TI SMJ34020AGB

SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
D
D
D
D
D
D
D
145-PIN GRID ARRAY PACKAGE
( BOTTOM VIEW )
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
A
B
C
D
E
F
G
H
J
K
L
M
N
P
R
132-PIN QUAD FLATPACK
( TOP VIEW )
100
D
Flexible Multi-Processor Interface
Packaging Options
– 145-Pin Grid Array Ceramic Package
(GB Suffix)
– 132-Pin Ceramic Quad Flat Pack
(Unformed Lead) (HT Suffix)
1
99
33
67
66
D
D
D
132
D
Class B High-Reliability Processing
1-µm CMOS Technology
Military Operating Temperature Range
55°C to 125°C
SMJ34020A-32/ 40
125 / 100-ns Instruction Cycle Time
Fully Programmable 32-Bit
General-Purpose Processor With
512-Megabyte Linear Address Range
(Bit Addressable)
Second-Generation Graphics System
Processor
– Object-Code Compatible With the
SMJ34010
– Enhanced Instruction Set
– Optimized Graphics Instructions
– Coprocessor Interface
Pixel Processing, XY Addressing, and
Window Checking Built Into the Instruction
Set
Programmable 1-, 2-, 4-, 8-, 16-, or 32-Bit
Pixel Size With 16 Boolean and Six
Arithmetic Pixel Processing Options
(Raster Ops)
512-Byte LRU On-Chip Instruction Cache
Optimized DRAM / VRAM Interface
– Page-Mode for Burst Memory Operations
– Dynamic Bus Sizing (16-Bit and
32-Bit Transfers)
– Byte-Oriented CAS Strobes
Flexible Host Processor Interface
– Supports Host Transfers
– Direct Access to All of the SMJ34020A
Address Space
– Implicit Addressing
– Prefetch for Enhanced Read Access
Programmable CRT Control
– Composite Sync Mode
– Separate Sync Mode
– Synchronization to External Sync
Direct Support for Special Features of
1M VRAMs
– Load Write Mask
– Load Color Mask
– Block Write
– Write Using the Write Mask
34
D
D
D
Please be aware that an important notice concerning availability, standard warranty, and use in critical applications of
Texas Instruments semiconductor products and disclaimers thereto appears at the end of this data sheet.
Copyright  1995, Texas Instruments Incorporated
PRODUCTION DATA information is current as of publication date.
Products conform to specifications per the terms of Texas Instruments
standard warranty. Production processing does not necessarily include
testing of all parameters.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
description
The SMJ34020A graphics system processor (GSP) is the second generation of an advanced high-performance
CMOS 32-bit microprocessor optimized for graphics display systems. With a built-in instruction cache, the ability
to simultaneously access memory and registers, and an instruction set designed to expedite raster graphics
operations, the SMJ34020A provides user-programmable control of the CRT interface as well as the memory
interface (both standard DRAM and multiport video RAM). The 4-gigabit (512-megabyte) physical address
space is addressable on bit boundaries using variable width data fields (1 to 32 bits). Additional graphics
addressing modes support 1-, 2-, 4-, 8-, 16- and 32-bit wide pixels.
architecture
The SMJ34020A is a CMOS 32-bit processor with hardware support for graphics operations such as pixel block
transfers (PIXBLTS) during raster operations and curve-drawing algorithms. Also included is a complete set of
general-purpose instructions with addressing modes tuned to support high-level languages. In addition to its
ability to address a large external memory range, the SMJ34020A contains 30 general-purpose 32-bit registers,
a hardware stack pointer, and a 512-byte instruction cache. On-chip functions include 64 programmable I/O
registers that control CRT timing, input / output control, and parameters required by some instructions. The
SMJ34020A directly interfaces to DRAMs and VRAMs and generates raster control signals. The SMJ34020A
can be configured to operate as a standalone processor, or it can be used as a graphics engine with a host
system. The host interface provides a generalized communication port for any standard host processor. The
SMJ34020A also accommodates a multiprocessing or direct memory access (DMA) environment through the
request / grant interface protocols. Virtual memory systems are supported through bus-fault detection and
instruction continuation.
The SMJ34020A provides single-cycle execution of general-purpose instructions and most common integer
arithmetic and Boolean operations from its instruction cache. Additionally, the SMJ34020A incorporates a
hardware barrel shifter that provides a single-state bidirectional shift-and-rotate function for 1 to 32 bits.
The local-memory controller is designed to optimize memory access operations. It also supports pipeline
memory write operations of variable-sized fields and allows memory access and instruction execution in
parallel.
The SMJ34020A graphics-processing hardware supports pixel and pixel-array processing capabilities for both
monochrome and color systems at a variety of pixel sizes. The hardware incorporates two-operand and
three-operand raster operations with Boolean and arithmetic operations, XY addressing, window clipping,
window-checking operations, 1 to n bits-per-pixel transforms, transparency, and plane masking. The
architecture further supports operations on single pixel transfer (PIXT) instructions or on two-dimensional
arrays of arbitrary size (PIXBLTS).
The SMJ34020A’s flexible graphics-processing capabilities allow software-based graphics algorithms without
sacrificing performance. These algorithms include clipping to arbitrary window size, custom incremental-curve
drawing, two-operand raster operations, and masked two-operand raster operations.
The SMJ34020A provides for extensions to the basic architecture through the coprocessor interface. Special
instructions and cycle timings are included to enhance data flow to coprocessors without requiring the
coprocessor to decode the instruction stream, generate system addresses, or move data for the coprocessor
through the SMJ34020A.
2
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Pin Assignments – 145-Pin Grid Array Package
PIN
PIN
PIN
PIN
NUMBER
NAME
NUMBER
NAME
NUMBER
NAME
NUMBER
NAME
A1
C9
RCA8
J1
EMU0
N15
LAD17
A2
VSS
ALTCH
C10
RCA12
J2
GI
P1
A3
CBLNK / VBLNK
C11
LAD30
J3
EMU1
P2
VCC
HWRITE
A4
HSYNC
C12
J13
LAD4
P3
HCS
A5
TR / QE
C13
VSS
VSS
J14
HA30
RCA2
C14
J15
P5
HA27
A7
RCA3
C15
VCC
LAD26
VCC
LAD5
P4
A6
K1
EMU2
P6
HA24
A8
VCC
RCA6
D1
RAS
K2
RESET
P7
HA22
D2
CAS2
K3
LINT2
P8
HA18
A10
RCA7
VSS
NU†
HA14
K14
VSS
LAD3
P9
RCA10
D3
D4†
K13
A11
P10
HA13
A12
SCLK
D13
LAD28
K15
LAD20
P11
HA10
A13
LAD15
D14
LAD11
L1
LINT1
P12
HA7
A14
LAD29
D15
LAD10
L2
CAMD
P13
HA5
A15
VSS
CAS3
E1
R1
L3
LRDY
P14
HBS0
B1
E2
L13
LAD1
P15
LAD0
B2
WE
E3
VCC
CAS1
L14
LAD2
R1
HREAD
B3
VSS
CSYNC / HBLNK
E13
LAD27
L15
LAD19
R2
HA31
B4
E14
LAD25
M1
BUSFLT
R3
HA28
B5
VSYNC
E15
LAD9
M2
PGMD
R4
HA26
B6
RCA0
F1
HRDY
M3
VCLK
R5
HA23
B7
RCA1
F2
R0
M13
R6
HA20
B8
RCA5
F3
M14
R7
HA19
B9
RCA9
F13
VSS
LAD24
VSS
LAD16
M15
LAD18
R8
HA17
B10
RCA11
F14
LAD8
N1
SIZE16
R9
HA16
VSS
HINT
N2
VCC
CLKIN
R10
HA15
R11
HA11
HA9
A9
B11
LAD31
F15
B12
LAD14
G1
B13
VCC
LAD13
G2
HOE
N4
G3
HDST
N5
VSS
HA29
R12
B14
R13
HA8
B15
LAD12
G13
LAD7
N6
HA25
R14
HBS3
C1
CAS0
G14
HA21
R15
VSS
G15
N8
C3
VCC
DDOUT
VSS
LAD23
N7
C2
H1
LCLK1
N9
VSS
VSS
C4
DDIN
H2
EMU3
N10
HA12
C5
H3
LCLK2
N11
HA6
C6
VSS
SF
H13
LAD22
N12
HBS2
C7
RCA4
H14
LAD21
N13
HBS1
N3
C8
VSS
H15
LAD6
N14
† This pin is provided for device orientation purpose only. Make no external connection.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
VCC
3
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Pin Assignments – 132-Pin Ceramic Quad Flatpack Package
PIN
PIN
PIN
NAME
NUMBER
NAME
NUMBER
NAME
NUMBER
NAME
1
CAS3
34
HCS
67
LAD0
100
LAD29
2
CAS2
35
HA31
68
LAD16
101
LAD14
3
CAS1
36
HA30
69
LAD1
102
LAD30
4
CAS0
37
HA29
70
LAD17
103
LAD15
5
VCC
RAS
38
HA28
71
LAD2
104
LAD31
39
HA27
72
LAD18
105
SCLK
VSS
R0
40
HA26
73
RCA12
41
HA25
74
VSS
LAD3
106
8
107
RCA11
9
R1
42
HA24
75
LAD19
108
RCA10
10
HOE
43
HA23
76
RCA9
HDST
44
HA22
77
VCC
LAD4
109
11
110
RCA8
12
HRDY
45
HA21
78
LAD20
111
RCA7
13
HINT
46
HA20
79
LAD5
112
RCA6
14
EMU3
47
HA19
80
LAD21
113
RCA5
15
LCLK1
48
HA18
81
LAD6
114
16
LCLK2
49
HA17
82
LAD22
115
VCC
VSS
17
EMU1
50
83
LAD7
116
RCA4
18
EMU0
51
VSS
HA16
84
LAD23
117
RCA3
19
EMU2
52
HA15
85
118
RCA2
20
GI
53
HA14
86
VSS
VSS
119
RCA1
21
RESET
54
HA13
87
LAD8
120
RCA0
22
LINT2
55
HA12
88
LAD24
121
SF
23
LINT1
56
HA11
89
LAD9
122
TR / QE
24
CAMD
57
HA10
90
LAD25
123
VSYNC
25
BUSFLT
58
HA9
91
LAD10
124
HSYNC
26
SIZE16
59
HA8
92
LAD26
125
CBLNK / VBLNK
27
PGMD
60
HA7
93
LAD11
126
CSYNC / HBLNK
28
LRDY
61
HA6
94
LAD27
127
29
VCC
VCLK
62
HA5
95
128
VSS
VSS
63
HBS3
96
VCC
LAD12
129
ALTCH
31
CLKIN
64
HBS2
97
LAD28
130
DDIN
32
HWRITE
65
HBS1
98
131
DDOUT
33
HREAD
66
HBS0
99
VSS
LAD13
132
WE
6
7
30
4
PIN
NUMBER
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Pin Functions
PIN
NAME
DESCRIPTION
I/O †
LOCAL MEMORY INTERFACE
ALTCH
BUSFLT
O
I
Address latch. The high-to-low transitions of ALTCH can be used to capture the address and status available on LAD.
A transparent latch (such as a 54ALS373) maintains the current address and status as long as ALTCH remains low.
Bus fault. External logic asserts BUSFLT high to the SMJ34020A to indicate that an error or fault has occurred on the
current bus cycle. BUSFLT is also used with LRDY to generate externally requested bus cycle retries so that the entire
memory address is presented again on LAD.
In the emulation mode, BUSFLT is used for write protecting mapped memory (by disabling CAS outputs for the current
cycle).
DDIN
O
Data bus direction in enable. DDIN is used to drive the active-high output enables on bidirectional transceivers (such
as the 54ALS623). The transceivers buffer data input and output on LAD0 – LAD31 when the SMJ34020A is interfaced
to several memories.
DDOUT
O
Data bus direction output enable. DDOUT drives the active-low output enables on bidirectional transceivers (such as
the 54ALS623). The transceivers buffer data input and output on LAD0 – LAD31.
LAD0 – LAD31
I/O
32-bit multiplexed local address/data bus. At the beginning of a memory cycle, the word address is output on
LAD4 – LAD31 and the cycle status is output on LAD0 – LAD3. After the address is presented, LAD0 – LAD31 are used
for transferring data within the SMJ34020A system. LAD0 is the LSB and LAD31 is the MSB.
I
Local ready. External circuitry drives LRDY low to inhibit the SMJ34020A from completing a local-memory cycle it has
initiated. While LRDY remains low, the SMJ34020A waits unless the SMJ34020A loses bus priority or is given an
external RETRY request (through BUSFLT). Wait states are generated in increments of one full LCLK1 cycle. LRDY can
be driven low to extend local memory-read and memory-write cycles, VRAM serial-data-register-transfer cycles, and
DRAM-refresh cycles. During internal cycles, the SMJ34020A ignores LRDY.
LRDY
PGMD
I
Page mode. The memory-decode logic asserts PGMD low if the currently addressed memory supports burst (page
mode) accesses. Burst accesses occur as a series of CAS cycles for a single RAS cycle to memory. LRDY is used with
BUSFLT to describe the cycle termination status for a memory cycle.
PGMD is also used in emulation mode for mapping memory.
SIZE16
I
Bus size. The memory-decode logic can pull SIZE16 low if the currently addressed memory or port supports only 16-bit
transfers. SIZE16 can also be used to determine which 16 bits of the data bus are used for a data transfer.
In the emulation mode, SIZE16 is used to select the size of mapped memory.
DRAM AND VRAM CONTROL
CAMD
I
Column-address mode. CAMD dynamically shifts the column address on the RCA0 – RCA12 bus to allow the mixing
of DRAM and VRAM address matrices using the same multiplexed address RCA0 – RCA12 signals.
CAS0 – CAS3
O
Four column-address strobes. CAS outputs drive the CAS inputs of DRAMs and VRAMs. CAS0 – CAS3 strobe the
column address on RCA0 – RCA12 to the memory. The four CAS strobes provide byte write-access to the memory.
RAS
O
Row-address strobe. RAS output drives the RAS inputs of DRAMs and VRAMs. RAS strobes the row address on
RCA0 – RCA12 to memory.
RCA0 – RCA12
O
Thirteen multiplexed row-address/column-address signals. At the beginning of a memory-access cycle, the row address
for DRAMs is present on RCA0 – RCA12. The row address contains the most significant address bits for the memory.
As the cycle progresses, the memory column address is placed on RCA0 – RCA12. The addresses that are actually
output during row and column times depend on the memory configuration (set by RCM0 and RCM1 in the CONFIG
register) and the state of CAMD during the access. RCA0 is the LSB, and RCA12 is the MSB.
SF
O
Special function pin. SF is the special-function signal to 1M VRAMs that allows the use of block write, load write mask,
load color mask, and write using write mask. SF is also used to differentiate instructions and addresses for the
coprocessor as part of the coprocessor interface.
TR / QE
O
Transfer/output-enable. TR / QE drives the TR / QE input of VRAMs. During a local memory-read cycle, TR / QE functions
as an active-low output enable to gate from memory to LAD0 – LAD31. During special VRAM function cycles, TR / QE
controls the type of cycle that is performed.
† I = input, O = output
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
5
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Pin Functions (Continued)
PIN
NAME
DESCRIPTION
I/O †
DRAM AND VRAM CONTROL (CONTINUED)
WE
O
Write enable. The active low WE drives the WE inputs of DRAMs and VRAMs. WE can also be used as the active
low write enable to static memories and other devices connected to the SMJ34020A local interface. During a
local-memory read cycle, WE remains inactive high while CAS is strobed active low. During a local-memory write
cycle, WE is strobed active low before CAS. During VRAM serial-data-register transfer cycles, the state of WE at
the falling edge of RAS controls the direction of the transfer.
HA5 – HA31
I
Twenty-seven host address input signals. A host can access a long word by placing the address on these lines.
HA5 – HA31 correspond to LAD5 – LAD31 that output the address to the local memory.
HBS0 – HBS3
I
Four host byte selects. HBS0 – HBS3 identify which bytes within the long word are being selected.
HCS
I
Host chip select. A host drives HCS low to latch the current host address present on HA5 – HA31 and the host byte
selects on HBS0 – HBS3. HCS also enables host access cycles to the SMJ34020A I/O registers or local memory.
During the low-to-high transition of RESET, the level on HCS determines whether the SMJ34020A is halted (HCS
is high for host-present mode) or whether it begins executing its reset service routine (HCS is low for self-bootstrap
mode).
HDST
O
Host data-latch strobe. The rising edge of HDST latches data from the SMJ34020A local address space to the
external host data latch on host read accesses. HDST can be used in conjunction with HRDY to indicate that data
is valid in the external data latch.
HINT
O
Host Interrupt. HINT allows the SMJ34020A to interrupt a host by setting the INTOUT bit in the HSTCTLL I/O register.
HINT can also be used to interrupt the host if a BUSFLT or RETRY occurs due to a host access cycle.
HOE
O
Host data latch output enable. HOE enables data from host data latches to the SMJ34020A local address space on
host write cycles. HOE can be used in conjunction with HRDY to indicate data has been written to memory from the
external data latch.
HRDY
O
Host ready. HRDY is normally low and goes high to indicate that the SMJ34020A is ready to complete a host-initiated
read or write cycle. If the SMJ34020A is ready to accept the access request, HRDY is driven high and the host can
proceed with the access. A host can use HRDY logically combined with HDST and HOE to determine when the local
bus access cycles have completed.
HREAD
I
Host read strobe. HREAD is driven low during a read request from a host processor. This notifies the SMJ34020A
that the host is requesting access to the I/O registers or to local memory. HREAD should not be asserted at the same
time that HWRITE is asserted.
HWRITE
I
Host write strobe. HWRITE is driven low to indicate a write request by a host processor. This notifies the SMJ34020A
that a write request is pending. The rising edge of HWRITE is used to indicate that the host has latched data to be
written in the external data transceivers. HWRITE should not be asserted at the same time HREAD is asserted.
HOST INTERFACE
SYSTEM CONTROL
CLKIN
I
Clock input. CLKIN generates LCLK1 and LCLK2, to which all processor functions in the SMJ34020A are
synchronous. A separate asynchronous input clock (VCLK) controls the video timing and video registers.
LCLK1, LCLK2
O
Local output clocks. LCLK1 and LCLK2 are 90 degrees out of phase with each other. They provide convenient
synchronous control of external circuitry to the internal timing. All signals output from the SMJ34020A (except the
CRT timing signals) are synchronous to LCLK1 and LCLK2.
LINT1, LINT2
I
Local interrupt requests. Interrupts from external devices are transmitted to the SMJ34020A on LINT1 and LINT2.
Each local interrupt signal activates the request for one of two interrupt request levels. An external device generates
an interrupt request by driving the appropriate interrupt request pin to its active-low state. LINT1, LINT2 should
remain low until the SMJ34020A recognizes it. LINT1, LINT2 can be applied asynchronously to the SMJ34020A as
they are synchronized internally before use.
RESET
I
System reset. During normal operation, RESET is driven low to reset the SMJ34020A. When RESET is asserted
low, the SMJ34020A’s internal registers are set to an initial known state and all output and bidirectional pins are driven
either to inactive levels or to the high-impedance state. The SMJ34020A’s behavior following reset depends on the
level of the HCS input just before the low-to-high transition of RESET. If HCS is low, the SMJ34020A begins executing
the instructions pointed to by the reset vector. If HCS is high, the SMJ34020A is halted until a host processor writes
a 0 to the HLT bit in the HSTCTLL register.
† I = input, O = output
6
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Pin Functions (Continued)
PIN
NAME
DESCRIPTION
I/O †
POWER
VCC‡
VSS‡
I
Nominal 5-V power supply inputs. Five pins on QFP; Nine pins on PGA.
I
Electrical ground inputs. Nine pins on QFP; 17 pins on PGA.
EMU0 – EMU 2
I
Emulation pins 0 – 2
EMU3
O
Emulation pin 3
GI
I
Bus grant input. External bus arbitration logic drives GI low to enable the SMJ34020A to gain access to the
local-memory bus. The SMJ34020A must release the bus if GI is high so that another device can access the bus.
R1, R0
O
Bus request and control. R1 and R0 indicate a request for use of the bus in a multiprocessor system; they are
decoded as shown below:
EMULATION CONTROL
MULTIPROCESSOR INTERFACE
R1
L
L
H
H
R0
L
H
L
H
Bus Request Type
High-priority bus request
Bus-cycle termination
Low-priority bus request
No bus request pending
A high-priority bus request provides for VRAM serial-data-register transfer cycles (midline or blanked), DRAM
refresh (when 12 or more refresh cycles are pending), or a host-initiated access. The external arbitration logic
should grant the request as soon as possible by asserting GI low.
A low-priority bus request is used to provide for CPU-requested access and DRAM refresh (when less than
12 refresh cycles are pending).
Bus-cycle termination status is provided so that the arbitration logic can determine that the device currently
accessing the bus is completing an access, and other devices can compete for the next bus cycle. A
no-bus-request-pending status is output when the currently active device does not require the bus on subsequent
cycles.
VIDEO INTERFACE
CBLNK / VBLNK
O
Composite blanking / vertical blanking. CBLNK / VBLNK can be programmed to select one of two blanking
functions:
Composite blanking for blanking the display during both horizontal and vertical retrace periods in
composite-sync-video mode
Vertical blanking for blanking the display during vertical retrace in separate-sync-video mode.
Immediately following reset, CBLNK / VBLNK is configured as a CBLNK output.
CSYNC / HBLNK
I/O
Composite sync / horizontal blanking. CSYNC / HBLNK can be programmed to select one of two functions:
Composite sync (either input or output as set by a control bit in the DPYCTL register) in
composite-sync-video mode:
As an input, extracts HSYNC and VSYNC from externally generated horizontal sync pulses
As an output, CSYNC / HBLNK generates active-low composite-sync pulses from either externally
generated HSYNC and VSYNC signals or signals generated by the SMJ34020A’s on-chip video timers
Horizontal blank (output only) for blanking the display during horizontal retrace in separate-sync-video mode.
Immediately following reset, CSYNC / HBLNK is configured as a CSYNC input.
HSYNC
I/O
Horizontal sync. HSYNC is the horizontal sync signal that controls external video circuitry. HSYNC can be
programmed to be either an input or an output by modifying a control bit in the DPYCTL register.
As an output, HSYNC is the active-low horizontal-sync signal generated by the SMJ34020A’s on-chip video
timers.
As an input, HSYNC synchronizes the SMJ34020A video-control registers to externally generated
horizontal-sync pulses. The actual synchronization can be programmed to begin at any VCLK cycle; this
allows for any external pipelining of signals.
Immediately following reset, HSYNC is configured as an input.
† I = input, O = output
‡ For proper SMJ34020A operation, all VCC and VSS pins must be connected externally.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
7
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Pin Functions (Continued)
PIN
NAME
DESCRIPTION
I/O †
VIDEO INTERFACE (CONTINUED)
SCLK
I
Serial data clock. SCLK is the same as the signal that drives VRAM serial data registers. SCLK allows the SMJ34020A
to track the VRAM serial-data-register count, providing serial-register transfer and midline-reload cycles. (SCLK can
be asynchronous to VCLK; however, it typically has a frequency that is a multiple of the VCLK frequency).
VCLK
I
Video clock. VCLK is derived from a multiple of the video system’s dot clock and is used internally to drive the video
timing logic.
I/O
Vertical sync. VSYNC is the vertical sync signal that controls external video circuitry. VSYNC can be programmed to
be either an input or an output by modifying a control bit in the DPYCTL register.
As an output, VSYNC is the active-low vertical-sync signal generated by the SMJ34020A’s on-chip video timers.
As an input, VSYNC synchronizes the SMJ34020A video-control registers to externally generated vertical-sync
pulses. The actual synchronization can be programmed to begin at any horizontal line; this allows for any
external pipelining of signals.
Immediately following reset, VSYNC is configured as an input.
VSYNC
† I = input, O = output
functional block diagram
HA5 – HA31
27
HBS0 – HBS3
4
HCS
HREAD
HWRITE
HINT
HRDY
HDST
HOE
GI
R0
R1
EMU0
EMU1
EMU2
EMU3
CLKIN
LCLK1
LCLK2
Buffer/
Page-mode
Register
Host
Address
Latch
MUX
I/O
Regs
PC
ST
Host
Interface
MultiProcessor
Interface
DRAM/
VRAM
Interface
Decode
SP
Local
Memory
and
Bus
Timing
Bus
Interface
ALU
Microcontrol ROM
Barrel
Shifter
Reset and Interrupts
RESET, LINT1,
LINT2
8
13
RCA0 – RCA12
DDIN
DDOUT
RAS
Bus
Control
LRU
Emulation
Interface
System
Clocks
LAD0 – LAD31
Cache
Register
File A
Register
File B
32
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Video
Timing
and
Control
4
CAS0 – CAS3
WE
TR / QE
ALTCH
SF
PGMD
SIZE16
LRDY
BUSFLT
CAMD
VSYNC
HSYNC
CSYNC / HBLNK
CBLNK / VBLNK
VCLK
SCLK
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
architecture (continued)
register files
Boolean, arithmetic, pixel-processing, byte, and field-move instructions operate on data within the
general-purpose register files. The SMJ34020A contains two register files of fifteen 32-bit registers and a
system stack pointer (SP). The SP is addressed in both register file A and register file B as a sixteenth register.
Transfers between registers and memory are facilitated using a complete set of field move instructions with
selectable field sizes.
The 15 general-purpose registers in register file A are used for high-level language support and
assembly-language programming. The 15 registers in register file B are dedicated to special functions during
PIXBLTS and other pixel operations but can be used as general-purpose registers at other times.
stack pointer (SP)
The stack pointer is a dedicated 32-bit internal register that points to the top of the system stack.
program counter (PC)
The SMJ34020A’s 32-bit program counter register points to the next instruction-stream word to be fetched.
Since instruction words are aligned to 16-bit boundaries, the four LSBs of the PC are always zero.
instruction cache
An on-chip cache contains 512 bytes of RAM and provides unimpeded access to instructions. The cache
operates automatically and is transparent to software. The cache is divided into four 128-byte segments.
Associated with each segment is a 22-bit segment start address register (SSA) to identify the addresses in
memory corresponding to the current contents of the cache segment. Each cache segment is further partitioned
into eight subsegments of four long words (32 bits) each. Each subsegment has an associated present (P) flag
to indicate whether or not the subsegment contains valid data.
The cache is loaded only when an instruction requested by the execution section of the SMJ34020A is not
already contained within the cache. A least-recently-used (LRU) algorithm determines which of the four
segments of the cache is overwritten with new data. For this purpose, an internal four-by-two LRU stack keeps
track of cache usage. Although the cache is loaded so as to always fill a subsegment completely, not all eight
subsegments within a segment are necessarily filled (this is dependent upon the instruction stream).
status register
The status register (ST) is a special purpose 32-bit register dedicated to status codes set by the results of implicit
and explicit compare operations and parameters used to specify the length and behavior of fields 0 and 1. During
an interrupt, when the IX bit in the ST is placed on the stack, it indicates that execution of an interruptable
instruction (PIXBLT, FILL or LINE) was halted to service the interrupt. The single-step bit causes a trap to the
single-step vector (located at address FFFF FBE0h) after the execution of one instruction when the bit is set
high. Normal program execution occurs when the bit is set low.
fields, bytes, words, long words, pixels and pixel arrays
The SMJ34020A outputs a 28-bit address on LAD4 – LAD31 that is valid at the falling edge of ALTCH. The most
significant 27 bits (LAD5 – LAD31) define a 32-bit-long word of physical memory; logically, however, the
SMJ34020A views memory data as fields addressable at the bit level. The least significant bit of the 28-bit
address (LAD4) is used to select the odd or even word when accessing 16-bit memories (indicated by SIZE16
asserted low). Primitive data types supported by the SMJ34020A include bytes, words, long words, pixels, two
independent fields of from 1 to 32 bits, and user-defined pixel arrays.
Words and long words, respectively, refer to 16- and 32-bit values that are aligned on 32-bit boundaries.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
9
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
fields, bytes, words, long words, pixels and pixel arrays (continued)
The two independent fields are referenced as field 0 and field 1. The attributes of these fields (field size and sign
extension within a register) are defined in the status register as FS0, FE0, FS1, and FE1. Fields 0 and 1 are
specified independently to be signed or unsigned and from 1 to 32 bits in length. Bytes are special 8-bit cases
of the field data type, while pixels are 1, 2, 4, 8, 16, or 32 bits in length. In general, fields (including bytes) can
start and terminate on arbitrary bit boundaries; however, pixels must pack evenly into 32-bit-long words.
pixel operations
Pixel arrays are two-dimensional data types of user-defined width, length, pixel depth (number of bits per pixel),
and pitch (distance between rows). A pixel or pixel array can be accessed by means of either its memory address
or its XY coordinates. Transfers of individual pixels or pixel blocks are influenced by the pixel processing,
transparency, window checking, plane masking, pixel masking, or corner adjustment operations selected. For
further information, see the TMS32020 User’s Guide, literature number SPVU019.
transparency
Transparency is a mechanism that allows the surrounding pixels in an array to be specified as invisible. This
is useful for ensuring that only the object and not the rectangle surrounding it are written to the display. The
SMJ34020A provides four transparency modes:
D
D
D
D
D
No transparency
Transparency on result equal zero
Transparency on source equal COLOR0
Transparency on destination equal COLOR0
Refer to the TMS34020 User’s Guide for more information.
I/O registers
The SMJ34020A contains an on-chip block of sixty-four 16-bit locations (mapped into the SMJ34020A’s
memory address space) that are used for I/O control registers. Eight of these are used by the host interface logic
and are not available to the user. Forty-seven I/O registers control parameters necessary to configure the
operation and report status of the following interfaces:
D
D
D
D
D
D
Host interface
Local memory
Video timing
Screen refresh
External interrupts
Internal interrupts
host interface registers
The host interface registers (HSTDATA, HSTADRL, HSTADRH, HSTCTLL, and HSTCTLH) are provided to
facilitate communications between the SMJ34020A and a host processor and maintain compatibility with the
SMJ34010. The registers are mapped into five of the I/O locations accessible to the SMJ34020A.
Two of these registers (HSTCTLL and HSTCTLH) are used to provide control by the host. This control consists
of the passing of interrupt requests, flushing the instruction cache, halting the SMJ34020A, transmitting a
non-maskable interrupt request to the SMJ34020A, enabling emulation interrupts, and setting host access
modes and configurations.
The other three registers are simple read/write registers to allow the SMJ34020A software to leave addresses
for the host at a known location and allow compatibility with some SMJ34010 software.
10
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
memory interface control registers
Some of the I/O registers are used to control various local memory interface functions, including:
D
D
D
D
D
Frequency of DRAM refresh cycles
Masking (read/write protection) of individual color planes
DRAM row/column addressing configuration
Accessing mode (big endian/little endian)
Bus fault and retry recovery
video timing and screen refresh
Twenty-eight I/O registers are dedicated to video timing and screen refresh functions. The SMJ34020A can be
configured to drive composite sync or separate sync displays.
In composite sync mode, the SMJ34020A can be set to extract VSYNC and HSYNC from an external CSYNC
or it can be used to generate CSYNC from separate VSYNC and HSYNC inputs. Internally, the SMJ34020A
can be set to preset the horizontal and vertical counts on receipt of an external sync signal. This allows
compensation for any combination of internal and external delays that occur in the video synchronization
process. The HCOUNT register is loaded from SETHCNT by an external HSYNC, VCOUNT is loaded from
SETVCNT on an external VSYNC, and an external CSYNC loads both HCOUNT and VCOUNT from SETHCNT
and SETVCNT, respectively.
The SMJ34020A directly supports VRAMs by generating the serial-data-register transfer cycles necessary to
refresh the display. The memory locations from which the display information is taken, as well as the number
of horizontal scan lines displayed between serial-data-register transfer cycles, are programmable.
The SMJ34020A supports various display resolutions and either interlaced or noninterlaced video. The
SMJ34020A can optionally be programmed to synchronize to externally generated sync signals so that images
created by the SMJ34020A can be superimposed upon images created externally. The external sync mode can
also be used to synchronize the video signals generated by two or more SMJ34020As in a multiple-SMJ34020A
graphics system.
CPU control registers
Five of the I/O registers (CONVDP, CONVMP, CONVSP, CONTROL, and PSIZE) provide CPU control to
configure the SMJ34020A for operation with specific characteristics. These characteristics include pitches for
pixel transfers, window checking mode, Boolean or arithmetic pixel processing operation, transparency mode,
PIXBLT direction control, and pixel size.
interrupt interface registers
Two dedicated I/O registers (INTENB and INTPEND) monitor and mask interrupt requests to the SMJ34020A,
including two externally generated interrupts and three internally generated interrupts. An internal interrupt
request can be generated on one of the following conditions.
D
D
D
D
D
Window violation: an attempt has been made to write a pixel to a location inside or outside a specified
window boundary.
Host interrupt: the host processor has set the interrupt request bit in the host control register.
Display interrupt: a specified horizontal line in the frame has been displayed on the screen.
Bus fault
Single-step emulator
A nonmaskable interrupt occurs when the host processor sets a control bit in the host interface register (NMI
in HSTCTLH). The host-initiated interrupt is associated with a mode bit (NMIM in HSTCTLH) that enables and
disables saving of the processor state on the stack when the interrupt occurs. This is useful if the host wishes
to use the host interrupt before releasing the SMJ34020A to execute instructions (that is, before the stack
pointer is initialized). A dedicated terminal controls the SMJ34020A reset function.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
11
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
memory controller/local-memory interface
The memory controller manages the SMJ34020A’s interface to the local memory and automatically performs
the bit alignment and masking necessary to access data located at arbitrary bit boundaries within memory. The
memory controller operates autonomously with respect to the CPU. It has a write queue one field (1 to 32 bits)
deep that permits it to complete those memory cycles necessary to insert a field into memory without delaying
the execution of subsequent instructions. Only when a second memory operation is required before completion
of the first operation is the SMJ34020A forced to defer execution of the subsequent instruction.
The SMJ34020A directly interfaces to standard DRAMs and in particular, to standard video RAMs (VRAMs)
such as the SMJ44C25x multiport VRAMs. The SMJ34020A memory interface consists of the local
address/data bus (LAD), the DRAM row/column address (RCA) bus, and associated control signals. The
currently selected word address (28 bits) and status (4 bits) are multiplexed with data on LAD. The RCA bus
allows direct connection to address/address multiplexed DRAMs from 64K to 16M. Refresh for DRAMs is
supported by CAS-before-RAS (CBR) refresh cycles.
BIT 232 –1
(Last Bit in Memory)
ADDRESS
FFFFFFF0h
68 Words
FFFFFBC0h
FFFFFBB0h
444 Words
FFFFE000h
FFFFDFF0h
65024 Words
FFFF0000h
FFEFFFF0h
226 – 66560 Words
(67 042 304 Words)
Interrupt Vectors and
Extended Trap Vectors
Reserved for Interrupt Vectors
and Extended Trap Vectors
General Use and
Extended Trap Vectors
General Use
C0004000h
C0003FF0h
512 Words
Reserved for System I/O
448 Words
Reserved for I/O Registers
64 Words
I/O Registers
(3 × 226) – 64K
(201 261 056 Words)
General Use
C0002000h
C0001FF0h
C0000400h
C00003F0h
C0000000h
BFFFFFF0h
00100000h
000FFFF0h
64K Words
General Use and Extended
Trap Vectors
00000000h
16
Bit 0
(First Bit in Memory)
Figure 1. Memory Map
12
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
reset
Reset puts the SMJ34020A into a known initial state. This state is entered when the input signal at RESET is
asserted low. While RESET remains asserted, all outputs are in a known state, no DRAM refresh cycles take
place, and no screen refresh cycles are performed.
The state of the HCS input on the CLKIN cycle before the low-to-high transition of RESET determines whether
the SMJ34020A is halted or begins executing instructions. The SMJ34020A can be in one of two modes,
host-present or self-bootstrap mode.
Host-present mode: if HCS is high at the end of reset, SMJ34020A instruction execution halts and remains
halted until the host clears the HLT (halt) bit in HSTCTLH (host control register). Following reset, the RAS cycles
required to initialize the dynamic RAMs are performed automatically by the GSP memory control logic. The host
can request a memory access after the eight RAS initialization cycles have completed. The SMJ34020A
automatically performs DRAM refresh cycles at regular intervals although the SMJ34020A remains halted until
the host clears the HLT bit. Only then does SMJ34020A fetch the level-0 vector address from location
FFFF FFE0h and begin executing the reset service routine.
Self-bootstrap mode: if HCS is low at the end of reset, the SMJ34020A first performs eight refresh cycles to
initialize the DRAMs. Immediately following the eight refresh cycles, the GSP fetches the level-0 vector address
from location FFFF FFE0h and begins executing the reset service routine.
At the time the SMJ34020A fetches the level-0 vector address (the reset vector), the least significant four bits
(bit address part) are used to load configuration data that establishes the initial condition of the
big-endian/little-endian mode and the current RCA bus configuration bits in the CONFIG register as described
in the I/O register section.
Unlike other interrupts and software traps, reset does not save the previous ST or PC values (this can also occur
on host initiated nonmaskable interrupts if the NMIM bit in HSTCTLH is set to a 1) because the value of the stack
pointer just before a reset is generally not valid. Saving these values on the stack could contaminate valid
memory locations. A TRAP 0 instruction, which uses the same vector address as reset, similarly does not save
the ST or PC values.
asserting reset
A reset is initiated by asserting RESET to its active-low level. To reset the SMJ34020A at powerup, RESET must
remain active low for a minimum of 40 local clock periods (LCLK1 and LCLK2) after power levels have become
stable. At times other than powerup, the SMJ34020A can be reset by holding RESET low for a minimum of four
local clock periods; the GSP enters an internal reset state for 34 local clock cycles. While in the internal reset
state and RESET is high, memory-refresh cycles occur.
reset and multiprocessor synchronization
The synchronization of multiple SMJ34020As sharing a local memory is done using the RESET input. In
systems where the multiprocessor interface is used to control the access to a common memory, the processors
must be synchronized. Synchronization is achieved by taking RESET high within a specific interval relative to
CLKIN. This can be done by using CLKIN to clock the RESET as received by the SMJ34020As. All SMJ34020As
to be synchronized should use the same CLKIN and RESET inputs. All of the local memory and bus control
signals should be connected in parallel (without buffers) between the processors. After powerup, the
processors are not necessarily synchronized with respect to the particular quarter cycle in progress. The rising
edge of RESET is used to set the SMJ34020A to a particular quarter cycle by adding Q1 cycles. All SMJ34020As
in a multiprocessor environment operate on the same quarter cycle within 10 quarter cycles after the rising edge
of RESET.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
13
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
reset and DRAM/VRAM initialization
The SMJ34020A drives its RAS signal inactive high as long as RESET remains low. The specifications for
certain DRAM and VRAM devices require that RAS be driven inactive-high for 1 millisecond after power is stable
to provide the proper conditions for the DRAMs. Typically, eight RAS cycles are also required to initialize the
DRAMs for proper operation. In general, holding RESET low for t microseconds ensures that RAS remains high
initially for t–(10tQ ) microseconds, tQ being the quarter-cycle time as defined by the input clock period, tc(CHI).
The SMJ34020A memory controller automatically inserts the required eight RAS cycles after all resets (after
powerup or after the internal reset state) by issuing CAS-before-RAS refresh cycles before it allows the CPU
access to memory. A host must delay requests to memory until the initialization cycles have had sufficient time
to complete. Immediately following reset, the SMJ34020A is set to perform a refresh sequence every eight
cycles.
At times other than powerup, to maintain the memory in DRAMs and do a reset, the RESET pulse must not
exceed the maximum refresh interval of the DRAMs minus the time for the SMJ34020A to refresh the memories.
On reset, the SMJ34020A is set to do a refresh cycle every eight local clock periods. A 30-MHz (CLKIN) system
with one (refresh) bank of D/VRAM would be completely refreshed in one sixteenth of the total memory refresh
interval. The reset pulse then should not exceed about fifteen-sixteenths of the total refresh interval required
by the DRAMs to maintain memory integrity.
If RESET remains low longer than the maximum refresh interval specified for the memory, the previous contents
of the local memory can not be valid after the reset.
initial state following reset
While RESET is asserted low (or while in the internal reset state), the SMJ34020A’s output and bidirectional
pins are forced to the states below.
INITIAL STATE OF PINS FOLLOWING A RESET (WITH GI LOW)†
OUTPUTS DRIVEN HIGH
OUTPUTS DRIVEN LOW
BIDIRECTIONALS DRIVEN TO
HIGH IMPEDANCE
RAS
HRDY
VSYNC
CAS0 – CAS3
CBLNK / VBLNK
HSYNC
WE
DDIN
CSYNC / HBLNK
TR / QE
LAD0 – LAD31
DDOUT
ALTCH
HINT
R0
R1
HOE
HDST
EMU3
RCA0 – RCA12
SF
† If GI is high, then all GI-controlled pins are high-impedance. GI-controlled pins are RAS, CAS0 – CAS3, WE, TR / QE, DDOUT, DDIN, ALTCH,
HOE, HDST, RCA0 – RCA12, LAD0 – LAD31, and SF.
Immediately following reset, all I/O registers are cleared (set to 0000) with the exception of the HLT bit in the
HSTCTLH register. The HLT bit is set to 1 if HCS is high just prior to the low-to-high transition of RESET;
otherwise, it is set to 0.
14
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
reset and DRAM/VRAM initialization (continued)
Just prior to the execution of the first instruction in the reset routine, the SMJ34020A’s internal registers are in
the following states:
D
D
D
D
D
General-purpose register files A and B are uninitialized.
The ST is set to 0000 0010h.
The PC contains the most-significant 28 bits of the vector fetched from memory address FFFF FFE0h (the
least significant four bits of the PC are set to zero).
The BEN bit in the I/O register CONFIG is set to the least significant bit read from the vector fetched from
memory address FFFF FFE0h.
The CBP, RCM0, and RCM1 bits in the I/O register CONFIG are set to the corresponding bits read from
the vector fetched from memory address FFFF FFE0h. The configuration byte protect bit (CBP) can be set
high to prevent further modification of the lower eight bits of the I/O register CONFIG.
The state of the instruction cache at this time is as follows:
D
D
D
The SSA (segment start address) registers are uninitialized.
The LRU (least recently used) stack is set to the initial sequence 0, 1, 2, 3, where 0 occupies the most
recently used position and 3 occupies the least recently used position.
All P (present) flags are cleared to 0s.
local memory and DRAM/VRAM interface
The SMJ34020A local memory interface consists of an address/data multiplexed bus on which addresses and
data are transmitted. The associated control signals support memory widths of 16 or 32 bits, burst (page-mode)
accesses, local memory-wait states, and optional external data bus buffers. The SMJ34020A DRAM / VRAM
interface consists of an address/address multiplexed bus and the control signals to interface directly to both
DRAMs and VRAMs. The local memory interface and the DRAM / VRAM interface are interrelated and,
therefore, considered together for this description. At the beginning of a typical memory cycle, the address and
status of the current cycle are output on LAD while the ROW address is output on the row/column address (RCA)
bus. ALTCH and RAS are used to latch the address/status and ROW address, respectively, on these two buses.
LAD is then used to transfer data to or from the memory while the RCA bus is set to the column address for the
memory. (NOTE: LAD31 is the most significant bit of the address or data).
LAD During the Address Cycle
5
31
Address
Address
W=0
W=1
STS
—
—
—
—
4
W
3
0
STS
Memory address (select for 128M 32-bit long-words)
Access to lower 16-bit word (even-addressed word or 32-bit boundary)
Access to upper 16-bit word (odd-addressed word)
Bus cycle status code
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
15
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
local memory and DRAM/VRAM interface (continued)
The address output on the row/column address (RCA) lines is determined by the row/column mode bits (RCM0
and RCM1 in the I/O registers CONFIG) and the state of column-address mode (CAMD) during each memory
cycle. The CAMD is sampled on the internal Q4 clock phase, which allows CAMD to be generated by static logic
wired to the local address/data (LAD) bus.
BASIC MEMORY ROW/COLUMN ACCESS MODES
BANKS§
CAMD SUPPORT MATRICES¶
ADDRS‡
RCM1
RCM
VRAM
MODE†
0
0
64K × N
8
16
0
1
256K × N
9
8
2564K × 16, 256K × 32, 1M × 16, 1M × 32, 4M × 32
1
0
1M × N
10
4
1M × 16, 1M × 32, 4M × 16, 4M × 32
1
1
4M × N
11
2
4M × 16, 4M × 32, 16M × 32
64K × 16, 64K × 32, 256K × 16, 256K × 32, 1M × 16, 1M × 32
† VRAM mode = basic size of VRAM addressing supported with CAMD = 0
‡ Addrs = number of RCA signals required to provide row/column addressing
§ Banks = number of possible interleaved 32-bit wide memory spaces
¶ CAMD support = possible sizes and configurations of DRAMs that can be supported within the basic VRAM mode
The following figure illustrates the actual logical address bits output on each of the RCA lines during row and
column intervals for each of the four VRAM modes and states of CAMD:
ROW TIME
COLUMN TIME
CAMD = 0
64K
CAMD = 1
256K
1M
RCA BIT
64K
256K
1M
4M
12
24
25
26
27
16
23
26
15
4M
28
11
23
24
25
26
15
22
14
14
14
10
22
23
24
25
14
13
13
13
13
9
21
22
23
24
13
12
12
12
12
8
20
21
22
23
12
11
11
11
11
7
19
20
21
22
11
10
10
10
10
6
18
19
20
21
10
9
9
9
9
5
17
18
19
20
9
8
8
8
8
4
16
17
18
19
8
7
7
7
7
3
15
16
17
18
7
6
6
6
6
2
14
15
16
17
6
5
5
5
5
1
13
14
15
16
5
4
4
4
4
0
12
13
14
15
4
4
4
4
16
In the 64K mode with CAMD=0, any eight adjacent RCA0 – RCA12 pins output 16 contiguous logical address
bits. The eight most significant addresses are output during row-address time while the least significant
addresses are output during column-address time. Logical addresses 12 through 16 are output twice during a
memory cycle (during both RAS and CAS falling edges) but at different pins. This allows a variety of VRAM
memory organizations and decoding schemes to be used. When CAMD = 1, the addresses output during
column-address time are changed such that a new logical address mapping occurs, allowing connection of RCA
directly to 256K or 1M DRAMs.
16
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
local memory and DRAM/VRAM interface (continued)
Similarly, for each of the other VRAM modes, direct connection is provided for other DRAM modes requiring
larger matrices than the configuration mode. The following table gives examples of the connections using this
feature.
RCA
CONNECTIONS TO RCA FOR CAMD = 1
256K†
64K†
12
1M × 32
11
1M × 16
1M†
4M
4M × 32
4M × 32
16M × 32
1M × 32
4M × 32
4M × NN
16M × 32
10
256K × 32
1M × 32
1M × NN
4M × 32
4M × NN
16M × 32
9
256K × NN
1M × NN
1M × NN
4M × 32
4M × NN
16M × 32
8
256K × NN
1M × NN
1M × NN
4M × 32
4M × NN
16M × 32
7
256K × NN
1M × NN
1M × NN
4M × 32
4M × NN
16M × 32
6
256K × NN
1M × NN
1M × NN
4M × 32
4M × NN
16M × 32
5
256K × NN
1M × NN
1M × NN
4M × 32
4M × NN
16M × 32
4
256K × NN
1M × NN
1M × NN
4M × 32
4M × NN
16M × 32
3
256K × NN
1M × NN
1M × NN
4M × 32
4M × NN
16M × 32
2
256K × NN
1M × NN
1M × NN
4M × 32
4M × NN
16M × 32
1
256K × 16
1M × 16
1M × 16
0
† NN is used for either 16-bit (× 16) or 32-bit (× 32) memory connections.
4M × 16
16M × 32
status codes
Status codes are output on LAD0 – LAD3 at the time of the falling edge of ALTCH and can be used to determine
the type of cycle that is being initiated. The following table lists the codes and their respective meanings.
CODE
STATUS
TYPE
0000
Coprocessor code
0001
Emulator operation
OTHER
0010
Host cycle
(00XX)
0011
DRAM refresh
0100
Video-generated DRAM serial register transfer
0101
CPU-generated VRAM serial register transfer
VRAM
0110
Write mask load
(01XX)
0111
Color latch load
1000
Data access
1001
Cache fill
1010
Instruction fetch
1011
Interrupt vector fetch
CPU
1100
Bus locked operation
(1XXX)
1101
Pixel operation
1110
Block write
1111
– RESERVED –
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
17
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
dynamic bus sizing
The SMJ34020A supports dynamic bus sizing between 16 and 32 bits on any local memory access. Any
port / memory that is only 16 bits wide must assert SIZE16 low during Q1 (to be valid at the start of Q2) of the
bus cycle accessing the even memory word (LAD4 = 0) corresponding to its address.The SMJ34020A then
performs another memory access to the next 16-bit (odd) word in memory. The SMJ34020A samples SIZE16
at the start of Q2 in the second cycle (access to odd word address) to determine to which half of LAD the port
or memory is aligned. If the port is on LAD0 – LAD15, SIZE16 should be low during the second cycle access (odd
word); otherwise, if the port is on LAD16 – LAD31, SIZE16 must be high at this time. The SMJ34020A always
performs two memory cycles to access the 16-bit wide memories, even when attempting only a 16-bit transfer.
The SMJ34020A outputs the four CAS strobes and LAD bus initially aligned for a 32-bit bus. If the memory is
16 bits wide, the two most significant CAS strobes are swapped with the two least significant strobes when it
accesses the second word and the halves of LAD are also swapped; therefore, 16-bit memories need to respond
only to the two CAS strobes corresponding to the upper or lower 16 bits of LAD to which they are connected.
Note that devices connected to LAD0 – LAD15 transfer the least significant word during the first cycle and the
most significant word during the second cycle. Data accesses on LAD16 – LAD31 transfer the most significant
word first, then the least significant word.
The second memory cycle forced by SIZE16 is performed as a page mode access if PGMD was low during the
first access. A read-write cycle to the 16-bit page-mode memory requires five bus cycles that occur as address,
read0, read1, write0, write1. If a 16-bit transfer is interrupted due to a bus fault, the restart causes the entire
access to be restarted.
For memory that supports page-mode accesses (PGMD low), SIZE16 is sampled during each access to
memory. If SIZE16 is high on the even word access, then a 32-bit transfer occurs over LAD0 – LAD31. If SIZE16
is low on the even word access (16-bit wide memory), then it is sampled again on the odd word access to
determine to which half of LAD the memory is connected (low for connection to LAD0 – LAD15 or high for
connection to LAD16 – LAD31).
special 1-M VRAM cycles
The SMJ34020A provides control for special function VRAM cycles that are available in the 1-M devices. These
cycles are obtained by the appropriate timing control of SF, CAS, TR / QE, and WE of the VRAMs at the falling
edge of RAS. The cycles include:
D
D
D
D
D
D
Load write mask
Load color mask
Block write (no mask)
Block write (current mask)
Write using mask
Alternate write transfer
In addition, other special modes can be implemented by using external logic.
multiprocessor arbitration
The multiprocessor interface allows multiple processors to operate in a system sharing the same local memory.
The use of the bus grant in GI and the priority request signals R0 and R1 allows a flexible method of passing
control from one processor to another. The control scheme allows local memory cycles to occur back-to-back,
even when passing control from one SMJ34020A to another. Synchronization of multiple SMJ34020As in a
system occurs at reset with the rising edge of RESET meeting the setup and hold requirements to CLKIN, so
all SMJ34020As are certain to respond to RESET during the same quarter cycle. RESET is not required to be
synchronous to CLKIN except to allow synchronization of multiple SMJ34020As in a system.
18
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
multiprocessor arbitration (continued)
The GI priority for multiprocessing environments is determined by arbitration logic external to the SMJ34020A.
If GI goes inactive-high, the SMJ34020A releases the bus on the next available cycle boundary. If the cycle in
progress has not successfully completed, the SMJ34020A restarts the cycle upon regaining control of the bus.
Normally, if the SMJ34020A asserts both R0 and R1 low, it should be given the control of the bus by the arbitrator.
host interface
The SMJ34020A host interface allows the local memory to be mapped into the host address space. The
SMJ34020A acts as a DRAM controller for the host. The address for the host access is latched within the
SMJ34020A; however, the data for the access is transferred using external transceivers. The host selects the
address of a 32-bit long word for an access using the 27 host address lines, HA5 – HA31. If the host desires byte
addressability, it can select the active bytes for the access by using HBS0 – HBS3. The SMJ34020A always
reads 32 bits from memory; however, on host writes, it uses the host byte selects to enable CAS0 – CAS3 to
memory. The address and byte selects are latched at the falling edge of HCS within the SMJ34020A. The host
indicates a read or write by asserting HREAD or HWRITE (as appropriate) either before or after HCS. (Note
that HREAD and HWRITE must never be asserted at the same time.)
The SMJ34020A responds to a host read request by latching the requested data in the external latches and
providing HRDY to the host, indicating that the read cycle is completing. The rising edge of HDST with HRDY
high indicates data is latched in the external transceivers.
The host indicates that a write to a particular location is required by providing the address and asserting
HWRITE. The host must maintain both HCS and HWRITE asserted until valid data is in the transceivers. (The
rising edge of HOE with HRDY high indicates that the data previously stored in the external transceivers has
been written to memory.) Typically, the rising edge of HWRITE is used to strobe the data into the latches and
signal the SMJ34020A that the write access can start. The SMJ34020A uses its byte-write capability to write
only to the selected bytes.
The SMJ34020A always accesses the required location as latched at the falling edge of HCS; however, in order
to increase the data rate, a look ahead mechanism is implemented. The host increment enable (HINC) and host
prefetch after write enable (HPFW) bits in the host control register (HSTCTLH) must be appropriately set to
make optimum use of this feature. These bits provide four modes of operation as indicated in the following table:
HINC
HPFW
HOST ACCESS MODE
DESCRIPTION
0
0
Random/Same
No increment, no prefetch
0
1
Random/Same
No increment, no prefetch
1
0
Block
Increment after read or write, prefetch after read
1
1
Read-Modify-Write
Increment after write, prefetch after write
When the SMJ34020A is programmed for block mode or read-modify-write accesses, the host can still do
random accesses because the SMJ34020A always uses the address provided at the falling edge of HCS;
however, there is a prefetch to the next sequential address. The prefetch occurs after reads in block mode and
after writes in read-modify-write mode. The SMJ34020A compares the address latched by HCS on host reads
to see if it is the same as that of the last prefetched data. If the addresses match, data is not re-accessed but
HRDY is set high to indicate that the data is presently available.
dynamic bus sizing on host accesses
If the host makes a read access to a 16-bit wide memory, the SMJ34020A automatically does the second cycle
required to read the rest of the 32-bit word (even if the host did not require a 32-bit cycle). The external logic
must comprehend the sense of SIZE16 or the CAS strobes during the accesses in order to route the data into
the proper external host data transceivers. The SMJ34020A uses the host byte selects HBS0 – HBS3 to enable
the CAS strobes when doing a host write.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
19
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
coprocessor interface
Support for coprocessors is provided through special instructions and bus cycles that allow communication with
the coprocessor. A coprocessor can be register based, depending on the SMJ34020A to do all address
calculations, or it can operate as its own bus controller, using the multiprocessor arbitration scheme. Five basic
cycles are provided for direct communication and control of coprocessors:
D
D
D
D
D
SMJ34020A to coprocessor
Coprocessor to SMJ34020A
Move memory to coprocessor
Move coprocessor to memory
Coprocessor internal command
The first four of these cycles provide for command of the coprocessor in addition to the movement of parameters
to and from the coprocessor. In this manner, parameters can be sent to the coprocessor and operated upon
without an explicit coprocessor command cycle.
instruction set
The SMJ34020A instruction set can be divided into five categories:
D
D
D
D
D
Graphics instructions
Coprocessor instructions
Move instructions
General-purpose instructions
Program control and context switching
Specialized graphics instructions manipulate pixel data that is accessed using memory addresses or
XY coordinates. These instructions include graphics operations, such as array and raster operations, pixel
processing, windowing, plane masking, pixel masking, and transparency. Coprocessor instructions allow for the
control and data flow to and from coprocessors that reside in the system. Move instructions comprehend the
bit-addressing and field operations, which manipulate fields of data using linear addressing for transfer to and
from memory and the register file. General-purpose instructions provide a complete set of arithmetic and
Boolean operations on the register file as well as general program control and data processing. Program control
and context switching instructions allow the user to control flow and to save and restore information using
instructions with both register-direct and absolute operands.
clock stretch
The SMJ34020A supports a clock stretching mechanism, that is described below.
With advances in semiconductor manufacturing, newer versions of the SMJ34020A can be made, each
supporting a higher CLKIN frequency. The increase in CLKIN frequency means that the SMJ34020A machine
cycles execute more quickly, with a consequent increase in code execution speed. However, there comes a
point when, as the machine cycle time becomes shorter, the local-memory control signals begin to violate DRAM
and VRAM timing parameters for certain types of memory access.
The clock-stretch mechanism allows the SMJ34020A to slow down and execute those critical local-memory
cycles while still benefiting from the accelerated processing allowed by higher CLKIN frequencies during
noncritical memory access cycles.
Exact timing issues vary from system to system, reflecting differences in bus buffering, etc., but, broadly
speaking, the clock-stretch mechanism allows the system designer to interface to slower memory devices than
the designer could use if no stretch mechanism was available.
A normal, unstretched machine cycle consists of four quarter cycles, Q1, Q2, Q3, and Q4. A stretched cycle
consists of five quarter cycles, Q1, Q2, Q3, Q4a, and Q4b.
20
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
clock stretch (continued)
When clock-stretch mode is enabled, the fourth machine quarter cycle can be stretched to twice its original
length. This stretching takes place only when the SMJ34020A attempts certain types of memory cycles.
Normal Sequence
Q1
Q2
Q3
Q4
Q1
Q1
Q2
Q3
Q3
Q4
Normal Cycle
Normal Cycle
Possible New Sequence
Q2
Q4a
Q4b
Q1
Stretched Cycle
Q2
Q3
Q4
Normal Cycle
The stretch is achieved by holding the internal SMJ34020A clocks in the Q4 state for an extra quarter cycle so
all of the device outputs remain unchanged during Q4a and Q4b. The SMJ34020A stretches only certain
machine cycles so that the execution of code is not slowed unnecessarily.
enabling clock stretch
Clock-stretch mode is enabled and disabled using a bit in the CONFIG register memory mapped to location
C00001A0h, see Figure 1.
7
31
6
5
4
3
2
1
0
C
S
E
Loaded at Reset from Reset Vector
Protected Byte
CONFIG register
CSE = 0: Disable stretch mode (normal operation)
CSE = 1: Enable stretch mode
Bit 4 of the CONFIG register is the clock-stretch-enable mode bit. A zero in this bit disables stretch mode and
a one in this bit enables stretch mode. The bit is cleared during reset; that is, stretch mode is disabled by default.
When stretch mode is enabled, the following machine cycles are stretched:
D
D
All address cycles of all memory-access sequences
Read data cycles in read-modify-write sequences
Notes:
a) The host default cycle shown in the TMS34020 User’s Guide is not stretched because it is not a true
address cycle; that is, RAS, etc., do not go low.
b) The CPU default cycle, which is similar to the host default cycle in that RAS, etc., do not go low, is
also not stretched.
c) Clock-stretch mode disregards the page-mode input so that read data cycles in nonpage-mode
read-modify-write sequences are stretched even though there are no timing constraints that require
a stretch.
d) All other memory subcycles are not stretched, even if the SMJ34020A is running with the CSE bit
set to 1.
The advantage of this implementation of clock-stretch mode is that the SMJ34020A can execute code at
maximum speed, slowing down only during certain parts of memory access sequences.
It is important to remember that a stretched cycle is 25% longer than a normal cycle and that the SMJ34020A
(with the exception of the video logic, which is clocked independently by VCLK) effectively slows down during
such a stretched cycle.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
21
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
enabling clock stretch (continued)
The following are examples of stretch-mode memory operations.
two 32-bit nonpage-mode reads
ADDR
1
2
3
READ
4
1
2
ADDR
3
ADDR
4
1
2
READ
3
READ
4
1
2
Stretch Mode Disabled
3
4
ADDR
READ
Stretch Mode Enabled
1
2
3
4
4
1
2
3
4
1
2
3
4
4
Stretch
1
2
3
4
Stretch
one 32-bit page-mode read-modify-write
ADDR
1
2
3
READ
4
1
2
3
ADDR
WRITE
4
1
2
3
Stretch Mode Disabled
4
READ
WRITE
Stretch Mode Enabled
1
2
3
4
4
1
2
3
4
4
Stretch
1
2
3
4
Stretch
three 32-bit page-mode reads
ADDR
1
2
3
READ
4
1
2
ADDR
3
READ
4
1
2
READ
READ
3
4
1
2
READ
Stretch Mode Disabled
3
4
READ
Stretch Mode Enabled
1
2
3
4
4
1
2
3
4
1
2
3
4
1
2
3
4
Stretch
The stretched cycles are designed to accommodate worst-case 32-bit page-mode accesses, so during some
nonpage-mode memory accesses stretches that are not essential can be generated. For example:
one 32-bit nonpage-mode read-write
ADDR
1
2
3
READ
4
1
2
3
ADDR
ADDR
4
1
2
3
WRITE
4
READ
1
2
Stretch Mode Disabled
3
4
ADDR
WRITE
Stretch Mode Enabled
1
2
3
4
4
1
Stretch
2
3
4
4
1
2
Stretch
3
4
4
1
2
3
4
Stretch
Stretches are inserted in read-modify-write accesses to help ease bus turn-around timings. In the above
example, the second stretch is not needed to help these timings because the read/write turn-around has the
whole of the address cycle to evaluate.
22
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
a clock-stretch timing example, SMJ34020A-32 and 150-ns DRAMs
This example analyzes a memory interface timing parameter. It shows that the clock-stretch mechanism can
be used to allow the SMJ34020A-32 to avoid a timing violation when interfaced to 100-ns VRAMs.
Consider a system with:
D
A SMJ34020A-32,
D
A SMJ44C251-10
which has a 32-MHz clock input frequency and hence a 125-ns cycle time, so
tQ = 31 ns. Timing parameters are taken from this data sheet.
1 megabit × 1 bit DRAM. Timing parameters are taken from the corresponding
Texas Instruments data sheet.
row address hold data after RAS low, th(ADV-REL)
Without clock stretch
SMJ4C1024
th(RA)
Hold time, row address valid after RAS low
Min = 20 ns
SMJ34020A
Parameter 88
Hold time, row address valid after RAS low
Min = tQ – 5 ns = 26 ns
If RAS is passed through a PAL with a delay of 7 ns, then th(RA) seen by the DRAM is 26 ns – 7 ns = 19 ns.
This violates the 20 ns minimum.
With clock stretch
SMJ34020A
Parameter 88
th(ADV - REL)
Hold time, row address valid after RAS low
Min = 2tQ – 5 ns = 57 ns
With the same 7-ns PAL delay, the DRAM sees th(RA) as 57 ns – 7ns = 50 ns, which does not violate the
20 ns minimum.
cycle timing examples
The following figures show examples of many of the basic cycles that the SMJ34020A uses for memory access,
VRAM control, multiprocessor bus control, and coprocessor communication. These figures should not be used
to determine specific signal timings, but can be used to see signal relationships for the various cycles. The
Q4 phases that could be stretched are marked with an * on the diagrams. The conditions required for the stretch
are:
D
D
D
The design uses a SMJ34020A.
The CONFIG register’s CSE bit is set to 1.
The SMJ34020A is doing either:
a) Any address cycle, or
b) A read data cycle in a read-modify-write sequence
The following remarks apply to memory timing in general. A row address is output on RCA0 – RCA12 at the start
of a cycle along with the full address and status on LAD0 – LAD31. These remain valid until after the fall of ALTCH
and RAS. The column address is then output on RCA0 – RCA12, and LAD0 – LAD31 are set to read or write data
for the memory access. During a write, the data and WE are set valid prior to the falling edge of CAS; the data
remains valid until after WE and CAS have returned high.
Large memory configurations can require external buffering of the address and data lines. DDIN and DDOUT
coordinate these external buffers with LAD.
During the address output to LAD by the SMJ34020A (Figure 2), the least significant four bits (LAD0 – LAD3)
contain a bus-status code. PGMD low at the start of Q2 after RAS low indicates that this memory supports
page-mode operation. LRDY high at the start of Q2 after RAS low indicates that the cycle can continue without
inserting wait states. DDOUT returns high after the initial address output on LAD (during Q4), indicating that
a memory read cycle is about to take place.
PAL is a trademark of Advanced Micro Devices, Inc.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
23
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Page-Mode Read
Standard Memory Read Cycle
Q4
Q1
Q2
Q3
Data Transfer
Subcycle
Data Transfer
Subcycle
Address Subcycle
Q4†
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD (SMJ34020A)
(see Note A)
Address
Data
LAD (Memory)
(see Note A)
Data
CAMD
RCA
Row
1st Column
2nd Column
ALTCH
RAS
CAS
WE
TR/ QE
SF
DDIN
DDOUT
LRDY
(see Note B)
PGMD
(see Note B)
SIZE16
(see Note B)
BUSFLT
(see Note B)
R0
R1
† See clock stretch, page 20.
NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A
LAD (memory): Output to LAD by the memory.
B. LRDY, PGMD, SIZE16, and BUSFLT are not sampled on subsequent page-mode cycle accesses to
32-bit-wide memory space.
Figure 2. Local-Memory Read-Cycle Timing (with page mode)
24
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
LRDY low at the start of the first Q2 after RAS low (Figure 3) indicates that the memory requires the addition
of wait states. LRDY high at the next Q2 indicates the cycle can continue without inserting more wait states.
PGMD high at the start of Q2 where LRDY is sampled high indicates that this memory does not support
page-mode operation.
Address Subcycle
Q4
Q1
Q2
Q3
Q4†
Wait State
Q1
Q2
Q3
Read Transfer
Q4
Q1
Q2
Q3
Q4
Q1
LCLCK1
LCLCK2
GI
LAD (SMJ34020A)
(see Note A)
Address
Data
LAD (Memory)
(see Note A)
CAMD
RCA
Row
Column
ALTCH
RAS
CAS
WE
TR/ QE
SF
DDIN
DDOUT
LRDY
PGMD
(see Note B)
SIZE16
(see Note B)
BUSFLT
R0
R1
† See clock stretch, page 20.
NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A
LAD (memory): Output to LAD by the memory.
B. Although they are not internally sampled, PGMD and SIZE16 must be held at a valid level at the
start of each Q2 until LRDY is sampled high.
Figure 3. Local-Memory Read-Cycle Timing (without page mode, with one wait state)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
25
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
During the address output to LAD by the SMJ34020A (Figure 4), the least significant four bits (LAD0 – LAD3)
contain a bus-status code. PGMD low at the start of Q2 after RAS low indicates that this memory supports
page-mode operation. LRDY high at the start of Q2 after RAS low indicates that the cycle can continue without
inserting wait states.
DDOUT remains low after the initial address output on LAD (during Q4 after RAS goes low), indicating that a
memory write cycle is about to take place.
26
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Standard Memory Write Cycle
Address Subcycle
Q4
Q1
Q2
Q3
Q4†
Page-Mode Write
Data Transfer
Subcycle
Data Transfer
Subcycle
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
LCLCK1
LCLCK2
GI
LAD
Address
Data Out 1
Data Out 2
CAMD
RCA
Row
1st Column
2nd Column
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
(see Note A)
PGMD
(see Note A)
SIZE16
(see Note A)
BUSFLT
(see Note A)
R0
R1
† See clock stretch, page 20.
NOTE A: LRDY, PGMD, SIZE16, and BUSFLT are not sampled on subsequent page-mode cycle
accesses to 32-bit-wide memory space.
Figure 4. Local-Memory Write-Cycle Timing (With Page Mode)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
27
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The local memory read-modify-write cycle (Figure 5) is used when inserting a field into memory that crosses
byte boundaries. This cycle is actually performed as a read access followed by a page-mode write cycle.
Standard Memory Write Cycle
Q1
Q2
Q3
Q4
Data Transfer
Subcycle
Data Transfer
Subcycle
Address Subcycle
Q4
Page-Mode Write
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
LCLCK1
LCLCK2
GI
Data Out
Address
LAD (SMJ34020A)
Data
LAD (Memory)
CAMD
Row
RCA
Column
ALTCH
RAS
CAS
WE
TR/ QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 5. Local-Memory Read-Modify-Write-Cycle Timing
28
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q4
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The refresh pseudo-address output to RCA0 – RCA12 and LAD0 – LAD31 comes from the 16-bit refresh
address register (I/O register C000 01F0h) that is incremented after each refresh cycle (Figure 6). The 16 bits
of address are placed on LAD16 – LAD31; all other LAD bus lines are zero. The logical addresses on
RCA0– RCA12 corresponding to LAD16 – LAD31 also output the address from the refresh-address register.
Although PGMD and SIZE16 are ignored during a refresh cycle, they should be held at valid levels. LRDY and
BUSFLT are not sampled until the start of the first Q2 cycle after RAS has gone low.
If a refresh cycle is aborted due to a high-priority bus request (assuming LRDY is low at Q2 after RAS low), a
bus fault, or an external retry, then the count of refreshes pending is not decremented and the same
pseudo-address is reissued when the refresh is restarted.
CBR
Refresh Status
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Refresh End
Q4
Q1
Q2
Q3
Q4
Q1
LCLCK1
LCLCK2
GI
LAD
Refresh Pseudo-Address
CAMD
RCA
Refresh Psuedo-Address
ALTCH
RAS
CAS
WE
TR/ QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 6. Refresh-Cycle Timing
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
29
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
When SIZE16 is selected low (Figure 7), the SMJ34020A performs a second cycle to read (or write) the
remaining 16 bits of the word. Reads always access all 32 bits (all CAS strobes are active). Internally, the
SMJ34020A latches both the high and the low words obtained on the first read cycle. The sense of SIZE16 on
the second (odd-word) access is used to determine which half of the bus is to be sampled to replace the data
word latched during the first cycle.
30
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Data Transfer
Subcycle
Address Subcycle
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Data Transfer
Subcycle
Q4
Q1
Q2
Q3
Q4
Q1
LCLCK1
LCLCK2
GI
LAD0 – LAD15
Low Address
LAD16 – LAD31
High Address
Low
Hi
CAMD
RCA
(see Note A)
Row
Column (S=0)
Column (S=1)
ALTCH
RAS
CAS0
CAS1
CAS2
CAS3
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
NOTE A: RCA0 can be used to determine accesses to odd or even words because it outputs the least significant bit
of the word address during the column-address time (except in 4-M mode with CAMD = 1).
Figure 7. Dynamic Bus Sizing, Read-Cycle Timing
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
31
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Write accesses to 16-bit memory are performed by swapping the data on upper and lower words of LAD and
exchanging data on CAS0 and CAS1 for data on CAS2 and CAS3, respectively (Figure 8). During the first cycle,
data is placed on LAD0 – LAD31 as in a normal write. The sampling of SIZE16 low during the first access
indicates that this is 16-bit-wide memory, so the SMJ34020A swaps data on the upper and lower halves of LAD.
Notice that during the first cycle, CAS0 is inactive (because this byte was not selected), and during the second
cycle, CAS2 is inactive due to the exchange of CAS0 for CAS2 and CAS1 for CAS3.
32
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Data Transfer
Subcycle
Address Subcycle
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Data Transfer
Subcycle
Q4
Q1
Q2
Q3
Q4
Q1
LCLCK1
LCLCK2
GI
LAD0 – LAD15
Low Address
Data Low
Data High
LAD16 – LAD31
High Address
Data High
Data Low
Row
Column (S=0)
Column (S=1)
CAMD
RCA
ALTCH
RAS
CAS0
CAS1
CAS2
CAS3
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 8. Dynamic Bus Sizing, Write-Cycle Timing
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
33
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Clock stretch is a special 1-megabit VRAM control cycle that is executed when VEN in the CONFIG I/O register
is set and PMASKL and/or PMASKH are written (Figure 9). This cycle is indicated by CAS, WE, TR / QE, and
SF high at the falling edge of RAS and SF low at the falling edge of CAS. As the plane mask is copied to the
PMASK register(s), it is also output on LAD to be written to a special register on the VRAM that is used in
subsequent cycles requiring a write mask. During the address portion of the cycle, the status on LAD0 – LAD3
indicates a write-mask load is being performed (status code = 0110). Although CAMD, PGMD, and SIZE16 are
ignored on this cycle, they should be held at valid levels as shown.
Load-Write-Mask Cycle
Write to the PMASK I/O Register
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
GI
LAD
PMASK Address
PMASK Data
Zero Address
Not PMASK Data
CAMD
RCA
PMASK Row
PMASK Column
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 9. Load-Write-Mask-Cycle Timing
34
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
All-Zero Address
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The clock stretch is generated by the VLCOL instruction and is indicated by CAS, WE, TR / QE, and SF high
at the falling edge of RAS and SF high at the falling edge of CAS (Figure 10). The data in the COLOR1 register
is output on LAD to be written to a special register on the VRAM that is used in subsequent cycles requiring a
color latch. During the address portion of the cycle, the status on LAD0 – LAD3 indicates a color-mask load is
being performed (status code = 0111). Although CAMD, PGMD, and SIZE16 are ignored on this cycle, they
should be held at valid levels as shown.
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
GI
Zero Address
LAD
Color Register Data
CAMD
RCA
All-Zero Address
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 10. Load-Color-Latch-Cycle Timing
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
35
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The clock stretch is also performed when a VBLT or VFILL instruction is executed and PMASKL and PMASKH
are set to zero (Figure 11). It is indicated by CAS, WE, TR / QE high and SF low at the falling edge of RAS and
by SF high at the falling edge of CAS. The data on LAD is used as an address mask, and the data stored in the
color latch is written to the VRAM. The address selects chosen by the two least significant bits of the column
addresses within the VRAM are replaced with the four DQ bits latched on the falling edge of CAS. A logic 1 on
each bit enables that nibble to be written, while a logic 0 disables the write from occurring. This cycle allows up
to 16 bits to be written into each VRAM (four adjacent nibbles, each set to the value in the color latch) for a total
of 128 bits. During the address portion of the cycle, the status on LAD0 – LAD3 indicates a block write is being
performed (status code = 1110). SIZE16 can be used with this cycle, but external multiplex logic is required to
map the data correctly to appropriate memories.
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
Q2
Q3
GI
LAD
Address
Data Out 1
Data Out 2
Row
1st Column
2nd Column
CAMD
RCA
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 11. Block-Write-Cycle Timing (Without Mask)
36
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q4
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The clock stretch is also performed when a VBLT or VFILL instruction is executed and PMASKL and PMASKH
are set to nonzero values (Figure 12). It is indicated by CAS, TR / QE, and SF high and WE low at the falling edge
of RAS and by SF high at the falling edge of CAS. The data on LAD is used as an address mask, and the data
stored in the color latch is written to the VRAM, just as in the block-write cycle without mask, except that the
data in the write mask is used to enable the bits from the color latch that are written to memory. This cycle allows
up to 16 bits to be written into each VRAM (four adjacent nibbles, each set to the value in the color latch as
enabled by the write mask) for a total of 128 bits. During the address portion of the cycle, the status on
LAD0– LAD3 indicates a block write is being performed (status code = 1110). SIZE16 can be used with this
cycle, but external multiplex logic is required to map the data correctly to appropriate memories.
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
GI
LAD
Address
Data Out 1
Data Out 2
Row
1st Column
2nd Column
CAMD
RCA
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 12. Block-Write-Cycle Timing (With Mask)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
37
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
As a special 1-megabit VRAM control cycle, the clock strech is also performed when the PMASKL and PMASKH
registers are set to nonzero values, CST in DPYCTL is cleared, VEN in CONFIG is set, and the byte-aligned
pixel-write instruction is executed (Figure 13). This cycle is indicated by CAS, TR / QE, and SF high and WE low
at the falling edge of RAS and by SF low at the falling edge of CAS. The data on LAD is written to memory just
as a normal DRAM write except that data in the write mask is used to enable DQs that are written to memory.
During the address portion of the cycle, the status on LAD0 – LAD3 indicates that a pixel operation is being
performed (status code = 1101).
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
Q2
Q3
GI
LAD
Address
Data Out 1
Data Out 2
Row
1st Column
2nd Column
CAMD
RCA
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 13. Write-Cycle Timing Using Mask
38
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q4
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The VRAM cycle shown in Figure 14 is issued in any of three ways:
D
D
D
Pixel operation instruction with CST in DPYCTL set
Horizontal blank reload cycle requested by the video-control logic with VCE in DPYCTL cleared
Video timeout due to SCOUNT match with the value in MLRNXT and VCE and SSV in DPYCTL cleared
This cycle is indicated by TR / QE and SF low and CAS and WE high at the time RAS goes low. The timing of
the low-to-high transition of TR / QE is dependent upon the timing of SCLK when doing a midline reload cycle.
During the address portion of the cycle, the status on LAD0 – LAD3 indicates either a video-initiated VRAM
memory-to-register transfer (status code = 0100), or a CPU-initiated VRAM memory-to-register transfer
(status code = 0101).
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
39
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Address
Subcycle
Q4
Q1
Q2
Q3
Wait
State
Q4†
Q1
Q2
Q3
Cycle
Completion
Q4
Q1
Q2
Q3
Q4
GI
LAD
Address
CAMD
Row
RCA
Tap Point
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 14. Memory-to-Serial-Data-Register-Cycle Timing (VRAM Read Transfer)
40
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
This VRAM cycle shown in Figure 15 is performed when a video timeout occurs due to a match of the MLRNXT
register, VCE in DPYCTL is cleared, and SSV in DPYCTL is set. This cycle is indicated by TR/ QE low and CAS,
SF, and WE high at the time RAS goes low. The timing of the low-to-high transition of TR/ QE is not dependent
upon the timing of SCLK because there is not as great a timing constraint to position the cycle as in midline
reload. During the address portion of the cycle, the status on LAD0 – LAD3 indicates a video-initiated VRAM
memory-to-register transfer (status code = 0100). Although PGMD and SIZE16 are ignored on this cycle, they
should be held at valid levels as shown.
Q4
Q1
Q2
Q4†
Q3
Q1
Q2
Q3
Q4
Q1
GI
LAD
Address
CAMD
RCA
Row
Tap Point
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 15. Memory-to-Split-Serial-Data-Register-Cycle Timing (VRAM Split-Register Read Transfer)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
41
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Figure 16 shows the VRAM cycle performed when a horizontal blank reload is requested by the video-control
logic and VCE and SRE in DPYCTL are both set. This cycle is indicated by TR/ QE, WE and SF low and CAS
high at the time RAS goes low. The SOE pin of the VRAMs is used to select between write transfer and
pseudo-write transfer cycles (SOE must be generated by logic external to the SMJ34020A). During the address
portion of the cycle, the status on LAD0 – LAD3 indicates that a video-initiated VRAM register-to-memory
transfer (status code = 0100) is being performed. Although PGMD and SIZE16 are ignored on this cycle, they
should be held at valid levels as shown.
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
GI
LAD
Address
CAMD
RCA
Row
Tap Point
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 16. Serial-Data-Register-to-Memory-Cycle Timing (VRAM-Write Transfer, Pseudo-Write Transfer)
42
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
This VRAM cycle (Figure 17) is performed when a pixel-write instruction is executed with the CST bit in DPYCTL
set. This cycle is indicated by TR/ QE and WE low and SF and CAS high at the time RAS goes low. This cycle
does not require the use of SOE of the VRAM and does not affect the status of the serial I / O pins. During the
address portion of the cycle, the status on LAD0 – LAD3 indicates that a CPU-initiated VRAM
register-to-memory transfer (status code = 0101) is being performed. Although PGMD and SIZE16 are ignored
on this cycle, they should be held at valid levels as shown.
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
GI
Address
LAD
CAMD
Row
RCA
Tap Point
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
† See clock stretch, page 20.
Figure 17. Serial-Data-Register-to-Memory Cycle Timing (VRAM-Alternate-Write Transfer)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
43
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
In Figure 18, transition points are shown for R0 and R1 to indicate where they occur relative to the other signals.
This example indicates that the SMJ34020A has control of the bus, yields control, and then regains control. The
SMJ34020A regains bus mastership as soon as GI is driven active (low). R0 and R1 could be outputting any
of the codes with the exception of the access-termination code. The bus arbitration logic must control the timing
of GI to all of the processors requiring the bus.
It is recommended that SMJ34020A clock stretch not be used in multiprocessor systems.
44
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Q4
Host-Default
Cycle
Q1 Q2 Q3 Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Beginning
of Cycle
Q1 Q2 Q3
Q4
LCLK1
LCLK2
Hi-Z
LAD
GI
R0
R1
CAMD
Hi-Z
RCA
Row
SF
ALTCH
Hi-Z
RAS
Hi-Z
CAS
Hi-Z
WE
Hi-Z
TR / QE
Hi-Z
DDIN
Hi-Z
DDOUT
Hi-Z
LRDY
SIZE16
PGMD
BUSFLT
HOE
Hi-Z
HDST
Hi-Z
Figure 18. Multiprocessor-Interface-Cycle Timing (High-Impedance Signals)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
45
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Two SMJ34020As use the multiprocessor interface to pass control of local memory from one to the other
(Figure 19). GSP1 completes a read cycle to the local memory and, although desiring another read, loses the
bus to GSP2, which does a single write cycle (perhaps a host-write access). GSP1 then regains control and
completes the read cycle (shown with a single wait state). Since no further memory-access requests are
present, GSP1 maintains control of the bus and holds all of the local-memory control signals at their inactive
levels. LRDY is a common input to both GSP1 and GSP2.
The host cycle timing diagrams shown in this data sheet are only a sample. For more information, see the
TMS34020 User’s Guide.
1
2
GSP1 Read
3 4 1 2
3 4
1
2
GSP2 Write
3 4 1 2
3 4
1
2
GSP1 Read With Wait
3 4 1 2 3 4 1
2
3 4
1
Bus Idle
2 3 4
1
GI
R0
R1
RAS
CAS
WE
GSP1
TR / QE
DDIN
DDOUT
ALTCH
LRDY
1
2
3 4
1
2
3 4
1
2
3 4
1
2
3 4
1
2
3 4
1
2
3 4
1
2
3 4
1
2
3 4
1
ALTCH
GI
R0
R1
RAS
GSP2
CAS
WE
TR / QE
DDIN
DDOUT
Figure 19. Multiprocessor-Interface Cycle Timing (Passing Control)
46
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
In Figure 20, the host-access request is synchronized to the SMJ34020A at the beginning of Q4 so that the
local-memory cycle can begin in Q1. If the external host-access request occurs after the setup time requirement
before Q4, the request is not considered until the next Q4 cycle. In order to provide back-to-back accesses as
indicated in this example, the host must remove HCS on receipt of HRDY and reassert it before Q4 (it can also
remove and reassert HREAD with HCS).
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
47
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
HA / HBS
HCS
HREAD
HWRITE
HRDY
(see Note A)
DATA
(out)
Previous Read
Valid
Local-Memory Host Read Cycle
Q4
Q1
Q2
Q3
Q4† Q1 Q2
Q3
Q4
Q1
HOE
HDST
LAD
GI
CAMD
Row
RCA
Column
SF
ALACH
RAS
CAS
WE
TR / QE
DDIN
DDOUT
LRDY
SIZE16
PGMD
BUSERR
R0
R1
† See clock stretch, page 20.
NOTE A: HRDY goes high at the start of Q2; however, data is not strobed into
the external latches until the start of Q4 when HDST goes high.
Figure 20. Host-Read-Cycle Timing (Random/Same Accesses, not From SMJ34020A I/O Registers)
48
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The host-access request is synchronized to the SMJ34020A at the beginning of Q4 so that the local memory
cycle can begin in Q1.
In block mode (prefetch after read), the SMJ34020A automatically initiates sequential read accesses as soon
as the host deasserts the current read request. In this example, the host reads a location and must wait for the
first access to complete. When the host removes HREAD (Figure 21), indicating the end of the first read, the
SMJ34020A starts to prefetch the next sequential location. When the host makes the next request, the
SMJ34020A has prefetched the data so that the host reads with no delay. While in block mode, the SMJ34020A
continues to prefetch data for the host read each time the host removes either HREAD or HCS. If the address
present and latched at the falling edge of HCS matches the previously prefetched address, HRDY is asserted
high so that the host can read with no delay.
In read-modify-write mode (prefetch after write), the SMJ34020A initiates the read access as soon as the current
write request is deasserted.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
49
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
HA / HBS
HCS
HREAD
HWRITE
HRDY
DATA
(out)
Previous Read
Local-Memory Host Read Cycle
Q4
Q1
Q2
Q3
Q4†
2nd
1st Read Valid
Q1
Q2
Local-Memory Host Prefetch Cycle
Q3
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
HOE
HDST
LAD
1st Address
2nd Address
GI
CAMD
RCA
Row
Column
Row
Column
SF
ALTCH
RAS
CAS
WE
TR / QE
DDIN
DDOUT
LRDY
SIZE16
PGMD
BUSERR
R0
R1
† See clock stretch, page 20.
Figure 21. Back-to-Back Host-Read Cycles With Implicit Addressing; HREAD as Strobe
50
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q4
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The host read of the SMJ34020A I/O registers (Figure 22) suppresses the generation of TR / QE and CAS so
that data is read from the SMJ34020A rather than from memory. DDOUT is enabled so that data can flow
through external buffers on LAD to the host data latches. The SMJ34020A I/O registers can be accessed in any
of the host access modes (random/same, block, or read-modify-write).
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
51
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
HA / HBS
HCS
HREAD
HWRITE
HRDY
DATA
(out)
Previous Read
Valid
Local-Memory Host Read I/O Cycle
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
HOE
HDST
I/O Data
LAD
GI
CAMD
Row
RCA
Column
SF
ALTCH
RAS
CAS
WE
TR / QE
DDIN
DDOUT
LRDY
SIZE16
PGMD
BUSERR
R0
R1
† See clock stretch, page 20.
Figure 22. Host-Read Cycle Timing From SMJ34020A I/O Registers
52
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
In Figure 23, SMJ34020A provides HRDY as soon as it recognizes the host write cycle (if no other host write
cycle is in progress), allowing the host to latch the data in the external data latches. The host then attempts a
second write but does not get an immediate HRDY because the SMJ34020A is still writing the first data to
memory. As soon as the memory write completes, HRDY goes high so that the host can latch the new data. The
SMJ34020A then writes the second data while the host continues other processing. The host access request
is synchronized to the SMJ34020A at the beginning of Q4 so that the local memory cycle can begin in Q1. If
the external host access request occurs after the setup time requirement before Q4, the request is not
considered until the next Q4 cycle. During a host write cycle DDIN is active so that if the write is to the
SMJ34020A I/O registers, the data can be required within the GSP.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
53
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
HA / HBS
HCS
HREAD
HWRITE
HRDY
(see Note A)
DATA
(in)
1st Write Valid
DATA
(out)
Previous Read
Valid
Local-Memory Host Write Cycle
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Local-Memory Host Prefetch
Q4
Q1
Q2
Q3
Q4† Q1
Q2
Q3
Q4
Q1
HOE
HDST
LAD
1st Address
2nd Address
GI
CAMD
RCA
Row
Column
Row
Column
SF
ALTCH
RAS
CAS
WE
TR / QE
DDIN
DDOUT
LRDY
SIZE16
PGMD
BUSERR
R0
R1
† See clock stretch, page 20.
NOTE A: HRDY goes high at the start of Q2; however, the memory cycle writing data to memory is not completed
until the start of Q4 when ALTCH, CAS, and HOE return high. The host must not strobe new data into
the external latch until just after the start of Q4.
Figure 23. Host-Write Cycle Back-to-Back With Prefetch of Next Word and Implicit Addressing; HREAD
and HWRITE Used as Strobes
54
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Although RESET is not normally required to be synchronous to CLKIN, in order to facilitate synchronization of
multiple SMJ34020As in a system, the rising edge of RESET must meet the setup and hold requirements to
CLKIN so that all GSPs are certain to respond to the RESET on the same quarter cycle (Figure 24). The four
possible conditions for the state of the SMJ34020A at the time RESET goes high are shown below. Quarter cycle
1 is extended accordingly to provide synchronization of the GSPs. All SMJ34020As to be synchronized must
share a common CLKIN and RESET. Within 10 CLKIN cycles after RESET goes high, all GSPs are
synchronized to the same quarter cycle through the extension of Q1 cycles.
It is recommended that SMJ34020A stretch mode not be used in multiprocessor systems.
CLKIN
RESET
Case 1
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q1
Q2
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q1
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q1
Q1
Q1
Q2
Q4
Q1
Q2
Q3
Q4
Q1
Q1
Q1
Q1
Q1
Q2
LCLK1
LCLK2
Case 2
LCLK1
LCLK2
Case 3
LCLK1
LCLK2
Case 4
LCLK1
LCLK2
NOTE A: No timing dependencies of LCLK1 and LCLK2 relative to CLKIN or RESET are to be implied from this figure.
Figure 24. Synchronization of Multiple SMJ34020As
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
55
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The timing example in Figure 25 is like a memory write cycle except that RAS and SF are high.
Command
Q4
Q1
Q2
Q3
Data Transfer
Q4
Q1
Q2
Q3
Data Transfer
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD (TMS34020A)
(see Note A)
Command
Operand 1
Operand 2
CAMD
RCA
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
NOTE A: LAD (SMJ34020A): Output to LAD by the SMJ34020A
Command:
Coprocessor ID, instruction and status code present on LAD
Operand n:
Data to or from the coprocessor
Figure 25. Transfer SMJ34020A Register(s) to Coprocessor (One or Two 32-Bit Values)
56
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The timing example in Figure 26 is like a memory write cycle except that RAS and SF are high.
Q4
Q1
Command
Q2
Q3
Q4
Q1
Data Transfer
Q2
Q3
Q4
Q1
Data Transfer
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD (TMS34020A)
(see Note A)
Command
LAD (Coprocessor)
(see Note A)
Operand 1
Operand 2
CAMD
RCA
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
NOTE A: LAD (SMJ34020A):
LAD (coprocessor):
Command:
Operand n:
Output to LAD by the SMJ34020A
Output to LAD by the coprocessor
Coprocessor ID, instruction and status code present on LAD
Data to or from the coprocessor
Figure 26. Transfer-Coprocessor Register to SMJ34020A (One or Two 32-Bit Values)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
57
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Data transfer from memory to a coprocessor requires an initialization cycle to inform the coprocessor what is
to be transferred and then a memory cycle to perform the actual transfer (Figure 27). The coprocessor can place
status information on LAD during the initialization cycle for the SMJ34020A. Two types of
memory-to-coprocessor instructions are supported: one provides a count (from 1 to 32) of data to be moved
in the instruction; the other specifies a register in the SMJ34020A to be used for the count. Both instructions
specify a register to be used as an index into memory. The index can be postincremented or predecremented
on each transfer cycle.
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Data Transfer
Data Tansfer
Address
Command Cycle
Q4
Q4†
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD
(TMS34020A)
(see Note A)
LAD
(memory)
(see Note A)
CAMD
Command
Address
Row
RCA
Data 1
Data 2
1st Column
2nd Column
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
(see Note B)
BUSFLT
R0
R1
† See clock stretch, page 20.
NOTES: A. LAD (SMJ34020A):
LAD (memory):
Command:
Address:
Data n:
Output to LAD by the SMJ34020A
Output to LAD by the memory
Coprocessor ID, instruction and status code present on LAD
Memory address for the data transfer with coprocessor status code
Data to or from the coprocessor (number of values transferred depends on a value in a register or count in
the instruction)
B. All coprocessor cycles are implemented as 32-bit operations; therefore SIZE16 should be high during these cycles.
Figure 27. Transfer Memory to Coprocessor Register(s)
58
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Data transfer from a coprocessor to memory requires an initialization cycle to inform the coprocessor what is
to be transferred and then a memory cycle to perform the actual transfer (Figure 28). The coprocessor can place
status information on LAD during the initialization cycle for the SMJ34020A. The memory cycle includes a dead
cycle to enable the SMJ34020A to take LAD drivers to the high-impedance state before the coprocessor
activates its LAD bus drivers to the memory. Two types of memory-to-coprocessor instructions are supported.
Both provide a count (from 1 to 32) of data to be moved in the instruction. Both also specify a register to be used
as an index into memory. One uses this index register with a postincrement and the other uses it with a
predecrement after each transfer cycle.
Address
Command Cycle
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q2
Q3
Data Transfer
Data Transfer
Spacer
Q4† Q1
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
LCLK1
LCLK2
GI
LAD
(TMS34020A)
(see Note A)
LAD
(Coprocessor)
(see Note A)
CAMD
Command
Address
Data 1
Status
Row
RCA
1st Column
Data 2
2nd Column
ALTCH
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
(see Note B)
BUSFLT
R0
R1
† See clock stretch, page 20.
NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A
LAD (coprocessor): Output to LAD by the coprocessor
Command:
Coprocessor ID, instruction and status code present on LAD
Address:
Memory address for the data transfer, with coprocessor status code
Data n:
Data from the coprocessor (number of values transferred depends on a count in the instruction)
Status:
Optional coprocessor status register output to LAD bus
B. All coprocessor cycles are implemented as 32-bit operations; therefore, SIZE16 should be high during these cycles.
Figure 28. Transfer-Coprocessor Register(s) to Memory (ALTCH High During Data Transfer)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
59
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
The timing example in Figure 29 is like a memory write cycle except that RAS and SF are high.
A coprocessor internal command assumes no transfer of operands or results but causes the coprocessor to
execute some internal function. The coprocessor can place status information on LAD during the cycle for the
SMJ34020A.
60
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
cycle timing examples (continued)
Q4
Q1
Command Cycle
Q3
Q4
Q1 Q2
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD
(TMS34020A)
(see Note A)
Command
CAMD
RCA
ALTCH
(see Note B)
RAS
CAS
WE
TR / QE
SF
DDIN
DDOUT
LRDY
PGMD
(see Note B)
SIZE16
(see Note C)
BUSFLT
R0
R1
NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A
LAD command:
Coprocessor ID, instruction and status
code present on LAD
B. Although the coprocessor internal command never requires the
use of page mode cycles, PGMD should be held at a valid level
during the start of Q2 after ALTCH has gone low.
C. All coprocessor cycles are implemented as 32-bit operations;
therefore, SIZE16 should be high during these cycles.
Figure 29. Coprocessor-Internal Operation Command Cycle
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
61
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
absolute maximum ratings over operating free-air temperature range†
Maximum supply voltage, VCC (see Note 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 V
Input voltage range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 0.3 V to 7 V
Off-state output voltage range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 2 V to 7 V
Operating termperature range, TA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 55°C to 125°C
Storage temperature range, Tstg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 65°C to 150°C
† Stresses beyond those listed under “absolute maximum ratings” may cause permanent damage to the device. These are stress ratings only, and
functional operation of the device at these or any other conditions beyond those indicated under “recommended operating conditions” is not
implied. Exposure to absolute-maximum-rated conditions for extended periods may affect device reliability.
NOTE 1: All voltage values are with respect to VSS.
recommended operating conditions
MIN
NOM
MAX
’34020A-32
4.5
5
5.5
’34020A-40
4.75
5
5.25
UNIT
VCC
Supply voltage
VSS
IOH
Supply voltage (see Note 2)
High-level output current
400
µA
IOL
TC
Low-level output current
2
mA
125
°C
0
Operating case temperature
V
TA
Operating temperature
– 55
125
NOTE 2: Ensure that a minimum inductance path between the VSS pins and system ground is provided to minimize noise on VSS.
62
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
V
°C
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
DC electrical characteristics over recommended range of supply voltage (see Note 3)
PARAMETER
TEST CONDITIONS
BUSFLT, LRDY, VCLK,
PGMD SIZE16,
PGMD,
SIZE16 CSYNC,
CSYNC
VSYNC, HSYNC
VIH
High-level input
voltage
lt
HWRITE HREAD
HWRITE,
HA5 – HA31,, HCS,,
HBS0 – HBS3
MIN
TYP†
MAX
GB PKG
2.2
VCC + 0.3
HT PKG
2.3
VCC + 0.3
GB PKG
2
HT PKG
2.3
VCC + 0.3
VCC + 0.3
GB PKG
2
HT PKG
2.3
CLKIN only
3
All other inputs
2
VIL
Low-level input voltage, HT only: HCS VIL = – 0.3 min, 0.7 V max
VOH
High-level output voltage
VCC + 0.3
VCC + 0.3
0.8
2.6
VOL
Low-level output
voltage
HT PKG
HYSNC, VSYNC
V
V
GB PKG
DDIN, HINT, HRDY, R0, R1,
EMU3
V
VCC + 0.3
VCC + 0.3
– 0.3
VCC = MIN,
IOH = MAX
UNIT
0.60
0.8
VCC = MAX,
IOL = MIN
V
0.8
All other outputs
0.6
GB PKG
HT PKG
IO
Output current,
current leakage (high impedance)
II
Input current (All inputs except EMU0 – EMU2,
HREAD, HWRITE‡)
ICC
Supply current
Ci
Input capacitance
GB PKG
HT PKG
’34020A-32
’34020A-40
VCC = MAX,,
VO = 2.8 V
20
VCC = MAX,,
VO = 0.6 V
– 20
VI = VSS to VCC
±20
VCC = MAX,,
Freq = MAX
265
20
µA
– 20
280
10
18
µA
mA
pF
Co
Output capacitance
18
25
pF
† All typical values are at VCC = 5 V, TA = 25° C.
‡ EMU0 – EMU2 are not connected in a typical configuration. Nominal pullup current for EMU0 – EMU2 and HREAD, HWRITE is 600 µA.
NOTE 3: HDST and HOE (output terminals) have internal pullup resistors that allow high logic levels to be maintained when the SMJ34020A is
not actually driving these pins.
signal transition levels
2V
(see Note A)
0.8 V
NOTE A: 2.2 V for BUSFLT, VCLK, LRDY, PGMD, SIZE16. 3V for CLKIN.
Figure 30. TTL-Level Inputs
For high-to-low and low-to-high transitions, the level at which the input timing is measured is 1.5 V.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
63
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
signal transition levels (continued)
2.6 V
2V
1.5 V
1V
0.6 V
Figure 31. TTL-Level Outputs
TTL-level outputs are driven to a minimum logic-high level of 2.6 V and to a maximum logic-low level of 0.6 V.
For a high-to-low transition on a TTL-compatible output signal, the level at which the output is said to be no
longer high is 2 V, and the level at which the output is said to be low is 1 V. For a low-to-high transition, the level
at which the output is said to be no longer low is 1 V, and the level at which the output is said to be high is 2 V.
A VOL trip level of 1.5 V is used for timing requirements for testing at – 55°C.
test measurement
The test load circuit shown in Figure 32 represents the programmable load of the tester pin electronics that is
used to verify timing parameters of SMJ34020A output signals.
IOL
Test
Point
From Output
Under Test
VLOAD
CLOAD
IOH
Where:
IOL = 2 mA (all outputs)
IOH = 400 µA (all outputs)
VLOAD = 1.5 V
CLOAD = 80 pF typical load circuit capacitance
NOTE: The load applied may be set higher than the values
indicated for IOL and IOH during timing tests in order to
reduce signal bounce induced by the tester hardware.
However the timing performance is assured at the stated
load values.
Figure 32. Test Load Circuit
64
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
timing parameter symbology
Timing parameter symbols used herein were created in accordance with JEDEC Standard 100. In order to
shorten the symbols, some of the pin names and other related terminology have been abbreviated as follows:
A
HA5– HA31 and HBS0 – HBS3
LINT
LINT1, LINT2
AD
LAD0– LAD31 and RCA0 – RCA12
OE
HOE
AL
ALTCH
RC
RCA0– RCA12
BC
Any of the bus control input signals
(LRDY, PGMD, SIZE16, or BUSFLT)
RD
HREAD
CE
CAS0– CAS3
RE
RAS
CK
LCLK1 and LCLK2
RQ
R0 or R1
CK1
LCLK1
RS
RESET
CK2
LCLK2
RY
HRDY
CKI
CLKIN
S
HSYNC, VSYNC, or CSYNC
CM
CAMD
SC
EMU3
CS
HCS
SCK
SCLK
CT
Any of the bus control output signals
(ALTCH, CAS0 – CAS3, RAS, WE,
TR / QE, HOE, or HDST)
SF
SF
DI
DDIN
SG
Any output signal
DO
DDOUT
SGV
Signal valid
EM
EMU0, EMU1, EMU2
ST
HDST
HI
HINT
TR
TR / QE
HS
HSYNC, VSYNC, CSYNC / HBLNK, or
CBLNK/ VBLNK
VCK
VCLK
GI
LA
GI
LAD0– LAD31
WR
HWRITE
Lowercase subscripts and their meaning are:
a
c
d
h
su
t
w
access time
cycle time (period)
delay time
hold time
setup time
transition time
pulse duration (width)
The following letters and symbols and their meaning are:
H
L
V
X
Z
High level
Low level
Valid level
Unknown, changing or don’t care level
High-impedance state of 3-state output
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
65
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
general notes on timing parameters
The period of the local clocks (LCLK1 and LCLK2) is four times the period of the input clock (CLKIN).
The quarter cycle time (tQ) that appears in the following tables is one quarter of a local output clock period or
equal to the input clock period, tc(CKI).
All output signals from the SMJ34020A are derived from an internal clock such that all output transitions for a
given quarter cycle occur with a minimum of skewing relative to each other. In the timing diagrams, the
transitions of all output signals are shown with respect to the local clocks (LCLK1 and LCLK2). The local clock
edge used as a reference occurs one internal clock cycle before the transition specified.
The signal combinations shown in the timing parameters are for timing reference only; they do not necessarily
represent actual cycles. For actual cycle descriptions, please refer to the cycle timings section of this
specification.
66
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
CLKIN and RESET timing requirements (see Figure 33)
SMJ34020A-32
NO
NO.
1
MIN
MAX
31.25
50
’34020A-40
MIN
UNIT
MAX
tc(CKI)
tw(CKIH)
Cycle time, period of CLKIN (4tQ)
Pulse duration, CLKIN high
10
tw(CKIL)
tt(CKI)
Pulse duration, CLKIN low
4
10
2†
5
th(CKI-RSL)
Hold time, RESET low after CLKIN high
15‡
8
2†
12‡
6
tsu(RSH-CKI)
Setup time, RESET high to CLKIN no longer
low
10‡
6‡
ns
7
tw(RSL)
(RSL)
Pulse duration,,
RESET low
160tQ – 40§
16tQ – 40§
160tQ – 40§
16tQ – 40§
ns
8
tsu(CSL-RSH)
Setup time, HCS low to RESET high to
configure self-bootstrap mode
8tQ + 55
8tQ + 55
ns
9
td(CSH-RSH)
Delay time, HCS no longer low to RESET high
to configure self-bootstrap mode
10
tw(CSL)
Pulse duration, HCS low to configure GSP in
self-bootstrap mode
2
3
Transition time, CLKIN
Initial reset during powerup
Reset during active operation
25
50
8
5†
4tQ – 50¶
ns
ns
5†
4tQ + 55
ns
ns
4tQ – 50¶
4tQ + 55
ns
ns
ns
† These values are based on computer simulation and are not tested.
‡ These timings are required only to synchronize the SMJ34020A to a particular quarter cycle.
§ The initial reset pulse on powerup must remain valid until all internal states have been initialized. Resets applied after the SMJ34020A has been
initialized need to be present only long enough to be recognized by the internal logic; the internal logic maintains an internal reset until all internal
states have been initialized (34 LCLK1 cycles).
¶ Parameter 9 is the maximum amount by which the RESET low-to-high transition can be delayed after the start of the HCS low-to-high transition
and still assure that the SMJ34020A is configured to run in the self-bootstrap mode (HLT bit = 0) following the end of reset.
1
4
3
4
2
CLKIN
6
RESET
5
7
8
9
10
HCS
Figure 33. CLKIN and RESET Timing Requirements
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
67
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
local-bus timing: output clocks (see Note 4 and Figure 34)
SMJ34020A-32
NO
NO.
11
12
12a
13
13a
14
15
16
17
18
19
20
21
MIN
tc(LCK)
tw(LCKH)
Cycle time, period of local clocks LCLK1, LCLK2
tw(LCKH)
tw(LCKL)
Pulse duration, LCLK1 high (see Note 5)
tw(LCKL)
tt(LCK)
Pulse duration, LCLK1 low (see Note 5)
th(CK1H-CK2L)
th(CK2H-CK1H)
Hold time, LCLK2 low after LCLK1 high
th(CK1L-CK2H)
th(CK2L-CK1L)
Hold time, LCLK2 high after LCLK1 low
th(CK1H-CK2H)
th(CK2H-CK1L)
Hold time, LCLK2 high after LCLK1 high
th(CK1L-CK2L)
th(CK2L-CK1H)
Hold time, LCLK2 low after LCLK1 low
Pulse duration, local clock high
Pulse duration, local clock low
’34020A-40
MAX
MIN
MAX
4tc(CKI)†+ s
2tQ – 15
4tc(CKI) + s†
2tQ – 13.5
ns
2tQ – 10
2tQ – 15+ s
2tQ–10+ s
2tQ –7
2tQ – 13.5 + s
2tQ –7 + s
ns
Transition time, LCLK1 or LCLK2
ns
ns
ns
15
Hold time, LCLK1 low after LCLK2 high
13.5
ns
tQ – 15
tQ – 15
tQ – 13.5
tQ – 13.5
ns
tQ – 15
tQ – 15+ s
3tQ – 15
tQ – 13.5
tQ – 13.5 + s
3tQ – 13.5
ns
3tQ – 15+ s
3tQ – 15+ s
3tQ – 13.5 + s
3tQ – 13.5 + s
ns
Hold time, LCLK1 high after LCLK2 high
Hold time, LCLK1 low after LCLK2 low
UNIT
ns
ns
ns
ns
22
Hold time, LCLK1 high after LCLK2 low
3tQ – 15+ s
3tQ – 13.5 + s
ns
† This is a functional minimum and is not tested. This parameter can also be specified as 4tQ.
NOTES: 4. s = tQ if using the clock stretch;
s = 0 otherwise
5. Parameters 12a and 13a are specified with 1.5 V timing levels (parameters 12 and 13 are specified with standard timing voltage
levels).
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
14
11
12
14
13
LCLK1
12a
19
13a
20
15
21
22
16
11
17
LCLK2
18
14
12
14
13
NOTE A: Although LCLK1 and LCLK2 are derived from CLKIN, no timing relationship between CLKIN and the local clocks is to be assumed,
except the period of the local clocks is four times the period of CLKIN.
Figure 34. Local-Bus Timing: Output Clocks
68
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
output signal characteristics†‡
The following general parameters are common to all output signals from the SMJ34020A unless otherwise
stated. They are intended as an aid in estimating the timing requirements. Please refer to the specific numbered
parameters for actual times. In the minimum and maximum values shown, “n” is an integral number of quarter
cycles.
SMJ34020A-32
PARAMETER
th(CK-SGNV)
td(CK-SGV)
td(SGNV-SGV)
td(SGV-SG)
MIN
Hold time, LCLKx to output signal not valid
tQ – 15
Delay time, LCLKx start of transition to output signal
valid
Delay time, output signal started transition to output
signal valid
Delay time, output signal valid to output signal not valid
tw(SGH)
tw(SGL)
Output signal transition time
Pulse duration, output signal high
Pulse duration, output signal low
UNIT
ns
Fast: RAS, CAS, ALTCH,
TR / QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
tQ + 15
ns
Slow: LAD, RCA, SF
tQ + 22
ns
Fast: RAS, CAS, ALTCH,
TR / QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
ntQ + 15
ns
Slow: LAD, RCA, SF
ntQ + 22
ns
Fast: RAS, CAS, ALTCH,
TR / QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
Slow: LAD, RCA, SF
tt(SG)
MAX
ntQ – 15
ntQ – 16§
ns
ntQ – 22
ns
Fast: RAS, CAS, ALTCH,
TR / QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
15
ns
Slow: LAD, RCA, SF
22
ns
Fast: RAS, CAS, ALTCH,
TR / QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
ntQ – 15
ns
Slow: LAD, RCA, SF
ntQ – 22
ns
Fast: RAS, CAS, ALTCH,
TR / QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
ntQ – 15
ns
Slow: LAD, RCA, SF
ntQ – 22
ns
† Also refer to Figure 34 on following page.
‡ For parameters on this page specifying minimum or maximum times between two output signals, the word fast or slow in column 2 refers to the
signal with a subscript of 1, regardless of the other signal. For example, if you are using the spec th(SG2NV-SG1V), use the slow value if the signal
becoming valid (SG1) is RCA, LAD, or SF; use the fast value otherwise. The pin referred to as SG2 does not determine fast or slow signal time.
§ See parameter 73 in “local-bus timing: bus control inputs” table.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
69
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
QW
QX
QY
QZ
See Note A
LCLKx
td(CK-SGV)
td(CK-SGV)
th(CK-SG)
tw(SGH)
SIGNALa
tt(SG)
th(CK-SG)
td(SG-SGV)
tt(SG)
td(SG-SGV)
td(SGV-SG)
td(SGV-SG)
tw(SGL)
SIGNALb
td(SGV-SG)
td(SG-SGV)
NOTE A: Any of these quarter phases could be 2tQ if they are stretched (see clock stretch,
page 20).
Figure 35. Output Signal Characteristics
example of how to use the general output signal characteristics
Assume a system is using a SMJ34020A-32. Determine the maximum time from the start of the falling edge
of ALTCH to the time when data must be valid on LAD for a local-memory write cycle.
From the local-memory read-modify-write-cycle timing diagram on page 28, the time from the falling edge of
ALTCH to valid data on LAD is roughly Q3 + Q4; i.e., 2tQ. A more precise value can be obtained by using the
table of output signal characteristics.
The parameter of interest is td(SG-SGV). Note that in Figure 35, there are two representations of td(SG-SGV) that
relate SIGNALa and SIGNALb (the third representation of this parameter relates SIGNALb to itself and is not
useful in this example). Let SIGNALa represent ALTCH because ALTCH is making a transition first. Let
SIGNALb represent LAD. By definition, the signal becoming valid (SGV) determines whether the fast value or
the slow value from the table is used.
In this case, for parameter td(SG-SGV), SGV is LAD. LAD is in the slow group, so the maximum value for td(SG-SGV)
is ntQ + 22. The value for n is 2 from the analysis of the diagram on page 28. Thus, the maximum time from the
start of the falling edge of ALTCH to the time when data must be valid on LAD for a local-memory write cycle
is 2tQ + 22 ns.
70
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
host-interface-cycle timing requirements (see Note 6 and Figure 36)
’34020-32
’34020A-32
NO.
23
MIN
MAX
’34020A-40
MIN
MAX UNIT
tsu(AV-CSL)
th(CSL-AV)
Setup time, address prior to HCS no longer high
12
10
ns
Hold time, address after HCS low
12
10
ns
tw(CSH)
tw(RDH)
Pulse duration, HCS high
28
25
ns
Pulse duration, HREAD high
28
25
ns
tw(WRH)
tsu(RDH-WRL)
Pulse duration, HWRITE high
28
25
ns
Setup time, HREAD high to HWRITE no longer high
28
25
ns
tsu(WRH-RDL)
tw(RDL)
Setup time, HWRITE high to HREAD no longer high
28
25
ns
Pulse duration, HREAD low
18
15
ns
tw(WRL)
tsu(CSL-WRH)
Pulse duration, HWRITE low
18
15
ns
32
tsu(RDL-CK2L)
Setup time, HCS low or HREAD low to LCLK2 no longer high
15
25†
ns
33
18
30†
34
tsu(WRH-CK2L)
Setup time, HWRITE high or HCS high to LCLK2 no longer
high
30†
25†
ns
35
th(CK2L-RDH)
th(CK2L-WRL)
0‡
0‡
0‡
0‡
ns
30†§
25†§
ns
24
25
26
27
28
29
30
31
36
37
tsu(RDH-CK2L)
Setup time, HCS low to HWRITE no longer low
Hold time, HREAD high after LCLK2 no longer high
Hold time, HWRITE low after LCLK2 no longer high
Setup time, HREAD high to LCLK2 no longer high, prefetch
read mode
ns
ns
38
tsu(CSL-RDH)
Setup time, HCS low to HREAD no longer low
18
15
ns
† Setup time to ensure recognition of input on this clock edge.
‡ Hold time required to assure response on next clock edge. These values are based on computer simulation and are not tested.
§ When the SMJ34020A is set for block reads, use the deassertion of HREAD to request a local memory cycle at the next sequential address
location.
NOTE 6: Although HCS, HREAD, and HWRITE can be totally asynchronous to the SMJ34020A, cycle responses to the signals are determined
by local memory cycles.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
71
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
HA0 – HA27
HBS0 – HBS3
23
25
24
HCS
32
26
38
30
HREAD
28
31
27
HWRITE
35
34
37
33
36
LCLK2
Figure 36. Host-Interface-Cycle Timing Requirements
72
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
29
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
host-interface-cycle timing responses (random read cycle) (see Note 4 and Figure 37)
’34020A -30
NO
NO.
26
MIN
tw(RDH)
Pulse duration, HREAD high
33
tsu(RDL-CK2L)
Setup time, HCS low or HREAD low to LCLK2 no
longer high
39
td(CK1H-RYH)
Delay time, LCLK1 going high to HRDY high (end of
read cycle)
40
td(RDH-RYL)
td(CK2L-STL)
Delay time, HREAD or HCS high to HRDY low
td(CK1L-STH)
tsu(STL-RYH)
Delay time, LCLK1 no longer high to HDST high
41
42
43
’34020A-40
MAX
MIN
UNIT
28
25
ns
30†
25†
ns
tQ + 20
tQ+18
ns
18
ns
tQ + 15
tQ + 15
tQ + 13.5+s
tQ –13.5
ns
ns
2tQ + 15
2tQ + 13.5
2tQ + 13.5
20
s+
Delay time, LCLK2 no longer high to HDST low
Setup time, HDST low to HRDY no longer low
MAX
tQ–15
44
td(RYH-STH)
Delay time, HRDY no longer low to HDST high
† Setup time to ensure recognition of input on this clock edge.
NOTE 4: s = tQ if using the clock stretch;
s = 0 otherwise
ns
ns
.
Q4
Q1
Q2
Q3
Q4‡
Q1
Q2
Q3
Q4
LCLK1
39
LCLK2
33
33
26
HCS/HREAD
40
41
HRDY
42
44
43
HDST
‡ See clock stretch, page 20.
Figure 37. Host-Interface-Cycle Timing Responses (Random Read Cycle)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
73
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
host-interface-cycle timing (block-read cycle) (see Notes 4 and 6 and Figure 38)
’34020A-32
NO
NO.
26
MIN
30
tw(RDH)
tw(RDL)
37
tsu(RDH-CK2L)
39
td(CK1H-RYH)
td(RDH-RYL)
Delay time, LCLK1 no longer low to HRDY high
td(CK2L-STL)
td(CK1L-STH)
Delay time, LCLK2 no longer high to HDST low
Setup time, HDST low to HRDY no longer low
44
tsu(STL-RYH)
td(RYH-STH)
45
td(RDL-RYH)
40
41
42
43
’34020A-40
MAX
MIN
MAX
UNIT
Pulse duration, HREAD high
28
25
ns
Pulse duration, HREAD low
18
15
ns
30†
25†
Setup time, HREAD high to LCLK2 no longer high,
prefetch read mode
Delay time, HREAD or HCS high to HRDY low
Delay time, LCLK1 no longer high to HDST high
ns
tQ + 20
20
tQ + 18
18
ns
tQ + 15+s
tQ + 15
tQ + 13.5 + s
tQ + 13.5
ns
tQ–15
tQ – 13.5
Delay time, HRDY no longer low to HDST high
ns
ns
2tQ + 15
Delay time, HREAD or HCS low to HRDY high after
prefetch
ns
2tQ + 13.5
ns
20
ns
25
50
th(STH-CTV)
Hold time, CAS, TR / QE, DDIN valid after HDST high
–2
–2
ns
† Setup time to ensure recognition of input on this clock edge. When the SMJ34020A is set for block reads, the deassertion of HREAD is used
to request a local memory cycle at the next sequential address location.
NOTES: 4. s = tQ if using the clock stretch;
s = 0 otherwise
6. Although HCS, HREAD, and HWRITE can be totally asynchronous to the SMJ34020A, cycle responses to the signals are
determined by local memory cycles.
Q4
Q1
Q2
Q3
Q4‡
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
LCLK1
LCLK2
37
37
HCS
26
HREAD
26
39
30
40
40
45
HRDY
43
41
42
44
HDST
50
CAS
TR / QE
DDIN
‡ See clock stretch, page 20.
Figure 38. Host-Interface-Cycle Timing (Block-Read Cycle)
74
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
host interface timing responses (write cycle) (see Note 4 and Figure 39)
NO.
’34020-32
’34020A-32
PARAMETER
MIN
27
31
tw(WRH)
tw(WRL)
’34020A-40
MAX
MIN
UNIT
MAX
Pulse duration, HWRITE high
28
25
ns
Pulse duration, HWRITE low
18
15
ns
30†
25†
ns
34
tsu(WRH-CK2L)
Setup time, HWRITE high or HCS high to LCLK2 no
longer high
39
td(CK1L-RYH)
Delay time from LCLK1↑ to HRDY high
46
td(WRL-RYH)
47
tQ + 20
tQ + 18
ns
Delay time from later of HCS or HWRITE low to HRDY
high (TMS34020 ready)
25
20
ns
td(WRH-RYL)
Delay time from earlier of HCS or HWRITE high to
HRDY low (end of write)
25
20
ns
48
td(CK2L-OEL)
Delay time from LCLK2↓ to HOE low
tQ + 15 + s
tQ + 13.5 + s
ns
49
td(CK1H-OEH)
Delay time from LCLK1↓ to HOE high
tQ + 15
tQ + 13.5
ns
51
td(RYH-OEH)
Delay time from HRDY↑ to HOE high
† Setup time to ensure recognition of input on this clock edge.
NOTE 4: s = tQ if using the clock stretch;
s = 0 otherwise
2tQ + 15
2tQ + 13.5
ns
Q4
Q1
Q2
Q3
Q4‡
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
LCLK1
LCLK2
34
34
27
31
HCS
HWRITE
39
47
47
HRDY
49
46
48
51
HOE
‡ See clock stretch, page 20.
Figure 39. Host-Interface-Cycle Timing Responses (Write-Cycle)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
75
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
local-bus timing: bus control inputs (see Note 4 and Figure 40)
’34020A-32
NO
NO.
MIN
ta(CMV-LAV)†
Access time, CAMD valid after address valid on LAD
53
th(LA-CMV)†
Hold time, CAMD valid after address no longer valid
on LAD
54
ta(BCV-ALL)†
Access time, control valid (LRDY, PGMD, SIZE16,
BUSFLT) after ALTCH low
55
th(CK2H-BCV)†
Hold time, control (LRDY, PGMD, SIZE16, BUSFLT)
valid after LCLK2 high
56
tsu(BCV-CK2H)†
Setup time, SIZE16 valid before LCLK2 no longer
low
52
’34020A-40
MAX
MIN
MAX
3tQ – 45
3tQ – 37
0
0
3tQ – 35 + s
UNIT
ns
ns
3tQ – 27 + s
ns
0
0
ns
20
15
ns
† CAMD, LRDY, PGMD, SIZE16, and BUSFLT are synchronous inputs. The specified setup, access and hold times must be met for proper device
operation.
NOTE 4: s = tQ if using the clock stretch;
s = 0 otherwise
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4‡
Q1
Q2
Q3
LCLK1
LCLK2
Address
LAD
52
53
CAMD
Valid
55
ALTCH
54
LRDY
Valid
PGMD
Valid
56
55
SIZE16
BUSFLT
Valid
Valid
Valid
‡ See clock stretch, page 20.
Figure 40. Local-Bus Timing: Bus Control Inputs
76
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q4
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
local-bus timing: bus control inputs (see Note 4 and Figure 41)
’34020A-32
NO
NO.
MIN
’34020A-40
MAX
MIN
MAX
UNIT
57
td(CK2H-ALL)
Delay time, ALTCH low after LCLK2 no
longer low
tQ + 15
tQ + 13.5
ns
58
td(CK1L-ALH)
Delay time, ALTCH high after LCLK1 no
longer high
tQ + 15
tQ + 13.5
ns
59
td(CK1H-LAV)
Delay time, LAD0 – LAD31 address valid
after LCLK1 no longer low
tQ + 22
tQ + 20
ns
60
th(LAV-CK2L)
Hold time, LAD0 – LAD31 address valid after
LCLK2 low
tQ – 15 + s
tQ – 12 + s
ns
61
td(CT-LAD)
Delay time, LAD0 – LAD31 driven after
earlier of DDIN no longer high or CAS no
longer low or TR/QE no longer low
tQ – 5 + s †
tQ – 5 + s †
ns
62
th(LAV-CTV)
Hold time, LAD0 – LAD31 read data valid
after earlier of DDIN low or RAS, CAS, or
TR/QE low
3.5
3.5
ns
63
td(CK2L-LAV)
Delay time, LAD0 – LAD31 data valid after
LCLK2 no longer high (write)
64
th(CK2L-LAV)
Hold time, LAD0 – LAD31 data valid after
LCLK2 low (write)
65
td(CK1H-RCV)
Delay time, RCA0 – RCA12 row address
valid after LCLK1 no longer low
tQ + 22
tQ + 22
ns
66
td(CK2L-RCV)
Delay time, LAD0 – LAD31 column address
valid after LCLK2 no longer high
tQ + 22 + s
tQ + 20+ s
ns
67
th(RCV-CK2L)
Hold time, RCA0-RCA12 address valid after
LCLK2 low
68
td(CK1H-DIH)
Delay time, DDIN high after LCLK1 no longer
low
tQ + 15
tQ + 13.5
ns
69
td(CK1L-DIL)
Delay time, DDIN low after LCLK1 no longer
high
tQ + 15
tQ + 13.5
ns
70
td(CK1H-DOL)
Delay time, DDOUT low after LCLK1 no
longer low
tQ + 15
tQ + 13.5
ns
71
td(CK1L-DOH)
Delay time, DDOUT high after LCLK1 no
longer high
tQ + 15
tQ + 13.5
ns
72
td(CK2L-DOL)
Delay time, DDOUT low after LCLK2 no
longer high
tQ + 15 + s
tQ + 13.5 + s
ns
73
tsu(LAV-ALL)
Setup time, LAD0 – LAD31 data valid before
ALTCH no longer high
74
ten(DAV-DIH)
Enable time, data valid after DDIN high
(see Note 7)
75
tdis(DAV-DIL)
Disable time, data in the high-impedance
state after DDIN low (see Note 7)
tQ + 22 + s
tQ – 15
tQ + 20 + s
tQ – 13.5
tQ – 15
ns
tQ – 13.5
tQ – 16
ns
ns
tQ – 13.5
ns
2tQ – 20
2tQ – 17
ns
tQ – 12 + s †
tQ – 10 + s†
ns
† These values are derived from characterization data and are not tested.
NOTES: 4. s = tQ if using the clock stretch;
s = 0 otherwise
7. DDIN is used to control LAD bus buffers between the SMJ34020A and local memory. Parameter 74 references the time for these
data buffers to go from the high-impedance state to an active level. Parameter 75 references the time for the buffers to go from an
active level to the high-impedance state.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
77
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
LCLK1
LCLK2
57
63
58
ALTCH
73
60
64
59
Address/Status
LAD0 – LAD31
Data In
66
Data Out
75
67
74
66
65
Row
RCA0 – RCA12
67
67
1st Column
68
2nd Column
62
69
DDIN
72
70
71
71
DDOUT
CAS
61
TR / QE
Figure 41. Local-Bus Timing: Bus Control Inputs (Continued)
78
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
local-bus timing: RAS, CAS0–CAS3, WE, TR/QE, and SF (see Notes 4 and 8 and Figure 42)
’34020A-32
NO
NO.
’34020A-40
MIN
MAX
MIN
MAX
UNIT
62
th(LAV-CTV)
Hold time, LAD0 – LAD31 read data valid
after earlier of DDIN, low after RAS,
CAS, or TR / QE high
76
td(CK1L-REL)
Delay time, RAS low after LCLK1 no
longer high
tQ + 12+s
tQ + 10 + s
ns
77
td(CK1L-REH)
Delay time, RAS high after LCLK1 no
longer high
tQ + 12
tQ + 10
ns
78
td(CK1H-CEL)
Delay time, CAS low after LCLK1 no
longer low
tQ + 12
tQ + 10
ns
79
td(CK1L-CEH)
Delay time, CAS high after LCLK1 no
longer high
tQ + 12
tQ + 10
ns
80
td(CK2L-WEL)
Delay time, WE low after LCLK2 no
longer high
tQ + 15+s
tQ + 13.5 + s
ns
81
td(CK1L-WEH)
Delay time, WE high after LCLK1 no
longer high
tQ + 15
tQ + 15
ns
82
td(CK2L-TRL)
Delay time, TR / QE low after LCLK2 no
longer high
tQ + 15+s
tQ + 13.5 + s
ns
83
td(CK1L-TRH)
Delay time, TR / QE high after LCLK1 no
longer high
tQ + 15
tQ + 13.5
ns
84
td(CK1H-SFV)
Delay time, SF valid after LCLK1 no
longer low
tQ + 22
tQ + 20
ns
85
td(CK2L-SFV)
Delay time, SF valid after LCLK2 no
longer high
tQ + 22+s
tQ + 20 + s
ns
86
td(CK2L-SFZ)
Delay time, SF in the high-impedance
state after LCLK2 no longer high
tQ + 22 †
tQ + 20 †
ns
87
tsu(ADV-REL)‡
Setup time, row address valid before
RAS no longer high
2tQ – 22
2tQ – 20
ns
88
th(ADV-REL)‡
Hold time, row address valid after RAS
low
tQ – 5+ s
tQ – 5 + s
ns
89
tsu(RCV-CEL)
Setup time, column address valid before
CAS no longer high
tQ – 22
tQ – 20
ns
90
th(RCV-CEH)
Hold time, column address valid after
CAS high
tQ – 15
tQ – 13.5
ns
91
tsu(CAV-CEL)
Setup time, write data valid before CAS
no longer high
tQ – 22
tQ – 20
ns
92
th(CAV-CEH)
Hold time, write data valid after CAS no
longer low
tQ – 15
tQ – 13.5
ns
93
ta(LAV-REL)
Access time, data-in valid after RAS low
(assuming maximum transition time)
4tQ – 8+s
4tQ – 8 + s
ns
94
ta(LAV-CEL)
Access time, data-in valid after CASL no
longer high
2tQ –8+s
2tQ – 8+s
ns
95
ta(LAV-RCV)
Access time, data-in valid after column
address valid
3tQ –20+s
3tQ – 12
ns
3.5
3.5
ns
† These values are derived from characterization data and are not tested.
‡ Parameters 87 and 88 also apply to WE, TR / QE, and SF relative to RAS.
NOTES: 4. s = tQ if using the clock stretch;
s = 0 otherwise
8. Parameter 96 has been eliminated.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
79
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
local-bus timing: RAS, CAS0–CAS3, WE, TR/QE, and SF (see Notes 4 and 8 and Figure 42)
(continued)
’34020A-32
NO
NO.
MIN
Setup time, write low before CAS no
longer high (on write cycles)
97
tsu(WEL-CEL)
98
tw(REH)
tw(REL)
Pulse duration, RAS high
tw(CEH)
tw(CEL)
Pulse duration, CAS high
99
100
101
102
td(REL-CEH)
Pulse duration, RAS low
Pulse duration, CAS low
Delay time, RAS low to CAS no longer
low
POST OFFICE BOX 1443
MIN
MAX
UNIT
tQ – 15
tQ – 13.5
ns
4tQ – 12 + s
4ntQ – 12 + s
4tQ – 10 + s
4n tQ – 4 + s
ns
2tQ – 12
2tQ – 12
2tQ – 10
2tQ – 8
ns
4tQ – 12 + s
4tQ – 4+s
ns
NOTES: 4. s = tQ if using the clock stretch;
s = 0 otherwise
8. Parameter 96 has been eliminated.
80
’34020A-40
MAX
• HOUSTON, TEXAS 77251–1443
ns
ns
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Q1
Q2
Q4†
Q3
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
76
77
LCLK2
98
99
RAS
87
88
Row
RCA
1st Column
93
2nd Column
90
62
89
LAD (READ)
Address/Status
Data In 1
Data In 2
95
LAD (WRITE)
Address/Status
1st Data Out
92
79
78
102
CAS0 – CAS3
62
2nd Data Out
94
91
100
79
101
80
81
81
97
WE
82
TR / QE
83
83
84
85
86
SF
ALTCH
† See clock stretch, page 20.
Figure 42. Local-Bus Timing: RAS, CAS0 – CAS3, WE, TR / QE, and SF
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
81
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
The refresh pseudo-address present on LAD0 – LAD31 is the output from the 16-bit refresh address
register([ I O] register located at C000 01F0h) on LAD16 – LAD31. LAD0 – LAD3 have the refresh status code
(status code = 0011), and LAD4 – LAD15 are held low.
CBR refresh: RAS and CAS0–CAS3 (see Note 4 and Figure 42)
’34020A-32
NO
NO.
MIN
76
td(CK1L-REL)
Delay time, RAS low after LCLK1 no
longer high
77
td(CK1L-REH)
78
MIN
MAX
UNIT
tQ + 12 + s
tQ + 10 + s
ns
Delay time, RAS high after LCLK1 no
longer high
tQ + 12
tQ + 10
ns
td(CK1H-CEL)
Delay time, CAS low after LCLK1 no
longer low
tQ + 12
tQ + 10
ns
79
td(CK1L-CEH)
Delay time, CAS high after LCLK1 no
longer high
tQ + 12
tQ + 10
ns
102
td(REL-CEH)
Delay time, RAS low to CAS no longer
low
4tQ – 12 + s
4tQ – 4 + s
ns
103
td(CEL-REL)
Delay time, CAS low to RAS no longer
high
2tQ – 15
2tQ – 13.5
ns
104
td(REH-CEL)
Delay time, RAS high to CAS no longer
high
2tQ – 15 + s
2tQ – 13.5 + s
ns
NOTE 4: s = tQ if using the clock stretch;
s = 0 otherwise
82
’34020A-40
MAX
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
76
RAS
77
78
79
103
104
102
CAS
ALTCH
(see Note A)
LAD
(see Note A)
Refresh Pseudo-Address
RCA
(see Note A)
Refresh Pseudo-Address
DDOUT
(see Note A)
† See clock stretch, page 20.
NOTE A: ALTCH, LAD, RCA, and DDOUT are shown for reference only.
Figure 43. CBR Refresh: RAS and CAS0 – CAS3
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
83
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
multiprocessor-interface timing: GI, ALTCH, RAS, R0 and R1 (see Figure 44)
’34020A-32
NO
NO.
MIN
105
ta(GIV-RQV)
Access time, GI valid after R0 and R1 valid
(see Note 9)
105.1
tsu(GIV-CK1H)
Setup time, GI valid before LCLK1 no longer low
(see Note 9)
106
th(CK1H-GIV)
td(CK2H-RQV)
107
MAX
’34020A-40
MIN
MAX
2tQ – 40
2tQ – 30
40
Hold time, GI valid after LCLK1 no longer low
35
0
ns
tQ + 15
108
td(CK2H-RQNV)
Delay time, LCLK2 high to R0 or R1 no longer valid
tQ – 15
NOTE 9: These timings must be met to ensure that GI is recognized on this clock cycle.
ns
ns
0
Delay time, LCLK2 no longer low to R0 or R1 valid
UNIT
tQ + 13.5
tQ – 13.5
ns
ns
For a SMJ34020A to gain control of the local bus during a given cycle, GI must be low at the start of Q1 (indicating
that the bus arbitration logic is granting the bus to this processor).
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
LCLK1
105.1
LCLK2
107
108
Valid
R0 – R1
106
105
Valid
GI
† See clock stretch, page 20.
†
Figure 44. Multiprocessor-Interface Timing: GI, ALTCH, RAS, R0 and R1
84
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q1
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
multiprocessor-interface timing: high-impedance signals (see Note 4 and Figure 45)
’34020A-32
NO
NO.
84
MIN
MAX
’34020A-40
MIN
MAX
UNIT
td(CK1H-SFV)
Delay time, SF valid after LCLK1 no longer low
tQ+ 22
tQ + 20
ns
86
td(CK2L-SFZ)
Delay time, SF in the high-impedance state after
LCLK2 no longer high
tQ+ 22+ s †
tQ + 20 + s†
ns
109
td(CK2L-ADZ)
Delay time, LAD and RCA in the high-impedance
state after LCLK2 no longer high
tQ+ 22+ s †
tQ+ 20 + s†
ns
110
td(CK1H-ADV)
Delay time, LAD and RCA valid after LCLK1 no longer
low
tQ+ 22
tQ + 20
ns
111
td(CK1H-CTZ)
Delay time, ALTCH, RAS, CAS, WE, TR / QE, HOE,
and HDST in the high-impedance state after LCLK1
no longer low
tQ+15 †
tQ +13.5†
ns
112
td(CK2L-CTH)
Delay time, ALTCH, RAS, CAS, WE, TR / QE, HOE,
and HDST in the high-impedance state after LCLK2
no longer high
tQ + 15+ s
tQ + 13.5 + s
ns
113
td(CK1H-DIZ)
Delay time, DDIN in the high-impedance state after
LCLK1 no longer low
tQ + 15 †
tQ + 13.5 †
ns
114
td(CK2L-DIL)
Delay time, DDIN low after LCLK2 no longer high
tQ + 15 + s
tQ + 13.5+ s
ns
td(CK2L-DOZ)
Delay time, DDOUT in the high-impedance state after
LCLK2 no longer high
tQ+15 + s†
tQ +13.5 + s†
ns
116
td(CK2L-DOH) Delay time, DDOUT high after LCLK2 no longer high
† These values are derived from characterization data and are not tested.
tQ + 15+ s
tQ + 13.5 + s
ns
115
NOTE 4: s = tQ if using the clock stretch;
s = 0 otherwise
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
85
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
Q4
Q1
Q2
Q3
Q4†
Q1
Q2
Q3
Q4
LCLK1
LCLK2
GI
R0 – R1
109
110
LAD0 – LAD31,
RCA0 – RCA12
Hi-Z
112
111
ALTCH, RAS,
CAS, WE,
TR / QE, HOE, HDST
Hi-Z
84
86
SF
Hi-Z
114
113
Hi-Z
DDIN
116
115
Hi-Z
DDOUT
† See clock stretch, page 20.
Figure 45. Multiprocessor-Interface Timing: High-Impedance Signals
86
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q1
Q2
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
video-shift-clock timing: SCLK (see Figure 46)
NO
NO.
117
118
119
120
’34020A-32
’34020A-40
MIN
MAX
MIN
MAX
50
25
50
tc(SCK)
tw(SCKH)
Cycle time, period of video serial clock SCLK
35
Pulse duration, SCLK high
12
tw(SCKL)
tt(SCK)
Pulse duration, SCLK low
12
2†
Transition time, (rise and fall) of SCLK
10
5†
UNIT
ns
ns
10
2†
ns
5†
ns
† This value is determined through computer simulation and is not tested.
117
120
118
120
119
SCLK
Figure 46. Video-Shift-Clock Timing: SCLK
video-interface timing: VCLK and video outputs (see Figure 47)
NO
NO.
123
124
125
126
’34020A-32
’34020A-40
MIN
MAX
MIN
MAX
62.5
100
62.5
100
tc(VCK)
tw(VCKH)
Cycle time, period of video input clock VCLK
Pulse duration, VCLK high
28
28
tw(VCKL)
tt(VCK)
Pulse duration, VCLK low
28
2†
28
2†
Transition time, (rise and fall) of VCLK
5†
UNIT
ns
ns
ns
5†
ns
127
td(VCKL-HSL)
Delay time, VCLK low to HSYNC, VSYNC, CSYNC / VBLNK or
CBLNK / VBLNK low
128
td(VCKL-HSH)
Delay time, VCLK low to HSYNC, VSYNC, CSYNC / HBLNK, or
CBLNK / VBLNK high
129
th(VCKL-HSL)
Hold time, VCLK no longer high to HSYNC, VSYNC,
CSYNC / HBLNK, or CBLNK / VBLNK no longer high
0†
0†
ns
130
th(VCKL-HSH)
Hold time, VCLK no longer high to HSYNC, VSYNC,
CSYNC / HBLNK, or CBLNK / VBLNK no longer low
0†
0†
ns
40
40
ns
40
40
ns
† This value is determined through computer simulation and is not tested.
123
124
126
126
125
VCLK
127
HSYNC
VSYNC
CSYNC / HBLNK
CBLNK / VBLNK
(outputs)
128
129
130
Figure 47. Video-Interface Timing: VCLK and Video Outputs
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
87
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
video-interface timing: external sync inputs (see Note 10 and Figure 48)
NO
NO.
131
’34020A-40
MIN
MIN
MAX
MAX
UNIT
Setup time, HSYNC, VSYNC, CSYNC low to VCLK no longer low
20
20
ns
Setup time, HSYNC, VSYNC, CSYNC high to VCLK no longer low
20
20
ns
133
th(VCKH-SV)
Hold time, HSYNC, VSYNC, CSYNC valid after VCLK high
20
20
NOTE 10: Setup and hold times on asynchronous inputs are required only to assure recognition at indicated clock edges.
ns
132
tsu(SL-VCKH)
tsu(SH-VCKH)
’34020A-32
A
B
C
D
VCLK
133
131
133
HSYNC
VSVNC
CSYNC
(inputs)
See Note A
132
See Note B
NOTES: A. If the falling edge of the sync signal occurs more than th(VCKH-SV) after VCLK edge A and at least tsu(SL-VCKH) before
edge B, the transition is detected at edge B instead of edge A.
B. If the rising edge of the sync signal occurs more than th(VCKH-SV) after VCLK edge C and at least tsu(SH-VCKH) before
edge D, the transition is detected at edge D instead of edge C.
Figure 48. Video-Interface Timing: External Sync Inputs
interrupt timing: LINT1 and LINT2 (see Figure 49)
’34020A-32
NO
NO.
134
MIN
tsu(LINTL-CK2H)
Setup time, LINT1 or LINT2 low before LCLK2 no longer
low
tQ + 45 †
MAX
’34020A-40
MIN
tQ + 40 †
MAX
UNIT
ns
135
tw(LINTL)
Pulse duration, LINT1 or LINT2 low
8tQ‡
8tQ‡
ns
† Although LINT1 and LINT2 can be asynchronous to the SMJ34020A, this setup ensures recognition of the interrupt on this clock edge.
‡ This pulse duration minimum ensures that the interrupt is recognized by internal logic; however, the level must be maintained until it has been
acknowledged by the interrupt service routine.
LCLK1
LCLK2
134
LINT1
LINT2
135
Figure 49. Interrupt Timing: LINT1 and LINT2
88
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
host-interrupt timing: HINT (see Figure 50)
’34020A-32
NO
NO.
136
td(CK1H-HINTV)
Delay time, LCLK1 no longer low to HINT valid
’34020A-40
30
UNIT
25
ns
LCLK1
136
LCLK2
136
HINT
Figure 50. Host-Interrupt Timing: HINT
emulator-interface timing (see Figure 51)
’34020A-32
NO
NO.
137
138
139
140
MIN
tsu(EMV-CK1H)
th(EMV-CK1H)
Setup time, EMU0 – EMU2 valid to LCLK1 no longer low
td(CK1L-SCV)
th(CK2H-SCNV)
Delay time, EMU3 valid after LCLK1 low
Hold time, EMU0 – EMU2 valid after LCLK1 no longer low
MAX
’34020A-40
MIN
UNIT
30
25
ns
0
0
ns
25
Hold time, LCLK2 high before EMU3 not valid
MAX
20
tQ – 15
ns
tQ – 13.5
ns
LCLK1
LCLK2
137
137
138
138
EMU0 – EMU2
139
139
EMU3
140
Figure 51. Emulator-Interface Timing
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
89
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
MECHANICAL DATA
GA-GB (S-CPGA-P15 X 15)
CERAMIC PIN GRID ARRAY PACKAGE
A or A1 SQ
1.400 (35,56) TYP
R
P
N
M
L
K
J
H
G
F
E
D
C
B
A
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
DIM
MIN
MAX
A
1.540 (39,12)
1.590 (40,38)
Large
Outline
A1
1.480 (37,59)
1.535 (38,99)
Small
Outline
B
0.110 (2,79)
0.205 (5,21)
Cavity
Up
B1
0.095 (2,41)
0.205 (5,21)
Cavity
Down
C
0.040 (1,02)
0.060 (1,52)
Cavity
Up
C1
0.025 (0,63)
0.060 (1,52)
Cavity
Down
B or B1
C or C1
0.050 (1,27) DIA
4 Places
0.022 (0,55)
0.016 (0,41)
0.140 (3,56)
0.120 (3,05)
0.100 (2,54)
DIA TYP
Notes
MAXIMUM PINS WITHIN MATRIX – 225
4040114-8 / B 10/94
NOTES: A.
B.
C.
D.
All linear dimensions are in inches (millimeters).
This drawing is subject to change without notice.
Index mark may appear on top or bottom depending on package vendor.
Pins are located within 0.005 (0,13) radius of true position relative to each other at maximum material condition and within
0.015 (0,38) radius relative to the center of the ceramic.
E. This package can be hermetically sealed with metal lids or with ceramic lids using glass frit.
F. The pins can be gold plated or solder dipped.
G. Falls within MIL-STD-1835 CMGA7-PN and CMGA19-PN and JEDEC MO-067AG and MO-066AG, respectively
90
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
SMJ34020A
GRAPHICS SYSTEM PROCESSOR
SGUS011B – APRIL 1991 – REVISED AUGUST 1995
MECHANICAL DATA
HT (S-CQFP-F132)
CERAMIC QUAD FLATPACK
1.525 (38,74)
SQ
1.480 (37,59)
0.960 (24,38)
SQ
0.935 (23,75)
66
34
33
67
0.800 (20,32) TYP
99
1
100
132
0.105 (2,67) MAX
132
0.013 (0,330)
0.006 (0,152)
0.025 (0,64)
0.018 (0,46) MAX
0.014 (0,36)
0.008 (0,20)
0.004 (0,10)
0.002 (0,05)
At Braze Pads
0.130 (3,30) MAX
4040169-3 / B 10/94
NOTES: A.
B.
C.
D.
E.
All linear dimensions are in inches (millimeters).
This drawing is subject to change without notice.
This package is hermetically sealed with a metal lid.
The terminals are gold plated.
Falls within JEDEC MO -090 AB
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
91
IMPORTANT NOTICE
Texas Instruments and its subsidiaries (TI) reserve the right to make changes to their products or to discontinue
any product or service without notice, and advise customers to obtain the latest version of relevant information
to verify, before placing orders, that information being relied on is current and complete. All products are sold
subject to the terms and conditions of sale supplied at the time of order acknowledgement, including those
pertaining to warranty, patent infringement, and limitation of liability.
TI warrants performance of its semiconductor products to the specifications applicable at the time of sale in
accordance with TI’s standard warranty. Testing and other quality control techniques are utilized to the extent
TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily
performed, except those mandated by government requirements.
CERTAIN APPLICATIONS USING SEMICONDUCTOR PRODUCTS MAY INVOLVE POTENTIAL RISKS OF
DEATH, PERSONAL INJURY, OR SEVERE PROPERTY OR ENVIRONMENTAL DAMAGE (“CRITICAL
APPLICATIONS”). TI SEMICONDUCTOR PRODUCTS ARE NOT DESIGNED, AUTHORIZED, OR
WARRANTED TO BE SUITABLE FOR USE IN LIFE-SUPPORT DEVICES OR SYSTEMS OR OTHER
CRITICAL APPLICATIONS. INCLUSION OF TI PRODUCTS IN SUCH APPLICATIONS IS UNDERSTOOD TO
BE FULLY AT THE CUSTOMER’S RISK.
In order to minimize risks associated with the customer’s applications, adequate design and operating
safeguards must be provided by the customer to minimize inherent or procedural hazards.
TI assumes no liability for applications assistance or customer product design. TI does not warrant or represent
that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other
intellectual property right of TI covering or relating to any combination, machine, or process in which such
semiconductor products or services might be or are used. TI’s publication of information regarding any third
party’s products or services does not constitute TI’s approval, warranty or endorsement thereof.
Copyright  1998, Texas Instruments Incorporated