TI TMS34020APCM40

TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
•
•
•
•
•
•
•
•
•
•
•
Instruction Cycle Time
– 100 ns . . . TMS34020A-40
– 125 ns . . . TMS34020-32
– 125 ns . . . TMS34020A-32
Fully Programmable 32-Bit
General-Purpose Processor With
512-Megabyte Linear Address Range
(Bit Addressable)
Second-Generation Graphics Processor
– Object Code Compatible With the
TMS34010
– Enhanced Instruction Set
– Optimized Graphics Instructions
– TMS34082 Graphics Floating-Point
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 6 Arithmetic
Pixel-Processing Options (Raster-Ops)
512-Byte LRU On-Chip Instruction Cache
Optimized DRAM/VRAM Interface
– Page-Mode for Burst Memory Operations
up to 40 Megabytes per Second
– Dynamic Bus Sizing
(16-Bit and 32-Bit Transfers)
– Byte-Oriented CAS Strobes
Flexible Host Processor Interface
– Supports Host Transfers at up to
20 Megabytes per Second
– Direct Access to All of the TMS34020
Address Space
– Implicit Addressing
– Prefetch for Enhanced Read Access
Flexible Multiprocessor Interface
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
145-PIN GB PACKAGE
( TOP VIEW )
A B C D E F G H J K L M N P R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
144-PIN PCM QUAD FLAT PACKAGE
( TOP VIEW )
108
109
72
144
37
1
36
Copyright  1993, 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
73
• HOUSTON, TEXAS 77251–1443
1
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
description
The TMS34020 and the TMS34020A graphics processors are 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 TMS34020 and TMS34020A provide 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.
The information contained in this data sheet is applicable to both the TMS34020 and the TMS34020A, except
that pertaining to clock stretch, which begins on page 21. Use of the term TMS34020 shall refer to both devices
except where noted.
architecture
The TMS34020 is a CMOS 32-bit processor with hardware support for graphics operations such as PixBlts
(raster ops) 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 TMS34020 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 TMS34020 directly interfaces to
dynamic RAMs and video RAMs and generates raster control signals. The TMS34020 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 TMS34020 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 TMS34020 provides single-cycle execution of general-purpose instructions and most common integer
arithmetic and Boolean operations from its instruction cache. Additionally, the TMS34020 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 TMS34020 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 pixels (PIXT instructions) or on two-dimensional arrays of
arbitrary size (PixBlts).
The TMS34020’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 TMS34020 provides for extensions to the basic architecture through the coprocessor interface. Special
instructions and cycle timings are included to enhance data flow to coprocessors, such as the TMS34082
floating-point unit, without requiring the coprocessor to decode the instruction stream, generate system
addresses, or move data for the coprocessor through the TMS34020.
2
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
pin assignments – GBL pin-grid-array package
PIN NO.
FUNCTION
PIN NO.
FUNCTION
PIN NO.
FUNCTION
PIN NO.
FUNCTION
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
LAD4
P3
HCS
TR/QE
C13
VSS
VSS
J13
A5
J14
P4
HA30
A6
RCA2
C14
J15
P5
HA27
A7
RCA3
C15
VCC
LAD26
VCC
LAD5
K1
EMU2
P6
HA24
A8
D1
RAS
K2
RESET
P7
HA22
A9
VCC
RCA6
D2
CAS2
K3
LINT2
P8
HA18
A10
RCA7
HA14
K14
VSS
LAD3
P9
RCA10
VSS
NC
K13
A11
D3
D4†
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
E2
L13
LAD1
P15
LAD0
B1
B2
WE
E3
VCC
CAS1
L14
LAD2
R1
HREAD
B3
E13
LAD27
L15
LAD19
R2
HA31
B4
VSS
CSYNC/HBLNK
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
HA20
RCA5
F3
M14
R7
HA19
B9
RCA9
F13
VSS
LAD24
VSS
LAD16
R6
B8
M15
LAD18
R8
HA17
B10
RCA11
F14
LAD8
N1
SIZE16
R9
HA16
B11
LAD31
F15
N2
HA15
LAD14
G1
VCC
CLKIN
R10
B12
VSS
HINT
R11
HA11
B13
G2
HOE
N4
HA9
G3
HDST
N5
VSS
HA29
R12
B14
VCC
LAD13
R13
HA8
B15
LAD12
G13
LAD7
N6
HA25
R14
HBS3
C1
CAS0
G14
N7
HA21
R15
VSS
C2
G15
N8
C3
VCC
DDOUT
VSS
LAD23
H1
LCLK1
N9
VSS
VSS
C4
DDIN
H2
EMU3
N10
HA12
C5
VSS
SF
H3
LCLK2
N11
HA6
C6
H13
LAD22
N12
HBS2
C7
RCA4
H14
LAD21
N13
HBS1
N3
C8
VSS
H15
LAD6
N14
† User must leave pin D4 unconnected. It is provided for device orientation purposes only.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
VCC
3
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
pin assignments – PCM quad flat package
PIN NO.
FUNCTION
PIN NO.
FUNCTION
PIN NO.
FUNCTION
PIN NO.
FUNCTION
1
37
74
VSS
VCC
109
38
VSS
HCS
73
2
VSS
VCC
110
VSS
VCC
3
CAS3
39
HA31
75
LAD0
111
LAD29
4
CAS2
40
HA30
76
LAD16
112
LAD14
5
CAS1
41
HA29
77
LAD1
113
LAD30
6
CAS0
42
HA28
78
LAD17
114
LAD15
7
VCC
RAS
43
HA27
79
LAD2
115
LAD31
44
HA26
80
LAD18
116
SCLK
VSS
R0
45
HA25
81
RCA12
46
HA24
82
VSS
LAD3
117
10
118
RCA11
11
R1
47
HA23
83
LAD19
119
RCA10
12
HOE
48
HA22
84
RCA9
HOST
49
HA21
85
VCC
LAD4
120
13
121
RCA8
14
HRDY
50
HA20
86
LAD20
122
RCA7
15
HINT
51
HA19
87
LAD5
123
RCA6
16
EMU3
52
HA18
88
LAD21
124
RCA5
17
LCLK1
53
HA17
89
LAD6
125
18
LCLK2
54
90
LAD22
126
VCC
VSS
19
EMU1
55
VSS
VSS
91
LAD7
127
RCA4
20
EMU0
56
HA16
92
LAD23
128
RCA3
21
EMU2
57
HA15
93
129
RCA2
22
GI
58
HA14
94
VSS
VSS
130
RCA1
23
RESET
59
HA13
95
LAD8
131
RCA0
24
LINT2
60
HA12
96
LAD24
132
SF
25
LINT1
61
HA11
97
LAD9
133
TR/QE
26
CAMD
62
HA10
98
LAD25
134
VSYNC
27
BUSFLT
63
HA9
99
LAD10
135
HSYNC
28
SIZE16
64
HA8
100
LAD26
136
CBLNK/VBLNK
29
PGMD
65
HA7
101
LAD11
137
CSYNC/HBLNK
30
LRDY
66
HA6
102
LAD27
138
31
67
HA5
103
VSS
VSS
68
HBS3
104
VCC
LAD12
139
32
VCC
VCC
140
ALTCH
33
VCLK
69
HBS2
105
LAD28
141
DDIN
34
CLKIN
70
HBS1
106
142
DDOUT
35
HWRITE
71
HBS0
107
VSS
LAD13
143
WE
36
HREAD
72
VSS
108
VSS
144
VSS
8
9
4
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Terminal Functions
LOCAL MEMORY INTERFACE
I/O
DESCRIPTION
ALTCH
NAME
O
Address latch. The high-to-low transitions of ALTCH can be used to capture the address and status present of the LAD
signals. A transparent latch (such as a 74ALS373) maintains the current address and status as long as ALTCH remains
low.
BUSFLT
I
Bus fault. External logic asserts BUSFLT high to the TMS34020 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 the LAD pins. In 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. This active-high output is used to drive the active-high output-enables on bidirectional
transceivers (such as the 74ALS623). The transceivers buffer data input and output on the LAD0 – LAD31 pins when
the TMS34020 is interfaced to several memories.
DDOUT
O
Data bus direction output-enable. This active-low signal drives the active-low output-enables on bidirectional
transceivers (such as the 74ALS623). The transceivers buffer data input and output on the LAD0 – LAD31 pins.
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 TMS34020 system. LAD0 is the LSB and LAD31 is the MSB.
LRDY
I
Local ready. External circuitry drives this signal low to inhibit the TMS34020 from completing a local-memory cycle it
has initiated. While LRDY remains low, the TMS34020 will wait unless the TMS34020 loses bus priority or is given an
external RETRY request (through the BUSFLT signal). Wait states are generated in increments of one full LCLK1 cycle.
LRDY can be driven low to extend local-memory read and write cycles, VRAM serial-data-register transfer cycles, and
DRAM refresh cycles. During internal cycles, the TMS34020 ignores LRDY.
PGMD
I
Page mode. The memory decode logic asserts this signal 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 this signal 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 emulation mode, SIZE16 is used to select the size of mapped memory.
CAMD
I
Column-address mode. This input 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
4 column-address strobes. The CAS outputs drive the CAS inputs of DRAMs and VRAMs. These signals 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. The RAS output drives the RAS inputs of DRAMs and VRAMs. This signal strobes the row
address on RCA0 – RCA12 to memory.
RCA0 – RCA12
O
13 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. This 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. This signal is also used to differentiate instructions and addresses
for the coprocessor as part of the coprocessor interface.
TR/QE
O
Transfer/output-enable. This signal 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.
WE
O
Write-enable. The active low WE output 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 TMS34020 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 is. During VRAM serial-data-register transfer cycles, the state of WE at the falling
edge of RAS controls the direction of the transfer.
DRAM AND VRAM CONTROL
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
5
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Terminal Functions (Continued)
NAME
I/O
DESCRIPTION
HA5 – HA31
I
27 host-address input signals. A host can access a long word by placing the address on these lines. HA5 – HA31
correspond to the LAD5 – LAD31 signals that output the address to the local memory.
HBS0 – HBS3
I
4 host byte selects. The byte selects identify which bytes within the long word are being selected.
HCS
I
Host chip select. A host drives this signal low to latch the current host address present on HA5 – HA31 and the host
byte selects on HBS0 – HBS3. This signal also enables host access cycles to the TMS34020 I/O registers or local
memory. During the low-to-high transition of RESET, the level on the HCS input determines whether the TMS34020
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 this signal latches data from the TMS34020 local address space to the
external host data latch on host read accesses. It can be used in conjunction with HRDY to indicate that data is valid
in the external data latch.
HINT
O
Host Interrupt. This signal allows the TMS34020 to interrupt a host by setting the INTOUT bit in the HSTCTLL I/O
register. This signal 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. This signal enables data from host data latches to the TMS34020 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. This signal is normally low and goes high to indicate that the TMS34020 is ready to complete a
host-initiated read or write cycle. If the TMS34020 is ready to accept the access request, HRDY is driven high and
the host proceeds 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. This signal is driven low during a read request from a host processor. This notifies the TMS34020
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. This signal is driven low to indicate a write request by a host processor. This notifies the
TMS34020 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.
CLKIN
I
Clock input. This system input clock generates the LCLK1 and LCLK2 outputs, to which all processor functions in
the TMS34020 are synchronous. A separate asynchronous input clock (VCLK) controls the video timing and video
registers.
LCLK1, LCLK2
O
Local output clocks. These two clocks 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 TMS34020 (except the CRT
timing signals) are synchronous to these clocks.
LINT1, LINT2
I
Local interrupt requests. Interrupts from external devices are transmitted to the TMS34020 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. The signal should remain
low until the TMS34020 recognizes it. These signals can be applied asynchronously to the TMS34020; they are
synchronized internally before use.
RESET
I
System reset. During normal operation, RESET is driven low to reset the TMS34020. When RESET is asserted
low, the TMS34020’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 TMS34020’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 TMS34020 begins executing
the instructions pointed to by the reset vector. If HCS is high, the TMS34020 is halted until a host processor writes
a 0 to the HLT bit in the HSTCTLL register.
HOST INTERFACE
SYSTEM CONTROL
6
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Terminal Functions (Continued)
NAME
I/O
DESCRIPTION
POWER
VCC†
VSS†
I
Nominal 5-volt power supply inputs. 9 pins.
I
Electrical ground inputs. 17 pins.
EMU0 – EMU2
I
Emulation pins 0 – 2.
EMU3
O
Emulation pin 3.
GI
I
Bus grant input. External bus artitration logic drives GI low to enable the TMS34020 to gain access to the
local-memory bus. The TMS34020 must release the bus if GI is high so that another device can access the bus.
R1, R0
O
Bus request and control. These two signals indicate a request for use of the bus in a multiprocessor system; they
are decoded as shown below:
EMULATION CONTROL
MULTIPROCESSOR INTERFACE
R1
0
0
1
1
R0
0
1
0
1
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. You can program this signal 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, this signal is configured as a CBLNK output.
CSYNC / HBLNK
I/O
Composite sync/horizontal blanking. You can program this signal 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:
Input: Extracts HSYNC and VSYNC from externally generated horizontal sync pulses.
Output: Generates active-low composite-sync pulses from either externally generated HSYNC and VSYNC
signals or signals generated by the TMS34020’s on-chip video timers.
Horizontal blank (output only) for blanking the display during horizontal retrace in separate-sync-video mode.
Immediately following reset, this signal is configured as a CSYNC input.
† For proper TMS34020 operation, all VCC and VSS pins must be connected externally.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
7
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Terminal Functions (Continued)
VIDEO INTERFACE
NAME
HSYNC
I/O
DESCRIPTION
I/O
Horizontal sync. HSYNC is the horizontal sync signal that controls external video circuitry. You can program this
signal 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 TMS34020’s on-chip video
timers.
As an input, HSYNC synchronizes the TMS34020 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.
SCLK
I
Serial data clock. This signal is the same as the signal that drives VRAM serial data registers. This allows the
TMS34020 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. This clock is derived from a multiple of the video system’s dotclock 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. You can program this signal
to be either an input or an output by modifying a control bit in the DPYCTL register.
VSYNC
As an output, VSYNC is the active-low vertical-sync signal generated by the TMS34020’s on-chip video timers.
As an input, VSYNC synchronizes the TMS34020 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.
8
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
functional block diagram
HA5 – HA31
HBS0 – HBS3
27
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
Host
Interface
MultiProcessor
Interface
ST
4
DRAM/
VRAM
Interface
Decode
SP
Local
Memory
and
Bus
Timing
Bus
Interface
ALU
Barrel
Shifter
Microcontrol ROM
Reset and Interrupts
RESET, LINT1,
LINT2
Video
Timing
and
Control
LAD0 – LAD31
RCA0 – RCA12
DDIN
DDOUT
RAS
Bus
Control
LRU
Emulation
Interface
System
Clocks
13
Cache
Register
File A
Register
File B
32
CAS0 – CAS3
WE
TR/QE
ALTCH
SF
PGMD
SIZE16
LRDY
BUSFLT
CAMD
VSYNC
HSYNC
CSYNC/HBLNK
CBLNK/VBLNK
VCLK
SCLK
3
architecture (continued)
register files
Boolean, arithmetic, pixel-processing, byte, and field-move instructions operate on data within the
general-purpose register files. The TMS34020 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 via 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 TMS34020’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
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
9
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 associated with it a present
(P) flag to indicate whether the subsegment contains valid data.
The cache is loaded only when an instruction requested by the execution section of the TMS34020 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.
In addition, during an interrupt, the IX bit in the ST placed on the stack provides indication that execution of an
interruptible 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 TMS34020 outputs a 28-bit address on LAD4 – LAD31, which 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
TMS34020 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 TMS34020 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.
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 TMS34020 User’s Guide, literature number SPVU019.
transparency
Transparency is a mechanism that allows 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 TMS34020
provides four transparency modes:
•
•
•
•
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.
10
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
I/O registers
The TMS34020 contains an on-chip block of sixty-four 16-bit locations (mapped into the TMS34020’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:
•
•
•
•
•
•
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 TMS34020 and a host processor and maintain compatibility with the
TMS34010. The registers are mapped into five of the I/O locations accessible to the TMS34020.
Two of these registers (HSTCTLL and HSTCTLH) are used to provide control by the host. This control consists
of passing of interrupt requests, flushing the instruction cache, halting the TMS34020, transmitting a
nonmaskable interrupt request to the TMS34020, enabling emulation interrupts, and setting host-access modes
and configurations.
The other three registers are simple read/write registers to allow the TMS34020 software to leave addresses
for the host at a known location and allow compatibility with some TMS34010 software.
memory interface control registers
Some of the I/O registers are used to control various local memory interface functions, including:
•
•
•
•
•
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 TMS34020 can be
configured to drive composite sync or separate sync displays.
In composite mode, the TMS34020 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 TMS34020 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 TMS34020 directly supports multiport video RAMs (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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
11
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
The TMS34020 supports various display resolutions and either interlaced or noninterlaced video. The
TMS34020 can optionally be programmed to synchronize to externally generated sync signals so that images
created by the TMS34020 may be superimposed upon images created externally. The external sync mode can
also be used to synchronize the video signals generated by two or more TMS34020s in a multiple-TMS34020
graphics system.
CPU control registers
Five of the I/O registers (CONVDP, CONVMP, CONVSP, CONTROL, and PSIZE) provide CPU control to
configure the TMS34020 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 TMS34020,
including two externally generated interrupts and three internally generated interrupts. An internal interrupt
request can be generated on one of the following conditions.
•
•
•
•
•
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 TMS34020 to execute instructions (i.e., before the stack pointer
is initialized). The TMS34020 reset function is controlled by a dedicated pin.
memory controller/local memory interface
The memory controller manages the TMS34020’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 the 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 TMS34020 forced to defer execution of the subsequent instruction.
The TMS34020 directly interfaces to standard dynamic RAMs and, in particular, to standard video RAMs such
as the TMS44C251 multiport VRAMs. The TMS34020 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 the LAD bus. The RCA bus allows direct
connection to address/address multiplexed DRAMs from 64K to 16M. Refresh for DRAMs is supported by
CAS-before-RAS refresh cycles.
12
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
BIT 232 –1
(Last Bit in Memory)
ADDRESS
FFFFFFF0h
Interrupt Vectors and
Extended Trap Vectors
68 Words
FFFFFBC0h
FFFFFBB0h
Reserved for Interrupt Vectors
and Extended Trap Vectors
444 Words
FFFFE000h
FFFFDFF0h
65024 Words
FFF0000h
FFEFFFF0h
226 – 66560 Words
(67 042 304 Words)
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
reset
Reset puts the TMS34020 into a known initial state. This state is entered when the input signal at the RESET
pin 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 the RESET signal determines
whether the TMS34020 will be halted or begin executing instructions. The TMS34020 may be in one of two
modes, host-present or self-bootstrap mode.
1. Host-Present Mode. If HCS is high at the end of reset, TMS34020 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 may request a memory access after the eight RAS initialization cycles have completed. The
TMS34020 automatically performs DRAM refresh cycles at regular intervals; although the TMS34020
remains halted until the host clears the HLT bit. Only then does TMS34020 fetch the level-0 vector address
from location FFFF FFE0h and begin executing the reset service routine.
2. Self-Bootstrap Mode. If HCS is low at the end of reset, the TMS34020 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
13
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
At the time the TMS34020 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), since the value of the stack
pointer just before a reset is generally not valid, and 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 the RESET input pin to its active-low level. To reset the TMS34020 at power up,
RESET must remain low for a minimum of 40 local clock periods (LCLK1 and LCLK2) after power levels have
become stable. At times other than power up, the TMS34020 may be reset by holding RESET low for a minimum
of four local clock periods; the GSP will enter 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 TMS34020s 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 may be done by using CLKIN to clock RESET as received by the TMS34020s. All TMS34020s 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 power up, 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 TMS34020 to a particular quarter cycle by adding Q1 cycles. All TMS34020s in a
multiprocessor environment operate on the same quarter cycle within 10 quarter cycles after the rising edge
of RESET.
reset and DRAM/VRAM initialization
The TMS34020 drives its RAS signal inactive (high) as long as RESET remains low. The specifications for
certain DRAM and VRAM devices require that the RAS signal be driven inactive 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–(10 tQ ) microseconds. The TMS34020 memory controller automatically inserts the required
eight RAS cycles after all resets (after power up 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 TMS34020 is set to
perform a refresh sequence every eight cycles.
At times other than power up, 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 TMS34020 to refresh the memories.
On reset, the TMS34020 is set to do a refresh cycle every eight local clock periods. A 32-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 the RESET signal remains low longer than the maximum refresh interval specified for the memory, the
previous contents of the local memory may not be valid after the reset.
14
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
initial state following reset
While the RESET pin is asserted low (or while in the internal reset state), the TMS34020’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
BIDIRECTIONAL 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
NOTE: If GI is high, then all GI-controlled pins will be in the high-impedance state. The 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 the HCS pin is high just prior to the low-to-high transition of RESET;
otherwise, it is set to 0.
Just prior to the execution of the first instruction in the reset routine, the TMS34020’s internal registers are in
the following states:
•
•
•
•
•
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:
•
•
•
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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
15
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
local memory and DRAM/VRAM interface
The TMS34020 local memory interface consists of an address/data multiplexed bus on which address 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 TMS34020 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 the LAD bus, 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.
The LAD bus is then used to transfer data to or from the memory while the RCA bus is set to the column address
for the memory. 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
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 register CONFIG) and the state of the column address mode pin (CAMD) during each
memory cycle. The CAMD signal 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
RCM1
RCM0
VRAM MODE
0
0
64K × N
ADDRS
8
BANKS
16
CAMD SUPPORT MATRICES
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 may be supported within the basic VRAM mode.
16
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
The following table 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
RCA BIT
64K
256K
1M
4M
CAMD = 0
12
24
25
26
27
11
23
24
25
26
10
22
23
24
9
21
22
23
8
20
21
7
19
6
18
5
4
CAMD = 1
64K
256K
1M
4M
16
23
26
15
28
15
22
14
14
14
25
14
13
13
13
13
24
13
12
12
12
12
22
23
12
11
11
11
11
20
21
22
11
10
10
10
10
19
20
21
10
9
9
9
9
17
18
19
20
9
8
8
8
8
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 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 the
RCA signals directly to 256K or 1M DRAMs.
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.
CONNECTIONS TO RCA FOR CAMD = 1
RCA
64K
256K
1M
4M
12
1M × 32
4M × 32
4M × 32
16M × 32
11
1M × 16
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
4M × 16
16M × 32
0
NOTE: NN is used for either 16-bit (× 16) or 32-bit (× 32) memory connections.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
17
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
status codes
Status codes are output on LAD0 – LAD3 at the time of the falling edge of ALTCH and may be used to determine
the type of cycle that is being initiated. The following table lists the codes and their respective meanings.
CODE
STATUS
0000
Coprocessor Code
0001
Emulator Operation
0010
Host Cycle
TYPE
Other
(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
dynamic bus sizing
The TMS34020 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 TMS34020 then
performs another memory access to the next 16-bit (odd) word in memory. The TMS34020 samples the SIZE16
pin at the start of Q2 in the second cycle (access to odd word address) to determine to which half of the LAD
bus the port or memory is aligned. If the port is on LAD0 – LAD15, the SIZE16 input should be low during the
second cycle access (odd word); otherwise, if the port is on LAD16 – LAD31, the SIZE16 input must be high at
this time. The TMS34020 always performs two memory cycles to access the 16-bit wide memories, even when
attempting only a 16-bit transfer.
The TMS34020 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 the LAD bus 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 the LAD bus 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 the SIZE16 pin is performed as a page-mode access if PGMD was low
during the first access. A read-write cycle to 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 the LAD bus the memory is connected (low for connection to LAD0 – LAD15 or high
for connection to LAD16 – LAD31).
18
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
special 1M VRAM cycles
The TMS34020 provides control for special-function VRAM cycles that are available in the 1M devices. These
cycles are obtained by the appropriate timing control of the SF, CAS, TR/QE, and WE pins of the VRAMs at the
falling edge of RAS. The cycles include:
•
•
•
•
•
•
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 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 on TMS34020 to another. Synchronization of multiple TMS34020s in a system
occurs at reset with the rising edge of RESET meeting the setup and hold requirements to CLKIN, so all
TMS34020s are certain to respond to the RESET during the same quarter cycle. RESET is not required to be
synchronous to CLKIN, except to allow synchronization of multiple TMS34020s in a system.
The GI priority for multiprocessing environments is determined by arbitration logic external to the TMS34020.
If GI goes inactive (high), the TMS34020 releases the bus on the next available cycle boundary. If the cycle in
progress has not successfully completed, the TMS34020 restarts the cycle upon regaining control of the bus.
Normally, if the TMS34020 asserts both R0 and R1 low, it should be given the control of the bus by the arbitrator.
host interface
The TMS34020 host interface allows the local memory to be mapped into the host address space. The
TMS34020 acts as a DRAM controller for the host. The address for the host access is latched within the
TMS34020; however, the data for the access is transferred via 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 TMS34020 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 TMS34020. The host indicates
a read or write by asserting HREAD or HWRITE (as appropriate) either before or after HCS. HREAD and
HWRITE must never be asserted at the same time.
The TMS34020 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 the
HWRITE signal. 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 TMS34020 that the write access can start. The TMS34020 uses its byte-write capability to write
only to the selected bytes.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
19
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
The TMS34020 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 HINC (host increment enable) and
HPFW (host prefetch after write enable) 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
0
0
RANDOM/SAME
HOST ACCESS MODE
No increment, no prefech
DESCRIPTION
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 TMS34020 is programmed for block-mode or read/modify/write accesses, the host can still do random
accesses because the TMS34020 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 TMS34020 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 reaccessed, 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 TMS34020 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 TMS34020 uses the host byte selects (HBS0 – HBS3) to enable
the CAS strobes when doing a host write.
coprocessor interface
Support for coprocessors is provided through special instructions and bus cycles that allow communication with
the coprocessor. A coprocessor may be register-based, depending on the TMS34020 to do all address
calculations, or it may operate as its own bus controller, using the multiprocessor arbitration scheme. Five basic
cycles are provided for direct communication and control of coprocessors.
1.
2.
3.
4.
5.
TMS34020 to coprocessor.
Coprocessor to TMS34020.
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.
20
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
instruction set
The TMS34020 instruction set can be divided into five categories:
1.
2.
3.
4.
5.
Graphics instructions.
Coprocessor instructions.
Move instructions.
General-purpose instructions.
Program control and context switching.
Specialized graphics instructions manipulate pixel data that is accessed via memory addresses or XY
coordinates. These instructions include graphics operations, such as array and raster ops, 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 TMS34020A supports a clock stretching mechanism, which is described in outline below.
With advances in semiconductor manufacturing, new versions of the TMS34020A can be made, each
supporting a higher CLKIN frequency. The increase in CLKIN frequency means that the TMS34020A 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 TMS34020A 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 will 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 (and hence cheaper
and more available) memory devices than the designer could use if no stretch mechanism were 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, and Q4a, and Q4b.
When clock-stretch mode is enabled, the fourth machine quarter cycle may be stretched to twice its original
length. This stretching takes place only when the TMS34020A attempts certain types of memory cycle.
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 TMS34020A clocks in the Q4 state for an extra quarter cycle so
all the device outputs remain unchanged during Q4a and Q4b. The TMS34020A stretches only certain machine
cycles so that the execution of code is not slowed unnecessarily.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
21
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
enabling clock stretch
Clock-stretch mode is enabled and disabled via a bit in the CONFIG register (C00001A0h).
31
7
6
5
4
3
2
1
0
C
S
E
Loaded at Reset from Reset Vector
Protected Byte
CONFIG register C00001A0h
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 0 in this bit will disable stretch mode and
a 1 in this bit will enable stretch mode. The bit is cleared during reset; i.e., stretch mode is disabled by default.
When stretch mode is enabled, the following machine cycles are stretched:
1. All address cycles of all memory access sequences.
2. Read data cycles in read-modify-write sequences.
Notes:
a) The host default cycle shown on page 8-49 of the TMS34020 User’s Guide is not stretched because
it is not a true address cycle; i.e., 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 TMS34020A is running with the CSE bit
set to 1.
The advantage of this implementation of clock-stretch mode is that the TMS34020A 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 TMS34020A
(with the exception of the video logic, which is clocked independently by VCLK) will effectively slow down during
such a stretched cycle.
22
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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
3
READ
READ
4
1
2
3
Stretch Mode Disabled
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
3
READ
4
1
2
READ
3
Stretch Mode Disabled
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 may 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
3
Stretch Mode Disabled
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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
23
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
a clock-stretch timing example, TMS34020A-40 and 100-ns VRAMs
This example analyzes a memory interface timing parameter. It shows that the clock stretch mechanism can
be used to allow the TMS34020A-40 to avoid a timing violation when interfaced to 100-ns VRAMs.
Consider a system with:
1. TMS34020A-40,
2. TMS44C251-10
which has a 40-MHz clock input frequency and hence a 100-ns cycle time, so
tQ = 25 ns. Timing parameters are taken from the TMS34020 data sheet.
256K × 4 bit VRAM. Timing parameters are taken from the appropriate section in
the Texas Instruments MOS Memory Data Book, 1989.
row address hold data after RAS low, th(ADV-REL)
Without clock stretch
TMS44C251 th(RA)
Row address hold time after RAS low.
Min = 15 ns
TMS34020A
Hold time, row address valid after RAS low.
Min = tQ – 5 ns = 20 ns.
Parameter 88
If RAS is passed through a PAL with a delay of 7 ns then th(RA) seen by the VRAM is 20 ns – 7 ns = 13 ns.
This violates the 15 ns minimum.
With clock stretch
TMS34020A
Parameter 88
Hold time, row address valid after RAS low.
Min = 2tQ – 5 ns = 45 ns.
With the same 7-ns PAL delay, the VRAM sees th(RA) as 45 ns – 7ns = 38 ns, which does not violate the VRAM
minimum of 15 ns.
cycle timing examples
The following figures show examples of many of the basic cycles that the TMS34020 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. Q4 phases
that could be stretched are marked with a * on the diagrams. The conditions required for the stretch are:
1. The design uses a TMS34020A
2. The CONFIG register’s CSE bit is set to 1
3. The TMS34020A 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 may require external buffering of the address and data lines. The DDIN and
DDOUT signals coordinate these external buffers with the LAD bus.
PAL is a trademark of Advanced Micro Devices, Inc.
24
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Page-Mode Read
Standard Memory Read Cycle
Address Subcycle
Q4
Q1
Q2
Q3
Q4*
Data Transfer
Subcycle
Data Transfer
Subcycle
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD (TMS34020)
Address
Data
LAD (Memory)
Data
CAMD
RCA
Row
1st Column
2nd Column
ALTCH
RAS
CAS
WE
TR/QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
*See clock stretch, page 21.
NOTES: A. LAD (TMS34020): Output to the LAD bus by the TMS34020.
LAD (memory): Output to the LAD bus 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)
During the address output to the LAD bus by the TMS34020, the least significant four bits (LAD0 – LAD2) 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
25
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Address Subcycle
Q4
Q1
Q2
Q3
Q4*
Wait State
Q1
Q2
Q3
Read Transfer
Q4
Q1
Q2
Q3
Q4
Q1
LCLCK1
LCLCK2
GI
LAD (TMS34020)
Address
Data
LAD (Memory)
CAMD
RCA
Row
Column
ALTCH
RAS
CAS
WE
TR/QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
*See clock stretch, page 21.
NOTES: A. LAD (TMS34020): Output to the LAD bus by the TMS34020. LAD (memory): Output to the LAD bus 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 the LRDY
is sampled high.
Figure 3. Local Memory Read Cycle Timing (Without Page Mode, With One Wait State)
LRDY low at the start of the first Q2 after RAS low 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.
26
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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
PGMD
SIZE16
BUSFLT
R0
R1
*See clock stretch, page 21.
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)
During the address output to the LAD bus by the TMS34020, 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
27
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 (TMS34020)
Data Out
Address
Data
LAD (Memory)
CAMD
RCA
Row
Column
ALTCH
RAS
CAS
WE
TR/QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
*See clock stretch, page 21.
Figure 5. Local Memory Read/Modify/Write Cycle Timing
The read/modify/write cycle 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.
28
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Refresh Status
Q4
Q1
Q2
Q3
Q4*
CAS-Before-RAS
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 21.
Figure 6. Refresh Cycle Timing
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. The 16 bits of address are placed
on LAD16 – LAD31; all other LAD bus lines will be 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
29
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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
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 21.
NOTE A: RCA0 may 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 4M-mode with CAMD = 1).
Figure 7. Dynamic Bus Sizing, Read Cycle
When SIZE16 is selected low, the TMS34020 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 TMS34020 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
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 21.
Figure 8. Dynamic Bus Sizing, Write Cycle
Write accesses to 16-bit memory are performed by swapping the data on upper and lower words of the LAD
bus and exchanging data on CAS0 and CAS1 for data on CAS2 and CAS3, respectively. This is illustrated in
the above example where the TMS34020 is writing the upper 24 bits (LAD8 – LAD31) of data to memory. 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 TMS34020 swaps data on the upper and lower halves
of the LAD bus. Notice that during the first cycle CAS0 is inactive (since 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
31
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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
Q1
GI
LAD
PMASK Address
PMASK Data
Zero Address
Not PMASK Data
CAMD
RCA
PMASK Row
PMASK Column
All-Zero Address
ALTCH
RAS
CAS
WE
TR/QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
*See clock stretch, page 21.
Figure 9. Load Write Mask Cycle
This special 1M VRAM control cycle is executed when the VEN bit in the CONFIG I/O register is set and PMASKL
and/or PMASKH are written. 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 the LAD bus 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.
32
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Q4
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4
Q1
GI
LAD
Zero Address
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 21.
Figure 10. Load-Color-Latch Cycle
This special 1M VRAM control cycle 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. 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
33
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 21.
Figure 11. Block-Write Cycle Without Mask
This special 1M VRAM control cycle is performed when a VBLT or VFILL instruction is executed and PMASKL
and PMASKH are set to zero. 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.
34
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 21.
Figure 12. Block-Write Cycle With Mask
This special 1M VRAM control cycle is performed when a VBLT or VFILL instruction is executed and PMASKL
and PMASKH are set to nonzero. 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
35
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 21.
Figure 13. Write Cycle Using Mask
This special 1M VRAM control cycle is performed when the PMASKL and PMASKH registers are set to nonzero,
the CST bit in DPYCTL is cleared, the VEN bit in CONFIG is set, and the byte-aligned pixel-write instruction is
executed. 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 the DQ bits 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).
36
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Address
Subcycle
Q4
Q1
Q2
Q3
Wait
State
Q4*
Q1
Q2
Q3
Cycle
Completion
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 21.
Figure 14. Memory-to-Serial-Data-Register Cycle (VRAM Read Transfer)
This VRAM cycle is issued in any of three ways:
1. Pixel operation instruction with the CST bit in DPYCTL set.
2. Horizontal blank reload cycle requested by the video-control logic with the VCE bit in DPYCTL cleared.
3. Video timeout due to SCOUNT match with the value in MLRNXT and the VCE and SSV bits 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
37
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 21.
Figure 15. Memory-to-Split-Serial-Data-Register Cycle (VRAM Split-Register Read Transfer)
This VRAM cycle is performed when a video timeout due to a match of the MLRNXT register occurs, the VCE
bit in DPYCTL is cleared, and the SSV bit 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.
38
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 21.
Figure 16. Serial-Data-Register-to-Memory Cycle (VRAM Write Transfer, Pseudo-Write Transfer)
This VRAM cycle is performed when a horizontal blank reload is requested by the video-control logic and the
VCE bit and the SRE bit 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 TMS34020). 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
39
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 21.
Figure 17. Serial-Data-Register-to-Memory Cycle (VRAM Alternate-Write Transfer)
This VRAM cycle 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 the SOE pin 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.
40
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 Timing (High-Impedance Signals)
Transition points are shown for R0 and R1 to indicate where they occur relative to the other signals.
This example indicates that the TMS34020 has control of the bus, yields control, and then regains control. The
TMS34020 regains bus mastership as soon as its GI pin 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 TMS34020A clock stretch not be used in multiprocessor systems.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
41
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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 Timing (Passing Control)
Two TMS34020s use the multiprocessor interface to pass control of local memory from one to the other. 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. If you want more information, see
the TMS34020 User’s Guide.
42
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
HA/HBS
HCS
HREAD
HWRITE
HRDY
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 21.
NOTE A: HRDY goes high at the start of Q2; however, data is not strobed into the external latches until the start of Q4 by HDST going high.
Figure 20. Host Read Cycle (Random/Same Accesses, not From TMS34020 I/O Registers)
The host-access request is synchronized to the TMS34020 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 may also remove and
reassert HREAD with HCS).
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
43
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
HA/HBS
HCS
HREAD
HWRITE
HRDY
DATA
(out)
Previous Read
1st Read Valid
Local-Memory Host Read Cycle
Q4
Q1
Q2
Q3
Q4*
Q1
Q2
2nd
Local-Memory Host Prefetch Cycle
Q3
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 21.
Figure 21. Back-to-Back Host Read Cycles With Implicit Addressing; HREAD as Strobe
The host-access request is synchronized to the TMS34020 at the beginning of Q4 so that the local memory cycle
may begin in Q1.
In block mode (prefetch after read), the TMS34020 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, indicating the end of the first read, the TMS34020
starts to prefetch the next sequential location. When the host makes the next request, the TMS34020 has
prefetched the data so that the host reads with no delay. While in block mode, the TMS34020 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, the HRDY signal is asserted high
so that the host can read with no delay.
In read/modify/write mode (prefetch after write), the TMS34020 initiates the read access as soon as the current
write request is deasserted.
44
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
HA/HBS
HCS
HREAD
HWRITE
HRDY
DATA
(out)
Prevoius 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 21.
Figure 22. Host Read Cycle From TMS34020 I/O Registers
The host read of the TMS34020 I/O registers suppresses the generation of TR/QE and CAS so that data is read
from the TMS34020 rather than from memory. DDOUT is enabled so that data CAN flow through external buffers
on the LAD bus to the host data latches. The TMS34020 I/O registers may be accessed in any of the host access
modes (random/same, block, or read/modify/write).
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
45
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
HA/HBS
HCS
HREAD
HWRITE
HRDY
DATA
(in)
1 st 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 21.
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
The TMS34020 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 TMS34020 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 TMS34020 then
writes the second data while the host continues other processing. The host access request is synchronized to
the TMS34020 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 TMS34020 I/O registers, the data can be
required within the GSP.
46
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
CLKIN
RESET
Case 1
LCLK1
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
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 TMS34020s
Although RESET is not normally required to be synchronous to CLKIN, in order to facilitate synchronization of
multiple TMS34020s 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. The four possible
conditions for the state of the TMS34020 at the time RESET goes high are shown above. Quarter cycle 1 is
extended accordingly to provide synchronization of the GSPs. All TMS34020s to be synchronized must share
a common CLKIN and RESET. Within 10 CLKIN cycles after RESET goes high, all GSPs will be synchronized
to the same quarter cycle through the extension of Q1 cycles.
It is recommended that TMS34020A stretch mode not be used in multiprocessor systems.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
47
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Command
Q4
Q1
Q2
Q3
Data Transfer
Q4
Q1
Q2
Q3
Data Transfer
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD (TMS34020)
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 (TMS34020): Output to the LAD bus by the TMS34020
Command:
Coprocessor ID, instruction and status code present on LAD
Operand n:
Data to or from the coprocessor
Figure 25. Transfer TMS34020 Register(s) to Coprocessor (One or Two 32-Bit Values)
This timing example is like a memory write cycle, except that RAS and SF are high.
48
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Command
Q4
Q1
Q2
Q3
Data Transfer
Q4
Q1
Q2
Q3
Data Transfer
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD (TMS34020)
Command
LAD (Coprocessor)
Operand 1
Operand 2
CAMD
RCA
ALTCH
RAS
CAS
WE
TR/QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
NOTE A: LAD (TMS34020):
LAD (coprocessor):
Command:
Operand n:
Output to the LAD bus by the TMS34020
Output to the LAD bus by the coprocessor
Coprocessor ID, instruction and status code present on LAD
Data to or from the coprocessor
Figure 26. Transfer Coprocessor Register to TMS34020 (One or Two 32-Bit Values)
This timing example is like a memory write cycle, except that RAS and SF are high.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
49
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
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
(TMS34020)
Command
Address
LAD
(memory)
Data 1
Data 2
1st Column
2nd Column
CAMD
Row
RCA
ALTCH
RAS
CAS
WE
TR/QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
*See clock stretch, page 21.
NOTES: A. LAD (TMS34020):
LAD (memory):
Command:
Address:
Data n:
Output to the LAD bus by the TMS34020
Output to the LAD bus 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)
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. The coprocessor can place status
information on the LAD bus during the initialization cycle for the TMS34020. 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 TMS34020 to be used for the count. Both instructions
specify a register to be used as an index into memory. The index may be post-incremented or pre-decremented
on each transfer cycle.
50
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Address
Command Cycle
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Data Transfer
Data Tansfer
Spacer
Q4
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
LCLK1
LCLK2
GI
LAD
(TMS34020)
LAD
Coprocessor)
Address
Command
Status
Data 1
Data 2
CAMD
Row
RCA
1st Column
2nd Column
ALTCH
RAS
CAS
WE
TR/QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
*See clock stretch, page 21.
NOTES: A. All coprocessor cycles are implemented as 32-bit operations; therefore, SIZE16 should be high during these cycles.
B. LAD (TMS34020):
Output to the LAD bus by the TMS34020
LAD (coprocessor): Output to the LAD bus 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
Figure 28. Transfer Coprocessor Register(s) to Memory (ALTCH High During Data Transfer)
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. The coprocessor can place status
information on the LAD bus during the initialization cycle for the TMS34020. The memory cycle includes a dead
cycle to enable the TMS34020 to take the LAD bus 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 post-increment and the other uses it with a
pre-decrement after each transfer cycle.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
51
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
Q4
Q1
Command Cycle
Q3
Q4
Q1 Q2
Q2
Q3
Q4
Q1
LCLK1
LCLK2
GI
LAD
(TMS34020)
Command
CAMD
RCA
ALTCH
RAS
CAS
WE
TR/QE
SF
DDIN
DDOUT
LRDY
PGMD
SIZE16
BUSFLT
R0
R1
NOTES: A. All coprocessor cycles are implemented as 32-bit operations; therefore, SIZE16 should be high during these cycles.
B. LAD (TMS34020):
Output to the LAD bus by the TMS34020
command:
Coprocessor ID, instruction and status code present on LAD
C. 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.
Figure 29. Coprocessor Internal Operation Command Cycle
This timing example 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 the LAD bus during the cycle
for the TMS34020.
52
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
absolute maximum ratings over operating free-air temperature range (see Note 1)†
Maximum supply voltage, VCC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 V
Input voltage range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 0.3 V to 7 V
Off-state output voltage range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 2 V to 7 V
Operating free-air temperature range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0°C to 70°C
Storage temperature range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – 10°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: Voltage values are with respect to VSS.
recommended operating conditions
VCC
VSS
Supply voltage
Supply voltage‡
VIH
HIgh level input voltage
HIgh-level
VIL
IOH
Low-level input voltage
IOL
TA
Low-level output current
MIN
NOM
4.75
5
5.25
V
0
0
0
V
VCC + 0.3
VCC + 0.3
V
All input pins except CLKIN
2
CLKIN only
3
MAX
– 0.3
High-level output current
UNIT
0.8
V
400
µA
2
mA
70
°C
TC
Operating case temperature
TMS34020APCM40
0
85
‡ Take care to provide a minimum inductance path between the VSS pins and system ground in order to minimize noise on VSS.
°C
Operating free-air temperature
TMS34020AGBL40
0
electrical characteristics over recommended ranges of supply voltage and operating free-air
temperature
PARAMETER
VOH
VOL
TEST CONDITIONS
High-level output voltage
Low-level output voltage
IO
impedance)
Output current
current, leakage (high
(high-impedance)
II
ICC
Input current, leakage (All input pins except EMU0 – EMU2¶)
Ci
Input capacitance
Supply current
VCC = MIN,
VCC = MAX,
IOH = MAX
IOL = MIN
VCC = MAX,
VCC = MAX,
VO = 2.8 V
VO = 0.6 V
MIN
TYP§
MAX
2.6
UNIT
V
0.6
20
– 20
VI = VSS to VCC
VCC = MAX
10
V
µA
± 20
µA
260
mA
pF
Co
Output capacitance
10
pF
§ All typical values are at VCC = 5 V, TA = 25° C.
¶ EMU0 – EMU2 will not be connected in a typical configuration. Nominal pullup current will be 500 µA.
NOTE 2: HDST and HOE (output pins) have internal pullup resistors that allow high logic levels to be maintained when the TMS34020 is not
actually driving these pins.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
53
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
signal transition levels
2V
0.8 V
Figure 30. TTL-Level Inputs
For a high-to-low transition on a TTL-compatible input signal, the level at which the input is said to be no longer
high is 2 V, and the level at which the input is said to be low is 0.8 V. For a low-to-high transition, the level at which
the input is said to be no longer low is 0.8 V, and the level at which the input is said to be high is
2 V.
2.6 V
2V
0.8 V
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 0.8 V. For a low-to-high transition, the level at
which the output is said to be no longer low is 0.8 V, and the level at which the output is said to be high is 2 V.
Timing parameters 12a and 13a do not follow these conventions. They are measured at a 1.5-V level.
test measurement
The test load circuit shown in Figure 32 represents the programmable load of the tester pin electronics, which
are used to verify timing parameters of TMS34020 output signals.
IOL
From Output
Under Test
Test
Point
VLOAD
CLOAD
IOH
Where:
IOL = 2 mA (all outputs)
IOH = 400 µA (all outputs)
VLOAD = 1.5 V
CLOAD = 65 pF typical load circuit distributed capacitance
Figure 32. Test Load Circuit
54
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
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
AD
AL
BC
CE
CK
CK1
CK2
CKI
CM
CS
CT
DI
DO
EM
HI
HS
HA5– HA31 and HBS0 – HBS3
LAD0– LAD31 and RCA0 – RCA12
ALTCH
Any of the bus control input signals
(LRDY, PGMD, SIZE16, or BUSFLT)
CAS0– CAS3
LCLK1 and LCLK2
LCLK1
LCLK2
CLKIN
CAMD
HCS
Any of the bus control output signals
(ALTCH, CAS0 – CAS3, RAS, WE,
TR/QE, HOE, or HDST)
DDIN
DDOUT
EMU0, EMU1, EMU2
HINT
HSYNC, VSYNC, CSYNC/HBLNK, or
CBLNK/ VBLNK
GI
LA
LINT
OE
GI
LAD0– LAD31
LINT1, LINT2
HOE
RC
RD
RE
RQ
RS
RY
RCA0– RCA12
HREAD
RAS
R0 or R1
RESET
HRDY
S
SC
SCK
SF
SG
ST
TR
VCK
WR
HSYNC, VSYNC, or CSYNC
EMU3
SCLK
SF
Any output signal
HDST
TR/QE
VCLK
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
NV
V
Z
↑
↓
D
High
Low
Not valid
Valid
High impedance
No longer low
No longer high
Driven
s=0
For (i) TMS34020
(ii) TMS34020A, if CSE bit in CONFIG register is
set to 0
(iii) TMS34020A, if CSE bit in CONFIG register is set to
1 and the cycle is not stretched. See page 21.
s = tQ For TMS34020A, if CSE bit in CONFIG register is set to 1
and the cycle is stretched. See page 21.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
55
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
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 TMS34020 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.
56
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
CLKIN and RESET timing requirements
NO.
’34020-32
’34020A-32
PARAMETER
MIN
1
2
3
4
5
6
MAX
31.25
MIN
Period of CLKIN (tQ)
Pulse duration, CLKIN high
10
8
tw(CKIL)
tt(CKI)
Pulse duration, CLKIN low
10
2†
8
2†
th(CKI-RSL)
tsu(RSH-CKI)
Hold time, RESET low after CLKIN high
Transition time, CLKIN
tw(RSL)
(RSL)
8
tsu(CSL-RSH)
Setup time of HCS low to RESET high to configure
self-bootstrap mode
9
td(CS-RSH)
Delay from HCS↑ to RESET high to configure selfbootstrap mode
10
tw(CSL)
Pulse duration, HCS low to configure GSP in selfbootstrap mode
25
5†
10‡
4‡
10‡
4‡
160tQ – 40§
160tQ – 40§
16tQ – 40§
16tQ – 40§
8tQ + 55
8tQ + 55
Setup time, RESET high to CLKIN↑
Initial reset during power up
50
Reset during active operation
UNIT
MAX
tc(CKI)
tw(CKIH)
Pulse
duration
duration,
RESET low
7
’34020A-40
4tQ – 50¶
50
ns
ns
5†
4tQ + 55
ns
ns
ns
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 TMS34020 to a particular quarter cycle.
§ The initial reset pulse on power up must remain valid until all internal states have been initialized. Resets applied after the TMS34020 has been
initialized need to be present only long enough to be recognized by the internal logic; the internal logic will maintain 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 TMS34020 is configured to run in the self-bootstrap mode (HLT bit = 0) following the end of reset.
1
4
3
4
2
CLKIN
6
5
RESET
7
8
9
10
HCS
Figure 33. CLKIN and RESET Timing Requirements
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
57
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
local bus timing: output clocks
NO.
’34020-32
’34020A-32
PARAMETER
MIN
11
12
12a
13
13a
14
15
16
17
18
19
20
21
tc(LCK)
tw(LCKH)
Period of local clocks LCLK1, LCLK2
tw(CK1H)
tw(LCKL)
Pulse duration, LCLK1 high
tw(CK1L)
tt(LCK)
Pulse duration, LCLK1 low
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
’34020A-40
MAX
MIN
UNIT
MAX
4tc(CKI) + s†
2tQ – 15
2tQ –10‡
4tc(CKI) + s†
2tQ – 13.5
2tQ –7‡
ns
2tQ – 15 + s
2tQ –10 + s‡
2tQ – 13.5 + s
2tQ –7 + s
ns
Pulse duration, local clock low
Transition time, LCLK1 or LCLK2
15
ns
ns
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
Hold time, LCLK1 low after LCLK2 high
ns
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 may also be specified as 4tQ.
‡ These parameters are specified with 1.5-V timing levels (parameters 12 and 13 are specified with standard timing voltage levels as detailed on
page 54).
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4*
Q1
Q2
14
11
12
14
13
LCLK1
12a
19
13a
20
21
15
22
16
11
17
14
18
14
12
13
LCLK2
*See clock stretch, page 21.
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
58
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
output signal characteristics†
The following general parameters are common to all output signals from the TMS34020 unless otherwise
specifically given. In the minimum and maximum values given, n is an integral number of quarter cycles.
PARAMETER
th(CKx-SGNV)
td(CKx-SGV)
th(SG1V-SG2NV)‡
td(SG2NV-SG1V)‡
tt(SG)
tw(SGH)
tw(SGL)
Hold time, LCLKx to output signal not valid
Delay time, LCLKx start of transition to output signal valid
Hold time, output signal valid to
output signal not valid
Delay time, output signal started
transition to output signal valid
Output signal transition time
Pulse duration, output signal high
Pulse duration, output signal low
’34020-32
’34020A-32
’34020A-40
MIN
MIN
MAX
tQ – 15
UNIT
MAX
tQ – 13.5
ns
Fast: RAS, CAS, ALTCH,
TR/QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
tQ + 15
tQ + 13.5
ns
Slow: LAD, RCA, SF
tQ + 22
tQ + 20
ns
Fast: RAS, CAS, ALTCH,
TR/QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
n tQ – 15
n tQ – 13.5
ns
Slow: LAD, RCA, SF
n tQ – 22
n tQ – 20
ns
Fast: RAS, CAS, ALTCH,
TR/QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
n tQ + 15
n tQ + 13.5
ns
Slow: LAD, RCA, SF
n tQ + 22
n tQ + 20
ns
Fast: RAS, CAS, ALTCH,
TR/QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
15
13.5
ns
Slow: LAD, RCA, SF
22
20
ns
Fast: RAS, CAS, ALTCH,
TR/QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
n tQ – 15
n tQ – 13.5
ns
Slow: LAD, RCA, SF
n tQ – 22
n tQ – 20
ns
Fast: RAS, CAS, ALTCH,
TR/QE, DDOUT, DDIN,
EMU3, HOE, R0, R1,
HDST, WE
n tQ – 15
n tQ – 13.5
ns
Slow: LAD, RCA, SF
n tQ – 22
n tQ – 20
ns
† Also refer to Figure 35 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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
59
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
QW*
QX*
QY*
QZ*
(see Note A)
LCLKx
td(CKx-SGV)
td(CKx-SGV)
th(CKx-SGNV)
tw(SGH)
SIGNAL a
tt(SG)
th(CKx-SGNV)
td(SGNV-SGV)
tt(SG)
td(SGNV-SGV)
th(SGV-SGNV)
th(SGV-SGNV)
tw(SGL)
SIGNAL b
th(SGV-SGNV)
td(SGNV-SGV)
*See clock stretch, page 21.
indicates the point at which the signal has attained a valid level.
NOTE A: Any of these quarter phases could be 2tQ if they are stretched. See clock stretch, page 21.
Figure 35. Output Signal Characteristics
All timing parameters relative to the circled points on the diagram have a fast and a slow value. Determine which
value to use for any parameter by the name of the signal that has the circle on it.
example of how to use the general output signal characteristics
Assume a system is using a TMS34020-32. Determine the maximum time from the start of the falling edge of
ALTCH to the time when data must be valid on the LAD bus for a local memory write cycle.
From the local memory write cycle diagram on page 27, the time from the falling edge of ALTCH to valid data
on the LAD bus 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(SGNV-SGV). Note that in the diagram above, there are two representations of
td(SGNV-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 the LAD bus. By definition, the signal becoming valid (SGV) determines whether
the fast value or the slow value from the table is used. On the diagram, the SGV points are marked with a circle.
In this case, for parameter td(SGNV-SGV), SGV is the LAD bus. LAD is in the slow group, so the maximum value
for td(SGNV-SGV) is n tQ + 22. The value for n is 2 from the analysis of the diagram on page 27. Thus, the maximum
time from the start of the falling edge of ALTCH to the time when data must be valid on the LAD bus for a local
memory write cycle is 2tQ + 22 ns.
60
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
host interface timing requirements (see Figure 36 and Note 3)
NO.
’34020-32
’34020A-32
PARAMETER
MIN
23
MAX
’34020A-40
MIN
UNIT
MAX
tsu(AV-CS↓)
th(CSL-AV)
Setup time of address prior to HCS↓
12
10
ns
Hold time of address after HCS low
12
10
ns
tw(CSH)
tw(RDH)
Pulse duration of HCS high
28
25
ns
Pulse duration of HREAD high
28
25
ns
Pulse duration of HWRITE high
28
25
ns
28
tw(WRH)
tsu(RDH-WR↓)
Setup time, HREAD high to HWRITE↓
28
25
ns
29
tsu(WRH-RD↓)
Setup time, HWRITE high to HREAD↓
28
25
ns
30
Pulse duration of HREAD low
18
15
ns
31
tw(RDL)
tw(WRL)
Pulse duration of HWRITE low
18
15
ns
32
tsu(CSL-WR↑)
Setup time, HCS low to HWRITE↑
18
15
ns
25†
ns
24
25
26
27
33
tsu(RDL-CK2↓)
Setup time, later of HCS low or HREAD low to LCLK2↓
30†
34
tsu(WRH-CK2↓)
Setup time, later of HWRITE high or HCS high to LCLK2↓
30†
25†
ns
0‡
ns
35
th(CK2↓-RDH)
Hold time, HREAD high after LCLK2↓
0‡
36
th(CK2↓-WRL)
Hold time, HWRITE low after LCLK2↓
0‡
0‡
ns
30†§
25†§
ns
37
tsu(RDH-CK2↓)
Setup time, HREAD high to LCLK2↓, prefetch read mode
38
tsu(CSL-RD↑)
Setup time, HCS low to HREAD↑
18
15
ns
† Setup time to insure 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 TMS34020 is set for block reads, use the deassertion of HREAD to request a local memory cycle at the next sequential address location.
NOTE 3: Although HCS, HREAD, and HWRITE can be totally asynchronous to the TMS34020, cycle responses to the signals are determined
by local memory cycles.
HA0 – HA27
HBS0 – HBS
3
23
25
24
HCS
32
26
38
30
HREAD
28
31
27
HWRITE
29
35
34
37
33
36
LCLK2
Figure 36. Host Interface Timing Requirements
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
61
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
host interface timing responses (random read cycle) (see Figure 37)
NO.
33
’34020-32
’34020A-32
PARAMETER
MIN
30†
’34020A-40
MAX
MIN
25†
tsu(RDL-CK2↓)
Setup time, later of HCS low or HREAD low to LCLK2↓
39
td(CK1↑-RYH)
Delay time from LCLK1↑ to HRDY high (end of read
cycle)
tQ + 20
tQ + 18
ns
40
td(RDH-RYL)
Delay time from earlier of HREAD or HCS high to HRDY
low
20
18
ns
41
td(CK2↓-STL)
Delay time from LCLK2↓ to HDST low
tQ + 15 + s
tQ + 13.5 + s
ns
42
td(CK1↓-STH)
Delay time from LCLK1↓ to HDST high
tQ + 13.5
ns
43
tsu(STL-RY↑)
Setup time of HDST low to HRDY↑
44
td(RY↑-STH)
Delay time from HRDY↑ to HDST high
Q1
Q2
tQ –13.5
ns
2tQ + 15
th(STH-CTV)
Hold time, CAS, TR/QE, DDIN valid after HDST high
† Setup time to insure recognition of input on this clock edge.
Q4*
ns
tQ + 15
tQ –15
50
Q3
Q4*
2tQ + 13.5
0
Q1
–2
Q2
Q3
Q4*
39
LCLK2
33
33
26
HCS/HREAD
40
41
HRDY
42
44
43
HDST
50
CAS
TR/QE
DDIN
*See clock stretch, page 21.
Figure 37. Host Interface Timing Responses (Random Read Cycle)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
ns
ns
LCLK1
62
UNIT
MAX
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
host interface timing (block read cycle) (see Figure 38 and Note 3)
NO.
’34020-32
’34020A-32
PARAMETER
MIN
37
tsu(RDH-CK2↓)
Setup time, HREAD high to LCLK2↓, prefetch read
mode
39
td(CK1↑-RYH)
Delay time from LCLK1↑ to HRDY high
40
td(RDH-RYL)
Delay time from earlier of HREAD or HCS high to
HRDY low
41
td(CK2↓-STL)
Delay time from LCLK2↓ to HDST low
42
td(CK1↓-STH)
Delay time from LCLK1↓ to HDST high
43
tsu(STL-RY↑)
Setup time of HDST low to HRDY↑
44
td(RY↑-STH)
Delay time from HRDY↑ to HDST high
td(RDL-RYH)
Delay time from later of HREAD or HCS low to HRDY
high after prefetch
45
’34020A-40
MAX
MIN
30†
UNIT
MAX
25†
ns
tQ + 20
tQ + 18
ns
20
18
ns
tQ + 15 + s
tQ + 13.5 + s
ns
tQ + 13.5
ns
tQ + 15
tQ – 15
tQ – 13.5
ns
2tQ + 15
2tQ + 13.5
ns
25
20
ns
† Setup time to insure recognition of input on this clock edge. When the TMS34020 is set for block reads, the deassertion of HREAD is used to
request a local memory cycle at the next sequential address location.
NOTE 3. Although HCS, HREAD, and HWRITE can be totally asynchronous to the TMS34020, 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
26
HREAD
39
30
40
40
45
HRDY
42
43
41
44
HDST
*See clock stretch, page 21.
Figure 38. Host Interface Timing (Block Read Cycle)
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
63
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
host interface timing responses (write cycle) (see Figure 39)
NO.
’34020-32
’34020A-32
PARAMETER
MIN
’34020A-40
MAX
MIN
UNIT
MAX
46
td(WRL-RYH)
Delay time from later of HCS or HWRITE low to HRDY
high (TMS34020 ready)
25
20
ns
47
td(WRH-RYL)
Delay time from earlier of HCS or HWRITE high to HRDY
low (end of write)
25
20
ns
48
td(CK2↓-OEL)
Delay time from LCLK2↓ to HOE low
tQ + 15 + s
tQ + 13.5 + s
ns
49
td(CK1↓-OEH)
Delay time from LCLK1↓ to HOE high
tQ + 15
tQ + 13.5
ns
39
td(CK1↑-RYH)
Delay time from LCLK1↑ to HRDY high
tQ + 18
ns
b
th(CEH-OEL)
Hold time, CAS valid high after HOE valid low
51
td(RY↑-OEH)
Delay time from HRDY↑ to HOE high
2tQ + 13.5
ns
Q4
Q1
Q2
Q3
Q4*
Q1
Q2
tQ + 20
tQ – 10
tQ – 7
2tQ + 15
Q3
Q4*
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
LCLK1
LCLK2
34
34
27
31
HCS
HWRITE
39
47
47
HRDY
49
46
48
51
HOE
b
CAS
*See clock stretch, page 21.
Figure 39. Host Interface Timing Responses (Write Cycle)
64
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q4*
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
local bus timing: bus control inputs (see Figure 40)
NO.
’34020-32
’34020A-32
PARAMETER
MIN
’34020A-40
MAX
MIN
UNIT
MAX
ta(CMV-LAV)†
Access time, CAMD valid after address valid on LAD
53
th(LANV-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
0
0
ns
56
tsu(BCV-CK2↑)†
Setup time, SIZE16, LRDY, PGMD, BUSFLT valid before LCLK2↑
20
15
ns
52
3tQ – 45
3tQ – 37
0
0
3tQ – 35 + s
ns
ns
3tQ – 27 + s
ns
† CAMD, LRDY, PGMD, SIZE16, and BUSFLT are synchronous inputs. The specified setup, access, and hold times must be met for proper device
operation.
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4*
Q1
LCLK1
LCLK2
LAD
Address
52
53
CAMD
Valid
ALTCH
55
54
LRDY
Valid
PGMD
Valid
56
55
SIZE16
BUSFLT
Valid
Valid
Valid
*See clock stretch, page 21.
Figure 40. Local Bus Timing: Bus Control Inputs
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
65
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
local bus timing: bus control inputs (see Figure 41)
NO.
’34020-32
’34020A-32
PARAMETER
MIN
’34020A-40
MAX
MIN
UNIT
MAX
57
td(CK2↑-ALL)
Delay time, ALTCH low after LCLK2↑
tQ + 15
tQ + 13.5
ns
58
td(CK1↓-ALH)
Delay time, ALTCH high after LCLK1↓
tQ + 15
tQ + 13.5
ns
59
td(CK1↑-LAV)
Delay time, LAD0 – LAD31 address valid after
LCLK1↑
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(CTNV-LAD)
Delay time, LAD0 – LAD31 driven after earlier
of DDIN↓ or CAS↑ or TR/QE↑
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
high
0
2
ns
63
td(CK2↓-LAV)
Delay time, LAD0 – LAD31 data valid after
LCLK2↓ (write)
64
th(CK2L-LAV)
Hold time, LAD0 – LAD31 data valid after
LCLK2 low (write)
65
td(CK1↑-RCV)
Delay time, RCA0 – RCA12 row address valid
after LCLK1↑
tQ + 22
tQ + 20
ns
66
td(CK2↓-RCV)
Delay time, RCA0 – RCA12 column address
valid after LCLK2↓
tQ + 22+ s
tQ + 20+ s
ns
67
th(RCV-CK2L)
Hold time, RCA0 – RCA12 address valid after
LCLK2 low
68
td(CK1↑-DIH)
Delay time, DDIN high after LCLK1↑
tQ + 15
tQ + 13.5
ns
69
td(CK1↓-DIL)
Delay time, DDIN low after LCLK1↓
tQ + 15
tQ + 13.5
ns
70
td(CK1↑-DOL)
Delay time, DDOUT low after LCLK1↑
tQ + 15
tQ + 13.5
ns
71
td(CK1↓-DOH)
Delay time, DDOUT high after LCLK1↓
tQ + 15
tQ + 13.5
ns
72
td(CK2↓-DOL)
Delay time, DDOUT low after LCLK2↓
tQ + 15 + s
tQ + 13.5 + s
ns
73
tsu(LAV-AL↓)
Setup time, LAD0 – LAD31 data valid before
ALTCH↓
74
ten(DAV-DIH)
Enable time, data valid after DDIN high
(see Note 4)
75
tdis(DAV-DIL)
Disable time, data high-impedance after DDIN
low (see Note 4)
e
th(REL-RCV)
Hold time, RAS valid low after column address
valid
3tQ – 22
3tQ – 12
ns
f
th(CEL-RCV)
Hold time, CAS valid low after column address
valid
3tQ – 22
3tQ – 12
ns
g
th(WEH-RCV)
Hold time, WE valid low after column address
valid
4tQ – 22 + s
4tQ – 18 + s
ns
tQ + 22 + s
tQ – 15
tQ + 20 + s
tQ – 13.5
tQ – 15
ns
tQ – 12
tQ – 16
ns
ns
tQ – 13
ns
2tQ – 20
2tQ – 17
ns
tQ – 12 + s
tQ – 10 + s
ns
h
th(LAV-WEH)
Hold time, LAD data valid after WE valid high
tQ – 15
tQ – 12
ns
† These values are derived from characterization data and are not tested.
NOTE 4: DDIN is used to control LAD bus buffers between the TMS34020 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.
66
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
57
63
58
ALTCH
73
Address/Status
LAD0 – LAD31
64
60
59
Data In
Data Out
66
75
67
74
65
Row
RCA0 – RCA12
66
67
67
1st Column
68
2nd Column
62
69
DDIN
72
70
71
71
DDOUT
99a
RAS
e
f
CAS
61
TR/QE
h
g
WE
*See clock stretch, page 21.
Figure 41. Local Bus Timing: Bus Control Inputs
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
67
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
local bus timing: RAS, CAS0–CAS3, WE, TR/QE, and SF
NO.
’34020-32
’34020A-32
PARAMETER
MIN
’34020A-40
MAX
MIN
UNIT
MAX
62
th(LAV-CTV)
Hold time, LAD0 – LAD31 read data
valid after earlier of DDIN, LCLK2 low
or RAS, CAS, or TR/QE high
76
td(CK1↓-REL)
Delay time, RAS low after LCLK1↓
77
td(CK1↓-REH)
78
td(CK1↑-CEL)
79
td(CK1↓-CEH)
Delay time, CAS high after LCLK1↓
80
td(CK2↓-WEL)
Delay time, WE low after LCLK2↓
81
td(CK1↓-WEH)
Delay time, WE high after LCLK1↓
82
td(CK2↓-TRL)
Delay time, TR/QE low after LCLK2↓
83
td(CK1↓-TRH)
Delay time, TR/QE high after LCLK1↓
84
td(CK1↑-SFV)
Delay time, SF valid after LCLK1↑
85
td(CK2↓-SFV)
Delay time, SF valid after LCLK2↓
86
td(CK2↓-SFZ)
Delay time, SF high-impedance after
LCLK2↓
87
tsu(ADV-RE↓)‡
Setup time, row address valid before
RAS↓
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-CE↓)
Setup time, column address valid
before CAS↓
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-CE↓)
Setup time, write data valid before
CAS↓
tQ – 22
tQ – 20
ns
92
th(CAV-CE↑)
Hold time, write data valid after CAS↑
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 CAS↓
2tQ – 8
2tQ – 8
ns
95
ta(LAV-RCV)
Access time, data in valid after column
address valid
3tQ – 17
3tQ – 2
ns
96
ta(LAV-LAV)
Access time, data in valid after
address valid on LAD bus
6tQ – 17 + s
6tQ – 17 + s
ns
97
tsu(WEL-CE↓)
Setup time, write low before CAS↓
(on write cycles)
98
tw(REH)
tw(REL)
99a
0
2
ns
tQ + 12 + s
tQ + 10 + s
ns
Delay time, RAS high after LCLK1↓
tQ + 12
tQ + 10
ns
Delay time, CAS low after LCLK1↑
tQ + 12
tQ + 10
ns
tQ + 12
tQ + 10
ns
tQ + 15 + s
tQ + 13.5 + s
ns
tQ + 15
tQ + 15
ns
tQ + 15 + s
tQ + 13.5 + s
ns
tQ + 15
tQ + 13.5
ns
tQ + 22
tQ + 20
ns
tQ + 22 + s
tQ + 20 + s
ns
tQ + 22 †
tQ + 20 †
ns
Pulse duration of RAS high
Pulse duration of RAS low
tQ – 15
tQ – 13.5
ns
4tQ – 12 + s
4n tQ – 12 + s′§¶
4tQ – 10 + s
4n tQ – 4 + s′ §¶
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.
§ s′ is 2tQ since both the address cycle and the read data cycle of a read-modify-write will be stretched. See clock stretch, page 21.
¶ n = number of GSP data cycles in in the memory access.
68
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
ns
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
local bus timing: RAS, CAS0–CAS3, WE, TR/QE, and SF
NO.
’34020-32
’34020A-32
PARAMETER
MIN
99b
100
101
102
tw(REL)
tw(CEH)
Pulse duration of RAS low
tw(CEL)
td(REL-CE↑)
Pulse duration of CAS low
Pulse duration of CAS high
Delay time RAS low to CAS↑
e
th(REL-RCV)
Hold time, RAS valid low after column address
valid
f
th(CEL-RCV)
Hold time, CAS valid low after column address
valid
h
th(LAV1-WEH) Hold time, LAD data valid after WE valid high
† n = number of GSP data cycles in in the memory access.
POST OFFICE BOX 1443
’34020A-40
MAX
MIN
UNIT
MAX
4n tQ – 12 + s †
4n tQ – 4 + s †
ns
2tQ – 12
2tQ – 8
2tQ – 10
2tQ – 8
ns
ns
4tQ – 12 + s
4tQ – 4 + s
ns
3tQ – 22
3tQ – 12
ns
3tQ – 22
3tQ – 12
ns
tQ – 15
tQ – 12
ns
• HOUSTON, TEXAS 77251–1443
69
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
76
77
LCLK2
98
e
99b
RAS
87
88
Row
RCA
1st Column
93
2nd Column
95
90
62
89
LAD (READ)
Address/Status
Data In 1
Data In 2
f
96
LAD (WRITE)
Address/Status
1st Data Out
92
79
78
102a
CAS0 – CAS3
62
2nd Data Out
94
91
100
79
101
80
h
h
81
81
97
WE
82
TR/QE
83
84
85
86
SF
ALTCH
*See clock stretch, page 21.
Figure 42. Local Bus Timing: RAS, CAS0 – CAS3, WE, TR/QE, and SF
70
83
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
CAS-before-RAS refresh: RAS and CAS0–CAS3
NO.
’34020-32
’34020A-32
PARAMETER
MIN
’34020A-40
MAX
MIN
UNIT
MAX
76
td(CK1↓-REL)
Delay time, RAS low after LCLK1↓
tQ + 12
tQ + 10
ns
77
td(CK1↓-REH)
Delay time, RAS high after LCLK1↓
tQ + 12
tQ + 10
ns
78
td(CK1↑-CEL)
Delay time, CAS low after LCLK1↑
tQ + 12
tQ + 10
ns
79
td(CK1↓-CEH)
Delay time, CAS high after LCLK1↓
tQ + 12
tQ + 10
ns
102a
td(REL-CE↑)
Delay time, RAS low to CAS↑
4tQ – 12 + s
4tQ – 4 + s
ns
102b
td(REL-CE↑)
Delay time, RAS low to CAS↑
4tQ – 12
4tQ – 4
ns
103
td(CEL-RE↓)
Delay time, CAS low to RAS↓
2tQ – 15
2tQ – 13.5
ns
104
td(REH-CE↓)
Delay time, RAS high to CAS↓
2tQ – 15 + s
2tQ – 13.5 + s
ns
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
Q1
LCLK1
LCLK2
76
RAS
77
78
79
103
104
102b
CAS
ALTCH
LAD
Refresh Pseudo-Address
RCA
Refresh Pseudo-Address
DDOUT
*See clock stretch, page 21.
NOTE A: ALTCH, LAD, RCA, and DDOUT are shown for reference only.
Figure 43. CAS-Before-RAS Refresh: RAS and CAS0 – CAS3
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.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
71
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
multiprocessor interface timing: GI, ALTCH, RAS, R0 and R1 (see Figure 44)
NO.
’34020-32
’34020A-32
’34020A-40
MIN
MIN
PARAMETER
ta(GIV-RQV)
Access time, GI valid after R0 and R1 valid
(see Notes 5 and 6)
j
tsu(GIV-CK1↑)
Setup time, GI valid before LCLK1 no longer low
(see Note 6)
106
th(CK1↑-GIV)
Hold time, GI valid after LCLK1↑ (see Note 5)
107
td(CK2↑-RQV)
Delay time, LCLK2↑ to R0 or R1 valid
105
MAX
2tQ – 40
2tQ – 30
40
35
0
ns
ns
0
tQ + 15
UNIT
MAX
ns
tQ + 13.5
ns
108
td(CK2H-RQNV) Delay time, LCLK2 high to R0 or R1 no longer valid
tQ – 15
tQ – 13.5
ns
NOTES: 5. These timings must be met to insure that the GI input is recognized on this clock cycle.
6. It is required that either specification j or specification 105 be observed for the multiprocessor interface to function correctly. It is not
necessary for both to be met.
Q4*
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4*
Q1
Q2
Q3
Q4*
LCLK1
j
LCLK2
107
108
Valid
R0 – R1
106
105
Valid
GI
*See clock stretch, page 21.
Figure 44. Multiprocessor Interface Timing: GI, ALTCH, RAS, R0 and R1
72
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q1
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
multiprocessor interface timing: high-impedance signals
NO.
’34020-32
’34020A-32
PARAMETER
MIN
MAX
’34020A-40
MIN
UNIT
MAX
84
td(CK1↑-SFV)
Delay time, SF valid after LCLK1↑
tQ + 22
tQ + 20
ns
86
td(CK2↓-SFZ)
Delay time, SF high-impedance after LCLK2↓
tQ + 22 + s†
tQ + 20 + s†
ns
109
td(CK2↓-ADZ)
Delay time, LAD and RCA high-impedance after
LCLK2↓
tQ + 22 + s†
tQ+ 20 + s†
ns
110
td(CK1↑-ADV)
Delay time, LAD and RCA valid after LCLK1↑
tQ + 22
tQ + 20
ns
111
td(CK1↑-CTZ)
Delay time, ALTCH, RAS, CAS, WE, TR/QE, HOE,
and HDST high-impedance after LCLK1↑
tQ + 15†
tQ +13.5†
ns
112
td(CK2↓-CTH)
Delay time, ALTCH, RAS, CAS, WE, TR/QE, HOE,
and HDST high-impedance after LCLK2↓
tQ + 15 + s
tQ + 13.5 + s
ns
113
td(CK1↑-DIZ)
Delay time, DDIN high-impedance after LCLK↑
tQ + 15†
tQ + 13.5 †
ns
114
td(CK2↓-DIL)
Delay time, DDIN low after LCLK2↓
115
td(CK2↓-DOZ)
Delay time, DDOUT high-impedance after LCLK2↓
116
td(CK2↓-DOH) Delay time, DDOUT high after LCLK2↓
† These values are derived from characterization data and are not tested.
POST OFFICE BOX 1443
tQ + 15 + s
tQ + 13.5+ s
ns
tQ + 15 +s †
tQ +13.5 + s†
ns
tQ + 15 + s
tQ + 13.5 + s
ns
• HOUSTON, TEXAS 77251–1443
73
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
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 21.
Figure 45. Multiprocessor Interface Timing: High-Impedance Signals
74
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
Q1
Q2
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
video shift clock timing: SCLK
NO.
117
118
119
’34020-32
’34020A-32
PARAMETER
’34020A-40
MIN
MAX
MIN
MAX
50
25
50
tc(SCK)
tw(SCKH)
Period of video serial clock SCLK
35
Pulse duration of SCLK high
12
10
tw(SCKL)
tt(SCK)
Pulse duration of SCLK low
12
2†
10
2†
120
Transition time (rise and fall) of SCLK
† This value is determined through computer simulation and is not tested.
5†
UNIT
ns
ns
ns
5†
ns
117
120
118
120
119
SCLK
Figure 46. Video Shift Clock Timing: SCLK
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
75
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
video interface timing: VCLK and video outputs
NO.
123
124
125
126
’34020-32
’34020A-32
PARAMETER
’34020A-40
MIN
MAX
MIN
MAX
62.5
100
62.5
100
tc(VCK)
tw(VCKH)
Period of video input clock VCLK
Pulse duration of VCLK high
28
28
tw(VCKL)
tt(VCK)
Pulse duration of 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(VCK↓-HS↓)
Hold time, VCLK↓ to HSYNC, VSYNC, CSYNC/HBLNK, or
CBLNK/VBLNK↓
0†
0†
ns
130
th(VCK↓-HS↑)
Hold time, VCLK↓ to HSYNC, VSYNC, CSYNC/HBLNK, or
CBLNK/VBLNK↑
0†
0†
ns
40
40
ns
40
40
ns
† This value is determined through computer simulation and is not tested.
123
124
126
125
126
VCLK
127
HSYNC
VSYNC
CSYNC/HBLNK
CBLNK/VBLNK
(OUTPUTS)
128
129
130
Figure 47. Video Interface Timing: VCLK and Video Outputs
76
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
video interface timing: external sync inputs (see Note 7)
NO.
’34020-32
’34020A-32
PARAMETER
MIN
MAX
’34020A-40
MIN
UNIT
MAX
131
tsu(SL-VCK↑)
Setup time, HSYNC, VSYNC, CSYNC low to VCLK↑
20
20
ns
132
tsu(SH-VCK↑)
Setup time, HSYNC, VSYNC, CSYNC high to VCLK↑
20
20
ns
133
th(VCKH-SV)
Hold time, HSYNC, VSYNC, CSYNC valid after VCLK high
20
20
NOTE 7: Setup and hold times on asynchronous inputs are required only to assure recognition at indicated clock edges.
ns
A
B
C
D
VCLK
133
131
133
HSYNC
VSYNC
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 will be 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 will be detected at edge D instead of edge C.
Figure 48. Video Interface Timing: External Sync Inputs
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
77
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
interrupt timing: LINT1 and LINT2
NO.
134
’34020-32
’34020A-32
PARAMETER
tsu(LINTL-CK2↑)
MIN
tQ + 45 †
Setup time, LINT1 or LINT2 low before LCLK2↑
’34020A-40
MAX
MIN
tQ + 40 †
UNIT
MAX
ns
tw(LINTL)
Pulse duration of LINT1 or LINT2 low
8tQ‡
8tQ‡
ns
† Although LINT1 and LINT2 can be asynchronous to the TMS34020, this setup insures recognition of the interrupt on this clock edge.
‡ This pulse duration minimum insures that the interrupt is recognized by internal logic; however, the level must be maintained until it has been
acknowledged by the interrupt service routine.
135
LCLK1
LCLK2
134
LINT1
LINT2
135
Figure 49. Interrupt Timing: LINT1 and LINT2
host interrupt timing: HINT
NO.
’34020-32
’34020A-32
PARAMETER
MIN
136
td(CK1↑-HINTV)
Delay time, LCLK1↑ to HINT valid
30
LCLK1
136
LCLK2
136
HINT
Figure 50. Host Interrupt Timing: HINT
78
POST OFFICE BOX 1443
MAX
• HOUSTON, TEXAS 77251–1443
’34020A-40
MIN
UNIT
MAX
25
ns
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
PARAMETER MEASUREMENT INFORMATION
emulator interface timing
NO.
’34020-32
’34020A-32
PARAMETER
MIN
137
tsu(EMV-CK1↑)
Setup time, EMU0 – EMU2 valid to LCLK1↑
138
th(EMV-CK1↑)
Hold time, EMU0 – EMU2 valid after LCLK1↑
139
td(CK1L-SCV)
th(CK2H-SCNV)
Delay time, EMU3 valid after LCLK1 low
140
’34020A-40
MAX
MIN
30
25
ns
0
0
ns
25
Hold time, LCLK2 high before EMU3 not valid
UNIT
MAX
tQ – 15
20
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
79
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
MECHANICAL DATA
GB/S-CPGA-P145
CERAMIC PIN GRID ARRAY PACKAGE, CAVITY UP
1.590 (40,38)
SQ
1.560 (39,62)
1.400 (35,56) TYP
R
P
N
M
L
K
J
H
G
Extra Pin
F
E
D
C
B
A
0.080 (2,03)
0.060 (1,53)
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15
SQ
0.090 (2,28)
0.016 (0,41) TYP
0.055 (1,39)
DIA TYP
0.045 (1,15)
3 Places
0.070 (1,79)
0.140 (3,55)
0.120 (3,05)
0.100 (2,54) TYP
0.020 (0,51)
DIA TYP
0.016 (0,41)
4040114-11/A–10/93
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
0.051 (0,38) radius relative to the center of the ceramic.
E. This is a hermetically sealed ceramic package with metal and gold-plated pins.
80
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
TMS34020, TMS34020A
GRAPHICS PROCESSORS
SPVS004D – MARCH 1990 – REVISED NOVEMBER 1993
MECHANICAL DATA
PCM/S-PQFP-G***
PLASTIC QUAD FLATPACK
144-PIN SHOWN
0,65 TYP
0,38
0,22
108
73
72
109
NO. OF PINS***
144
144
22,75 TYP
160
25,35 TYP
37
1
36
A
0,16 TYP
3,67
3,17
A
28,10
SQ
27,90
31,45
SQ
30,95
0,25 MIN
0°–7°
0,95
0,65
Seating Plane
0,10
4,07 MAX
4040024/A–10/93
NOTES: A.
B.
C.
D.
All linear dimensions are in millimeters.
This drawing is subject to change without notice.
Falls within JEDEC MO-108.
The 144PCM is identical to 160PCM except for 4 leads per corner are removed.
POST OFFICE BOX 1443
• HOUSTON, TEXAS 77251–1443
81
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