
The SCSI Data register is simply a transparent latch
between the SCSI bus and the processor. The REQ
and ACK signals are not automatically asserted REQ
and ACK (and other SCSI control signals) may be modified by the processor through the SCSI Control register.
Parity is not verified when reading data from the SCSI
bus. Parity is always generated when writing to the
SCSI bus.
For information transfers, it is not recommended to use
HE e 0, because parity is not checked.
TL/F/11212 – 33
FIGURE 4.20. Synchronous Transfer Period
Offset
The offset value specifies the number of outstanding bytes
allowed before handshaking is inhibited. A value of zero results in an offset of 16.
If the Block Count register is greater than one, or if the
Block Count is pipelined, the offset value must not be greater than the Block Size registers. Otherwise, data may be
written to buffer memory incorrectly. If the Block Count is
one, and there is no pipelining, then there is no restriction
on the value of the offset.
Transfer Period
This field determines the time between adjacent Synchronous Data transfers. It is also used to choose between Synchronous and Asynchronous Transfer modes.
A value of zero will give Asynchronous Transfers with fast
handshaking between REQ and ACK. A value of one will
give Asynchronous Transfers an extra BCLK period per
REQ and ACK handshake. (It is also possible for the user to
set the setup time delay between REQ and ACK and the
SCSI data signals by setting the SCK1 and SCK0 bits of the
Differential SCSI 2 register.)
A value of two to seven enables Synchronous Transfers
with a Transfer Period (TP) equal to BCLK period multiplied
by the value programmed (2–7).
While in the Target Mode, if a non-Data Phase is executing,
Asynchronous transfers will be used, regardless of the value
in this field. This is done because Synchronous transfers are
only allowed during the Data In Phase and the Data Out
Phase.
Caution: While in Initiator Mode, the DP8496/7 will NOT
automatically switch between Synchronous and Asynchronous modes. This must be switched by the processor.
Table 4.23 summarizes the transfer mode used based on
this field and the phase in progress.
The value in this field, along with the BCLK frequency, also
determines the DP8496/7s adherence to other synchronous SCSI timing specificationsÐREQ assertion period
(90 ns, 30 ns for Fast), REQ negation period (90 ns, 30 ns
for Fast), DATA setup time (55 ns, 25 ns for Fast), and Data
Hold Time (100 ns, 35 ns for Fast). The equations for these
times are specified in the A.C. Specification section of this
datasheetÐin subsections 6.29 thru 6.32. Maximum BCLK
is specified in subsection 6.1.
Identify (IDENT)
7
6
I
Current SCSI Phase of DP8496/DP8497
Target, Data In
Target, Data Out
Initiator, All Phases
Target, All Phases
EXCEPT Data In and
Data Out
0
1
2
3
4
5
6
7
Async
Slow Async
Sync, TP e 2*BCLK
Sync, TP e 3*BCLK
Sync, TP e 4*BCLK
Sync, TP e 5*BCLK
Sync, TP e 6*BCLK
Sync, TP e 7*BCLK
Async
Slow Async
Async
Async
Async
Async
Async
Async
4
TAR
Reserved
3
2
1
R/W
0
LUN
The Identify register is used only with Reselect combination
and ‘‘Wait for Select’’ type commands. This is actually a
simple 8-bit register with no fixed structure. All of the bits
may be read and written by the processor. However, in most
applications the bit definitions should correspond with the
ANSI SCSI specification.
The Reselect combination commands use the contents of
the Identify register during the Message In phase following
Reselection. The processor must load the Identify register
before a Reselect combination command is issued.
The ‘‘Wait for Select’’ commands modify the Identify register. This byte is received from the Message Out phase after
a connection. The data here is valid as soon as the message phase is completed, although the processor is not notified with an interrupt until the entire command is complete.
The Identify register is overwritten upon the next ‘‘Wait for
Select’’ operation or when the processor loads the Identify
register preceding a Reselect combination command.
Caution: This register must be initialized with a valid value
after a reset before any command is written to the SCSI
Command register. Otherwise an ‘‘Uncompleted Command’’ interrupt will be generated. A valid value is any value
with bit 7 set and bits 3 and 4 cleared.
TABLE 4.23. Transfer Mode and Period
Trns.
Prd.
DISC
45h
5
Notes: TP e Transfer Period.
See AC Timing for Async vs. Slow Async.
55
4.0 Functional Description (Continued)
MCH: Match
1 Looks for a match with the indicated Phase Code. If
found, a ‘‘Phase Compare’’ interrupt will be generated.
0 Looks for a mismatch of the current SCSI bus phase
with the indicated Phase Code. This code should be
loaded after the Target’s phase is stable (REQ is asserted) if the DP8496/7 is in Initiator mode. If the phase
changes, a ‘‘Phase Compare’’ interrupt will be generated.
I: Identify Message
Since this register is used only for the Identify Message, this
bit must always be set to ‘‘1’’. If it is not set to ‘‘1’’, no new
command can be executed.
This bit will always be a ‘‘1’’ when read following a successful ‘‘Wait for Select’’ operation which included a Message
Out phase. Otherwise, an ‘‘Uncompleted Command’’ interrupt would be generated.
DISC: Disconnect Privilege
When this register is used for the Message In phase during
a Reselect combination command, this bit has no meaning.
Following a ‘‘Wait for Select’’ command which included a
Message Out phase, this bit indicates the ability of the Initiator to support disconnection and reconnection. A ‘‘1’’ indicates this support. A ‘‘0’’ indicates no support.
Destination ID
This ID is OR’ed with SCSI ID from the Setup 2 register. This
OR’ed value is asserted on the SCSI bus during Selection or
Reselection phases. Parity is also generated.
Source ID (SID)
7
6
TAR: Logical Unit Target
When this register is used for the Message In phase during
a Reselect combination command, this bit has no meaning.
Following a ‘‘Wait for Select’’ command which included a
Message Out phase, this bit specifies where the Identify
Message is directed to. A ‘‘0’’ indicates a logical unit (see
the LUN field). A ‘‘1’’ indicates a target routine. See the
SCSI-2 specification.
VID
Phase Code
46h
3
MCH
2
1
3
x
x
x
SCSI Status (SSTAT)
7
6
5
DBR
2
1
R Only
0
Source ID
BMR
48h
4
Error Code
3
2
1
R Only
0
Status Code
DBR: Data Buffer Ready
In Manual mode with the HE (Handshake Enable) bit set to
‘‘1’’, DBR is used to indicate when to send or receive the
next byte.
1 The SCSI Data register is full. If receiving, the SCSI
Data register should be read. If transmitting, the processor should wait until DBR e 0.
0 The SCSI Data register is empty. If receiving, the processor should wait for another byte to arrive. If transmitting, the SCSI Data register should be written.
The DBR bit is reset whenever an error is cleared in the
SCSI Interrupt register. An error is defined as a Completion
Interrupt Code value between 2h and 7h in the SCSI Interrupt register.
LUN: Logical Unit Number
Before a Reselect combination operation, this field should
be set with the Logical Unit Number from which the data is
being read from or being written to.
Following a ‘‘Wait for Select’’ command which included a
Message Out phase, this field indicates the Logical Unit
Number associated with the command to follow. They
should match the LUN bits contained in the Command Descriptor Block.
4
4
While in Target mode during the Selection Phase, or while in
Initiator mode during the Reselection Phase, the Initiator’s
ID and the Target ID are both present on the SCSI bus. Our
Target ID (SCSI ID in Setup 2 register) is masked off, and
the resulting Initiator ID can be read from Source ID field.
The VID (Valid ID) bit will be ‘‘1’’.
If there is no Initiator ID on the SCSI bus during Selection,
the VID bit will be ‘‘0’’. If the DP8496/7 is not properly Selected (ex. three or more bits set), the DP8496/7 will not
respond and the VID bit will be ‘‘0’’.
Reserved
Since these bits are reserved, it is recommended that these
bits be set to ‘‘0’’ before a Reselect combination command
is issued.
These bits will always be a ‘‘0’’ when read following a successful ‘‘Wait for Select’’ operation which included a Message Out phase. Otherwise, an ‘‘Uncompleted Command’’
interrupt would be generated.
Destination ID (DID)
7
6
5
x
47h
5
R/W
0
Destination ID
Phase Code
These four bits set up the SCSI bus phase for which the
DP8496/7 will match or not match to generate the Phase
Compare interrupt. Comparison is qualified with the active
edge of the REQ signal in both match and mismatch cases.
The Phase Code bit patterns are exactly the same as the
Status Code patterns defined in Table 4.25 in the SCSI
Status register.
BME: Buffer Memory Parity Error
A parity error was detected while writing data to the SCSI
port from buffer memory. This indicates a parity error in buffer memory. This will terminate any SCSI command at the
end of the current transfer phase.
The ‘‘Buffer Memory Parity Error’’ completion interrupt will
be generated.
This bit is cleared to zero when a new command is downloaded and executed.
56
4.0 Functional Description (Continued)
Important: In Initiator mode, information transfer phases are
updated on the leading edge of REQ. If, during a Transfer
Info or Transfer Pad command, the phase changes from
that detected on the first REQ, the command will be aborted
with an ‘‘Uncompleted Command’ interrupt if there is no
parity error.
Error Code
The error code is used in conjunction with the Status Codes
to determine the present state of the SCSI interface. In the
case of a SCSI Data Overrun error occurring at the same
time as a SCSI parity error, the more serious error, SCSI
Data Overrun, will be encoded. In the case where the SE bit
of the SCSI Operation register is set and therefore a SCSI
parity error causes the current command to terminate immediately, a SCSI data overrun may be caused. In this case
also the error code for SCSI Data Overrun will be reported,
effectively masking the original SCSI parity error from being
reported.
TABLE 4.25. Status Codes
Status
Code
3 2 1 0
0
0
0
0
Bus Free or Idle due to Reset command,
CRST pin, SCSI bus reset, normal
disconnection for initiator, or Abort
(Re)Selection command.
0
0
0
1
Bus Free due to command complete
(except after Abort (Re)Select command
or Reset command). Target mode only.
0
0
1
0
A DP8496/7 initiated Arbitration phase is
pending or in progress.
0
0
1
1
A DP8496/7 initiated (Re)Selection phase
is in progress.
0
1
0
0
Selected. The DP8496/7 has been
Selected as a Target.
0
1
0
1
Reselected. The DP8496/7 has been
Reselected as an Initiator and the
‘‘Reselected’’ interrupt completion code
has not yet been cleared.
0
1
1
0
Unexpected bus free in Initiator mode.
Command has been aborted.
0
1
1
1
Connected, but no pending command
1
0
0
0
Data Out Phase
1
0
0
1
Unspecified Info Out Phase
1
0
1
0
Command Phase
1
0
1
1
Message Out Phase
1
1
0
0
Data In Phase
1
1
0
1
Unspecified Info In Phase
1
1
1
0
Status Phase
1
1
1
1
Message In Phase
TABLE 4.24. Error Codes
Error Code
5 4
0
0
1
0
1
1
Error
No Error
SCSI Parity Error
SCSI Data Overrun
The Error Code field is cleared to zero when a new command is downloaded and executed.
SCSI Parity Error (0 1)
This code is present in any Selection or Information transfer
phase upon detection of a parity error from the SCSI bus in
either Auto or Manual mode.
SCSI Data Overrun (1 1)
This code is set if the DP8496/7 receives too many bytes to
transfer in the available bandwidth to buffer memory. Data
loss has occurred! If this error occurs in any transfer mode,
it indicates a lack of available buffer memory bandwidth. If it
occurs during an Asynchronous transfer, some AC timing
specification has been violatedÐsuch as too fast a REQ/
ACK period for a given BCLK. There must be at least 2
BCLK periods per REQ or ACK period.
Also set if 2 REQ pulses are received before a Transfer Info
or Transfer Pad command in Initiator mode is issued by the
processor. This indicates that the Target device is in Synchronous Transfer mode and the DP8496/7 is not set up to
respond in this mode.
During synchronous transfers, if the number of outstanding
REQ or ACK pulses received is greater than the programmed offset (offset overflow), this error will be set.
Also set if there are any abnormal handshakes between
REQ and ACK. Examples may be an ACK which deasserts
before REQ or any non-match of REQ/ACK pulses following a Synchronous mode transfer.
Comment
While in the Initiator mode, the Status Code will be frozen at
the time that an error completion code is generated (not
code ‘‘001’’). When the completion interrupt is cleared, the
Status Code will return to normal operation. A completion
code of ‘‘001’’ will not freeze the Status Code.
Status Code
This field indicates present bus condition and in the case of
bus free phase, how the DP8496/7 arrived in the phase. All
bits are latched on the leading edge of the RD strobe and
will not change during the read cycle.
Note that the state of the ATN signal can be read at any
time in any mode through the SCSI Control register.
The Status Code is not valid while in the Manual mode. The
Status Code is also not valid after an ‘‘Invalid Command’’
interrupt has been generated.
SCSI Interrupt (SINT)
7
6
5
PCMP
ATN
GTC
49h
4
3
BTC
LCI
2
1
R/W
0
Completion Interrupt Code
This register indicates interrupts that have occurred. There
are two types of interrupts, checkpoint and completion.
Checkpoint interrupts (bits 3 – 7) are separated into separate
bits and may be asserted simultaneously. These interrupts
may be cleared only by writing a ‘‘1’’ into the corresponding
bit location. These bits may be cleared individually or simultaneously.
57
4.0 Functional Description (Continued)
LCI: Last Command Ignored
Completion interrupts are encoded into bits 0–2. Only one
completion interrupt may be read at a time. Completion interrupts may be buffered behind each other if more than
one completion interrupt has occurred. See Table 4.27 for a
description of buffered interrupts. Completion interrupts may
be cleared only by writing the same three bit pattern back to
the SCSI Interrupt register. This will clear that interrupt and
allow a new or buffered completion interrupt to occur. A
completion interrupt may be cleared simultaneously with
other checkpoint interrupts.
Interrupts will be reported in the SCSI Interrupt register always, independent of the SCSI Interrupt Enable register.
If enabled interrupts remain after writing to the SCSI Interrupt register, the SINT pin will deassert momentarily to retrigger an edge sensitive interrupt input to the processor.
This register is latched to prevent the contents from changing while reading the register.
Can occur on the following conditions:
1. When a ‘‘Wait for Select’’ type command is overwritten
with another command, the new command will not be
executed if the DP8496/7 has started to respond to the
anticipated Selection from another SCSI device.
2. A (Re)Select or Reselect combination command may be
ignored if the DP8496/7 has been or is in the process of
being (Re)Selected by another SCSI device. This will
also produce a ‘‘(Re)Selected’’ completion interrupt.
3. If the Abort (Re)Select command is issued and the
DP8496/7 has already started a (Re)Selection phase
and received a response from the other device, the
(Re)Selection phase will not be aborted and the LCI bit
will be set.
4. Any ‘‘Wait for Select’’ type command is ignored if the
DP8496/7 has been or is in the process of being reselected by another SCSI device. This will also produce a
‘‘(Re)Selected’’ completion interrupt.
5. A new command is loaded before a ‘‘(Re)Selected’’ interrupt is cleared.
PCMP: Phase Compare
Set when the Phase Code selected in the Destination ID
register either matches (MCH e 1) or differs (MCH e 0)
from the current phase of the SCSI bus. The MCH (Match)
bit is also found in the Destination ID register.
Application Hint: One use of this interrupt is to inform the
processor of the start of data transfer (a Match of Data In or
Out phase) and thus the point at which a new SCSI pointer
and block count can be loaded on a combination type command. If in Initiator mode, the mismatch mode could be set
up to notify the processor if the phase changes from the
current phase.
Completion Interrupt Code
Table 4.26 lists the possible completion interrupts. While in
Manual mode, only codes ‘‘000’’, ‘‘010’’, and ‘‘111’’ may be
generated.
If an error completion code is generated, even if it is buffered, no new operations will be performed. When the error
completion code is cleared, new SCSI commands may be
executed. See the end of this register description for an
explantion of buffered interrupts.
ATN: Attention Detected
1 Set on assertion of the ATN signal when DP8496/7 is
configured as a Target in Automatic mode during an
information transfer phase. Also set on assertion of the
ATN line during selection phase unless executing a
‘‘Wait for Select’’ type command.
The current command will terminate immediately or at
the end of the current phase depending on the state of
the SE (Stop Enable) bit in the SCSI Operation register.
The SCSI Control register can be used to monitor when
the ATN signal goes away.
0 ATN signal not detected.
TABLE 4.26. Completion Interrupt Codes
Comp. Int.
Code
2 1 0
0
0
0
0
1
1
1
1
GTC: Group Transfer Complete Interrupt
This interrupt indicates that the SCSI Block Count register
may be reloaded. This occurs at the end of a block transfer
where the block counter has reached zero, and a new block
count has been downloaded. This interrupt will only be set if
the SCSI Block Count register has been written to since the
last download of the SCSI Block Count.
This bit is not asserted during a Transfer Pad command.
In addition, while executing a Reselect Combination command that ends with a Disconnect or Send End option, the
Group Complete Interrupt will be asserted after the last
block of data has been transferred (before sending any
Status or Message information).
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
Cause
No Interrupt
No Error
SCSI Bus Error*
Buffer Memory Parity Error
Invalid Command
Uncompleted Command
(Re)Selected
SCSI Bus Reset Received or Generated
*SCSI Bus Error can either be parity or data lost. Read SCSI Status register
to determine which.
No Error (001)
This is the normal completion interrupt of SCSI commands.
This is presented upon the successful completion of a SCSI
command. The phase of the SCSI bus depends on the command executed and can be found by reading the SCSI
Status register.
SCSI Bus Error (010)
While in Automatic mode, this interrupt indicates one of two
possible conditions. Either the DP8496/7 observed a SCSI
parity error, or data was lost in the synchronous transfer
mode due to lack of bandwidth to the buffer memory. The
exact error can be determined by reading the Error Code
field in the SCSI Status register.
BTC: Block Transfer Complete Interrupt
This interrupt indicates the Complete transfer of one block,
as defined by the SCSI Block Size registers reaching zero.
This bit will not be set when the GTC (Group Transfer Complete) bit or any completion interrupt code is asserted. It is
also not asserted for a Transfer Pad command in initiator
mode.
58
4.0 Functional Description (Continued)
# While executing a Reselect combination command, if the
This interrupt will be presented after the command has completed.
While in Manual mode, this error can only indicate a SCSI
parity error.
ATN signal is asserted during the Reselection phase, if
the SE (Stop Enable) bit in the SCSI Operation register is
set to ‘‘1’’.
# While in a Message Out phase, if the ATN signal be-
Buffer Memory Parity Error (011)
A parity error was detected while writing data to the SCSI
port from buffer memory. This indicates a parity error in buffer memory. This will terminate any SCSI command at the
end of the current transfer phase.
The BME (Buffer Memory Parity Error) bit in the SCSI Status
register should be set to ‘‘1’’ also.
comes deasserted after REQ is false and while ACK is
true.
# While in a Message Out phase, if the ATN signal becomes deasserted after both REQ and ACK are false.
In the Initiator mode, this interrupt will occur for an unexpected bus free condition. It will also occur if the bus phase
changes after the first REQ is received. The DP8496/7 records the SCSI bus phase when the Target asserts REQ. Any
change of the bus phase with REQ asserted will immediately terminate the current command with this interrupt.
(Re)Selected (110)
This interrupt could be generated without any SCSI command being issued. The moment Automatic mode is enabled, the DP8496/7 is susceptible to being (re)selected. It
is also possible to (re)select the DP8496/7 after a SCSI
command has been issued by the processor. If the
DP8496/7 loses Arbitration and the winning device chooses
to select the DP8496/7, the chip will allow this and respond
properly.
This interrupt will usually be unexpected and tell the processor that the DP8496/7 has responded to a Selection. As a
Target, the ATN line must be checked after receiving this
interrupt to determine if a Message Out phase is required.
As an Initiator, this interrupt would indicate a Reselection
from a Target.
If a ‘‘Wait for Select’’ type command is currently executing
and waiting to be Selected, it will terminate with an ‘‘Operation Complete’’ interrupt, not a ‘‘(Re)Selected’’ interrupt.
Invalid Command (100)
An invalid command for the present mode of the chip was
received in the SCSI Command register. It is not guaranteed
that all invalid commands will be detected. You should NEVER issue an invalid command.
Clearing an ‘‘Invalid Command’’ interrupt is a two-step process. First, the APE (Arbitration Phase Enable) bit in the
SCSI Operation register should be modified twice (set, then
reset, or reset, then set). Second, clear the interrupt by writing the three-bit pattern (100) back to the SCSI Interrupt
register.
Uncompleted Command (101)
In Target mode, this interrupt occurs when an Initiator sends
something unexpected. For example, the ATN signal remains asserted after the first mesage byte in combination
commands, or something other than a SCSI read command
is received in a Wait for Select/Disconnect command.
When this completion interrupt is generated, the SCSI
Status register will freeze in the state in which the SCSI bus
was in when the ‘‘Uncompleted Command’’ interrupt was
generated. While in Initiator mode, the actual phase may
now differ.
When the ‘‘Uncompleted Command’’ interrupt is cleared,
the SCSI Status register will return to normal operation and
will reflect the current status of the SCSI bus.
The unexpected events include:
SCI Bus Reset (111)
This interrupt could be generated without any SCSI command being issued. A reset signal greater than one BCLK
period in length was received on the SCSI bus. The
DP8496/7 is now in Bus Free phase and any command
pending or executing has been terminated.
# While executing a ‘‘Wait for Select/Disconnect’’ command, if a non-read SCSI Command Descriptor Block is
received.
Buffered Interrupts
The 3-bit Interrupt Code field can only present one type of
Interrupt at one time. There may be occasions where more
than one interrupt has occurred. If this is true, then the first
interrupt that occurred will be reported by the Interrupt Code
field. The second interrupt will be buffered internally. When
the first completion code is cleared, the buffered completion
code will be reported in the Interrupt Code field. The SINT
pin will pulse inactive for a short period of time between
interrupts.
The exception to this is the ‘‘No Error’’ completion interrupt.
If this interrupt is generated, it need not be cleared before
another completion interrupt is generated. A new completion interrupt will overwrite any pending ‘‘No Error’’ completion interrupt. This is summarized in Table 4.27.
# While executing a ‘‘Wait for Select’’ type command, if a
message byte other than Identify is received or if a reserved bit is not zero.
# While executing a ‘‘Wait for Select’’ type command, if the
first byte of the Command Descriptor Block does not
map to a Group 0, 1, 2, or 5 type command.
# While executing any command, if the ATN signal remains
asserted after the trailing edge of ACK on the last byte of
the Message Out phase.
# While executing a Message Out phase, if the ATN signal
is deasserted before the leading edge of the REQ for the
last byte.
# While executing any information transfer phase, if the
ATN signal is asserted.
59
4.0 Functional Description (Continued)
TABLE 4.27. Interrupt Buffering
TL/F/11212 – 34
SCSI Interrupt Enable (SINTE)
4Ah
7
6
5
4
3
PCMP
ATN
GTC
BTC
LCI
2
1
R/W
0
x
x
CI
SCSI Block SizeÐLow (SBSL)
4Dh
7
6
5
4
3
4Bh
4
3
2
1
Differential SCSI 1 (SDIF1)
7
6
5
1
4Eh
4
3
2
1
R/W
0
DCSDB7 DCSDB6 DCSDB5 DCSDB4 DCSDB3 DCSDB2 DCSDB1 DCSDB0
Differential SCSI 1 (SDIF1)
7
6
5
SCK1
SCK0
4Fh
4
3
2
1
R/W
0
DCBSY DCSEL DCTARG DCINIT DCRST DCSDBP
(Other than the SCK1 and SCK0 bits, all other bits are meaningful only on
the DP8497.)
R/W
0
DCSDB(7:0), DCSDBP
The polarity of each of these bits sets the polarity of the
DSDB(7:0) and DSDBP pins of the DP8497 and also the
direction of the on-chip transceiver at the associated signal
pin; thus providing differential transceiver direction control
to the user in manual mode operation. A ‘‘1’’ in any of these
bits, and therefore a high level on the corresponding direction control pin, means the transceivers are in ‘‘drive’’
mode; while a ‘‘0’’ signifies the ‘‘receive’’ mode.
In effect only during manual mode.
The value programmed in this register sets the number of
blocks to be transferred. A value of 00h means 256 blocks.
This value, in conjunction with the block size value in SBSH
and SBSL registers, determines the total number of bytes to
be transferred.
2
R/W
0
The size of the user’s SCSI blocks is programmed in these
two registers. This value, in conjunction with the block count
value in the SBC register, determines the total number of
bytes to be transferred.
Count
SCSI Block SizeÐHigh (SBSH)
4Ch
7
6
5
4
3
1
SizeÐLow Order Byte
This register simply enables selected interrupts to physically
assert the SINT pin. If an interrupt is disabled via this register, all the effects of the interrupt remain the same except it
will not affect the SINT pin.
Bits 3 – 7 enable individual checkpoint type interrupts. A ‘‘1’’
allows the individual interrupt to assert the interrupt pin. A
‘‘0’’ prevents it.
The CI (Completion Interrupts) bit allows all completion type
interrupts to assert the interrupt pin. The completion interrupts cannot be enabled individually.
Interrupts should be cleared from the SCSI Interrupt register
whether or not they are enabled in the SCSI Interrupt Enable register.
SCSI Block Count (SBC)
7
6
5
2
R/W
0
SizeÐHigh Order Byte
60
4.0 Functional Description (Continued)
The SCSI Pointer can also be modified when extending the
current executing SCSI commandÐallowing for noncontiguous buffer memory transfers. If the current executing SCSI
command is a single phase information transfer command,
the SCSI Pointer can be modified after 10 BCLK periods
have passed since the SCSI command was issued. If the
current executing SCSI command is a reselect combination
command, the SCSI Pointer can be modified after detecting
the SCSI data transfer phase. The SCSI pointer should not
be updated again (after the first pipelined value) until the
previous pipelined value has been downloaded. If the SCSI
pointer have been modified, the ‘‘Group Complete’’ interrupt
indicates when the pipelined SCSI pointer is downloaded.
If the SCSI pointer has not been modified when extending
the current executing SCSI command, a SCSI pointer download will not occur and contiguous buffer memory transfers
will result. Although ‘‘Group Complete’’ interrupts will still
occur at the completion of each transfer.
The ‘‘Group Complete’’ interrupt indicates when the pipelined Block Count and the SCSI Pointer (if modified) are
downloaded. After this interrupt is received, the Block Count
register and SCSI Pointer registers may be updated again.
A SCSI command may be extended multiple times if desired. Simply wait for the ‘‘Group Complete’’ interrupt between each update of the Block Count register.
Table 4.29 summarizes these interrupts.
DCBSY
Same as above. Affects the DBSY pin. In effect only during
manual mode.
DCSEL
Same as above. Affects the SEL pin. In effect only during
manual mode.
DCTARG
Same as above. Affects the DTARG pin which controls the
direction of target signals: C/D, I/O, REQ, MSG. In effect
only during manual mode.
DCINIT
Same as above. Affects the DINIT pin which controls the
direction of initiator signals: ACK, ATN. In effect only during
manual mode.
DCRST
Same as above. Affects the DRST pin. In effect only during
manual mode.
SCK(1:0)
By programming these two bits the user can adjust the setup time delay of REQ and ACK signals with respect to the
SCSI Data signals. Delay values of approximately 1, 1.5, or
2 Bus Clock periods are possible, depending on the values
of these bits and the CLK(1:0) bits of the Setup 1 register.
Default values for these bits are 00.
Pipelining SCSI Commands
There are several combinations of SCSI commands that
can be pipelined:
Receive Data followed by Send Disconnect
Receive Data followed by Send End
Send Data followed by Send Disconnect
Send Data followed by Send End
Reselect-Receive Data followed by Send Disconnect
Reselect-Receive Data followed by Send End
Reselect-Send Data followed by Send Disconnect
Reselect-Send Data followed by Send End
Except for the Clear Pending command only these command pairs can be pipelined. The Send End, Send Disconnect and Clear Pending commands are the only commands
allowed to be pipelined. No additional commands can be
written to the SCSI Command register until the pipelined
command has completed its operation.
The Clear Pending Command can be issued to clear the
pipelined command if the first command is still executing
and the pipelined command has not yet begun.
In addition, changing the SCSI pointer or the Block Count
register does not affect the pipelined Send End or Send
Disconnect command. If the Block Count register is modified after the Send End or Send Disconnect command has
been pipelined the current executing SCSI command may
possibly be extended.
A pipelined command should not be issued before 10 BCLK
periods have passed since the first command was issued.
It is important to understand how the DP8496/7 determines
completion interrupts when pipelining commands. It is possible to completely miss a completion interrupt if errors do
not occur for either of the pipelined commands. Refer to
Table 4.27 for a description of when completion interrupts
are generated and how they are buffered.
TABLE 4.28. SCSI Strobe Time Settings
SCSI Clock Bits (1:0)
BCLK
Freq
(1:0)
00
01
11
10
00
1.0
1.0
2.0
1.5
01
1.5
1.0
2.0
1.5
10
1.5
1.0
2.0
1.5
11
1.5
1.0
2.0
1.5
Where:
1.0 is defined as bcp b k; k is a characterization constant
1.5 is defined as bcp a bc b k; bc is either bcl or bch
2.0 is defined as 2*bcp b k; k is a characterization constant
bcp, bch and bcl refer to the BCLK period, high pulse width, and low pulse
width, respectively.
4.4.3. SCSI Operations
Extending the Current SCSI Command
SCSI commands may be programmed to transfer more
blocks than originaly programmed. If a new value is written
to the Block Count holding register while a SCSI command
is executing, the new value will be downloaded from the
holding register to the active register when the active Block
Count reaches zero. This new Block Count must be written
to the Block Count holding register any time before the active Block Count reaches zero. The Block Count register
should not be updated again (after the first pipelined value)
until the previous pipelined value has been downloaded.
The ‘‘Group Complete’’ interrupt indicates when the pipelined Block Count is downloaded.
61
4.0 Functional Description (Continued)
TABLE 4.29. SCSI Bufer Management Interrupts
Block
Count
Reached
Zero?
New SCSI
Command
Been
Loaded?
New Block
Count Been
Loaded
Interrupt
Type
Comments
No
X
X
BTC
Single block transfer complete.
Yes
X
Yes
GTC
The last block of the Group has been transferred to/from the SCSI Bus.
Command is extended with new Block Count.
Yes
Yes
No
CCI
Previous SCSI command has completed. New pipelined command is
starting.
Yes
No
No
CCI
Previous SCSI command has completed. No new command.
Note: BTC e Single Complete Interrupt, GTC e Group Complete Interrupt, CCI e Command Complete Interrupt.
Data written to the SCSI Data register will be asserted immediately on the SCSI bus and data read from the SCSI
Data register will be an exact and immediate reflection of
the SCSI bus.
By setting the HE bit to a ‘‘1’’, the SCSI Data register becomes latched and the processor can easily transfer single
bytes of data. The DBR (Data Buffer Ready) bit in the SCSI
Status register should be polled to pace the data transfer. If
sending data out of the DP8496/7, the DBR bit should be
polled for a ‘‘0’’. This indicates the SCSI Data register is
empty and ready for another byte to be sent. If receiving
data into the DP8496/7, the DBR bit should be polled for a
‘‘1’’. This indicates the SCSI Data register is full and a byte
is ready to be read by the processor.
Automatic Mode
The Automatic Mode transfers data between the SCSI bus
and external buffer memory using internal DMA. The Automatic Mode is enabled by setting the A/M (Automatic/Manual) bit in the SCSI Operation register.
If the Synchronous Transfer register is set for synchronous
transfers, the transfer mode will automatically change between synchronous and asynchronous depending on the information transfer phase, if the DP8496/7 is configured as a
Target. If configured as an Initiator, the Synchronous Transfer register must be setup correctly for the phase prior to
issuing the Transfer Info command.
Transfers of odd byte blocks are allowed in both Asynchronous and Synchronous modes. However, the SCSI FIFO will
be flushed after each block transferred if the block size is
odd in word mode. Therefore, the maximum data transfer
rate will be achieved only for transfers of even byte counts
while in word mode. While in byte mode, odd and even
transfers are equally fast.
In both synchronous and asynchronous transfer modes,
successive requests or acknowledges cannot occur faster
than two bus clock cycles apart. If this condition occurs, a
‘‘SCSI Bus Error’’ interrupt will be generated and a ‘‘SCSI
Data Overrun’’ error code will be set in the SCSI Status
register.
A Few Cautions (while in Manual Mode):
If the DP8496/7 is a target receiving information from the
SCSI Bus then the processor should switch the I/O line
before reading the last byte. This is to prevent the
DP8496/7 from issuing another REQ and causing a DMA
overrun condition on the Initiator side.
If in Manual Mode and not active on the SCSI bus, several
conditions must be met to prevent the DP8496/7 from driving the SCSI bus or causing unexpected error conditions
within itself. The HE (Handshake Enable) bit in the Synchronous Transfer register must be 0, the SCSI Data register
must be 00, the SPP (SCSI Parity Polarity) bit in the Setup 2
register must be 1 (even parity). These conditions are all set
up upon assertion of the CRST pin or a SCSI Reset command. Be sure that these conditions are set up before
switching the A/M bit from Automatic to Manual Mode.
If connected as a Target in Manual Mode, REQ is asserted
as soon as the HE (Handshake Enable) bit in the Synchronous Transfer register is switched from ‘‘0’’ to ‘‘1’’.
Note that no DMA transfers occur between the SCSI bus
and buffer memory while in Manual Mode.
While in Manual Mode, there is no automatic protection
against generating illegal SCSI operations. Care should be
taken to ensure that other devices on the SCSI bus are not
adversely affected by custom sequences.
Manual Mode Information Transfer
Manual Mode is enabled by clearing the A/M (Automatic/
Manual) bit in the SCSI Operation register. This is the default mode after a CRST or a SCSI Reset Command. While
in this mode all the commands in the SCSI Command register become unavailable. This mode is intended to accommodate lower level processor controlled transfers which
may be unique to an application. In other words, the SCSI
protocol becomes completely processor controlled.
A typical sequence may be to issue a (Re)Select command
in the Automatic Mode, since this is not easy or fast in the
Manual Mode, then switch to Manual Mode for further, perhaps unique, transfers or phase sequences.
Data can then be transferred in two different ways while in
Manual Mode. By setting the HE (Handshake Enable) bit in
the Synchronous Transfer register to a ‘‘0’’, the SCSI data
bus appears as a simple transceiver without a latch. This
mode is intended for unique applications which require processor control over the Arbitration and Selection phases.
Differential Transceiver Control in Manual Mode
On the DP8497, fourteen pins are dedicated to the task of
controlling the direction of off-chip differential transceivers
required for Fast SCSI type applications. Control of these
62
4.0 Functional Description (Continued)
transfer sequence. Simply issue the appropriate Reselect
combination command and the Reselection sequence will
automatically be skipped. The command will start by sending the Idenfity message.
The ‘‘Wait-for-Select’’ type commands are terminated either
by completing their execution, being overwritten by another
command, or aborted by an unexpected condition.
However, if a ‘‘Wait for Select’’ command is overwritten by
another command after it has started responding to a selection, but before an interrupt has been issued, a ‘‘Last Command Ignored’’ interrupt will be generated through the SCSI
Interrupt register.
It is important to consider the SCSI Pointer while overwriting
a ‘‘Wait for Select’’ with another command such as a Reselect combination command. Typically two different pointer
locations will be necessary; one for the Command Descriptor Block of the ‘‘Wait for select’’ command and one for the
data block(s) of the Reselect command. The software must
make sure there is no chance the ‘‘Wait for Select’’ command is executing before changing the SCSI Pointer. The
following procedure will ensure proper SCSI Pointer modification.
1. Issue a Clear Pending Command to overwrite the ‘‘Wait
for Select’’ command.
2. Read the SCSI Status register and check for the Selection in Process code.
3. If the SCSI Bus is free (Status Code e 0000 or 0001),
then the ‘‘Wait for Select’’ command has been successfully aborted. You are free to update the SCSI Pointer.
4. If any other Status Code, wait for an interrupt to indicate
the completion of the ‘‘Wait for Select’’ command (or
perhaps an unexpected reselection). Then, update the
SCSI Pointer.
signals in manual mode can be accomplished by writing to
the Differential SCSI 1 and 2, SDIF1 and SDIF2 (4Eh and
4Fh), registers.
SCSI Bus Reset
Normally, if the RST signal on the SCSI bus becomes asserted, the DP8496/7 will deassert all the SCSI bus signals.
A ‘‘SCSI Bus Reset Received’’ interrupt will be generated. If
a SCSI command is executing, it will be terminated immediately with a ‘‘SCSI Bus Reset Received’’ interrupt. This interrupt may be buffered after other pending interrupts as
defined in Table 4.27.
The RST signal on the SCSI bus will be ignored by the
DP8496/7 for only one condition. This condition is while in
the Manual Mode with the HE (Handshake Enable) bit in the
Synchronous Transfer register is set to ‘‘0’’. In this situation
the DP8496/7 will not be affected by the RST signal. This is
a dangerous mode to be in on a normally operating SCSI
bus! The intended use of this mode is to allow manipulation
of SCSI bus signals while SCSI Bus Reset shuts down all
other devices on the SCSI bus.
Target and Initiator Modes
The DP8496/7 has commands in the Automatic Mode for
both the Initiator and Target roles. However, since the chip
will primarily be used in the Target Mode, there are combination commands which optimize performance for this role.
Single- and multi-phase commands are available for the Initiator role, for example: Select, Transfer Info and Transfer
Pad. Single- and multi-phase commands are also available
for the Target role. But, combination commands are also
available in the Target mode which can minimize processor
overhead. Manual mode can be used with either Target or
Initiator modes.
SCSI Parity
Odd or even parity generation and checking on SCSI transfers is offered. The user enables parity and selects parity
polarity by setting the SPE and SPP bits of the Setup 2
register. Refer to Setup 2 register (61h) description in Section 3.2 and Table 4.10 in Section 4.2.9 for more details.
When in Initiator mode, the DP8496/7 will assert ATN
whenever it detects a parity error on the received data, if
SCSI parity is enabled.
Unexpected (Re)Selection
Even if no command is currently executing, the DP8496/7
will still respond to a (Re)Selection from an Initiator or Target broadcasting the proper ID while in Automatic Mode.
This will generate a ‘‘(Re)Selected’’ interrupt.
Asserting CRST or issuing a SCSI Reset command will set
the Manual Mode to avoid a Selection response before the
system is initialized.
A (re)selection is usually asynchronous to normal operations. Therefore, new commands may be written to the SCSI
Command register at critical times related to a (re)selection.
Here is a summary of responses by the DP8496/7 based on
the relationship between writing to the SCSI Command register and a (re)Selection attempt. This assumes that the
DP8496/7 is not connected and there is no command currently executing.
1. A new command is written just before (re)selection. Assuming the DP8496/7 losses arbitration and the winning
initiator (or Target) selects the DP8496/7, a
‘‘(Re)Selected’’ interrupt will be generated and a ‘‘Last
Command Ignored’’ status will be reported.
2. A new comand is written during a (re)selection. In this
case the DP8496/7 will generate a ‘‘(Re)Selected’’ interrupt and ‘‘Last Command Ignored’’ interrupt.
3. A new command is written after a (re)selection. Since the
(re)selection is complete, a ‘‘(Re)Selected’’ interrupt will
be generated and a ‘‘Last Command Ignored’’ status will
be reported.
Combination Commands
Combination commands bring together most of the commonly used sequences for Target operation. Their primary
purpose is to reduce the number of interrupts and thus the
processor overhead required in SCSI transactions.
Most combination commands may be issued before or after
a SCSI connection. However, two commands (Send Disconnect and Send End) can only be issued after a SCSI connection has been established.
The other combination commands are of two classes: those
which wait to be selected and those which reselect. Both of
these classes can be issued before or after a SCSI connection has been established. For example, an Initiator may
have selected the Target before any ‘‘Wait for Select’’ command was issued by the local processor. This ‘‘Wait for Select’’ command can still be issued after the selection is complete and the specified transfer phases will be accomplished. Likewise, the Target may have used the standalone Reselect command to re-establish a connection with
some Initiator and now wish an Identify message and data
63
4.0 Functional Description (Continued)
4.5 TIMER
DCOZ: Drive Command On Zero
Since there are time-outs of various lengths used throughout SCSI bus transactions and time periods that need to be
measured for other external and internal system timing, a
general purpose Timer is put on-chip. It can be programmed
to generate interrupts at a constant rate. In addition to Selection/Reselection time-outs and monitoring for a ‘‘hung’’
bus, this Timer can be reset by the index pulse and used to
check disk rotational speed, seek time-outs, head position
and many other events. Execution of disk commands can
also be delayed until a time-out for positioning of certain
operations on disk data.
The Timer interrupt is observable in the Disk Interrupt register and is maskable in the Disk Interrupt Enable register.
Two registers control the timer, the Timer Prescale register
and the Timer Count register. Timer Prescale not only controls the divisor, but three set/reset functions. Timer Count
is simply the number of periods in the interval and it reads
back the current count. Timer Count is loaded last and
starts the timer.
1
0
Prescale Code
TABLE 4.31. Prescale Code
Prescale
Code
3 2 1 0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
TABLE 4.30. Timer Periods
BCLK
Min Period
Max Period
10 MHz
400 ns
3.36 sec.
14 MHz
286 ns
2.40 sec.
18 MHz
222 ns
1.86 sec.
22 MHz
182 ns
1.53 sec.
4.5.1 Timer Register Descriptions
Timer Prescale (TPRE)
7
6
5
LOI
ROZ
DCOZ
63h
4
x
3
2
1
Drive commands loaded to Disk Command register will
not be started until the Timer Count register counts
down to zero. The Timer Count register should be loaded after the Disk Command register with a value calculated to position the command at the track position desired. Typically the LOI (Load On Index) bit will be set
and command would be a Read or Write Unformatted.
Drive commands are executed normally.
R/W
0
0
0
0
0
1
1
1
1
0
0
0
0
1
1
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
0
1
1
Prescale
Code (Hex)
Prescale
Value
0
1
2
3
4
5
6
7
8
9
A
B
C
D
E
F
4
8
16
32
64
128
256
512
1024
2048
4096
8192
16k
32k
64k
128k
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
Prescale Code
LOI and ROZ Bits
Load On Index and Restart On Zero bits can be used in any
combination. When both the LOI and ROZ bits are set, the
LOI function takes priority. If the Index pulse is received, a
reload is done regardless of the count. The interesting case
is when the Timer Count register reaches zero before the
next index pulse. Multiple interrupts can be generated during the disk revolution with each revolution synchronized to
the index pulse. A useful setup might be to program the
prescale and count to interrupt for each sector on a track.
This register is reset by the CRST pin and the reset commands.
LOI: Load On Index
1 Restarts counting sequence when leading edge of index pulse is received. This works independently of enabling the Index interrupt. This bit reloads the timer
count and leaves the prescale value unchanged. The
next interrupt generated will be [1/BCLK * PRESCALE
* TIMER COUNT] seconds away. No interrupt will be
generated until the Index pulse has been received and
the count value has elapsed.
0 The index pulse has no effect on the Timer.
Timer Count (TCNT)
7
6
5
64h
4
3
2
1
R/W
0
Count
ROZ: Reload On Zero
1 When the Timer Count reaches zero, it is automatically
reloaded with the initial value loaded by the processor.
An interrupt is generated, if enabled, when the Timer
Count register reaches zero. This mode can be used as
a ‘‘rate generator’’ giving an interrupt at a periodic rate.
0 No reloading of the Timer Count register occurs. An
interrupt occurs, if enabled, when the Timer Count register reaches zero. The Timer Interrupt will not occur
again until a new value is loaded into Timer Counter.
The action of loading the register starts the counting process. Timer Count is loaded by the processor and specifies
the number of [1/BCLK * PRESCALE] periods to count. It
decrements to zero and creates an interrupt if enabled. The
Timer Count is readable at any time and contains the current count as it is decrementing.
64
5.0 Application Information: Buffer Memory Interfacing
in a Fast SCSI Implementation
have transferred 4 words in 9 BCLK cycles that same four
word transfer now takes 5 (1 word) a 8 (three words) or 13
BCLK cycles. The system will only cross a page boundary at
the most once every page size. In the DP8496/7 this is 256
addresses. If transfers are never aligned to page boundaries, the following sequence of events will occur for a 256
word transfer:
3 blocks of 54 words each, terminated with a refresh.
1 block of 54 words, terminated with a refresh and a
page boundary.
1 block of 40 words, to complete the 256 word
transfer.
Today’s hard disk systems require high throughput in each
stage from the magnetic media to the CPU. Additionally,
users wish to have wide access to a great many different
drive types and manufacturers. This was the motivation for
the SCSI interface. To achieve high throughput, SCSI-2 defines a differential cable transmission scheme and a Fast
option that allows transfers at 10 MBytes/sec. At the same
time faster serial data rate from the disk drive is required in
order to utilize this increased bus bandwidth. Thus disk data
rates of 33 Mbits/sec, (or 4.2 MBytes/sec) are becoming
common. Into this fray of information exchange is the added
requirements of error correction and drive control, hence
the need for a processor to also have access to the data.
This example application, shown in Figure 5.1, illustrates
how the DP8497 can be used to implement a drive design
that achieves 10 MBytes/sec SCSI transfer rate and a
33 Mbit/sec disk data rate while using only low-cost 100 ns
variety DRAMs. The key feature of the DP8497 that makes
this possible is its word-wide buffer memory port. This example application requires the buffer memory port to provide enough bandwidth to support 10 MBytes/sec (SCSI) a
4.2 MBytes/sec (Disk) a Processor Accesses a DRAM
Refreshes under the maximum load. The DP8496/7 automatically handles refresh if DRAMs are used. The refresh
occurs once every 128 BCLK cycles and takes 5 BCLK cycles to complete. Hence a refresh cycle starts after 123
BCLK cycles. Under the maximum load, the FIFOs will
transfer 4 words in 9 BLCK cycles. This gives over 13 transfers per refresh. Using a word length of 2 bytes, this means
that 2 c 13 c 4 or 104 bytes will be transferred in 123
BLCK cycles. If BCLK is running at 20 MHz, then there is at
least 16.91 MBytes/sec available for disk, SCSI and processor accesses. With the above constraints this allows the
processor to have access to a bandwidth of 2.7 MBytes/sec
under the worst case.
Standard 100 ns fast page mode DRAMs have a cycle time
of 180 ns for a random access. Page mode DRAMs, the
most common type, allow a faster access to data that
shares the same ‘‘page’’ as the previous access. Each of
these ‘‘same page’’ accesses takes 55 ns. Thus with a burst
of 4 words under a high load condition, there is one random
access (usually) followed by 3 fast accesses to the same
page. The minimum time in which these accesses could be
achieved would be 180 ns a 3 c 55 ns or 345 ns. The time
for a burst read of 4 words is 9 BCLKs; thus at a BCLK of
20 MHz, the total time taken is 450 ns. This is more than
enough time for the DRAM. With this ability it is possible to
have a sustained memory bandwidth of 16.9 MBytes/sec
with 100 ns Page mode DRAMs. However, the random and
page access times, cycle time, and setup and hold times
should be checked against the buffer memory timing spec
for the particular type of DRAMs used.
Since page mode DRAMs have a limited page size, there is
often a need to change pages. This action requires a new
row address to be presented to the DRAMs. If a burst will
require the address to change its page, then the burst is
truncated to fit up to the end of the page, and a new burst
transfer occurs at the beginning of the page boundary. The
worst case is when there is only one word that can be transferred at the end of the page. Thus when the burst should
Total of 256 words transferred.
This requires the following number of corresponding cycles:
3 blocks of 128 BCLK cycles.
1 block of 132 cycles, because of the page boundary.
1 block of 90 cycles for the remaining 40 words.
Total of 606 BCLK cycles taken.
Therefore the corresponding transfer rate becomes 16.898
MBytes/sec. Note that the reduction in bandwidth due to
page effects is only 0.7%. Figure 5.3 illustrates the net, attainable bandwidth as a function of BCLK frequency.
The 512 KByte buffer memory shown in the schematic
makes possible large cache buffers for the purpose of reducing effective access times. The 256k c 18 DRAM chip
was chosen to minimize the number of chips, while still
maintaining the word-wide bandwidth. In this way, only one
DRAM chip is used. However, cheaper x4 DRAM chips may
be used at the cost of increased board space. The
DP8496/7s support other widths of DRAMs such as 256k x
4s and 1 Meg x 1s. The SDDC can support the load of up to
12 DRAMs. The example also uses parity to help ensure
that the data from the host computer is not errantly changed
on its way to the magnetic media, or visa versa.
The Rest of the Circuit:
To start, the connection between the SCSI cable and the
differential transceivers requires termination. The termination resistors are not shown in the schematic; however, they
should be the same as defined in the SCSI standard. This is
shown in Figure 5.2.
The DB8497 has the differential transceiver enable controls
on chip. These connect directly to the drive enable pins on
the DS36954As. In this way, complete differential SCSI can
be achieved by only adding the transceiver chips and termination resistors.
The mController used is the HPC46003. This 16-bit processor is capable of achieving high instruction rates and uses a
very compact coding scheme. Other versions of the mController have internal ROM that may be mask programmed.
All versions have at least 256 bytes of internal RAM.
This circuit has an external 32 KBytes of ROM for prototyping. An additional 32 KBytes of RAM is provided to allow the
HPC to download programs through its UART, again for prototyping, and also for additional storage. Since the mController has been forced into 8-bit access, only a single latch
is required to hold the address for the RAM and ROM.
65
5.0 Application Information: Buffer Memory Interfacing
in a Fast SCSI Implementation (Continued)
TL/F/11212 – 7
FIGURE 5.1. Typical Application Diagram
66
5.0 Application Information: Buffer Memory Interfacing
in a Fast SCSI Implementation (Continued)
The setup register SUP2 is concerned with how the processor accesses are performed and the enabling and type of
parity for the buffer and SCSI interfaces. Here it is loaded as
follows:
There is also a PAL programmed to provide the chip select
signals. Additionally, the HPC46003 has many input and
output pins available for drive control.
The disk connections are straight forward. The disk controller accepts the unencoded NRZ from any encoder/decoder,
here the 1,7 ENDEC is shown. The Read channel is the last
block before the head’s preamp. This schematic shows only
those read-channel connections that apply to the disk controller portion of the disk.
The ‘96B/97 have an advanced SCSI controller. This
means the on-board mController does not need extremely
sophisticated firmware for controlling the SCSI interface.
They also have high level disk control commands. The
mController can just issue simple commands with no further
need for intervention.
To set up the DP8496/DP8497 a few registers need to be
initialized. The first setup register is SUP1 at 60h. The bits in
this application are defined as:
BitÝ
Name
Description
7, 6
CLK (1:0) These two bits are set to 01 for a
20 MHz bus clock.
5
SWS
This bit is for SRAM wait states, it does
not apply to systems using DRAMs for
their buffer memory. Here it is set to 0 as
a default.
4
DFP
This bit when set disables the use of
fast page mode DRAMs. Here it is set to
0 because fast page mode is necessary
to acheive the high bandwidth.
3, 2
DD (1:0)
These bits select the DRAM size. Since
256k x 16’s are used, the values for
these bits should be 01.
1
RSEL
This is the RAM select bit. Here it is 1 to
select DRAMs.
0
RWID
RAM Data Path Width. Here it is 0 for
word mode.
BitÝ Name
Description
7
AI
Auto Increment. Depends on how the processor wants to access the buffer memory.
Since it depends on the software, it is set to
0 here for example.
6
BPP
Buffer Parity Polarity. Even parity is chosen
arbitrarily, so this bit is set to 1.
5
BPE
Buffer Parity Enable. This is a 1 to enable the
buffer memory parity checks.
4
SPP
SCSI Parity Polarity. Here Odd parity is chosen, so the bit is set to 0.
3
SPE
SCSI Parity Enable. This is set to 1 to check
the parity on the SCSI bus when the conditions are met.
2–0
SID
SCSI ID. This is set in this example to be 110
(6) to make it the next highest priority after
the initiator. This is a good priority for a boot
drive. But typically, the mC’s firmware will set
this value based on switches or jumpers set
by the user.
Thus the value to be written to the SUP2 register at 61h is
01101110 or 7Eh.
The third setup register SUP3 has only one bit to set for this
configuration. The bit is the CI bit, the Combine Interrupts
bit. It is 0 to make the disk related interrupts appear on the
DINT pin and the SCSI related interrupts to appear on the
SINT pin. Therefore a value of 00 should be written to the
SUP3 register at 62h.
The DP8496/DP8497 devices are extremely efficient in providing high speed disk control and a SCSI interface. The
disk control and interface functions in this application of a
512 KByte buffer, Fast 10 MBytes/sec differential SCSI,
and a high performance mController can be made using only
12 chips.
Therefore the value to load at 60h in the DP8496/DP8497 is
01000110 or 46h.
TL/F/11212 – 8
FIGURE 5.2
67
5.0 Application Information: Buffer Memory Interfacing
in a Fast SCSI Implementation (Continued)
TL/F/11212 – 73
FIGURE 5.3. Net, Attainable Bandwidth of DP8496/DP8497
68
6.0 D.C. Specifications
Operating Conditions
Absolute Maximum Ratings (Notes 1 and 2)
If Military/Aerospace specified devices are required,
please contact the National Semiconductor Sales
Office/Distributors for availability and specifications.
b 0.5V to a 7.0V
Supply Voltage (VCC)
Supply Voltage (VCC)
Operating Temperature (TA)
ESD Tolerance: CZAP e 100 pF
RZAP e 1.5 kX
(Note 4)
b 0.5V to VCC a 0.5V
DC Input Voltage (VIN)
b 0.5V to VCC a 0.5V
DC Output Voltage (VOUT)
b 65§ C to a 165§ C
Storage Temperature (TSTG) Range
1.5W
Package Power Dissipation (PD)
Lead Temperature (TL)
(Soldering, 10 Seconds)
260§ C
Min
4.5
0
2000
Max
5.5
a 70
Units
V
§C
V
DC Electrical Characteristics VCC e 5V g 10%, TA e 0§ C to 70§ C unless otherwise specified. (Note 3)
Symbol
Parameter
Conditions
Min
Max
Units
0.8
V
VIH
High Level Input Voltage
VIL
Low Level Input Voltage
IIH
Input Current
VIN e VCC or GND
g 20
mA
IOZ
Output TRI-STATE Leakage Current
VOUT e VCC or GND
g 20
mA
Average VCC Supply Curent
VIN e VIH or VIL, IOUT e 0 mA,
BCLK e 20 MHz, RCLK e 33 MHz
50
mA
ICC
2.0
V
BUFFER MEMORY PINS (DB1, ADB2, AB3, AB4, ADSo, RDo, WRo)
VOH
High Level Output Voltage
VIN e VIH or VIL, lIOUTl e 20 mA
l
VCC b 0.1
l
VIN e VIH or VIL, IOUT e 8.0 mA
VOL
Low Level Output Voltage (SCSI)
V
3.5
V
VIN e VIH or VIL, lIOUTl e 20 mA
0.1
V
VIN e VIH or VIL, lIOUTl e 8.0 mA
0.4
V
SCSI BUS PINS (SCSI Data and SCSI Control Pins on DP8496 Only)
ILKG
Output High Leakage Current
VOUT e VCC or GND
g 20
mA
VOL
Low Level Output Voltage (SCSI)
VIN e VIH or VIL, lIOUTl e 20 mA
0.1
V
VIN e VIH or VIL, lIOUTl e 48.0 mA
0.5
V
b 500
mA
RDY/MODE PIN IN MODE CONFIGURATION
IIL
Input Low Current
VIN e GND
ALL OTHER PINS (Including SCSI Data, Control, and Transceiver Direction Control Pins on DP8497)
VOH
VOL
High Level Output Voltage
Low Level Output Voltage (SCSI)
VIN e VIH or VIL, lIOUTl e 20 mA
VCC b 0.1
V
VIN e VIH or VIL, lIOUTl e 2.0 mA
3.5
V
l
l
VIN e VIH or VIL, IOUT e 20 mA
0.1
V
VIN e VIH or VIL, lIOUTl e 2.0 mA
0.4
V
SCSI INPUT PINS CHARACTERISTICS (SCSI Data and SCSI Control Pins)
Symbol
Parameter
Conditions
Typ
Units
VHYS
Input Hysteresis
0.2
V
VTH
Nominal Switching Threshold
1.4
V
Note 1: Absolute Maximum Ratings are those values beyond which damage to the device may occur.
Note 2: Unless otherwise specified, all voltages are referenced to ground.
Note 3: These DC Electrical Characteristics are measured staticly, and not under dynamic conditions.
Note 4: Value based on test complying with NSC SOP-5-028 human body model ESD testing using the ETS-910 tester.
69
7.0 A.C. Specifications
Refer to Section 8.0 for AC Timing Test Conditions
The timing values and formulae specified in this preliminary document are based on the chip’s design values; and therefore are
intended to assist the user in making initial design choices. Final AC Specifications will be determined after a thorough characterization of the product has been performend.
7.1 CLOCK TIMING
TL/F/11212 – 35
IDÝ
Symbol
Parameter
DP8496/7-33
DP8496/7-50
Min
Min
Max
Units
1.1
bcl
BCLK Low (Note 1)
18
18
60
ns
1.2
bch
BCLK High (Note 1)
18
18
60
ns
1.3
bcp
BCLK Period (Note 1)
40
40
100
ns
1.4
rcl
RCLK Low
13
9
600
ns
1.5
rch
RCLK High
13
9
600
ns
1.6
rcp
RCLK Period
30
20
1000
ns
Note 1: BCLK frequency should not be less than 10 MHz to ensure proper DRAM refresh and to guarantee proper SCSI bus timing. Please refer to Table 4.21A for
Synchronous Information Transfers.
Note 2: BCLK can be as low as 2 MHz to save power while the chip is idle, however, the DRAM contents will be invalid.
7.2 REGISTER READ (NSC/Intel Mode)
TL/F/11212 – 36
IDÝ
Symbol
2.1
aswi
Address Strobe Width
Parameter
Min
Max
Units
2.2
asdv
Address Strobe to Data Valid
120
ns
2.3
asas
Address Setup to Address Strobe
2.4
ahas
Address Hold from Address Strobe
10
ns
2.5
asrd
Address Strobe to RD Strobe
20
ns
2.6
dvcs
Data Valid from RD
2.7
rdwi
RD Strobe Width
50
2.8
dhcs
Data Hold from RD
20
20
ns
9
ns
50
70
ns
ns
30
ns
7.0 A.C. Specifications (Continued)
7.3 REGISTER WRITE (NSC/Intel Mode)
TL/F/11212 – 37
IDÝ
Symbol
3.1
aswi
Address Strobe Width
Parameter
Min
20
Max
Units
ns
3.2
aswr
Address Strobe to Write Strobe
20
ns
3.3
asas
Address Setup to Address Strobe
9
ns
3.4
ahas
Address Hold from Address Strobe
10
ns
3.5
dvwr
Data Valid to WR Strobe
100
ns
3.6
dhwr
Data Hold from WR Strobe
10
ns
3.7
wrwi
WR Strobe Width
100
ns
7.4 READY PIN
TL/F/11212 – 38
IDÝ
Symbol
4.1
srdyt
Strobe to Assertion of RDY
Parameter
Min
4.2
rdys
RDY Released to End of Strobe
Max
Units
55
ns
55
ns
7.5 MODE PIN
TL/F/11212 – 39
IDÝ
Symbol
Parameter
Min
5.1
crstwi
Chip Reset Strobe Width
5 bcp
5.1
mds
Mode Setup to CRST
100
5.1
mdh
Mode Hold from CRST
50
Note 1: The RDY/MODE pin will be driven low internally. It no longer functions as an output after reset goes inactive.
71
Max
Units
ns
ns
7.0 A.C. Specifications (Continued)
7.6 REGISTER READ (Zilog/Motorola Mode)
TL/F/11212 – 40
IDÝ
Symbol
6.1
aswi
Address Strobe Width
Parameter
Min
Max
6.2
asdv
Address Strobe to Data Valid
6.3
asas
Address Setup to Address Strobe
9
ns
6.4
ahas
Address Hold from Address Strobe
10
ns
6.5
asds
Address Strobe to Data Strobe
20
ns
6.6
dvds
Data Valid from Data Strobe
6.7
dswi
Data Strobe Width
6.8
dhds
Data Hold from Data Strobe
20
6.9
rwas
R/W Strobe to Address Strobe
30
ns
6.10
dsrw
Data Strobe to R/W
30
ns
20
Units
ns
120
50
50
ns
ns
ns
30
ns
7.7 REGISTER WRITE (Zilog/Motorola Mode)
TL/F/11212 – 41
IDÝ
Symbol
7.1
aswi
Address Strobe Width
Parameter
20
ns
7.2
asds
Address Strobe to Write Strobe
20
ns
7.3
asas
Address Setup to Address Strobe
9
ns
7.4
ahas
Address Hold from Address Strobe
10
ns
7.5
dvds
Data Valid to WR Strobe
100
ns
7.6
dhds
Data Hold from WR Strobe
10
ns
7.7
dswi
WR Strobe Width
100
ns
7.8
rwds
R/W Strobe to Data Strobe
10
ns
7.9
dsrw
Data Strobe to R/W Strobe
10
ns
72
Min
Max
Units
7.0 A.C. Specifications (Continued)
7.8 INTERRUPTS (Note 1)
TL/F/11212 – 42
IDÝ
Symbol
Parameter
Formula
BCLK e 20 MHz
Min
8.1
wraint
WR Strobe Assertion to INT Deassertion
8.1
wrdint
WR Strobe Deassertion to INT Assertion
Min e 2bcp,
Max e 3bcp a 30
Units
Max
100
30
ns
180
ns
Note 1: This waveform represents when the Disk Interrupt or SCSI Interrupt register is written to, but interrupts are still pending.
7.9 BUFFER MEMORY STATIC RAM READ
TL/F/11212 – 43
IDÝ
Symbol
Parameter
Formula
BCLK e 20 MHz
Min
Units
Max
9.1
cash
BCLK to Address Strobe High
30
9.2
casl
BCLK to Address Strobe Low
30
9.3
asw
Address Strobe Width
bcp b 5
45
ns
9.4
asrd
Address Strobe to RD Strobe
bch b 5
18
ns
9.5
rdw
RD Strobe Width
W a bcp a bcl b 5
68
9.6
rdd
RD Strobe to Data
W a bcp a bcl b 13
9.7
ds
Data Setup to RD Strobe
9.8
dh
Data Hold from RD Strobe
9.9
adsas
Address Setup to Address Strobe
9.10
adsd
Address Setup to Data
9.11
adhrd
Address Hold from RD Strobe
9.12
adhas
Address Hold from Address
Strobe (16-Bit Mode)
bch b 15
ns
ns
60
13
ns
ns
ns
10
ns
bcl b 5
18
ns
W a 2bcp a bcl b 13
110
ns
bch b 15
8
ns
8
ns
Note 1: While in the 8-bit mode, the last waveform is not used. While in the 16-bit mode, the last waveform is used and the previous waveform represents AB3 and
AB4 only.
Note 2: bcp, bcl & bch refer to the BCLK frequency in use.
Note 3: If the Wait States field in the Setup 1 register is ‘‘00’’, then W e 0. Otherwise, W e bcp.
73
7.0 A.C. Specifications (Continued)
7.10 BUFFER MEMORY STATIC RAM WRITE
TL/F/11212 – 44
IDÝ
Symbol
Parameter
Formula
BCLK e 20 MHz
Min
Units
Max
10.1
cash
BCLK to Address Strobe High
30
10.2
casl
BCLK to Address Strobe Low
30
10.3
asw
Address Strobe Width
10.4
aswr
Address Strobe to WR Strobe
ns
ns
bcp b 5
45
ns
bch b 5
18
ns
68
ns
10.5
wrw
WR Strobe Width
W a bcp a bcl b 5
10.6
dswr
Data Setup to WR Strobe
W a bcp a bcl b 15
58
ns
10.7
dhwr
Data Hold from WR Strobe
bch b 15
8
ns
10.8
adsas
Address Setup to Address Strobe
ns
10.9
adswr
Address Setup to WR Strobe
10.10
adhwr
10.11
adhas
bcl b 5
18
W a 2bcp a bcl b 5
118
ns
Address Hold from WR Strobe
bcl b 15
8
ns
Address Hold from Address Strobe (16-Bit Mode)
bch b 15
8
ns
Note 1: While in the 8-bit mode, the last waveform is not used. While in 16-bit mode, the last waveform is used and the previous waveform represents AB3 and AB4
only.
Note 2: bcp, bcl & bch refer to he BCLK frequency in use.
Note 3: If the Wait States field in the Setup 1 register is ‘‘00’’, then W e 0. Otherwise, W e bcp.
74
7.0 A.C. Specifications (Continued)
7.11 BUFFER MEMORY DYNAMIC RAM READ/WRITE
TL/F/11212 – 45
Timing parameter descriptions and specifications are shown on next page.
Note 1: This diagram shows a fast page mode type operation. Normal one transfer operation would not have multiple CAS cycles and associated column address
and data changes as shown here.
Note 2: RAS may be asserted on either edge of BCLK.
Note 3: WR remains deasserted during a read operation.
Note 4: ADB2 is used for 16-bit mode only.
Note 5: Transition edges shown in dashed line would occur in the normal mode (non fast page mode) transfers.
Note 6: For a Read operation, the last waveform is not used. For a Write operation the second from last waveform is not used.
75
7.0 A.C. Specifications (Continued)
IDÝ
Symbol
Formula
Parameter
Min
BCLK e 20 MHz
Max
Min
Units
Max
11.1
crasa
BCLK to RAS Asserted
30
ns
11.2
crasd
BCLK to RAS and CAS Deasserted
30
ns
11.3
rc
11.4
11.5
ras
rp
Read/Write Cycle Time
RAS/WR Pulse Width
Single xfr
4x Burst
6x Burst
5*bcp
9*bcp
12*bcp
Single xfr
4x Burst
6x Burst
2*bcp a bcl b 10
7*bcp b 5
10*bcp b 10
RAS Precharge Time
11.6
rcd
RAS to CAS Delay Time
11.7
cas
CAS Pulse Width
11.8
cp/cpn
11.9
asr
CAS Precharge Time
250
450
600
3*bcp b 10
7*bcp a 5
10*bcp a 10
2*bcp b 5
112
345
490
ns
140
355
510
95
ns
ns
bcp a bcl b 10
bcp a bcl b 10
62
68
ns
bcp b 5
bcp a 5
45
55
ns
bcl b 5
17
ns
Row Address Setup Time
bch b 10
12
ns
11.10
rah
Row Address Hold Time
bcp b 10
40
ns
11.11
asc
Column Address Setup Time
bch b 10
12
ns
11.12
cah
Column Address Hold Time
bcp b 10
40
11.13
rac
Access Time from RAS (Read Only)
2*bcp a bcl b 10
11.14
cac
Access Time from CAS (Read Only)
bcp b 10
11.15
off
Data Hold from CAS High (Read)
11.16
ds
Data Setup (Write Only)
11.17
dhr
0
bcp a bcl
0
ns
112
ns
40
ns
78
ns
bcmin b 10
32
ns
Data Hold from RAS (Write Only)
2*bcp a bcl b 10
112
ns
bcp b 10
40
ns
bcp a bcl b 10
62
11.18
dh
Data Hold from CAS (Write Only)
11.19
pc
Fast Page Mode Cycle Time
11.20
aa
Access Time from Column Address
11.21
rad
RAS to Column Addr. Delay Time
11.22
rsh
RAS Hold Time
bcp a bcmax b 10
bcp b 10
ns
68
ns
40
ns
bcp b 10
40
ns
68
ns
11.23
ral
Column Addr. to RAS Lead Time
bcp a bch b 10
11.24
wcs
Write Command Setup Time
bcp a bcl b 10
68
ns
11.25
wch
Write Command Hold Time
bcp b 10
40
ns
11.26
ar
Column Add. Hold Time from RAS
2*bcp a bcl b10
112
ns
2*bcp a bcl b 10
112
ns
2*bcp b 10
90
ns
11.27
csh
CAS Hold Time
11.28
crp
CAS to RAS Precharge Time
Note: bcp means Bus Clock period being used. bch and bcl refer to the positive and negative pulse widths of Bus Clock, respectively.
bcmax means the greater of the bch or bcl, depending on the Bus Clock’s duty cycle; bcmin means the lesser of the bch or bcl, depending on the Bus Clock’s duty
cycle.
76
7.0 A.C. Specifications (Continued)
7.12 BUFFER MEMORY DYNAMIC RAM REFRESH
TL/F/11212 – 46
IDÝ
Symbol
Formula
Parameter
Min
BCLK e 20 MHz
Max
Min
Units
Max
12.1
crasa
BCLK to RAS Asserted (See 11.1)
30
ns
12.2
crasd
BCLK to RAS and CAS Deasserted (11.2)
30
ns
12.3
rcr
Read/Write Cycle Time for Refresh
12.4
rasr
12.5
12.6
bcp a bcl
72
RAS Pulse Width for Refresh
2*bcp a bch b 5
117
ns
rpr
RAS Precharge Time for Refresh
2*bcp a bcl b 5
117
ns
csr
CAS to RAS Setup Time
bcp b 5
2*bcp a bch b 5
bcp a 5
ns
45
55
ns
12.7
chr
CAS Hold Time
117
ns
12.8
cpn
CAS Precharge Time
bcp a bcl b 5
67
ns
12.9
rpc
RAS to CAS Precharge Time
bcp a bcl b 5
67
ns
Note 1: bcp means Bus Clock Period being used. bch and bcl refer to the positive and negative pulse widths of Bus Clock in use.
Note 2: RAS may be asserted on either edge of BCLK.
7.13 DISK READ DATA AND READ GATE TIMING
TL/F/11212 – 47
IDÝ
Symbol
Parameter
DP8496/7-33
DP8496/7-50
Min
Min
Max
Units
13.1
rds
Read Data Setup to RCLK
7
5
ns
13.2
rdh
Read Data Hold from RCLK
10
9
ns
13.3
is
Index Setup to RCLK
10
10
ns
13.4
ih
Index Hold from RCLK
10
10
ns
13.5
crg
RCLK to Read Gate
15
77
ns
7.0 A.C. Specifications (Continued)
7.14 DISK WRITE DATA AND WRITE GATE TIMING
TL/F/11212 – 48
IDÝ
Symbol
Max
Units
14.1
rcwch
RCLK to WCLK High (Note 1)
Parameter
Min
15
ns
14.2
rcwcl
RCLK to WCLK Low (Note 1)
15
14.3
rcwga
RCLK to Write Gate Asserted
14.4
rcwgd
18
ns
14.5
wds
Write Data Setup to WCLK (Note 2)
rcl b 5
ns
14.6
wdh
Write Data Hold from WCLK (Note 2)
rch b 5
ns
RCLK to Write Gate Deasserted
15
ns
ns
Note 1: The absolute value of rcwch–rcwcl is 5 ns (max).
Note 2: rcl refers to the RCLK low time in use. rch refers to the RCLK high time in use. This formula is tested ony at the frequencies listed in Table 8.1.
Note 3: The WCLK transitions start before the WGATE is asserted and three cycles occur before WGATE assertion.
7.15 DISK ADDRESS MARK (PSEUDO-HARD SECTORED FORMAT)
TL/F/11212 – 49
IDÝ
Symbol
Parameter
15.1
came
RCLK to Address Mark Enable
15.2
amfs
Address Mark Found Setup to RCLK
Min
Max
Units
15
ns
10
ns
7.16 SCSI RESET
TL/F/11212 – 50
IDÝ
Symbol
Parameter
Min
16.1
srst
SCSI RST to All SCSI Lines Undriven
16.2
rstw
SCSI RST Width (Note 1)
600
2*bcp
Note 1: bcp refers to the BCLK period in use. This formula is tested only at the frequencies listed in Table 8.1.
All SCSI signals are active-high on the DP8497.
78
Max
Units
ns
ns
7.0 A.C. Specifications (Continued)
7.17 SCSI ARBITRATION
TL/F/11212 – 51
IDÝ
Symbol
17.1
Optimal BCLK
Parameter
Worst Case
Min
Max
Min
Max
Units
bfbl
Bus Free to BSY Out Asserted
1.2
1.5
1.2
2.2
17.2
blsl
BSY Out Asserted to SEL Out Asserted
2.4
4.0
2.4
5.0
ms
17.3
sea
SEL Out Asserted to SCSI Data Change
1.2
1.4
1.2
2.1
ms
17.4
blidv
BSY Out Asserted to Valid SCSI ID
b 30
30
b 30
30
ns
17.5
bfidi
Bus Free to Valid SCSI ID
1.2
1.5
1.2
2.2
ms
ms
Note: BCLK is defined in the Setup 1 register. Optimal BCLK is defined as follows:
Bclk Freq Code
7
6
0
1
0
1
0
0
1
1
Optimal Frequency
15 MHz
17.5 MHz
20 MHz
25 MHz
ALL SCSI Signals are active-high on the DP8497.
7.18 SCSI SELECTION AS INITIATOR (W/O Arbitration)
TL/F/11212 – 52
IDÝ
Symbol
Optimal BCLK
Parameter
Min
Worst Case
Max
90
Min
Units
Max
18.1
bsel
BSY In Asserted to SEL Out Deasserted
18.2
idsel
Valid SCSI ID Out to SEL Out Asserted
90
230
90
350
ns
18.3
selid
SEL Out Deasserted to SCSI ID Out Invalid
b 30
30
b 30
30
ns
18.4
idatn
Valid SCSI ID Out to ATN Out Asserted (if Enabled)
b 30
30
b 30
30
ns
18.5
bidi
BSY In Asserted to SCSI ID Out Invalid
90
Note: ALL SCSI Signals are active-high on the DP8497.
79
90
90
ns
ns
7.0 A.C. Specifications (Continued)
7.19 SCSI SELECTION AS INITIATOR (With Arbitration)
TL/F/11212 – 53
IDÝ
Symbol
Optimal BCLK
Parameter
Min
Worst Case
Max
19.1
bobi
BSY Out Released to BSY In Asserted (Note 1)
400
19.2
bisel
BSY In Asserted to SEL Out Deasserted
90
19.3
selid
SEL Out Asserted to Valid SCSI ID Out
1.2
1.4
19.4
selbo
Valid SCSI ID Out to BSY Out Deasserted
90
19.5
seld
SEL Out Deasserted to SCSI ID Out Invalid
b 30
19.6
datn
Valid SCSI ID Out to ATN Out Asserted (If Enabled)
b 30
19.7
bIIDI
BSY In Asserted to SCSI ID Out Invalid
Min
Units
Max
400
ns
90
ns
1.2
2.1
230
90
350
ns
30
b 30
30
ns
30
b 30
30
ns
90
90
ms
ns
Note 1: BSY may be asserted sooner, but it will not be detected until (bobi) time has passed.
All SCSI signals are active-high on the DP8497.
6.20 SCSI RESELECTION AS INITIATOR
TL/F/11212 – 54
IDÝ
Symbol
Optimal BCLK
Parameter
BSY In Deasserted to BSY Out Asserted (Note 1)
Worst Case
Min
Max
Min
Max
400
800
400
1300
Units
20.1
bihbol
20.2
blsh
BSY Out Asserted to SEL In Hold
0
0
ns
20.3
blih
BSY In Asserted to SCSI ID Hold
0
0
ns
Note 1: This time starts when BSY is deasserted AND SEL is asserted AND I/O is asserted AND the logical OR of SCSI ID is valid.
All SCSI signals are active-high on the DP8497.
80
ns
7.0 A.C. Specifications (Continued)
7.21 SCSI SELECTION AS TARGET
TL/F/11212 – 55
IDÝ
Symbol
Optimal BCLK
Parameter
BSY In Deasserted to BSY Out Asserted (Note 1)
Worst Case
Min
Max
Min
Max
400
800
400
1300
Units
21.1
bihbol
21.2
blsh
BSY Out Asserted to SEL In Hold
0
0
21.3
blih
BSY In Asserted to SCSI ID Hold
0
0
ns
21.4
shiov
100
100
ns
SEL In Deasserted to I/O, C/D, MSG Asserted
ns
ns
Note 1: This time starts when BSY is deasserted AND SEL is asserted AND I/O is asserted AND the logical OR of SCSI ID is valid.
All SCSI signals are active-high on the DP8497.
7.22 SCSI RESELECTION AS TARGET
TL/F/11212 – 56
IDÝ
Symbol
Optimal BCLK
Parameter
Min
Max
Worst Case
Min
Units
Max
22.1
bobv
BSY Out Deasserted to BSY In Valid (Note 1)
400
400
ns
22.2
blsu
BSY In Asserted to SEL Out Deasserted
90
90
ns
22.3
sliv
SEL Out Asserted to SCSI ID Out Valid
1.2
1.4
1.2
2.1
22.4
ivbo
SCSI ID Out Valid to BSY Out Deasserted
90
230
90
350
ns
22.5
shdz
SEL Out Deasserted to SCSI Data Deasserted
b 30
30
b 30
30
ns
22.6
sliol
SEL Out Asserted to I/O Asserted
1.2
1.4
1.2
2.1
ms
22.7
blidi
BSY in Asserted to SCSI ID Out In Valid
90
Note 1: BSY may be asserted sooner, but it will not be detected until (bobv) time has passed.
All SCSI signals are active-high on the DP8497.
81
90
ms
ns
7.0 A.C. Specifications (Continued)
7.23 SCSI LOST ARBITRATION
TL/F/11212 – 57
IDÝ
Symbol
Parameter
Min
Max
Units
23.1
slbf
SEL In Low to Bus Free
bcp
400
ns
Note: All SCSI signals are active-high on the DP8497.
7.24 SCSI ABORT (RE)SELECTION
TL/F/11212 – 58
IDÝ
Symbol
Optimal BCLK
Parameter
24.1
bss
BSY Setup to SEL (Last Chance) (Note 1)
24.2
iuso
SCSI ID Released to SEL Out High
82
Units
Max
Min
Max
1 bcp
4 bcp
1 bcp
4 bcp
ns
201
250
201
400
ms
Note 1: bcp refers to the BCLK period in use. This formula is tested only at the frequencies listed in Table 8.1.
All SCSI signals are active-high on the DP8497.
Worst Case
Min
7.0 A.C. Specifications (Continued)
7.25 SCSI ASYNCHRONOUS INFORMATION IN PHASE AS INITIATOR
TL/F/11212 – 59
IDÝ
Symbol
25.1
iodz
25.2
pcreq
25.3
sds
25.4
reqlackl
25.5
sdh
25.6
reqhackh
25.7
phack
Parameter
Min
Typ
I/O Valid to SCSI Data Released
Phase Change to REQ
50
Data Setup to REQ
0
REQ Low to ACK Low (Notes 1 and 2)
Data Hold from ACK
Max
4 bcp
ns
185
0
Phase Hold from ACK High
0
Note 2: Transfer Period (in Synchronous Transfer register) e 0. If Transfer Period e 1 then add an additional bcp.
Note 3: This value is not valid for very last transfer. ACK will stay asserted.
All SCSI signals are active-high on the DP8497.
83
ns
ns
40
Note 1: bcp is the BCLK period in use.
ns
ns
90
REQ High to ACK High (Note 3)
Units
60
ns
ns
7.0 A.C. Specifications (Continued)
7.26 SCSI ASYNCHRONOUS INFORMATION IN PHASE AS TARGET
TL/F/11212 – 60
IDÝ
Symbol
Optimal BCLK
Parameter
Min
26.1
iodz
26.2
pcreq
Typ
Worst Case
Max
Min
Max
Units
I/O Valid to SCSI Data Driven
800
1100
800
1700
ns
Phase Change to REQ
400
550
400
800
ns
26.3
sds
Data Setup to REQ
(Note 3)
(Note 3)
ns
26.4
sdh
Data Hold from ACK
0
0
ns
26.5
acklreqh
ACK Low to REQ High
30
45
TBD
ns
26.6
ackhreql
ACK High to REQ Low (Notes 1 and 2)
115
185
TBD
ns
26.7
phack
Phase Hold from ACK High
40
40
Note 1: bcp refers to the BCLK period in use. This formula is tested only at the frequencies listed in Table 8.1.
Note 2: Transfer Period (in Synchronous Transfer register) e 0. If Transfer Period e 1 then add an additional bcp.
Note 3: This value is programmable. It is a functon of the BCLK Frequency bits in SUP1 register and the SCSI clock bits in SDIF2 register.
The SCSI 1 specification (including cable skew) is 55 ns. The SCSI 2 specification (including cable skew) is 25 ns.
SCSI Clock Bits (1:0)
BCLK
Freq
(1:0)
00
01
11
10
00
1.0
1.0
2.0
1.5
01
1.5
1.0
2.0
1.5
10
1.5
1.0
2.0
1.5
11
1.5
1.0
2.0
1.5
Where 1.0 is defined as bcp b k; k is a characterization constant
Where 1.5 is defined as bcp a bc b k; bc is either bcl or bch
Where 2.0 is defined as 2*bcp b k; k is a characterization constant
All SCSI signals are active-high on the DP8497.
84
ns
7.0 A.C. Specifications (Continued)
7.27 SCSI ASYNCHRONOUS INFORMATION OUT PHASE AS INITIATOR
TL/F/11212 – 61
IDÝ
Symbol
27.1
iodz
27.2
pcreq
Parameter
27.3
sds
27.4
reqlackl
27.5
sdh
27.6
reqhackh
27.7
phack
Min
Typ
Max
45
ns
Phase Change to REQ
50
ns
Data Setup to ACK
(Note 4)
REQ Low to ACK Low (Note 1)
Data Hold from REQ
ns
90
235
0
REQ High to ACK High (Note 3)
Phase Hold from ACK High
40
60
0
Note 2: bcp refers to the BCLK period in use. This formula is tested only at the frequencies listed in Table 8.1.
Note 3: This value is not valid for the very last transfer. ACK will stay asserted.
Note 4: This value is programmable. It is a functon of the BCLK Frequency bits in SUP1 register and the SCSI clock bits in SDIF2 register.
The SCSI 1 specification (including cable skew) is 55 ns. The SCSI 2 specification (including cable skew) is 25 ns.
SCSI Clock Bits (1:0)
Freq
00
00
01
11
10
1.0
1.0
2.0
1.5
01
1.5
1.0
2.0
1.5
10
1.5
1.0
2.0
1.5
11
1.5
1.0
2.0
1.5
(1:0)
Where 1.0 is defined as bcp b k; k is a characterization constant
Where 1.5 is defined as bcp a bc b k; bc is either bcl or bch
Where 2.0 is defined as 2*bcp b k; k is a characterization constant
All SCSI signals are active-high on the DP8497.
85
ns
ns
Note 1: Transfer Period (in Synchronous Transfer register) e 0. If Transfer Period e 1 then add an additional bcp.
BCLK
Units
I/O Valid to SCSI Data Valid
ns
ns
7.0 A.C. Specifications (Continued)
7.28 SCSI ASYNCHRONOUS INFORMATION OUT PHASE AS TARGET
TL/F/11212 – 62
IDÝ
Symbol
Optimal BCLK
Parameter
Min
Typ
Worst Case
Max
I/O Valid to SCSI Data Released
Max
45
ns
400
800
ns
28.1
iodz
28.2
pcreq
45
28.3
sds
Data Setup to ACK
0
28.4
sdh
Data Hold from REQ
0
28.5
acklreqh
ACK Low to REQ High
30
45
TBD
28.6
ackhreql
ACK High to REQ Low (Notes 1 and 2)
70
135
TBD
28.7
phack
Phase Change to REQ
400
Phase Hold from ACK High
550
Units
Min
0
ns
0
40
ns
ns
ns
40
ns
Note 1: bcp refers to the BCLK period in use. This formula is tested only at the frequencies listed in Table 8.1.
Note 2: Transfer Period (in Synchronous Transfer register) e 0. If Transfer Period e 1 then add an additional bcp.
All SCSI signals are active-high on the DP8497.
7.29 SCSI SYNCHRONOUS INFORMATION IN PHASE AS INITIATOR
TL/F/11212 – 63
IDÝ
Symbol
29.1
syds
Parameter
Synchronous Data Setup
Min
0
Max
Units
ns
29.2
sydh
Synchronous Data Hold (Note 1)
20
ns
29.3
reqdp
REQ Deassertion Period
20
ns
29.4
reqap
REQ Assertion Period
20
ns
29.5
ackap
ACK Assertion Period (Note 2)
90
ns
29.6
ackdp
ACK Deassertion Period (Note 2)
90
ns
Note 1: This parameter meets the SCSI 1 specification of 45 ns. Characterization data will determine if the SCSI 2 specification of 10 ns can be obtained.
Note 2: The ACK assertion and deassertion time is programmable. The following equations determine these times:
For n even: 0.5 * n * bcp-k
For n odd: (0.5 * (n b 1) * bcp a bc)-k; where bc is bcl or bch
k is a characterization constantÐtarget value of which is 10 ns.
All SCSI signals are active-high on the DP8497.
86
7.0 A.C. Specifications (Continued)
7.30 SCSI SYNCHRONOUS INFORMATION IN PHASE AS TARGET
TL/F/11212 – 64
IDÝ
Symbol
Parameter
Min
30.1
syds
Synchronous Data Setup (Note 1)
(Note 1)
ns
30.2
sydh
Synchronous Data Hold (Note 2)
(Note 2)
ns
30.3
reqdp
REQ Deassertion Period (Note 3)
(Note 3)
ns
30.4
reqap
REQ Assertion Period (Note 3)
(Note 3)
ns
30.5
ackap
ACK Assertion Period
20
ns
30.6
ackdp
ACK Deassertion Period
20
ns
Note 1: The setup time is programmable. The following equations determine these typical times:
For n e 2: bcp-k
For n e 4, 6: (0.5 * (n b 2) * bcp a bc)-k; where bc is bcl or bch
For n Odd: (0.5 * (n b 1) * bcp)-k
Note 2: The hold time is programmable. The following equations determine these typical times:
For n e 2: bcp-k
For n e 4, 6: (0.5 * n * bcp a bc)-k; where bc is bcl or bch
For n Odd: (0.5 * (n a 1) * bcp)-k
Note 3: The ACK assertion and deassertion time is programmable. The following equations determine these typical times:
For n Even: (0.5 * n * bcp)-k
For n Odd: (0.5 * (n b 1) * bcp a bc)-k; where bc is bcl or bch
k is a characterization constantÐtarget value of which is 10 ns.
All SCSI signals are active-high on the DP8497.
87
Max
Units
7.0 A.C. Specifications (Continued)
7.31 SCSI SYNCHRONOUS INFORMATION OUT PHASE AS INITIATOR
TL/F/11212 – 65
IDÝ
Symbol
Parameter
Min
31.1
reqap
REQ Assertion Period
20
Max
Units
ns
31.2
reqdp
REQ Deassertion Period
20
ns
31.3
sydh
Synchronous Data Hold (Note 2)
(Note 2)
ns
31.4
syds
Synchronous Data Setup (Note 1)
(Note 1)
ns
31.5
ackap
ACK Assertion Period (Note 3)
(Note 3)
ns
31.6
ackdp
ACK Deassertion Period (Note 3)
(Note 3)
ns
Note 1: The setup time is programmable. The following equations determine these typical times:
For n e 2: bcp-k
For n e 4, 6: (0.5 * (n b 2) * bcp a bc)-k; where bc is bcl or bch
For n Odd: (0.5 * (n b 1) * bcp)-k
Note 2: The hold time is programmable. The following equations determine these typical times:
For n e 2: bcp-k
For n e 4, 6: (0.5 * n * bcp a bc)-k; where bc is bcl or bch
For n Odd: (0.5 * (n a 1) * bcp)-k
Note 3: The ACK assertion and deassertion time is programmable. The following equations determine these typical times:
For n Even: (0.5 * n * bcp)-k
For n Odd: (0.5 * (n b 1) * bcp a bc)-k; where bc is bcl or bch
k is a characterization constantÐtarget value for which is 10 ns.
All SCSI signals are active-high on the DP8497.
7.32 SCSI SYNCHRONOUS INFORMATION OUT PHASE AS TARGET
TL/F/11212 – 66
IDÝ
Symbol
32.1
reqap
REQ Assertion Period
Parameter
(Note 1)
Min
32.2
reqdp
REQ Deassertion Period
(Note 1)
ns
32.3
sydh
Synchronous Data Hold
20
ns
32.4
syds
Synchronous Data Setup
0
ns
32.5
ackap
ACK Assertion Period
20
ns
32.6
ackdp
ACK Deassertion Period
20
ns
Note 1: The REQ assertion and deassertion time is programmable. The following equations determine these typical times:
For n even: (0.5 * n * bcp)-k
For n odd: (0.5 * (n b 1) * bcp a bc)-k; where bc is bcl or bch
k is a characterization constantÐtarget value for which is 10 ns.
88
Max
Units
ns
7.0 A.C. Specifications (Continued)
7.33 SCSI BUS FREE FROM INFORMATION PHASE AS INITIATOR
TL/F/11212 – 67
IDÝ
33.1
Symbol
Optimal BCLK
Parameter
bihbf
BSY Deasserted to Bus Free
Worst Case
Units
Min
Max
Min
Max
400
800
400
1200
ns
7.34 SCSI BUS FREE FROM INFORMATION PHASE AS TARGET
TL/F/11212 – 68
IDÝ
Symbol
Parameter
34.1
bohbf
BSY Deasserted to Bus Free
Min
Max
Units
200
ns
Note: All SCSI signals are active-high on the DP8497.
7.35 DIFFERENTIAL TRANSCEIVER DIRECTION CONTROL (DP8497 Only)
TL/F/11212 – 69
IDÝ
Symbol
Parameter
Arbitration
Selection
Info Transfer
Min
Max
Min
Max
Min
Units
Max
35.1
ddsad
Data Direction Signal Assertion
to Data Valid
b 30
30
b 30
30
0
ns
35.2
pdsap
Parity Direction Signal Assertion
to Parity Valid
N/A
N/A
b 30
30
0
ns
35.3
diddsd
Data Invalid to Data Direction
Signal Deassertion
b 30
30
b 30
30
0
ns
35.4
pipdsd
Parity Invalid to Parity Direction
Signal Deassertion
N/A
N/A
b 30
30
0
ns
Note: All SCSI signals are active-high on the DP8497.
89
7.0 A.C. Specifications (Continued)
7.35 DIFFERENTIAL TRANSCEIVER DIRECTION CONTROL (DP8497 Only) (Continued)
TL/F/11212 – 70
IDÝ
Symbol
Min
Max
Units
35.5
bdsab
sdsas
rdsar
BSY Direction Signal Assertion to BSY Assertion
SEL Direction Signal Assertion to SEL Assertion
RST Direction Signal Assertion to RST Assertion
b 30
30
ns
bdbdsd
sdsdsd
rdrdsd
BSY Deassertion to BSY Direction Signal Deassertion
SEL Deassertion to SEL Direction Signal Deassertion
RST Deassertion to RST Direction Signal Deassertion
b 30
30
ns
35.6
Parameter
Note: All SCSI signals are active-high on the DP8497.
7.36 DIFFERENTIAL TRANSCEIVER DIRECTION CONTROL FOR TARGET AND INITIATOR SIGNALS (DP8497 Only)
a) DP8496/DP8497 Attempting SELECT or RESELECT
TL/F/11212 – 71
b) DP8496/DP8497 Being SELECTed or RESELECTed
TL/F/11212 – 72
IDÝ
Symbol
Parameter
Min
Max
Units
36.1
soatds
soaids
SEL Output Assertion to TARG Direction Signal Assertion
SEL Output Assertion to INIT Direction Signal Assertion
bcp b 30
bcp a 30
ns
36.2
sidtds
sidids
SEL Input Assertion to TARG Direction Signal Assertion
SEL Input Assertion to INIT Direction Signal Assertion
bcp
6*bcp
ns
Note 1: DTARG and DINIT are deasserted in the BUS FREE phase.
Note 2: bcp refers to the Bus Clock Period in use.
All SCSI signals are active-high on the DP8497.
90
8.0 A.C. Test Conditions
The A.C. Characteristics in Section 7.0 are tested under the
conditions described below. If a variable is included in the
Min or Max column, the value of that variable should be
based on the conditions used in the current applicationÐ
not the Min or Max specification from a table.
The AC specifications are tested at different BCLK and
RCLK frequencies. The frequencies tested are listed in Table 8.1. If a variable is included in the Min or Max column,
you may calculate an estimate for that characteristic for any
frequency. However, only the frequencies listed in Table 8.1
are guaranteed.
TABLE 8.3. Capacitance
Typ
Units
CIN
Symbol
Input Capacitance
Parameter
5
pF
COUT
Output Capacitance
8
pF
Note 1: TA e 25§ C, f e 1 MHz.
Note 2: These parameters are not 100% tested.
TABLE 8.1. Guaranteed Frequencies
BCLK
15 MHz
17.5 MHz
20 MHz
TL/F/11212 – 9
TABLE 8.2. Test Conditions
Input Pulse Levels
Input Rise and Fall Times
Input and Output Reference Levels
TRI-STATE Reference Levels
FIGURE 8.1. Test Jig
GND to 3V
5 ns
1.3V
Active High b 0.5V
Active Low a 0.5V
Note 1: CL e 110 pF for Buffer Memory Interface pins; 50 pF for all other
pins. Includes jig and scope capacitance.
Note 2: S1 e open for push-pull outputs. S1 e VCC for high impedance to
active low and active low to high impedance measurements. S1 e GND for
high impedance to active high and active high to high impedance measurements. RL e 1.0 kX (See Note 3).
Note 3: For the SCSI interface pins (SCSI Data bus and SCSI Control lines)
S1 e VCC and RL e 150X.
91
DP8496/DP8497 SCSI-2 Disk Data Controller
Physical Dimensions inches (millimeters)
Plastic Quad Flatpack, EIAJ, 100 Lead
Order Number DP8496VF or DP8497VF
NS Package Number VF100B
LIFE SUPPORT POLICY
NATIONAL’S PRODUCTS ARE NOT AUTHORIZED FOR USE AS CRITICAL COMPONENTS IN LIFE SUPPORT
DEVICES OR SYSTEMS WITHOUT THE EXPRESS WRITTEN APPROVAL OF THE PRESIDENT OF NATIONAL
SEMICONDUCTOR CORPORATION. As used herein:
1. Life support devices or systems are devices or
systems which, (a) are intended for surgical implant
into the body, or (b) support or sustain life, and whose
failure to perform, when properly used in accordance
with instructions for use provided in the labeling, can
be reasonably expected to result in a significant injury
to the user.
National Semiconductor
Corporation
1111 West Bardin Road
Arlington, TX 76017
Tel: 1(800) 272-9959
Fax: 1(800) 737-7018
2. A critical component is any component of a life
support device or system whose failure to perform can
be reasonably expected to cause the failure of the life
support device or system, or to affect its safety or
effectiveness.
National Semiconductor
Europe
Fax: (a49) 0-180-530 85 86
Email: cnjwge @ tevm2.nsc.com
Deutsch Tel: (a49) 0-180-530 85 85
English Tel: (a49) 0-180-532 78 32
Fran3ais Tel: (a49) 0-180-532 93 58
Italiano Tel: (a49) 0-180-534 16 80
National Semiconductor
Hong Kong Ltd.
13th Floor, Straight Block,
Ocean Centre, 5 Canton Rd.
Tsimshatsui, Kowloon
Hong Kong
Tel: (852) 2737-1600
Fax: (852) 2736-9960
National Semiconductor
Japan Ltd.
Tel: 81-043-299-2309
Fax: 81-043-299-2408
National does not assume any responsibility for use of any circuitry described, no circuit patent licenses are implied and National reserves the right at any time without notice to change said circuitry and specifications.