Datasheet SM34020AGBM32, SM34020AGBM40, SM34020AHTM32, SM34020AHTM40, SMJ34020AGBM40 Datasheet (Texas Instruments)

...

Download

Page 1

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

Class B High-Reliability Processing

1-µm CMOS Technology

Military Operating Temperature Range

55°C to 125°C

SMJ34020A-32/40

125/100-ns Instruction Cycle Time

Fully Programmable 32-Bit General-Purpose Processor With 512-Megabyte Linear Address Range (Bit Addressable)

Second-Generation Graphics System Processor – Object-Code Compatible With the

SMJ34010 – Enhanced Instruction Set – Optimized Graphics Instructions – Coprocessor Interface

Pixel Processing, XY Addressing, and Window Checking Built Into the Instruction Set

Programmable 1-, 2-, 4-, 8-, 16-, or 32-Bit Pixel Size With 16 Boolean and Six Arithmetic Pixel Processing Options (Raster Ops)

512-Byte LRU On-Chip Instruction Cache

Optimized DRAM/VRAM Interface – Page-Mode for Burst Memory Operations – Dynamic Bus Sizing (16-Bit and

32-Bit Transfers) – Byte-Oriented CAS

Strobes

Flexible Host Processor Interface – Supports Host Transfers – Direct Access to All of the SMJ34020A

Address Space – Implicit Addressing – Prefetch for Enhanced Read Access

Programmable CRT Control – Composite Sync Mode – Separate Sync Mode – Synchronization to External Sync

Direct Support for Special Features of 1M VRAMs – Load Write Mask – Load Color Mask – Block Write – Write Using the Write Mask

Flexible Multi-Processor Interface

Packaging Options – 145-Pin Grid Array Ceramic Package

(GB Suffix)

– 132-Pin Ceramic Quad Flat Pack

(Unformed Lead) (HT Suffix)

145-PIN GRID ARRAY PACKAGE

(BOTTOM VIEW)

A B C D E

F G H

L M N

123456789101112131415

132-PIN QUAD FLATPACK

(TOP VIEW)

132

10066

33 67

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.

Please be aware that an important notice concerning availability, standard warranty, and use in critical applications of Texas Instruments semiconductor products and disclaimers thereto appears at the end of this data sheet.

Page 2

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

description

The SMJ34020A graphics system processor (GSP) is the second generation of an advanced high-performance CMOS 32-bit microprocessor optimized for graphics display systems. With a built-in instruction cache, the ability to simultaneously access memory and registers, and an instruction set designed to expedite raster graphics operations, the SMJ34020A provides user-programmable control of the CRT interface as well as the memory interface (both standard DRAM and multiport video RAM). The 4-gigabit (512-megabyte) physical address space is addressable on bit boundaries using variable width data fields (1 to 32 bits). Additional graphics addressing modes support 1-, 2-, 4-, 8-, 16- and 32-bit wide pixels.

architecture

The SMJ34020A is a CMOS 32-bit processor with hardware support for graphics operations such as pixel block transfers (PIXBL TS) during raster operations and curve-drawing algorithms. Also included is a complete set of general-purpose instructions with addressing modes tuned to support high-level languages. In addition to its ability to address a large external memory range, the SMJ34020A contains 30 general-purpose 32-bit registers, a hardware stack pointer, and a 512-byte instruction cache. On-chip functions include 64 programmable I/O registers that control CRT timing, input / output control, and parameters required by some instructions. The SMJ34020A directly interfaces to DRAMs and VRAMs and generates raster control signals. The SMJ34020A can be configured to operate as a standalone processor, or it can be used as a graphics engine with a host system. The host interface provides a generalized communication port for any standard host processor. The SMJ34020A also accommodates a multiprocessing or direct memory access (DMA) environment through the request /grant interface protocols. Virtual memory systems are supported through bus-fault detection and instruction continuation.

The SMJ34020A provides single-cycle execution of general-purpose instructions and most common integer arithmetic and Boolean operations from its instruction cache. Additionally, the SMJ34020A incorporates a hardware barrel shifter that provides a single-state bidirectional shift-and-rotate function for 1 to 32 bits.

The local-memory controller is designed to optimize memory access operations. It also supports pipeline memory write operations of variable-sized fields and allows memory access and instruction execution in parallel.

The SMJ34020A graphics-processing hardware supports pixel and pixel-array processing capabilities for both monochrome and color systems at a variety of pixel sizes. The hardware incorporates two-operand and three-operand raster operations with Boolean and arithmetic operations, XY addressing, window clipping, window-checking operations, 1 to

bits-per-pixel transforms, transparency, and plane masking. The architecture further supports operations on single pixel transfer (PIXT) instructions or on two-dimensional arrays of arbitrary size (PIXBLTS).

The SMJ34020A ’ s flexible graphics-processing capabilities allow software-based graphics algorithms without sacrificing performance. These algorithms include clipping to arbitrary window size, custom incremental-curve drawing, two-operand raster operations, and masked two-operand raster operations.

The SMJ34020A provides for extensions to the basic architecture through the coprocessor interface. Special instructions and cycle timings are included to enhance data flow to coprocessors without requiring the coprocessor to decode the instruction stream, generate system addresses, or move data for the coprocessor through the SMJ34020A.

Page 3

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

Pin Assignments – 145-Pin Grid Array Package

PIN PIN PIN PIN

NUMBER NAME NUMBER NAME NUMBER NAME NUMBER NAME

A1 V

C9 RCA8 J1 EMU0 N15 LAD17

A2 ALTCH C10 RCA12 J2 GI P1 V

A3 CBLNK/VBLNK C11 LAD30 J3 EMU1 P2 HWRITE A4 HSYNC C12 V

J13 LAD4 P3 HCS

A5 TR/QE C13 V

J14 V

P4 HA30

A6 RCA2 C14 V

J15 LAD5 P5 HA27 A7 RCA3 C15 LAD26 K1 EMU2 P6 HA24 A8 V

D1 RAS K2 RESET P7 HA22

A9 RCA6 D2 CAS2 K3 LINT2 P8 HA18

A10 RCA7 D3 V

K13 V

P9 HA14

A11 RCA10 D4

†

K14 LAD3 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 V

E1 R1 L3 LRDY P14 HBS0

B1 CAS3 E2 V

L13 LAD1 P15 LAD0 B2 WE E3 CAS1 L14 LAD2 R1 HREAD B3 V

E13 LAD27 L15 LAD19 R2 HA31 B4 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 V

R6 HA20

B8 RCA5 F3 V

M14 LAD16 R7 HA19

B9 RCA9 F13 LAD24 M15 LAD18 R8 HA17

B10 RCA11 F14 LAD8 N1 SIZE16 R9 HA16 B11 LAD31 F15 V

N2 V

R10 HA15 B12 LAD14 G1 HINT N3 CLKIN R11 HA11 B13 V

G2 HOE N4 V

R12 HA9 B14 LAD13 G3 HDST N5 HA29 R13 HA8 B15 LAD12 G13 LAD7 N6 HA25 R14 HBS3

C1 CAS0 G14 V

N7 HA21 R15 V

C2 V

G15 LAD23 N8 V

C3 DDOUT H1 LCLK1 N9 V

C4 DDIN H2 EMU3 N10 HA12 C5 V

H3 LCLK2 N11 HA6 C6 SF H13 LAD22 N12 HBS2 C7 RCA4 H14 LAD21 N13 HBS1 C8 V

H15 LAD6 N14 V

†

This pin is provided for device orientation purpose only. Make no external connection.

Page 4

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

Pin Assignments – 132-Pin Ceramic Quad Flatpack Package

PIN PIN PIN PIN

NUMBER NAME NUMBER NAME NUMBER NAME NUMBER NAME

1 CAS3 34 HCS 67 LAD0 100 LAD29 2 CAS2 35 HA31 68 LAD16 101 LAD14 3 CAS1 36 HA30 69 LAD1 102 LAD30 4 CAS0 37 HA29 70 LAD17 103 LAD15 5 V

38 HA28 71 LAD2 104 LAD31 6 RAS 39 HA27 72 LAD18 105 SCLK 7 V

40 HA26 73 V

106 RCA12 8 R0 41 HA25 74 LAD3 107 RCA11 9 R1 42 HA24 75 LAD19 108 RCA10

10 HOE 43 HA23 76 V

109 RCA9

11 HDST 44 HA22 77 LAD4 110 RCA8 12 HRDY 45 HA21 78 LAD20 111 RCA7 13 HINT 46 HA20 79 LAD5 112 RCA6 14 EMU3 47 HA19 80 LAD21 113 RCA5 15 LCLK1 48 HA18 81 LAD6 114 V

16 LCLK2 49 HA17 82 LAD22 115 V

17 EMU1 50 V

83 LAD7 116 RCA4 18 EMU0 51 HA16 84 LAD23 117 RCA3 19 EMU2 52 HA15 85 V

118 RCA2

20 GI 53 HA14 86 V

119 RCA1 21 RESET 54 HA13 87 LAD8 120 RCA0 22 LINT2 55 HA12 88 LAD24 121 SF 23 LINT1 56 HA11 89 LAD9 122 TR/QE 24 CAMD 57 HA10 90 LAD25 123 VSYNC 25 BUSFLT 58 HA9 91 LAD10 124 HSYNC 26 SIZE16 59 HA8 92 LAD26 125 CBLNK /VBLNK 27 PGMD 60 HA7 93 LAD11 126 CSYNC/HBLNK 28 LRDY 61 HA6 94 LAD27 127 V

29 V

62 HA5 95 V

128 V

30 VCLK 63 HBS3 96 LAD12 129 ALTCH 31 CLKIN 64 HBS2 97 LAD28 130 DDIN 32 HWRITE 65 HBS1 98 V

131 DDOUT 33 HREAD 66 HBS0 99 LAD13 132 WE

Page 5

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

Pin Functions

NAME I/O

†

DESCRIPTION

LOCAL MEMORY INTERF ACE

ALTCH O

Address latch. The high-to-low transitions of ALTCH can be used to capture the address and status available on LAD. A transparent latch (such as a 54ALS373) maintains the current address and status as long as ALTCH

remains low.

BUSFLT I

Bus fault. External logic asserts BUSFLT high to the SMJ34020A to indicate that an error or fault has occurred on the current bus cycle. BUSFLT is also used with LRDY to generate externally requested bus cycle retries so that the entire memory address is presented again on LAD.

In the emulation mode, BUSFL T is used for write protecting mapped memory (by disabling CAS

outputs for the current

cycle).

DDIN O

Data bus direction in enable. DDIN is used to drive the active-high output enables on bidirectional transceivers (such as the 54ALS623). The transceivers buffer data input and output on LAD0–LAD31 when the SMJ34020A is interfaced to several memories.

DDOUT O

Data bus direction output enable. DDOUT drives the active-low output enables on bidirectional transceivers (such as the 54ALS623). The transceivers buffer data input and output on LAD0–LAD31.

LAD0–LAD31 I/O

32-bit multiplexed local address/data bus. At the beginning of a memory cycle, the word address is output on LAD4–LAD31 and the cycle status is output on LAD0–LAD3. After the address is presented, LAD0–LAD31 are used for transferring data within the SMJ34020A system. LAD0 is the LSB and LAD31 is the MSB.

LRDY I

Local ready. External circuitry drives LRDY low to inhibit the SMJ34020A from completing a local-memory cycle it has initiated. While LRDY remains low, the SMJ34020A waits unless the SMJ34020A loses bus priority or is given an external RETRY request (through BUSFL T). Wait states are generated in increments of one full LCLK1 cycle. LRDY can be driven low to extend local memory-read and memory-write cycles, VRAM serial-data-register-transfer cycles, and DRAM-refresh cycles. During internal cycles, the SMJ34020A ignores LRDY.

PGMD I

Page mode. The memory-decode logic asserts PGMD low if the currently addressed memory supports burst (page mode) accesses. Burst accesses occur as a series of CAS

cycles for a single RAS cycle to memory. LRDY is used with

BUSFLT to describe the cycle termination status for a memory cycle.

PGMD

is also used in emulation mode for mapping memory.

SIZE16 I

Bus size. The memory-decode logic can pull SIZE16 low if the currently addressed memory or port supports only 16-bit transfers. SIZE16

can also be used to determine which 16 bits of the data bus are used for a data transfer.

In the emulation mode, SIZE16

is used to select the size of mapped memory.

DRAM AND VRAM CONTROL

CAMD I

Column-address mode. CAMD dynamically shifts the column address on the RCA0–RCA12 bus to allow the mixing of DRAM and VRAM address matrices using the same multiplexed address RCA0–RCA12 signals.

CAS0–CAS3 O

Four column-address strobes. CAS outputs drive the CAS inputs of DRAMs and VRAMs. CAS0 –CAS3 strobe the column address on RCA0–RCA12 to the memory. The four CAS

strobes provide byte write-access to the memory.

RAS O

Row-address strobe. RAS output drives the RAS inputs of DRAMs and VRAMs. RAS strobes the row address on RCA0–RCA12 to memory.

RCA0–RCA12 O

Thirteen multiplexed row-address/column-address signals. At the beginning of a memory-access cycle, the row address for DRAMs is present on RCA0–RCA12. The row address contains the most significant address bits for the memory. As the cycle progresses, the memory column address is placed on RCA0–RCA12. The addresses that are actually output during row and column times depend on the memory configuration (set by RCM0 and RCM1 in the CONFIG register) and the state of CAMD during the access. RCA0 is the LSB, and RCA12 is the MSB.

SF O

Special function pin. SF is the special-function signal to 1M VRAMs that allows the use of block write, load write mask, load color mask, and write using write mask. SF is also used to differentiate instructions and addresses for the coprocessor as part of the coprocessor interface.

TR/QE O

Transfer/output-enable. TR/QE drives the TR/QE input of VRAMs. During a local memory-read cycle, TR/QE functions as an active-low output enable to gate from memory to LAD0–LAD31. During special VRAM function cycles, TR

/QE

controls the type of cycle that is performed.

†

I = input, O = output

Page 6

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

Pin Functions (Continued)

NAME I/O

†

DESCRIPTION

DRAM AND VRAM CONTROL (CONTINUED)

WE O Write enable. The active low WE drives the WE inputs of DRAMs and VRAMs. WE can also be used as the active

low write enable to static memories and other devices connected to the SMJ34020A local interface. During a local-memory read cycle, WE

remains inactive high while CAS is strobed active low. During a local-memory write

cycle, WE

is strobed active low before CAS. During VRAM serial-data-register transfer cycles, the state of WE at

the falling edge of RAS

controls the direction of the transfer.

HOST INTERFACE

HA5–HA31 I Twenty-seven host address input signals. A host can access a long word by placing the address on these lines.

HA5–HA31 correspond to LAD5–LAD31 that output the address to the local memory. HBS0–HBS3 I Four host byte selects. HBS0 –HBS3 identify which bytes within the long word are being selected. HCS I Host chip select. A host drives HCS low to latch the current host address present on HA5–HA31 and the host byte

selects on HBS0–HBS3. HCS

also enables host access cycles to the SMJ34020A I/O registers or local memory.

During the low-to-high transition of RESET

, the level on HCS determines whether the SMJ34020A is halted (HCS is high for host-present mode) or whether it begins executing its reset service routine (HCS is low for self-bootstrap mode).

HDST O Host data-latch strobe. The rising edge of HDST latches data from the SMJ34020A local address space to the

external host data latch on host read accesses. HDST can be used in conjunction with HRDY to indicate that data is valid in the external data latch.

HINT O Host Interrupt. HINT allows the SMJ34020A to interrupt a host by setting the INTOUT bit in the HSTCTLL I/O register.

HINT

can also be used to interrupt the host if a BUSFLT or RETRY occurs due to a host access cycle.

HOE O Host data latch output enable. HOE enables data from host data latches to the SMJ34020A local address space on

host write cycles. HOE

can be used in conjunction with HRDY to indicate data has been written to memory from the

external data latch.

HRDY O Host ready. HRDY is normally low and goes high to indicate that the SMJ34020A is ready to complete a host-initiated

read or write cycle. If the SMJ34020A is ready to accept the access request, HRDY is driven high and the host can proceed with the access. A host can use HRDY logically combined with HDST and HOE

to determine when the local

bus access cycles have completed.

HREAD I Host read strobe. HREAD is driven low during a read request from a host processor. This notifies the SMJ34020A

that the host is requesting access to the I/O registers or to local memory. HREAD

should not be asserted at the same

time that HWRITE

is asserted.

HWRITE I Host write strobe. HWRITE is driven low to indicate a write request by a host processor . This notifies the SMJ34020A

that a write request is pending. The rising edge of HWRITE

is used to indicate that the host has latched data to be

written in the external data transceivers. HWRITE

should not be asserted at the same time HREAD is asserted.

SYSTEM CONTROL

CLKIN I Clock input. CLKIN generates LCLK1 and LCLK2, to which all processor functions in the SMJ34020A are

synchronous. A separate asynchronous input clock (VCLK) controls the video timing and video registers.

LCLK1, LCLK2 O Local output clocks. LCLK1 and LCLK2 are 90 degrees out of phase with each other. They provide convenient

synchronous control of external circuitry to the internal timing. All signals output from the SMJ34020A (except the CRT timing signals) are synchronous to LCLK1 and LCLK2.

LINT1, LINT2 I Local interrupt requests. Interrupts from external devices are transmitted to the SMJ34020A on LINT1 and LINT2.

Each local interrupt signal activates the request for one of two interrupt request levels. An external device generates an interrupt request by driving the appropriate interrupt request pin to its active-low state. LINT1

, LINT2 should

remain low until the SMJ34020A recognizes it. LINT1

, LINT2 can be applied asynchronously to the SMJ34020A as

they are synchronized internally before use.

RESET I System reset. During normal operation, RESET is driven low to reset the SMJ34020A. When RESET is asserted

low, the SMJ34020A ’s internal registers are set to an initial known state and all output and bidirectional pins are driven either to inactive levels or to the high-impedance state. The SMJ34020A’ s behavior following reset depends on the level of the HCS

input just before the low-to-high transition of RESET . If HCS is low, the SMJ34020A begins executing

the instructions pointed to by the reset vector. If HCS

is high, the SMJ34020A is halted until a host processor writes

a 0 to the HLT bit in the HSTCTLL register.

†

I = input, O = output

Page 7

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

Pin Functions (Continued)

NAME I/O

†

DESCRIPTION

POWER

‡

I Nominal 5-V power supply inputs. Five pins on QFP; Nine pins on PGA.

‡

I Electrical ground inputs. Nine pins on QFP; 17 pins on PGA.

EMULATION CONTROL

EMU0–EMU2 I Emulation pins 0–2 EMU3 O Emulation pin 3

MULTIPROCESSOR INTERFACE

GI I Bus grant input. External bus arbitration logic drives GI low to enable the SMJ34020A to gain access to the

local-memory bus. The SMJ34020A must release the bus if GI

is high so that another device can access the bus.

R1, R0 O Bus request and control. R1 and R0 indicate a request for use of the bus in a multiprocessor system; they are

decoded as shown below:

R1 R0 Bus Request Type

L L High-priority bus request L H Bus-cycle termination H L Low-priority bus request H H No bus request pending

A high-priority bus request provides for VRAM serial-data-register transfer cycles (midline or blanked), DRAM refresh (when 12 or more refresh cycles are pending), or a host-initiated access. The external arbitration logic should grant the request as soon as possible by asserting GI

low.

A low-priority bus request is used to provide for CPU-requested access and DRAM refresh (when less than 12 refresh cycles are pending).

Bus-cycle termination status is provided so that the arbitration logic can determine that the device currently accessing the bus is completing an access, and other devices can compete for the next bus cycle. A no-bus-request-pending status is output when the currently active device does not require the bus on subsequent cycles.

VIDEO INTERFACE

CBLNK / VBLNK O Composite blanking/vertical blanking. CBLNK / VBLNK can be programmed to select one of two blanking

functions:

Composite blanking for blanking the display during both horizontal and vertical retrace periods in composite-sync-video mode Vertical blanking for blanking the display during vertical retrace in separate-sync-video mode.

Immediately following reset, CBLNK

/ VBLNK is configured as a CBLNK output.

CSYNC / HBLNK I/O Composite sync/horizontal blanking. CSYNC / HBLNK can be programmed to select one of two functions:

Composite sync (either input or output as set by a control bit in the DPYCTL register) in composite-sync-video mode: As an input, extracts HSYNC

and VSYNC from externally generated horizontal sync pulses

As an output,

CSYNC / HBLNK generates active-low composite-sync pulses from either externally

generated HSYNC

and VSYNC signals or signals generated by the SMJ34020A’ s on-chip video timers

Horizontal blank (output only) for blanking the display during horizontal retrace in separate-sync-video mode.

Immediately following reset, CSYNC

/ HBLNK is configured as a CSYNC input.

HSYNC I/O Horizontal sync. HSYNC is the horizontal sync signal that controls external video circuitry. HSYNC can be

programmed to be either an input or an output by modifying a control bit in the DPYCTL register.

As an output, HSYNC

is the active-low horizontal-sync signal generated by the SMJ34020A’s on-chip video timers. As an input, HSYNC

synchronizes the SMJ34020A video-control registers to externally generated horizontal-sync pulses. The actual synchronization can be programmed to begin at any VCLK cycle; this allows for any external pipelining of signals.

Immediately following reset, HSYNC

is configured as an input.

†

I = input, O = output

‡

For proper SMJ34020A operation, all VCC and VSS pins must be connected externally.

Page 8

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

Pin Functions (Continued)

NAME I/O

†

DESCRIPTION

VIDEO INTERFACE (CONTINUED)

SCLK I Serial data clock. SCLK is the same as the signal that drives VRAM serial data registers. SCLK allows the SMJ34020A

to track the VRAM serial-data-register count, providing serial-register transfer and midline-reload cycles. (SCLK can be asynchronous to VCLK; however, it typically has a frequency that is a multiple of the VCLK frequency).

VCLK I Video clock. VCLK is derived from a multiple of the video system’s dot clock and is used internally to drive the video

timing logic.

VSYNC I/O Vertical sync. VSYNC is the vertical sync signal that controls external video circuitry . VSYNC can be programmed to

be either an input or an output by modifying a control bit in the DPYCTL register.

As an output, VSYNC

is the active-low vertical-sync signal generated by the SMJ34020A’ s on-chip video timers.

As an input, VSYNC

synchronizes the SMJ34020A video-control registers to externally generated vertical-sync pulses. The actual synchronization can be programmed to begin at any horizontal line; this allows for any external pipelining of signals.

Immediately following reset, VSYNC

is configured as an input.

†

I = input, O = output

functional block diagram

HA5–HA31

HBS0–HBS3

HCS

HREAD

HWRITE

HINT

HRDY

HDST

HOE

GI R0 R1

EMU0 EMU1 EMU2 EMU3

CLKIN LCLK1 LCLK2

RESET

, LINT1,

LINT2

LAD0–LAD31

RCA0–RCA12 DDIN

DDOUT RAS

CAS0–CAS3 WE

TR/QE ALTCH SF

PGMD SIZE16 LRDY BUSFLT CAMD

VSYNC HSYNC CSYNC/HBLNK CBLNK/VBLNK VCLK SCLK

Host

Address

Latch

Host

Interface

Multi-

Processor

Interface

Emulation

Interface

System

Clocks

Buffer/

MUX

Bus

DRAM/

VRAM

Interface

Bus

Interface

Video

Timing

and

Control

Local

Memory

and Bus

Timing

I/O

LRU

Regs

ALU

Barrel

Shifter

Microcontrol ROM

Reset and Interrupts

Control

Page-mode

Cache

PC ST

Decode

File A

File B

Page 9

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

architecture (continued)

Boolean, arithmetic, pixel-processing, byte, and field-move instructions operate on data within the general-purpose register files. The SMJ34020A contains two register files of fifteen 32-bit registers and a system stack pointer (SP). The SP is addressed in both register file A and register file B as a sixteenth register. Transfers between registers and memory are facilitated using a complete set of field

move

instructions with

selectable field sizes. The 15 general-purpose registers in register file A are used for high-level language support and

assembly-language programming. The 15 registers in register file B are dedicated to special functions during PIXBLTS and other pixel operations but can be used as general-purpose registers at other times.

stack pointer (SP)

The stack pointer is a dedicated 32-bit internal register that points to the top of the system stack.

program counter (PC)

The SMJ34020A’s 32-bit program counter register points to the next instruction-stream word to be fetched. Since instruction words are aligned to 16-bit boundaries, the four LSBs of the PC are always zero.

instruction cache

An on-chip cache contains 512 bytes of RAM and provides unimpeded access to instructions. The cache operates automatically and is transparent to software. The cache is divided into four 128-byte segments. Associated with each segment is a 22-bit segment start address register (SSA) to identify the addresses in memory corresponding to the current contents of the cache segment. Each cache segment is further partitioned into eight subsegments of four long words (32 bits) each. Each subsegment has an associated present (P) flag to indicate whether or not the subsegment contains valid data.

The cache is loaded only when an instruction requested by the execution section of the SMJ34020A is not already contained within the cache. A least-recently-used (LRU) algorithm determines which of the four segments of the cache is overwritten with new data. For this purpose, an internal four-by-two LRU stack keeps track of cache usage. Although the cache is loaded so as to always fill a subsegment completely , not all eight subsegments within a segment are necessarily filled (this is dependent upon the instruction stream).

status register

The status register (ST) is a special purpose 32-bit register dedicated to status codes set by the results of implicit and explicit compare operations and parameters used to specify the length and behavior of fields 0 and 1. During an interrupt, when the IX bit in the ST is placed on the stack, it indicates that execution of an interruptable instruction (PIXBLT, FILL or LINE) was halted to service the interrupt. The single-step bit causes a trap to the single-step vector (located at address FFFF FBE0h) after the execution of one instruction when the bit is set high. Normal program execution occurs when the bit is set low.

fields, bytes, words, long words, pixels and pixel arrays

The SMJ34020A outputs a 28-bit address on LAD4–LAD31 that is valid at the falling edge of ALTCH

. The most significant 27 bits (LAD5–LAD31) define a 32-bit-long word of physical memory; logically, however, the SMJ34020A views memory data as fields addressable at the bit level. The least significant bit of the 28-bit address (LAD4) is used to select the odd or even word when accessing 16-bit memories (indicated by SIZE16 asserted low). Primitive data types supported by the SMJ34020A include bytes, words, long words, pixels, two independent fields of from 1 to 32 bits, and user-defined pixel arrays.

Words and long words, respectively, refer to 16- and 32-bit values that are aligned on 32-bit boundaries.

Page 10

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

fields, bytes, words, long words, pixels and pixel arrays (continued)

The two independent fields are referenced as field 0 and field 1. The attributes of these fields (field size and sign extension within a register) are defined in the status register as FS0, FE0, FS1, and FE1. Fields 0 and 1 are specified independently to be signed or unsigned and from 1 to 32 bits in length. Bytes are special 8-bit cases of the field data type, while pixels are 1, 2, 4, 8, 16, or 32 bits in length. In general, fields (including bytes) can start and terminate on arbitrary bit boundaries; however, pixels must pack evenly into 32-bit-long words.

pixel operations

Pixel arrays are two-dimensional data types of user-defined width, length, pixel depth (number of bits per pixel), and pitch (distance between rows). A pixel or pixel array can be accessed by means of either its memory address or its XY coordinates. Transfers of individual pixels or pixel blocks are influenced by the pixel processing, transparency , window checking, plane masking, pixel masking, or corner adjustment operations selected. For further information, see the

TMS32020 User’s Guide

, literature number SPVU019.

transparency

Transparency is a mechanism that allows the surrounding pixels in an array to be specified as invisible. This is useful for ensuring that only the object and not the rectangle surrounding it are written to the display. The SMJ34020A provides four transparency modes:

No transparency

Transparency on result equal zero

Transparency on source equal COLOR0

Transparency on destination equal COLOR0

Refer to the

TMS34020 User’s Guide

for more information.

I/O registers

The SMJ34020A contains an on-chip block of sixty-four 16-bit locations (mapped into the SMJ34020A’s memory address space) that are used for I/O control registers. Eight of these are used by the host interface logic and are not available to the user. Forty-seven I/O registers control parameters necessary to configure the operation and report status of the following interfaces:

Host interface

Local memory

Video timing

Screen refresh

External interrupts

Internal interrupts

host interface registers

The host interface registers (HSTDATA, HSTADRL, HSTADRH, HSTCTLL, and HSTCTLH) are provided to facilitate communications between the SMJ34020A and a host processor and maintain compatibility with the SMJ34010. The registers are mapped into five of the I/O locations accessible to the SMJ34020A.

Two of these registers (HSTCTLL and HSTCTLH) are used to provide control by the host. This control consists of the passing of interrupt requests, flushing the instruction cache, halting the SMJ34020A, transmitting a non-maskable interrupt request to the SMJ34020A, enabling emulation interrupts, and setting host access modes and configurations.

The other three registers are simple read/write registers to allow the SMJ34020A software to leave addresses for the host at a known location and allow compatibility with some SMJ34010 software.

Page 11

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

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 SMJ34020A can be configured to drive composite sync or separate sync displays.

In composite sync mode, the SMJ34020A can be set to extract VSYNC

and HSYNC from an external CSYNC or it can be used to generate CSYNC from separate VSYNC and HSYNC inputs. Internally, the SMJ34020A can be set to preset the horizontal and vertical counts on receipt of an external sync signal. This allows compensation for any combination of internal and external delays that occur in the video synchronization process. The HCOUNT register is loaded from SETHCNT by an external HSYNC

, VCOUNT is loaded from

SETVCNT on an external VSYNC

, and an external CSYNC loads both HCOUNT and VCOUNT from SETHCNT

and SETVCNT, respectively. The SMJ34020A directly supports VRAMs by generating the serial-data-register transfer cycles necessary to

refresh the display . The memory locations from which the display information is taken, as well as the number of horizontal scan lines displayed between serial-data-register transfer cycles, are programmable.

The SMJ34020A supports various display resolutions and either interlaced or noninterlaced video. The SMJ34020A can optionally be programmed to synchronize to externally generated sync signals so that images created by the SMJ34020A can be superimposed upon images created externally . The external sync mode can also be used to synchronize the video signals generated by two or more SMJ34020As in a multiple-SMJ34020A graphics system.

CPU control registers

Five of the I/O registers (CONVDP, CONVMP, CONVSP, CONTROL, and PSIZE) provide CPU control to configure the SMJ34020A for operation with specific characteristics. These characteristics include pitches for pixel transfers, window checking mode, Boolean or arithmetic pixel processing operation, transparency mode, PIXBLT direction control, and pixel size.

interrupt interface registers

Two dedicated I/O registers (INTENB and INTPEND) monitor and mask interrupt requests to the SMJ34020A, including two externally generated interrupts and three internally generated interrupts. An internal interrupt request can be generated on one of the following conditions.

Window violation: an attempt has been made to write a pixel to a location inside or outside a specified window boundary.

Host interrupt: the host processor has set the interrupt request bit in the host control register.

Display interrupt: a specified horizontal line in the frame has been displayed on the screen.

Bus fault

Single-step emulator

A nonmaskable interrupt occurs when the host processor sets a control bit in the host interface register (NMI in HSTCTLH). The host-initiated interrupt is associated with a mode bit (NMIM in HSTCTLH) that enables and disables saving of the processor state on the stack when the interrupt occurs. This is useful if the host wishes to use the host interrupt before releasing the SMJ34020A to execute instructions (that is, before the stack pointer is initialized). A dedicated terminal controls the SMJ34020A reset function.

Page 12

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

memory controller/local-memory interface

The memory controller manages the SMJ34020A’s interface to the local memory and automatically performs the bit alignment and masking necessary to access data located at arbitrary bit boundaries within memory . The memory controller operates autonomously with respect to the CPU. It has a write queue one field (1 to 32 bits) deep that permits it to complete those memory cycles necessary to insert a field into memory without delaying the execution of subsequent instructions. Only when a second memory operation is required before completion of the first operation is the SMJ34020A forced to defer execution of the subsequent instruction.

The SMJ34020A directly interfaces to standard DRAMs and in particular, to standard video RAMs (VRAMs) such as the SMJ44C25x multiport VRAMs. The SMJ34020A memory interface consists of the local address/data bus (LAD), the DRAM row/column address (RCA) bus, and associated control signals. The currently selected word address (28 bits) and status (4 bits) are multiplexed with data on LAD. The RCA bus allows direct connection to address/address multiplexed DRAMs from 64K to 16M. Refresh for DRAMs is supported by CAS

-before-RAS (CBR) refresh cycles.

BIT 232 –1

(Last Bit in Memory)

68 Words

226 –66560 Words

(67 042 304 Words)

(3 × 226) –64K

(201 261 056 Words)

444 Words

65024 Words

512 Words

448 Words

64 Words

64K Words

Interrupt Vectors and Extended Trap V ectors

Reserved for Interrupt Vectors and Extended Trap Vectors

General Use and Extended Trap V ectors

General Use and Extended Trap Vectors

Bit 0 (First Bit in Memory)

General Use

Reserved for System I/O

Reserved for I/O Registers

I/O Registers

General Use

ADDRESS

FFFFFFF0h

FFFFFBC0h FFFFFBB0h

FFFFE000h

FFFFDFF0h

FFFF0000h

FFEFFFF0h

C0004000h C0003FF0h

C0002000h C0001FF0h

C0000400h C00003F0h

C0000000h

BFFFFFF0h

00100000h

000FFFF0h

00000000h

Figure 1. Memory Map

Page 13

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

reset

Reset puts the SMJ34020A into a known initial state. This state is entered when the input signal at RESET

asserted low. While RESET

remains asserted, all outputs are in a known state, no DRAM refresh cycles take

place, and no screen refresh cycles are performed. The state of the HCS

input on the CLKIN cycle before the low-to-high transition of RESET determines whether the SMJ34020A is halted or begins executing instructions. The SMJ34020A can be in one of two modes, host-present or self-bootstrap mode.

Host-present mode: if HCS

is high at the end of reset, SMJ34020A instruction execution halts and remains

halted until the host clears the HL T (halt) bit in HSTCTLH (host control register). Following reset, the RAS

cycles required to initialize the dynamic RAMs are performed automatically by the GSP memory control logic. The host can request a memory access after the eight RAS

initialization cycles have completed. The SMJ34020A automatically performs DRAM refresh cycles at regular intervals although the SMJ34020A remains halted until the host clears the HLT bit. Only then does SMJ34020A fetch the level-0 vector address from location FFFFFFE0h and begin executing the reset service routine.

Self-bootstrap mode: if HCS

is low at the end of reset, the SMJ34020A first performs eight refresh cycles to initialize the DRAMs. Immediately following the eight refresh cycles, the GSP fetches the level-0 vector address from location FFFFFFE0h and begins executing the reset service routine.

At the time the SMJ34020A fetches the level-0 vector address (the reset vector), the least significant four bits (bit address part) are used to load configuration data that establishes the initial condition of the big-endian/little-endian mode and the current RCA bus configuration bits in the CONFIG register as described in the I/O register section.

Unlike other interrupts and software traps, reset does not save the previous ST or PC values (this can also occur on host initiated nonmaskable interrupts if the NMIM bit in HSTCTLH is set to a 1) because the value of the stack pointer just before a reset is generally not valid. Saving these values on the stack could contaminate valid memory locations. A TRAP 0 instruction, which uses the same vector address as reset, similarly does not save the ST or PC values.

asserting reset

A reset is initiated by asserting RESET to its active-low level. T o reset the SMJ34020A at powerup, RESET must remain active low for a minimum of 40 local clock periods (LCLK1 and LCLK2) after power levels have become stable. At times other than powerup, the SMJ34020A can be reset by holding RESET

low for a minimum of four local clock periods; the GSP enters an internal reset state for 34 local clock cycles. While in the internal reset state and RESET

is high, memory-refresh cycles occur.

reset and multiprocessor synchronization

The synchronization of multiple SMJ34020As sharing a local memory is done using the RESET input. In systems where the multiprocessor interface is used to control the access to a common memory , the processors must be synchronized. Synchronization is achieved by taking RESET

high within a specific interval relative to

CLKIN. This can be done by using CLKIN to clock the RESET

as received by the SMJ34020As. All SMJ34020As

to be synchronized should use the same CLKIN and RESET

inputs. All of the local memory and bus control signals should be connected in parallel (without buffers) between the processors. After powerup, the processors are not necessarily synchronized with respect to the particular quarter cycle in progress. The rising edge of RESET

is used to set the SMJ34020A to a particular quarter cycle by adding Q1 cycles. All SMJ34020As in a multiprocessor environment operate on the same quarter cycle within 10 quarter cycles after the rising edge of RESET

Page 14

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

reset and DRAM/VRAM initialization

The SMJ34020A drives its RAS signal inactive high as long as RESET remains low. The specifications for certain DRAM and VRAM devices require that RAS

be driven inactive-high for 1 millisecond after power is stable

to provide the proper conditions for the DRAMs. Typically, eight RAS

cycles are also required to initialize the

DRAMs for proper operation. In general, holding RESET

low for t microseconds ensures that RAS remains high

initially for

t–(10tQ)

microseconds, tQ being the quarter-cycle time as defined by the input clock period, t

c(CHI)

The SMJ34020A memory controller automatically inserts the required eight RAS

cycles after all resets (after

powerup or after the internal reset state) by issuing CAS

-before-RAS refresh cycles before it allows the CPU access to memory . A host must delay requests to memory until the initialization cycles have had sufficient time to complete. Immediately following reset, the SMJ34020A is set to perform a refresh sequence every eight cycles.

At times other than powerup, to maintain the memory in DRAMs and do a reset, the RESET

pulse must not exceed the maximum refresh interval of the DRAMs minus the time for the SMJ34020A to refresh the memories. On reset, the SMJ34020A is set to do a refresh cycle every eight local clock periods. A 30-MHz (CLKIN) system with one (refresh) bank of D/VRAM would be completely refreshed in one sixteenth of the total memory refresh interval. The reset pulse then should not exceed about fifteen-sixteenths of the total refresh interval required by the DRAMs to maintain memory integrity.

If RESET

remains low longer than the maximum refresh interval specified for the memory , the previous contents

of the local memory can not be valid after the reset.

initial state following reset

While RESET is asserted low (or while in the internal reset state), the SMJ34020A’s output and bidirectional pins are forced to the states below.

INITIAL STATE OF PINS FOLLOWING A RESET (WITH GI LOW)

†

OUTPUTS DRIVEN HIGH OUTPUTS DRIVEN LOW

BIDIRECTIONALS DRIVEN TO

HIGH IMPEDANCE

RAS HRDY VSYNC CAS0–CAS3 CBLNK/VBLNK HSYNC WE DDIN CSYNC/HBLNK TR/QE LAD0–LAD31 DDOUT ALTCH HINT R0 R1 HOE HDST EMU3 RCA0–RCA12 SF

†

If GI is high, then all GI-controlled pins are high-impedance. GI-controlled pins are RAS, CAS0–CAS3, WE, TR/QE, DDOUT, DDIN, ALTCH, HOE

, HDST, RCA0–RCA12, LAD0–LAD31, and SF .

Immediately following reset, all I/O registers are cleared (set to 0000) with the exception of the HLT bit in the HSTCTLH register. The HLT bit is set to 1 if HCS

is high just prior to the low-to-high transition of RESET;

otherwise, it is set to 0.

Page 15

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

reset and DRAM/VRAM initialization (continued)

Just prior to the execution of the first instruction in the reset routine, the SMJ34020A’ s internal registers are in the following states:

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.

local memory and DRAM/VRAM interface

The SMJ34020A local memory interface consists of an address/data multiplexed bus on which addresses and data are transmitted. The associated control signals support memory widths of 16 or 32 bits, burst (page-mode) accesses, local memory-wait states, and optional external data bus buffers. The SMJ34020A DRAM /VRAM interface consists of an address/address multiplexed bus and the control signals to interface directly to both DRAMs and VRAMs. The local memory interface and the DRAM/VRAM interface are interrelated and, therefore, considered together for this description. At the beginning of a typical memory cycle, the address and status of the current cycle are output on LAD while the ROW address is output on the row/column address (RCA) bus. AL TCH

and RAS are used to latch the address/status and ROW address, respectively , on these two buses. LAD is then used to transfer data to or from the memory while the RCA bus is set to the column address for the memory. (NOTE: LAD31 is the most significant bit of the address or data).

543 0

Address

STS

LAD During the Address Cycle

Address — Memory address (select for 128M 32-bit long-words) W = 0 — Access to lower 16-bit word (even-addressed word or 32-bit boundary) W = 1 — Access to upper 16-bit word (odd-addressed word) STS — Bus cycle status code

Page 16

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

local memory and DRAM/VRAM interface (continued)

The address output on the row/column address (RCA) lines is determined by the row/column mode bits (RCM0 and RCM1 in the I/O registers CONFIG) and the state of column-address mode (CAMD) during each memory cycle. The CAMD is sampled on the internal Q4 clock phase, which allows CAMD to be generated by static logic wired to the local address/data (LAD) bus.

BASIC MEMORY ROW/COLUMN ACCESS MODES

RCM1 RCM VRAM

MODE

†

ADDRS‡BANKS

CAMD SUPPORT MATRICES

0 0 64K × N 8 16 64K × 16, 64K × 32, 256K × 16, 256K × 32, 1M × 16, 1M × 32 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

†

VRAM mode = basic size of VRAM addressing supported with CAMD = 0

‡

Addrs = number of RCA signals required to provide row/column addressing

Banks = number of possible interleaved 32-bit wide memory spaces

CAMD support = possible sizes and configurations of DRAMs that can be supported within the basic VRAM mode

The following figure illustrates the actual logical address bits output on each of the RCA lines during row and column intervals for each of the four VRAM modes and states of CAMD:

ROW TIME COLUMN TIME

CAMD = 0 CAMD = 1

RCA BIT 64K 256K 1M 4M 64K 256K 1M 4M

12 24 25 26 27 16 23 26 15 28 11 23 24 25 26 15 22 14 14 14 10 22 23 24 25 14 13 13 13 13

9 21 22 23 24 13 12 12 12 12 8 20 21 22 23 12 11 11 11 11 7 19 20 21 22 11 10 10 10 10 6 18 19 20 21 10 9 999 517 18 19 20 9 8 888 416 17 18 19 8 7 777 315 16 17 18 7 6 666 214 15 16 17 6 5 555 113 14 15 16 5 4 444 012 13 14 15 4 4 4 4 16

In the 64K mode with CAMD=0, any eight adjacent RCA0–RCA12 pins output 16 contiguous logical address bits. The eight most significant addresses are output during row-address time while the least significant addresses are output during column-address time. Logical addresses 12 through 16 are output twice during a memory cycle (during both RAS

and CAS falling edges) but at different pins. This allows a variety of VRAM memory organizations and decoding schemes to be used. When CAMD = 1, the addresses output during column-address time are changed such that a new logical address mapping occurs, allowing connection of RCA directly to 256K or 1M DRAMs.

Page 17

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

local memory and DRAM/VRAM interface (continued)

Similarly, for each of the other VRAM modes, direct connection is provided for other DRAM modes requiring larger matrices than the configuration mode. The following table gives examples of the connections using this feature.

CONNECTIONS TO RCA FOR CAMD = 1

RCA 64K

†

256K

†

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 0 16M × 32

†

NN is used for either 16-bit (× 16) or 32-bit (× 32) memory connections.

status codes

Status codes are output on LAD0–LAD3 at the time of the falling edge of ALTCH

and can be used to determine

the type of cycle that is being initiated. The following table lists the codes and their respective meanings.

CODE STATUS TYPE

0000 Coprocessor code 0001 Emulator operation OTHER 0010 Host cycle (00XX) 0011 DRAM refresh 0100 Video-generated DRAM serial register transfer 0101 CPU-generated VRAM serial register transfer VRAM 0110 Write mask load (01XX) 0111 Color latch load 1000 Data access 1001 Cache fill 1010 Instruction fetch 1011 Interrupt vector fetch CPU 1100 Bus locked operation (1XXX) 1101 Pixel operation 1110 Block write

1111 – RESERVED –

Page 18

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

dynamic bus sizing

The SMJ34020A supports dynamic bus sizing between 16 and 32 bits on any local memory access. Any port/memory that is only 16 bits wide must assert SIZE16

low during Q1 (to be valid at the start of Q2) of the bus cycle accessing the even memory word (LAD4 = 0) corresponding to its address.The SMJ34020A then performs another memory access to the next 16-bit (odd) word in memory . The SMJ34020A samples SIZE16 at the start of Q2 in the second cycle (access to odd word address) to determine to which half of LAD the port or memory is aligned. If the port is on LAD0–LAD15, SIZE16

should be low during the second cycle access (odd

word); otherwise, if the port is on LAD16–LAD31, SIZE16

must be high at this time. The SMJ34020A always performs two memory cycles to access the 16-bit wide memories, even when attempting only a 16-bit transfer.

The SMJ34020A outputs the four CAS

strobes and LAD bus initially aligned for a 32-bit bus. If the memory is

16 bits wide, the two most significant CAS

strobes are swapped with the two least significant strobes when it accesses the second word and the halves of LAD are also swapped; therefore, 16-bit memories need to respond only to the two CAS

strobes corresponding to the upper or lower 16 bits of LAD to which they are connected.

Note that devices connected to LAD0–LAD15 transfer the least significant word during the first cycle and the most significant word during the second cycle. Data accesses on LAD16–LAD31 transfer the most significant word first, then the least significant word.

The second memory cycle forced by SIZE16

is performed as a page mode access if PGMD was low during the first access. A read-write cycle to the 16-bit page-mode memory requires five bus cycles that occur as address, read0, read1, write0, write1. If a 16-bit transfer is interrupted due to a bus fault, the restart causes the entire access to be restarted.

For memory that supports page-mode accesses (PGMD

low), SIZE16 is sampled during each access to

memory . If SIZE16

is high on the even word access, then a 32-bit transfer occurs over LAD0–LAD31. If SIZE16 is low on the even word access (16-bit wide memory), then it is sampled again on the odd word access to determine to which half of LAD the memory is connected (low for connection to LAD0 – LAD15 or high for connection to LAD16–LAD31).

special 1-M VRAM cycles

The SMJ34020A provides control for special function VRAM cycles that are available in the 1-M devices. These cycles are obtained by the appropriate timing control of SF, CAS

, TR/QE, and WE of the VRAMs at the falling

edge of RAS

. The cycles include:

Load write mask

Load color mask

Block write (no mask)

Block write (current mask)

Write using mask

Alternate write transfer

In addition, other special modes can be implemented by using external logic.

multiprocessor arbitration

The multiprocessor interface allows multiple processors to operate in a system sharing the same local memory . The use of the bus grant in GI

and the priority request signals R0 and R1 allows a flexible method of passing control from one processor to another. The control scheme allows local memory cycles to occur back-to-back, even when passing control from one SMJ34020A to another. Synchronization of multiple SMJ34020As in a system occurs at reset with the rising edge of RESET

meeting the setup and hold requirements to CLKIN, so

all SMJ34020As are certain to respond to RESET

during the same quarter cycle. RESET is not required to be

synchronous to CLKIN except to allow synchronization of multiple SMJ34020As in a system.

Page 19

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

multiprocessor arbitration (continued)

The GI

priority for multiprocessing environments is determined by arbitration logic external to the SMJ34020A.

If GI

goes inactive-high, the SMJ34020A releases the bus on the next available cycle boundary . If the cycle in progress has not successfully completed, the SMJ34020A restarts the cycle upon regaining control of the bus. Normally , if the SMJ34020A asserts both R0

and R1 low, it should be given the control of the bus by the arbitrator .

host interface

The SMJ34020A host interface allows the local memory to be mapped into the host address space. The SMJ34020A acts as a DRAM controller for the host. The address for the host access is latched within the SMJ34020A; however, the data for the access is transferred using external transceivers. The host selects the address of a 32-bit long word for an access using the 27 host address lines, HA5–HA31. If the host desires byte addressability, it can select the active bytes for the access by using HBS0 –HBS3. The SMJ34020A always reads 32 bits from memory; however, on host writes, it uses the host byte selects to enable CAS0

–CAS3 to

memory . The address and byte selects are latched at the falling edge of HCS

within the SMJ34020A. The host

indicates a read or write by asserting HREAD

or HWRITE (as appropriate) either before or after HCS. (Note

that HREAD

and HWRITE must never be asserted at the same time.)

The SMJ34020A responds to a host read request by latching the requested data in the external latches and providing HRDY to the host, indicating that the read cycle is completing. The rising edge of HDST with HRDY high indicates data is latched in the external transceivers.

The host indicates that a write to a particular location is required by providing the address and asserting HWRITE

. The host must maintain both HCS and HWRITE asserted until valid data is in the transceivers. (The

rising edge of HOE

with HRDY high indicates that the data previously stored in the external transceivers has

been written to memory .) Typically, the rising edge of HWRITE

is used to strobe the data into the latches and signal the SMJ34020A that the write access can start. The SMJ34020A uses its byte-write capability to write only to the selected bytes.

The SMJ34020A always accesses the required location as latched at the falling edge of HCS

; however, in order to increase the data rate, a look ahead mechanism is implemented. The host increment enable (HINC) and host prefetch after write enable (HPFW) bits in the host control register (HSTCTLH) must be appropriately set to make optimum use of this feature. These bits provide four modes of operation as indicated in the following table:

HINC HPFW HOST ACCESS MODE DESCRIPTION

0 0 Random/Same No increment, no prefetch 0 1 Random/Same No increment, no prefetch 1 0 Block Increment after read or write, prefetch after read 1 1 Read-Modify-Write Increment after write, prefetch after write

When the SMJ34020A is programmed for block mode or read-modify-write accesses, the host can still do random accesses because the SMJ34020A always uses the address provided at the falling edge of HCS

; however, there is a prefetch to the next sequential address. The prefetch occurs after reads in block mode and after writes in read-modify-write mode. The SMJ34020A compares the address latched by HCS

on host reads to see if it is the same as that of the last prefetched data. If the addresses match, data is not re-accessed but HRDY is set high to indicate that the data is presently available.

dynamic bus sizing on host accesses

If the host makes a read access to a 16-bit wide memory , the SMJ34020A automatically does the second cycle required to read the rest of the 32-bit word (even if the host did not require a 32-bit cycle). The external logic must comprehend the sense of SIZE16

or the CAS strobes during the accesses in order to route the data into

the proper external host data transceivers. The SMJ34020A uses the host byte selects HBS0

–HBS3 to enable

the CAS

strobes when doing a host write.

Page 20

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

coprocessor interface

Support for coprocessors is provided through special instructions and bus cycles that allow communication with the coprocessor. A coprocessor can be register based, depending on the SMJ34020A to do all address calculations, or it can operate as its own bus controller, using the multiprocessor arbitration scheme. Five basic cycles are provided for direct communication and control of coprocessors:

SMJ34020A to coprocessor

Coprocessor to SMJ34020A

Move memory to coprocessor

Move coprocessor to memory

Coprocessor internal command

The first four of these cycles provide for command of the coprocessor in addition to the movement of parameters to and from the coprocessor. In this manner, parameters can be sent to the coprocessor and operated upon without an explicit coprocessor command cycle.

instruction set

The SMJ34020A instruction set can be divided into five categories:

Graphics instructions

Coprocessor instructions

Move instructions

General-purpose instructions

Program control and context switching

Specialized graphics instructions manipulate pixel data that is accessed using memory addresses or XY coordinates. These instructions include graphics operations, such as array and raster operations, pixel processing, windowing, plane masking, pixel masking, and transparency . Coprocessor instructions allow for the control and data flow to and from coprocessors that reside in the system. Move instructions comprehend the bit-addressing and field operations, which manipulate fields of data using linear addressing for transfer to and from memory and the register file. General-purpose instructions provide a complete set of arithmetic and Boolean operations on the register file as well as general program control and data processing. Program control and context

switching instructions allow the user to control flow and to save and restore information using

instructions with both register-direct and absolute operands.

clock stretch

The SMJ34020A supports a clock stretching mechanism, that is described below. With advances in semiconductor manufacturing, newer versions of the SMJ34020A can be made, each

supporting a higher CLKIN frequency . The increase in CLKIN frequency means that the SMJ34020A machine cycles execute more quickly, with a consequent increase in code execution speed. However, there comes a point when, as the machine cycle time becomes shorter, the local-memory control signals begin to violate DRAM and VRAM timing parameters for certain types of memory access.

The clock-stretch mechanism allows the SMJ34020A to slow down and execute those critical local-memory cycles while still benefiting from the accelerated processing allowed by higher CLKIN frequencies during noncritical memory access cycles.

Exact timing issues vary from system to system, reflecting differences in bus buffering, etc., but, broadly speaking, the clock-stretch mechanism allows the system designer to interface to slower memory devices than the designer could use if no stretch mechanism was available.

A normal, unstretched machine cycle consists of four quarter cycles, Q1, Q2, Q3, and Q4. A stretched cycle consists of five quarter cycles, Q1, Q2, Q3, Q4a, and Q4b.

Page 21

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

clock stretch (continued)

When clock-stretch mode is enabled, the fourth machine quarter cycle can be stretched to twice its original length. This stretching takes place only when the SMJ34020A attempts certain types of memory cycles.

Q4Q3Q2Q1Q4bQ4aQ3Q2Q1

Stretched Cycle

Normal Cycle

Q4Q3Q2Q1Q4Q3Q2Q1

Normal Sequence

Possible New Sequence

The stretch is achieved by holding the internal SMJ34020A clocks in the Q4 state for an extra quarter cycle so all of the device outputs remain unchanged during Q4a and Q4b. The SMJ34020A stretches only certain machine cycles so that the execution of code is not slowed unnecessarily.

enabling clock stretch

Clock-stretch mode is enabled and disabled using a bit in the CONFIG register memory mapped to location C00001A0h, see Figure 1.

01234567

C S E

Loaded at Reset from Reset Vector

Protected Byte

CONFIG register

CSE= 0: Disable stretch mode (normal operation) CSE= 1: Enable stretch mode

Bit 4 of the CONFIG register is the clock-stretch-enable mode bit. A zero in this bit disables stretch mode and a one in this bit enables stretch mode. The bit is cleared during reset; that is, stretch mode is disabled by default.

When stretch mode is enabled, the following machine cycles are stretched:

All address cycles of all memory-access sequences

Read data cycles in read-modify-write sequences

Notes: a) The host default cycle shown in the

TMS34020 User’s Guide

is not stretched because it is not a true

address cycle; that is, RAS

, etc., do not go low.

b) The CPU default cycle, which is similar to the host default cycle in that RAS

, etc., do not go low, is

also not stretched.

c) Clock-stretch mode disregards the page-mode input so that read data cycles in nonpage-mode

read-modify-write sequences are stretched even though there are no timing constraints that require a stretch.

d) All other memory subcycles are

not

stretched, even if the SMJ34020A is running with the CSE bit

set to 1.

The advantage of this implementation of clock-stretch mode is that the SMJ34020A can execute code at maximum speed, slowing down only during certain parts of memory access sequences.

It is important to remember that a stretched cycle is 25% longer than a normal cycle and that the SMJ34020A (with the exception of the video logic, which is clocked independently by VCLK) effectively slows down during such a stretched cycle.

Page 22

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

enabling clock stretch (continued)

The following are examples of stretch-mode memory operations.

two 32-bit nonpage-mode reads

READADDRREAD

ADDR

4321432143214321

READADDRREADADDR

43214431432443211

Stretch Stretch

Stretch Mode Enabled

Stretch Mode Disabled

one 32-bit page-mode read-modify-write

WRITEREADADDR

432143214321

WRITEREADADDR

4324432443211

Stretch Stretch

Stretch Mode Enabled

Stretch Mode Disabled

three 32-bit page-mode reads

READREADREADADDR

4321432143214321

READREADREADADDR

4321431432443211

Stretch

Stretch Mode Enabled

Stretch Mode Disabled

The stretched cycles are designed to accommodate worst-case 32-bit page-mode accesses, so during some nonpage-mode memory accesses stretches that are not essential can be generated. For example:

one 32-bit nonpage-mode read-write

WRITEADDRREADADDR

4321432143214321

WRITEADDRREADADDR

43214434432443211

Stretch Stretch

Stretch Mode Enabled

Stretch Mode Disabled

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.

Page 23

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

a clock-stretch timing example, SMJ34020A-32 and 150-ns DRAMs

This example analyzes a memory interface timing parameter. It shows that the clock-stretch mechanism can be used to allow the SMJ34020A-32 to avoid a timing violation when interfaced to 100-ns VRAMs.

Consider a system with:

A SMJ34020A-32, which has a 32-MHz clock input frequency and hence a 125-ns cycle time, so

= 31 ns. Timing parameters are taken from this data sheet.

A SMJ44C251-10 1 megabit × 1 bit DRAM. Timing parameters are taken from the corresponding

Texas Instruments data sheet.

row address hold data after RAS low, t

h(ADV-REL)

Without clock stretch

SMJ4C1024 t

h(RA)

Hold time, row address valid after RAS low Min = 20 ns

SMJ34020A Parameter 88 Hold time, row address valid after RAS

low Min = tQ – 5 ns = 26 ns

If RAS

is passed through a P AL with a delay of 7 ns, then t

h(RA)

seen by the DRAM is 26 ns – 7 ns = 19 ns.

This violates the 20 ns minimum.

With clock stretch

SMJ34020A

Parameter 88 t

h(ADV-REL)

Hold time, row address valid after RAS low Min = 2tQ – 5 ns = 57 ns

With the same 7-ns PAL delay, the DRAM sees t

h(RA)

as 57 ns – 7ns = 50 ns, which does not violate the

20 ns minimum.

cycle timing examples

The following figures show examples of many of the basic cycles that the SMJ34020A uses for memory access, VRAM control, multiprocessor bus control, and coprocessor communication. These figures should not be used to determine specific signal timings, but can be used to see signal relationships for the various cycles. The Q4 phases that

could

be stretched are marked with an * on the diagrams. The conditions required for the stretch

are:

The design uses a SMJ34020A.

The CONFIG register’s CSE bit is set to 1.

The SMJ34020A is doing either:

a) Any address cycle, or b) A read data cycle in a read-modify-write sequence

The following remarks apply to memory timing in general. A row address is output on RCA0–RCA12 at the start of a cycle along with the full address and status on LAD0–LAD31. These remain valid until after the fall of ALTCH and RAS. The column address is then output on RCA0–RCA12, and LAD0–LAD31 are set to read or write data for the memory access. During a write, the data and WE

are set valid prior to the falling edge of CAS; the data

remains valid until after WE

and CAS have returned high.

Large memory configurations can require external buffering of the address and data lines. DDIN and DDOUT coordinate these external buffers with LAD.

During the address output to LAD by the SMJ34020A (Figure 2), the least significant four bits (LAD0–LAD3) contain a bus-status code. PGMD low at the start of Q2 after RAS low indicates that this memory supports page-mode operation. LRDY high at the start of Q2 after RAS

low indicates that the cycle can continue without

inserting wait states. DDOUT

returns high after the initial address output on LAD (during Q4), indicating that

a memory read cycle is about to take place.

PAL is a trademark of Advanced Micro Devices, Inc.

Page 24

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

BUSFLT

(see Note B)

LRDY

(see Note B)

DDIN

RCA

CAMD

LCLK2

LCLK1

ALTCH

RAS

CAS

TR/QE

DDOUT

PGMD

(see Note B)

SIZE16

(see Note B)

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

Address

Row 1st Column

2nd Column

Address Subcycle

Data Transfer

Subcycle

Data Transfer

Subcycle

Data Data

Standard Memory Read Cycle

Page-Mode Read

LAD (SMJ34020A)

(see Note A)

LAD (Memory)

(see Note A)

†

See clock stretch, page 20.

NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A

LAD (memory): Output to LAD by the memory.

B. LRDY, PGMD

, SIZE16, and BUSFLT are not sampled on subsequent page-mode cycle accesses to

32-bit-wide memory space.

Figure 2. Local-Memory Read-Cycle Timing (with page mode)

Page 25

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

LRDY low at the start of the first Q2 after RAS low (Figure 3) indicates that the memory requires the addition of wait states. LRDY high at the next Q2 indicates the cycle can continue without inserting more wait states. PGMD

high at the start of Q2 where LRDY is sampled high indicates that this memory does not support

page-mode operation.

LCLCK1

LCLCK2

LAD (SMJ34020A)

LAD (Memory)

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

Address

Column

Address Subcycle Wait State Read Transfer

Data

Row

†

(see Note A)

(see Note B)

†

See clock stretch, page 20.

NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A

LAD (memory): Output to LAD by the memory.

B. Although they are not internally sampled, PGMD

and SIZE16 must be held at a valid level at the

start of each Q2 until LRDY is sampled high.

Figure 3. Local-Memory Read-Cycle Timing (without page mode, with one wait state)

Page 26

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

During the address output to LAD by the SMJ34020A (Figure 4), the least significant four bits (LAD0–LAD3) contain a bus-status code. PGMD

low at the start of Q2 after RAS low indicates that this memory supports

page-mode operation. LRDY high at the start of Q2 after RAS

low indicates that the cycle can continue without

inserting wait states. DDOUT

remains low after the initial address output on LAD (during Q4 after RAS goes low), indicating that a

memory write cycle is about to take place.

Page 27

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

LCLCK1

LCLCK2

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

Address

Row 1st Column 2nd Column

Address Subcycle

Data Transfer

Subcycle

Data Transfer

Subcycle

Data Out 1 Data Out 2

Standard Memory Write Cycle Page-Mode Write

†

(see Note A)

†

See clock stretch, page 20.

NOTE A: LRDY, PGMD

, SIZE16, and BUSFLT are not sampled on subsequent page-mode cycle

accesses to 32-bit-wide memory space.

Figure 4. Local-Memory Write-Cycle Timing (With Page Mode)

Page 28

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The local memory read-modify-write cycle (Figure 5) is used when inserting a field into memory that crosses byte boundaries. This cycle is actually performed as a read access followed by a page-mode write cycle.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Address

Row Column

Address Subcycle

Data Transfer

Subcycle

Data Transfer

Subcycle

Data

Data Out

Standard Memory Write Cycle Page-Mode Write

LCLCK1

LCLCK2

LAD (SMJ34020A)

LAD (Memory)

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

Q4 Q4

†

See clock stretch, page 20.

Figure 5. Local-Memory Read-Modify-Write-Cycle Timing

Page 29

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The refresh pseudo-address output to RCA0–RCA12 and LAD0– LAD31 comes from the 16-bit refresh address register (I/O register C000 01F0h) that is incremented after each refresh cycle (Figure 6). The 16 bits of address are placed on LAD16–LAD31; all other LAD bus lines are zero. The logical addresses on RCA0–RCA12 corresponding to LAD16–LAD31 also output the address from the refresh-address register.

Although PGMD

and SIZE16 are ignored during a refresh cycle, they should be held at valid levels. LRDY and

BUSFLT are not sampled until the start of the first Q2 cycle after RAS

has gone low.

If a refresh cycle is aborted due to a high-priority bus request (assuming LRDY is low at Q2 after RAS

low), a bus fault, or an external retry, then the count of refreshes pending is not decremented and the same pseudo-address is reissued when the refresh is restarted.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

Refresh Pseudo-Address

Refresh Psuedo-Address

Refresh Status Refresh End

CBR

LCLCK1

LCLCK2

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

R0 R1

†

See clock stretch, page 20.

Figure 6. Refresh-Cycle Timing

Page 30

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

When SIZE16 is selected low (Figure 7), the SMJ34020A performs a second cycle to read (or write) the remaining 16 bits of the word. Reads always access all 32 bits (all CAS

strobes are active). Internally, the

SMJ34020A latches both the high and the low words obtained on the first read cycle. The sense of SIZE16

on the second (odd-word) access is used to determine which half of the bus is to be sampled to replace the data word latched during the first cycle.

Page 31

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Row Column (S=0) Column (S=1)

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

Low Address

Address Subcycle

Data Transfer Data Transfer

Subcycle Subcycle

High Address

HiLow

LCLCK1

LCLCK2

LAD0–LAD15

LAD16–LAD31

CAMD

RCA

ALTCH

RAS

CAS0

CAS1

CAS2

CAS3

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

(see Note A)

†

See clock stretch, page 20.

NOTE A: RCA0 can be used to determine accesses to odd or even words because it outputs the least significant bit

of the word address during the column-address time (except in 4-M mode with CAMD = 1).

Figure 7. Dynamic Bus Sizing, Read-Cycle Timing

Page 32

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Write accesses to 16-bit memory are performed by swapping the data on upper and lower words of LAD and exchanging data on CAS0

and CAS1 for data on CAS2 and CAS3, respectively (Figure 8). During the first cycle,

data is placed on LAD0 – LAD31 as in a normal write. The sampling of SIZE16

low during the first access indicates that this is 16-bit-wide memory , so the SMJ34020A swaps data on the upper and lower halves of LAD. Notice that during the first cycle, CAS0

is inactive (because this byte was not selected), and during the second

cycle, CAS2

is inactive due to the exchange of CAS0 for CAS2 and CAS1 for CAS3.

Page 33

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Row Column (S=0) Column (S=1)

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

Low Address

Address Subcycle

Data Transfer Data Transfer

Subcycle Subcycle

High Address

Data Low

Data High

Data Low

LCLCK1

LCLCK2

LAD0–LAD15

LAD16–LAD31

CAMD

RCA

ALTCH

RAS

CAS0

CAS1

CAS2

CAS3

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 8. Dynamic Bus Sizing, Write-Cycle Timing

Page 34

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Clock stretch is a special 1-megabit VRAM control cycle that is executed when VEN in the CONFIG I/O register is set and PMASKL and/or PMASKH are written (Figure 9). This cycle is indicated by CAS

, WE, TR/QE, and

SF high at the falling edge of RAS

and SF low at the falling edge of CAS. As the plane mask is copied to the PMASK register(s), it is also output on LAD to be written to a special register on the VRAM that is used in subsequent cycles requiring a write mask. During the address portion of the cycle, the status on LAD0–LAD3 indicates a write-mask load is being performed (status code = 01 10). Although CAMD, PGMD

, and SIZE16 are

ignored on this cycle, they should be held at valid levels as shown.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q1

PMASK Row

PMASK Address PMASK Data Zero Address

Q2 Q3 Q4 Q1

Not PMASK Data

PMASK Column All-Zero Address

Write to the PMASK I/O Register

Load-Write-Mask Cycle

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 9. Load-Write-Mask-Cycle Timing

Page 35

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The clock stretch is generated by the VLCOL instruction and is indicated by CAS, WE, TR/QE, and SF high at the falling edge of RAS

and SF high at the falling edge of CAS (Figure 10). The data in the COLOR1 register is output on LAD to be written to a special register on the VRAM that is used in subsequent cycles requiring a color latch. During the address portion of the cycle, the status on LAD0–LAD3 indicates a color-mask load is being performed (status code = 0111). Although CAMD, PGMD

, and SIZE16 are ignored on this cycle, they

should be held at valid levels as shown.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

All-Zero Address

Zero Address

Color Register Data

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 10. Load-Color-Latch-Cycle Timing

Page 36

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The clock stretch is also performed when a VBL T or VFILL instruction is executed and PMASKL and PMASKH are set to zero (Figure 1 1). 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 = 1 110). SIZE16

can be used with this cycle, but external multiplex logic is required to

map the data correctly to appropriate memories.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row

Address Data Out 1

Q2 Q3 Q4 Q1

Data Out 2

1st Column 2nd Column

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 11. Block-Write-Cycle Timing (Without Mask)

Page 37

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The clock stretch is also performed when a VBL T or VFILL instruction is executed and PMASKL and PMASKH are set to nonzero values (Figure 12). It is indicated by CAS

, TR/QE, and SF high and WE low at the falling edge

of RAS

and by SF high at the falling edge of CAS. The data on LAD is used as an address mask, and the data stored in the color latch is written to the VRAM, just as in the block-write cycle without mask, except that the data in the write mask is used to enable the bits from the color latch that are written to memory . This cycle allows up to 16 bits to be written into each VRAM (four adjacent nibbles, each set to the value in the color latch as enabled by the write mask) for a total of 128 bits. During the address portion of the cycle, the status on LAD0 – LAD3 indicates a block write is being performed (status code = 1110). SIZE16

can be used with this

cycle, but external multiplex logic is required to map the data correctly to appropriate memories.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row

Address Data Out 1

Q2 Q3 Q4 Q1

Data Out 2

1st Column 2nd Column

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 12. Block-Write-Cycle Timing (With Mask)

Page 38

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

As a special 1-megabit VRAM control cycle, the clock strech is also performed when the PMASKL and PMASKH registers are set to nonzero values, CST in DPYCTL is cleared, VEN in CONFIG is set, and the byte-aligned pixel-write instruction is executed (Figure 13). This cycle is indicated by CAS

, TR/QE, and SF high and WE low

at the falling edge of RAS

and by SF low at the falling edge of CAS. The data on LAD is written to memory just as a normal DRAM write except that data in the write mask is used to enable DQs that are written to memory . During the address portion of the cycle, the status on LAD0 –LAD3 indicates that a pixel operation is being performed (status code = 1101).

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row

Address Data Out 1

Q2 Q3 Q4 Q1

Data Out 2

1st Column 2nd Column

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 13. Write-Cycle Timing Using Mask

Page 39

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The VRAM cycle shown in Figure 14 is issued in any of three ways:

Pixel operation instruction with CST in DPYCTL set

Horizontal blank reload cycle requested by the video-control logic with VCE in DPYCTL cleared

Video timeout due to SCOUNT match with the value in MLRNXT and VCE and SSV in DPYCTL cleared

This cycle is indicated by TR

/QE and SF low and CAS and WE high at the time RAS goes low. The timing of

the low-to-high transition of TR

/QE is dependent upon the timing of SCLK when doing a midline reload cycle. During the address portion of the cycle, the status on LAD0 –LAD3 indicates either a video-initiated VRAM memory-to-register transfer (status code = 0100), or a CPU-initiated VRAM memory-to-register transfer (status code = 0101).

Page 40

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

Row

Address

Tap Point

Address

Subcycle

Wait

State

Cycle

Completion

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 14. Memory-to-Serial-Data-Register-Cycle Timing (VRAM Read Transfer)

Page 41

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

This VRAM cycle shown in Figure 15 is performed when a video timeout occurs due to a match of the MLRNXT register, VCE in DPYCTL is cleared, and SSV in DPYCTL is set. This cycle is indicated by TR

/QE low and CAS,

SF , and WE

high at the time RAS goes low. The timing of the low-to-high transition of TR/QE is not dependent upon the timing of SCLK because there is not as great a timing constraint to position the cycle as in midline reload. During the address portion of the cycle, the status on LAD0–LAD3 indicates a video-initiated VRAM memory-to-register transfer (status code = 0100). Although PGMD

and SIZE16 are ignored on this cycle, they

should be held at valid levels as shown.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row

Address

Tap Point

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 15. Memory-to-Split-Serial-Data-Register-Cycle Timing (VRAM Split-Register Read Transfer)

Page 42

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Figure 16 shows the VRAM cycle performed when a horizontal blank reload is requested by the video-control logic and VCE and SRE in DPYCTL are both set. This cycle is indicated by TR

/QE, WE and SF low and CAS high at the time RAS goes low. The SOE pin of the VRAMs is used to select between write transfer and pseudo-write transfer cycles (SOE

must be generated by logic external to the SMJ34020A). During the address portion of the cycle, the status on LAD0 – LAD3 indicates that a video-initiated VRAM register-to-memory transfer (status code = 0100) is being performed. Although PGMD

and SIZE16 are ignored on this cycle, they

should be held at valid levels as shown.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row

Address

Tap Point

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 16. Serial-Data-Register-to-Memory-Cycle Timing (VRAM-Write Transfer, Pseudo-Write Transfer)

Page 43

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

This VRAM cycle (Figure 17) is performed when a pixel-write instruction is executed with the CST bit in DPYCTL set. This cycle is indicated by TR

/QE and WE low and SF and CAS high at the time RAS goes low. This cycle

does not require the use of SOE

of the VRAM and does not affect the status of the serial I/O pins. During the address portion of the cycle, the status on LAD0–LAD3 indicates that a CPU-initiated VRAM register-to-memory transfer (status code = 0101) is being performed. Although PGMD

and SIZE16 are ignored

on this cycle, they should be held at valid levels as shown.

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row

Address

Tap Point

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

PGMD

SIZE16

BUSFLT

†

See clock stretch, page 20.

Figure 17. Serial-Data-Register-to-Memory Cycle Timing (VRAM-Alternate-Write Transfer)

Page 44

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

In Figure 18, transition points are shown for R0 and R1 to indicate where they occur relative to the other signals. This example indicates that the SMJ34020A has control of the bus, yields control, and then regains control. The

SMJ34020A regains bus mastership as soon as GI

is driven active (low). R0 and R1 could be outputting any of the codes with the exception of the access-termination code. The bus arbitration logic must control the timing of GI

to all of the processors requiring the bus.

It is recommended that SMJ34020A clock stretch not be used in multiprocessor systems.

Page 45

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4

Row

Hi-Z

Host-Default

Cycle

Beginning

of Cycle

LCLK1

LCLK2

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

SIZE16

PGMD

BUSFLT

HOE

HDST

Figure 18. Multiprocessor-Interface-Cycle Timing (High-Impedance Signals)

Page 46

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Two SMJ34020As use the multiprocessor interface to pass control of local memory from one to the other (Figure 19). GSP1 completes a read cycle to the local memory and, although desiring another read, loses the bus to GSP2, which does a single write cycle (perhaps a host-write access). GSP1 then regains control and completes the read cycle (shown with a single wait state). Since no further memory-access requests are present, GSP1 maintains control of the bus and holds all of the local-memory control signals at their inactive levels. LRDY is a common input to both GSP1 and GSP2.

The host cycle timing diagrams shown in this data sheet are only a sample. For more information, see the

TMS34020 User’s Guide

123412 34 12341234 12341234123412 34 1

GSP1 Read GSP2 Write GSP1 Read With Wait Bus Idle

GSP1

RAS

CAS

TR/QE

DDIN

DDOUT

ALTCH

LRDY

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

GSP

Figure 19. Multiprocessor-Interface Cycle Timing (Passing Control)

Page 47

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

In Figure 20, the host-access request is synchronized to the SMJ34020A at the beginning of Q4 so that the local-memory cycle can begin in Q1. If the external host-access request occurs after the setup time requirement before Q4, the request is not considered until the next Q4 cycle. In order to provide back-to-back accesses as indicated in this example, the host must remove HCS

on receipt of HRDY and reassert it before Q4 (it can also

remove and reassert HREAD

with HCS).

Page 48

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row Column

HOE

HDST

LAD

CAMD

RCA

ALACH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

SIZE16

PGMD

BUSERR

Previous Read Valid

Local-Memory Host Read Cycle

HA/HBS

HCS

HREAD

HWRITE

HRDY

DATA

(out)

†

(see Note A)

†

See clock stretch, page 20.

NOTE A: HRDY goes high at the start of Q2; however, data is not strobed into

the external latches until the start of Q4 when HDST goes high.

Figure 20. Host-Read-Cycle Timing (Random/Same Accesses, not From SMJ34020A I/O Registers)

Page 49

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The host-access request is synchronized to the SMJ34020A at the beginning of Q4 so that the local memory cycle can begin in Q1.

In block mode (prefetch after read), the SMJ34020A automatically initiates sequential read accesses as soon as the host deasserts the current read request. In this example, the host reads a location and must wait for the first access to complete. When the host removes HREAD

(Figure 21), indicating the end of the first read, the SMJ34020A starts to prefetch the next sequential location. When the host makes the next request, the SMJ34020A has prefetched the data so that the host reads with no delay . While in block mode, the SMJ34020A continues to prefetch data for the host read each time the host removes either HREAD

or HCS. If the address

present and latched at the falling edge of HCS

matches the previously prefetched address, HRDY is asserted

high so that the host can read with no delay. In read-modify-write mode (prefetch after write), the SMJ34020A initiates the read access as soon as the current

write request is deasserted.

Page 50

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

†

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row Column

Previous Read 1st Read Valid

1st Address

Row Column

2nd Address

Q2 Q3 Q1 Q2 Q3 Q4 Q1

2nd

Local-Memory Host Read Cycle Local-Memory Host Prefetch Cycle

HA/HBS

HCS

HREAD

HWRITE

HRDY

HOE

HDST

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

SIZE16

PGMD

BUSERR

DATA

(out)

†

See clock stretch, page 20.

Figure 21. Back-to-Back Host-Read Cycles With Implicit Addressing; HREAD as Strobe

Page 51

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The host read of the SMJ34020A I/O registers (Figure 22) suppresses the generation of TR/QE and CAS so that data is read from the SMJ34020A rather than from memory. DDOUT

is enabled so that data can flow through external buffers on LAD to the host data latches. The SMJ34020A I/O registers can be accessed in any of the host access modes (random/same, block, or read-modify-write).

Page 52

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row Column

Previous Read Valid

I/O Data

Local-Memory Host Read I/O Cycle

HA/HBS

HCS

HREAD

HWRITE

HRDY

HOE

HDST

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

SIZE16

PGMD

BUSERR

DATA

(out)

†

See clock stretch, page 20.

Figure 22. Host-Read Cycle Timing From SMJ34020A I/O Registers

Page 53

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

In Figure 23, SMJ34020A provides HRDY as soon as it recognizes the host write cycle (if no other host write cycle is in progress), allowing the host to latch the data in the external data latches. The host then attempts a second write but does not get an immediate HRDY because the SMJ34020A is still writing the first data to memory . As soon as the memory write completes, HRDY goes high so that the host can latch the new data. The SMJ34020A then writes the second data while the host continues other processing. The host access request is synchronized to the SMJ34020A at the beginning of Q4 so that the local memory cycle can begin in Q1. If the external host access request occurs after the setup time requirement before Q4, the request is not considered until the next Q4 cycle. During a host write cycle DDIN is active so that if the write is to the SMJ34020A I/O registers, the data can be required within the GSP.

Page 54

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Q4 Q1 Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row Column

1st Write Valid

Local-Memory Host Write Cycle

Q2 Q3 Q1 Q2 Q3 Q4 Q1

Row Column

1st Address 2nd Address

Local-Memory Host Prefetch

ValidPrevious Read

HA/HBS

HCS

HREAD

HWRITE

HRDY

HOE

HDST

LAD

CAMD

RCA

ALTCH

RAS

CAS

TR/QE

DDIN

DDOUT

LRDY

SIZE16

PGMD

BUSERR

DATA

(out)

DATA

(in)

†

(see Note A)

†

See clock stretch, page 20.

NOTE A: HRDY goes high at the start of Q2; however , the memory cycle writing data to memory is not completed

until the start of Q4 when ALTCH

, CAS, and HOE return high. The host must not strobe new data into

the external latch until just after the start of Q4.

Figure 23. Host-Write Cycle Back-to-Back With Prefetch of Next Word and Implicit Addressing; HREAD

and HWRITE Used as Strobes

Page 55

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Although RESET is not

normally

required to be synchronous to CLKIN, in order to facilitate synchronization of

multiple SMJ34020As in a system, the rising edge of RESET

must meet the setup and hold requirements to

CLKIN so that all GSPs are certain to respond to the RESET

on the same quarter cycle (Figure 24). The four

possible conditions for the state of the SMJ34020A at the time RESET

goes high are shown below. Quarter cycle 1 is extended accordingly to provide synchronization of the GSPs. All SMJ34020As to be synchronized must share a common CLKIN and RESET

. Within 10 CLKIN cycles after RESET goes high, all GSPs are

synchronized to the same quarter cycle through the extension of Q1 cycles.

It is recommended that SMJ34020A stretch mode not be used in multiprocessor systems.

CLKIN

RESET

Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q1 Q1 Q2

Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q1 Q1 Q1 Q2

Q4 Q1 Q2 Q3 Q4 Q1 Q1 Q1 Q1 Q1 Q2

LCLK1

LCLK2

LCLK1

LCLK2

LCLK1

LCLK2

LCLK1

LCLK2

Case 1

Case 2

Case 3

Case 4

Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q1 Q2Q1

NOTE A: No timing dependencies of LCLK1 and LCLK2 relative to CLKIN or RESET

are to be implied from this figure.

Figure 24. Synchronization of Multiple SMJ34020As

Page 56

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The timing example in Figure 25 is like a memory write cycle except that RAS and SF are high.

LCLK1

LCLK2

ALTCH

RAS

Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

LAD (TMS34020A)

Command

Q2 Q3 Q4 Q1

CAMD

RCA

CAS

TR/QE

DDIN

DDOUT

Operand 1 Operand 2

PGMD

SIZE16

LRDY

BUSFLT

Command Data Transfer Data Transfer

(see Note A)

NOTE A: LAD (SMJ34020A): Output to LAD by the SMJ34020A

Command: Coprocessor ID, instruction and status code present on LAD Operand n: Data to or from the coprocessor

Figure 25. Transfer SMJ34020A Register(s) to Coprocessor (One or Two 32-Bit Values)

Page 57

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The timing example in Figure 26 is like a memory write cycle except that RAS and SF are high.

LCLK1

LCLK2

ALTCH

RAS

Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

LAD (TMS34020A)

Command

Q2 Q3 Q4 Q1

LAD (Coprocessor)

CAMD

CAS

TR/QE

DDIN

DDOUT

Operand 1 Operand 2

PGMD

SIZE16

LRDY

BUSFLT

Command Data Transfer Data Transfer

RCA

(see Note A)

NOTE A: LAD (SMJ34020A): Output to LAD by the SMJ34020A

LAD (coprocessor): Output to LAD by the coprocessor Command: Coprocessor ID, instruction and status code present on LAD Operand n: Data to or from the coprocessor

Figure 26. Transfer-Coprocessor Register to SMJ34020A (One or Two 32-Bit Values)

Page 58

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Data transfer from memory to a coprocessor requires an initialization cycle to inform the coprocessor what is to be transferred and then a memory cycle to perform the actual transfer (Figure 27). The coprocessor can place status information on LAD during the initialization cycle for the SMJ34020A. Two types of memory-to-coprocessor instructions are supported: one provides a count (from 1 to 32) of data to be moved in the instruction; the other specifies a register in the SMJ34020A to be used for the count. Both instructions specify a register to be used as an index into memory . The index can be postincremented or predecremented on each transfer cycle.

LAD

Command Address

ALTCH

RAS

Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1R1Q2 Q3 Q1

CAMD

CAS

TR /QE

DDIN

DDOUT

Data 1

PGMD

SIZE16

LRDY

BUSFLT

RCA

LAD

Data 2

2nd Column1st ColumnRow

Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

LCLK1

LCLK2

(TMS34020A)

(memory)

Command Cycle

Address

Data Tansfer

Data Transfer

†

(see Note A)

(see Note B)

†

See clock stretch, page 20.

NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A

LAD (memory): Output to LAD by the memory Command: Coprocessor ID, instruction and status code present on LAD Address: Memory address for the data transfer with coprocessor status code Data n: 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)

Page 59

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

Data transfer from a coprocessor to memory requires an initialization cycle to inform the coprocessor what is to be transferred and then a memory cycle to perform the actual transfer (Figure 28). The coprocessor can place status information on LAD during the initialization cycle for the SMJ34020A. The memory cycle includes a dead cycle to enable the SMJ34020A to take LAD drivers to the high-impedance state before the coprocessor activates its LAD bus drivers to the memory . T wo types of memory-to-coprocessor instructions are supported. Both provide a count (from 1 to 32) of data to be moved in the instruction. Both also specify a register to be used as an index into memory. One uses this index register with a postincrement and the other uses it with a predecrement after each transfer cycle.

LCLK1

LCLK2

CAMD

RCA

LAD

Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

LAD

Command

Q2 Q3 Q1

DDIN

Data 1

LRDY

BUSFLT

Address

Data 2

2nd Column1st ColumnRow

Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4

(TMS34020A)

(Coprocessor)

Status

ALTCH

RAS

CAS

TR /QE

DDOUT

PGMD

SIZE16

(see Note B)

Command Cycle

Address

Data Transfer

Spacer

†

(see Note A)

†

See clock stretch, page 20.

NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A

LAD (coprocessor): Output to LAD by the coprocessor Command: Coprocessor ID, instruction and status code present on LAD Address: Memory address for the data transfer, with coprocessor status code Data n: Data from the coprocessor (number of values transferred depends on a count in the instruction) Status: Optional coprocessor status register output to LAD bus

B. All coprocessor cycles are implemented as 32-bit operations; therefore, SIZE16

should be high during these cycles.

Figure 28. Transfer-Coprocessor Register(s) to Memory (ALTCH High During Data Transfer)

Page 60

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

The timing example in Figure 29 is like a memory write cycle except that RAS and SF are high. A coprocessor internal command assumes no transfer of operands or results but causes the coprocessor to

execute some internal function. The coprocessor can place status information on LAD during the cycle for the SMJ34020A.

Page 61

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

cycle timing examples (continued)

ALTCH

(see Note B)

RAS

Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

LAD

Command

CAMD

CAS

TR/QE

DDIN

DDOUT

PGMD

(see Note B)

SIZE16

(see Note C)

LRDY

BUSFLT

RCA

LCLK1

LCLK2

(TMS34020A)

Command Cycle

(see Note A)

NOTES: A. LAD (SMJ34020A): Output to LAD by the SMJ34020A

LAD command: Coprocessor ID, instruction and status code present on LAD

B. Although the coprocessor internal command never requires the

use of page mode cycles, PGMD

should be held at a valid level

during the start of Q2 after AL TCH

has gone low.

C. All coprocessor cycles are implemented as 32-bit operations;

therefore, SIZE16

should be high during these cycles.

Figure 29. Coprocessor-Internal Operation Command Cycle

Page 62

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

absolute maximum ratings over operating free-air temperature range

†

Maximum supply voltage, V

(see Note 1) 7 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Input voltage range – 0.3 V to 7 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Off-state output voltage range – 2 V to 7 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Operating termperature range, T

– 55°C to 125°C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Storage temperature range, T

stg

–65°C to 150°C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

†

Stresses beyond those listed under “absolute maximum ratings” may cause permanent damage to the device. These are stress ratings only, and functional operation of the device at these or any other conditions beyond those indicated under “recommended operating conditions” is not implied. Exposure to absolute-maximum-rated conditions for extended periods may affect device reliability.

NOTE 1: All voltage values are with respect to VSS.

recommended operating conditions

MIN NOM MAX UNIT

’34020A-32 4.5 5 5.5

VCCSuppl

y v

oltage

’34020A-40 4.75 5 5.25

Supply voltage (see Note 2) 0 V

High-level output current 400 µA

Low-level output current 2 mA

Operating case temperature 125 °C

Operating temperature – 55 125 °C

NOTE 2: Ensure that a minimum inductance path between the VSS pins and system ground is provided to minimize noise on VSS.

Page 63

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

DC electrical characteristics over recommended range of supply voltage (see Note 3)

PARAMETER TEST CONDITIONS MIN TYP

†

MAX UNIT

BUSFLT, LRDY , VCLK,

GB PKG 2.2 VCC+0.3

PGMD, SIZE16, CSYNC

VSYNC

, HSYNC

HT PKG 2.3 VCC+0.3 GB PKG 2 VCC+0.3

High-level input

HWRITE, HREAD

HT PKG 2.3 VCC+0.3

voltage

HA5–HA31, HCS,

GB PKG 2 VCC+0.3

HBS0–HBS3

HT PKG 2.3 VCC+0.3 CLKIN only 3 VCC+0.3 All other inputs 2 VCC+0.3

Low-level input voltage, HT only: HCS VIL = – 0.3 min, 0.7 V max –0.3 0.8 V

High-level output voltage

VCC = MIN, IOH = MAX

2.6 V

GB PKG 0.60

Low-level output

DDIN, HINT, HRDY, R0, R1, EMU3

VCC = MAX,

0.8 V

voltage

HYSNC, VSYNC

HT PKG

MIN

0.8

All other outputs 0.6

GB PKG

= MAX,

HT PKG

VO = 2.8 V

IOOutput current, leakage (high impedance)

GB PKG

= MAX,

– 20

HT PKG

VO = 0.6 V

–20

Input current (All inputs except EMU0–EMU2, HREAD

, HWRITE‡)

VI = VSS to V

±20 µA

’34020A-32

= MAX,

265

ICCSupply current

’34020A-40

Freq = MAX

280

Input capacitance 10 18 pF

Output capacitance 18 25 pF

†

All typical values are at VCC = 5 V, TA = 25° C.

‡

EMU0–EMU2 are not connected in a typical configuration. Nominal pullup current for EMU0–EMU2 and HREAD

, HWRITE is 600 µA.

NOTE 3: HDST and HOE

(output terminals) have internal pullup resistors that allow high logic levels to be maintained when the SMJ34020A is

not actually driving these pins.

signal transition levels

2 V

(see Note A)

0.8 V

NOTE A: 2.2 V for BUSFLT, VCLK, LRDY, PGMD

, SIZE16. 3V for CLKIN.

Figure 30. TTL-Level Inputs

For high-to-low and low-to-high transitions, the level at which the input timing is measured is 1.5 V.

Page 64

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

signal transition levels (continued)

2.6 V 2 V

1 V

0.6 V

1.5 V

Figure 31. TTL-Level Outputs

TTL-level outputs are driven to a minimum logic-high level of 2.6 V and to a maximum logic-low level of 0.6 V . For a high-to-low transition on a TTL-compatible output signal, the level at which the output is said to be no longer high is 2 V , and the level at which the output is said to be low is 1 V . For a low-to-high transition, the level at which the output is said to be no longer low is 1 V , and the level at which the output is said to be high is 2 V. A V

trip level of 1.5 V is used for timing requirements for testing at – 55°C.

test measurement

The test load circuit shown in Figure 32 represents the programmable load of the tester pin electronics that is used to verify timing parameters of SMJ34020A output signals.

LOAD

Test

Point

From Output

Under Test

Where: IOL = 2 mA (all outputs)

IOH = 400 µA (all outputs) V

LOAD

= 1.5 V

LOAD

= 80 pF typical load circuit capacitance

NOTE: The load applied may be set higher than the values

indicated for IOL and IOH during timing tests in order to reduce signal bounce induced by the tester hardware. However the timing performance is assured at the stated load values.

Figure 32. Test Load Circuit

Page 65

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

timing parameter symbology

Timing parameter symbols used herein were created in accordance with JEDEC Standard 100. In order to shorten the symbols, some of the pin names and other related terminology have been abbreviated as follows:

A HA5–HA31 and HBS0

–HBS3 LINT LINT1, LINT2 AD LAD0–LAD31 and RCA0–RCA12 OE HOE AL AL TCH RC RCA0–RCA12 BC Any of the bus control input signals

(LRDY, PGMD, SIZE16, or BUSFLT)

RD HREAD

CE CAS0–CAS3 RE RAS CK LCLK1 and LCLK2 RQ R0 or R1 CK1 LCLK1 RS RESET CK2 LCLK2 RY HRDY CKI CLKIN S HSYNC

, VSYNC, or CSYNC CM CAMD SC EMU3 CS HCS

SCK SCLK

CT Any of the bus control output signals

(ALTCH, CAS0–CAS3, RAS, WE, TR

/QE, HOE, or HDST)

SF SF

DI DDIN SG Any output signal DO DDOUT

SGV Signal valid EM EMU0, EMU1, EMU2 ST HDST HI HINT

TR TR/QE HS HSYNC, VSYNC, CSYNC/HBLNK, or

CBLNK

/VBLNK

VCK VCLK

GI GI

WR HWRITE LA LAD0–LAD31

Lowercase subscripts and their meaning are: a access time

c cycle time (period) d delay time h hold time su setup time t transition time w pulse duration (width)

The following letters and symbols and their meaning are: H High level

L Low level V Valid level X Unknown, changing or don’t care level Z High-impedance state of 3-state output

Page 66

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

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 (t

) that appears in the following tables is one quarter of a local output clock period or

equal to the input clock period, t

c(CKI)

All output signals from the SMJ34020A are derived from an internal clock such that all output transitions for a given quarter cycle occur with a minimum of skewing relative to each other. In the timing diagrams, the transitions of all output signals are shown with respect to the local clocks (LCLK1 and LCLK2). The local clock edge used as a reference occurs one internal clock cycle before the transition specified.

The signal combinations shown in the timing parameters are for timing reference only; they do not necessarily represent actual cycles. For actual cycle descriptions, please refer to the cycle timings section of this specification.

Page 67

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

CLKIN and RESET timing requirements (see Figure 33)

SMJ34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

1 t

c(CKI)

Cycle time, period of CLKIN (4tQ) 31.25 50 25 50 ns

2 t

w(CKIH)

Pulse duration, CLKIN high 10 8 ns

3 t

w(CKIL)

Pulse duration, CLKIN low 10 8 ns

4 t

t(CKI)

Transition time, CLKIN 2

†

5 t

h(CKI-RSL)

Hold time, RESET low after CLKIN high 15

‡

6 t

su(RSH-CKI)

Setup time, RESET high to CLKIN no longer low

‡

Pulse duration,

Initial reset during powerup 160tQ–40

160tQ–40

w(RSL)

RESET low

Reset during active operation

16tQ–40

8 t

su(CSL-RSH)

Setup time, HCS low to RESET high to configure self-bootstrap mode

8tQ+55 8tQ+55 ns

9 t

d(CSH-RSH)

Delay time, HCS no longer low to RESET high to configure self-bootstrap mode

4tQ–50

4tQ–50¶ns

10 t

w(CSL)

Pulse duration, HCS low to configure GSP in self-bootstrap mode

4tQ+55 4tQ+55 ns

†

These values are based on computer simulation and are not tested.

‡

These timings are required only to synchronize the SMJ34020A to a particular quarter cycle.

The initial reset pulse on powerup must remain valid until all internal states have been initialized. Resets applied after the SMJ34020A has been initialized need to be present only long enough to be recognized by the internal logic; the internal logic maintains an internal reset until all internal states have been initialized (34 LCLK1 cycles).

Parameter 9 is the maximum amount by which the RESET

low-to-high transition can be delayed after the start of the HCS low-to-high transition

and still assure that the SMJ34020A is configured to run in the self-bootstrap mode (HLT bit = 0) following the end of reset.

CLKIN

RESET

HCS

Figure 33. CLKIN and RESET Timing Requirements

Page 68

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

local-bus timing: output clocks (see Note 4 and Figure 34)

SMJ34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

11 t

c(LCK)

Cycle time, period of local clocks LCLK1, LCLK2 4t

c(CKI)

†

c(CKI)

†

12 t

w(LCKH)

Pulse duration, local clock high 2tQ–15 2tQ–13.5 ns

12a t

w(LCKH)

Pulse duration, LCLK1 high (see Note 5) 2tQ–10 2tQ–7 ns

13 t

w(LCKL)

Pulse duration, local clock low 2tQ–15+

2tQ–13.5+

13a t

w(LCKL)

Pulse duration, LCLK1 low (see Note 5) 2tQ–10+

2tQ–7+

14 t

t(LCK)

Transition time, LCLK1 or LCLK2 15 13.5 ns

15 t

h(CK1H-CK2L)

Hold time, LCLK2 low after LCLK1 high tQ–15 tQ–13.5 ns

16 t

h(CK2H-CK1H)

Hold time, LCLK1 high after LCLK2 high tQ–15 tQ–13.5 ns

17 t

h(CK1L-CK2H)

Hold time, LCLK2 high after LCLK1 low tQ–15 tQ–13.5 ns

18 t

h(CK2L-CK1L)

Hold time, LCLK1 low after LCLK2 low tQ–15+

tQ–13.5+

19 t

h(CK1H-CK2H)

Hold time, LCLK2 high after LCLK1 high 3tQ–15 3tQ–13.5 ns

20 t

h(CK2H-CK1L)

Hold time, LCLK1 low after LCLK2 high 3tQ–15+

3tQ–13.5+

21 t

h(CK1L-CK2L)

Hold time, LCLK2 low after LCLK1 low 3tQ–15+

3tQ–13.5+

22 t

h(CK2L-CK1H)

Hold time, LCLK1 high after LCLK2 low 3tQ–15+

3tQ–13.5+

†

This is a functional minimum and is not tested. This parameter can also be specified as 4tQ.

NOTES: 4.

= tQ if using the clock stretch;

= 0 otherwise

5. Parameters 12a and 13a are specified with 1.5 V timing levels (parameters 12 and 13 are specified with standard timing voltage levels).

LCLK1

LCLK2

Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2

12a

13a

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

Page 69

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

output signal characteristics

†‡

The following general parameters are common to all output signals from the SMJ34020A unless otherwise stated. They are intended as an aid in estimating the timing requirements. Please refer to the specific numbered parameters for actual times. In the minimum and maximum values shown, “n” is an integral number of quarter cycles.

SMJ34020A-32

PARAMETER

MIN MAX

UNIT

h(CK-SGNV)

Hold time, LCLKx to output signal not valid tQ – 15 ns

d(CK-SGV)

Delay time, LCLKx start of transition to output signal valid

Fast: RAS, CAS, ALTCH, TR

/QE, DDOUT, DDIN,

EMU3, HOE

, R0, R1,

HDST, WE

tQ+15 ns

Slow: LAD, RCA, SF tQ+22 ns

d(SGNV-SGV)

Delay time, output signal started transition to output signal valid

Fast: RAS, CAS, ALTCH, TR

/QE, DDOUT, DDIN,

EMU3, HOE

, R0, R1,

HDST, WE

tQ+15 ns

Slow: LAD, RCA, SF

tQ+22 ns

d(SGV-SG)

Delay time, output signal valid to output signal not valid

Fast: RAS, CAS, ALTCH, TR

/QE, DDOUT, DDIN,

EMU3, HOE

, R0, R1,

HDST, WE

tQ–15

tQ–16

Slow: LAD, RCA, SF

tQ–22 ns

t(SG)

Output signal transition time

Fast: RAS, CAS, ALTCH, TR

/QE, DDOUT, DDIN,

EMU3, HOE

, R0, R1,

HDST, WE

15 ns

Slow: LAD, RCA, SF 22 ns

w(SGH)

Pulse duration, output signal high

Fast: RAS, CAS, ALTCH, TR

/QE, DDOUT, DDIN,

EMU3, HOE

, R0, R1,

HDST, WE

tQ–15 ns

Slow: LAD, RCA, SF

tQ–22 ns

w(SGL)

Pulse duration, output signal low

Fast: RAS, CAS, ALTCH, TR

/QE, DDOUT, DDIN,

EMU3, HOE

, R0, R1,

HDST, WE

tQ–15 ns

Slow: LAD, RCA, SF

tQ–22 ns

†

Also refer to Figure 34 on following page.

‡

For parameters on this page specifying minimum or maximum times between two output signals, the word fast or slow in column 2 refers to the signal with a subscript of 1, regardless of the other signal. For example, if you are using the spec t

h(SG2NV-SG1V)

, use the slow value if the signal

becoming valid (SG1) is RCA, LAD, or SF; use the fast value otherwise. The pin referred to as SG2 does not determine fast or slow signal time.

See parameter 73 in “local-bus timing: bus control inputs” table.

Page 70

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

t(SG)

d(SG-SGV)

d(SGV-SG)

h(CK-SG)

d(CK-SGV)

h(CK-SG)

LCLKx

SIGNALa

SIGNALb

w(SGH)

w(SGL)

d(SGV-SG)

d(SG-SGV)

d(CK-SGV)

d(SGV-SG)

d(SG-SGV)

t(SG)

See Note A

NOTE A: Any of these quarter phases could be 2tQ if they are stretched (see clock stretch,

page 20).

Figure 35. Output Signal Characteristics

example of how to use the general output signal characteristics

Assume a system is using a SMJ34020A-32. Determine the maximum time from the start of the falling edge of ALTCH

to the time when data must be valid on LAD for a local-memory write cycle.

From the local-memory read-modify-write-cycle timing diagram on page 28, the time from the falling edge of ALTCH

to valid data on LAD is roughly Q3 + Q4; i.e., 2tQ. A more precise value can be obtained by using the

table of output signal characteristics. The parameter of interest is t

d(SG-SGV)

. Note that in Figure 35, there are two representations of t

d(SG-SGV)

that relate SIGNALa and SIGNALb (the third representation of this parameter relates SIGNALb to itself and is not useful in this example). Let SIGNALa represent ALTCH

because ALTCH is making a transition first. Let SIGNALb represent LAD. By definition, the signal becoming valid (SGV) determines whether the fast value or the slow value from the table is used.

In this case, for parameter t

d(SG-SGV)

, SGV is LAD. LAD is in the slow group, so the maximum value for t

d(SG-SGV)

is ntQ + 22. The value for n is 2 from the analysis of the diagram on page 28. Thus, the maximum time from the start of the falling edge of ALTCH

to the time when data must be valid on LAD for a local-memory write cycle

is 2t

+ 22 ns.

Page 71

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

host-interface-cycle timing requirements (see Note 6 and Figure 36)

’34020-32 ’34020A-32

’34020A-40

NO. MIN MAX MIN MAX UNIT

23 t

su(AV-CSL)

Setup time, address prior to HCS no longer high 12 10 ns

24 t

h(CSL-AV)

Hold time, address after HCS low 12 10 ns

25 t

w(CSH)

Pulse duration, HCS high 28 25 ns

26 t

w(RDH)

Pulse duration, HREAD high 28 25 ns

27 t

w(WRH)

Pulse duration, HWRITE high 28 25 ns

28 t

su(RDH-WRL)

Setup time, HREAD high to HWRITE no longer high 28 25 ns

29 t

su(WRH-RDL)

Setup time, HWRITE high to HREAD no longer high 28 25 ns

30 t

w(RDL)

Pulse duration, HREAD low 18 15 ns

31 t

w(WRL)

Pulse duration, HWRITE low 18 15 ns

32 t

su(CSL-WRH)

Setup time, HCS low to HWRITE no longer low 18 15 ns

33 t

su(RDL-CK2L)

Setup time, HCS low or HREAD low to LCLK2 no longer high 30

†

34 t

su(WRH-CK2L)

Setup time, HWRITE high or HCS high to LCLK2 no longer

high

†

35 t

h(CK2L-RDH)

Hold time, HREAD high after LCLK2 no longer high 0

‡

36 t

h(CK2L-WRL)

Hold time, HWRITE low after LCLK2 no longer high 0

‡

37 t

su(RDH-CK2L)

Setup time, HREAD high to LCLK2 no longer high, prefetch

read mode

†§

38 t

su(CSL-RDH)

Setup time, HCS low to HREAD no longer low 18 15 ns

†

Setup time to ensure recognition of input on this clock edge.

‡

Hold time required to assure response on next clock edge. These values are based on computer simulation and are not tested.

When the SMJ34020A is set for block reads, use the deassertion of HREAD

to request a local memory cycle at the next sequential address

location.

NOTE 6: Although HCS

, HREAD, and HWRITE can be totally asynchronous to the SMJ34020A, cycle responses to the signals are determined

by local memory cycles.

Page 72

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

HA0–HA27

HBS0–HBS3

38 30

HCS

HREAD

HWRITE

LCLK2

Figure 36. Host-Interface-Cycle Timing Requirements

Page 73

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

host-interface-cycle timing responses (random read cycle) (see Note 4 and Figure 37)

’34020A-30 ’34020A-40

MIN MAX MIN MAX

UNIT

26 t

w(RDH)

Pulse duration, HREAD high 28 25 ns

33 t

su(RDL-CK2L)

Setup time, HCS low or HREAD low to LCLK2 no longer high

†

39 t

d(CK1H-RYH)

Delay time, LCLK1 going high to HRDY high (end of read cycle)

tQ+20 tQ+18 ns

40 t

d(RDH-RYL)

Delay time, HREAD or HCS high to HRDY low 20 18 ns

41 t

d(CK2L-STL)

Delay time, LCLK2 no longer high to HDST low

+ tQ+15 tQ+13.5+sns

42 t

d(CK1L-STH)

Delay time, LCLK1 no longer high to HDST high tQ+15 tQ–13.5 ns

43 t

su(STL-RYH)

Setup time, HDST low to HRDY no longer low tQ–15 2tQ+13.5 ns

44 t

d(RYH-STH)

Delay time, HRDY no longer low to HDST high 2tQ+15 2tQ+13.5 ns

†

Setup time to ensure recognition of input on this clock edge.

NOTE 4:

= tQ if using the clock stretch;

= 0 otherwise

Q4 Q1 Q2 Q3 Q4

‡

Q1 Q2 Q3 Q4

LCLK1

LCLK2

HCS

/HREAD

HRDY

HDST

‡

See clock stretch, page 20.

Figure 37. Host-Interface-Cycle Timing Responses (Random Read Cycle)

Page 74

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

host-interface-cycle timing (block-read cycle) (see Notes 4 and 6 and Figure 38)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

26 t

w(RDH)

Pulse duration, HREAD high 28 25 ns

30 t

w(RDL)

Pulse duration, HREAD low 18 15 ns

37 t

su(RDH-CK2L)

Setup time, HREAD high to LCLK2 no longer high, prefetch read mode

†

39 t

d(CK1H-RYH)

Delay time, LCLK1 no longer low to HRDY high tQ+20 tQ+18 ns

40 t

d(RDH-RYL)

Delay time, HREAD or HCS high to HRDY low 20 18 ns

41 t

d(CK2L-STL)

Delay time, LCLK2 no longer high to HDST low tQ+15+

tQ+13.5+

42 t

d(CK1L-STH)

Delay time, LCLK1 no longer high to HDST high tQ+15 tQ+13.5 ns

43 t

su(STL-RYH)

Setup time, HDST low to HRDY no longer low tQ–15 tQ–13.5 ns

44 t

d(RYH-STH)

Delay time, HRDY no longer low to HDST high 2tQ+15 2tQ+13.5 ns

45 t

d(RDL-RYH)

Delay time, HREAD or HCS low to HRDY high after prefetch

25 20 ns

50 t

h(STH-CTV)

Hold time, CAS, TR/QE, DDIN valid after HDST high – 2 – 2 ns

†

Setup time to ensure recognition of input on this clock edge. When the SMJ34020A is set for block reads, the deassertion of HREAD is used to request a local memory cycle at the next sequential address location.

NOTES: 4.

= tQ if using the clock stretch;

= 0 otherwise

6. Although HCS

, HREAD, and HWRITE can be totally asynchronous to the SMJ34020A, cycle responses to the signals are

determined by local memory cycles.

HDST

HRDY

HREAD

HCS

LCLK2

LCLK1

Q4 Q1 Q2 Q3 Q4‡Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4

41 44

‡

See clock stretch, page 20.

CAS

TR/QE

DDIN

Figure 38. Host-Interface-Cycle Timing (Block-Read Cycle)

Page 75

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

host interface timing responses (write cycle) (see Note 4 and Figure 39)

NO. PARAMETER

’34020-32 ’34020A-32

’34020A-40

UNIT

MIN MAX MIN MAX

27 t

w(WRH)

Pulse duration, HWRITE high 28 25 ns

31 t

w(WRL)

Pulse duration, HWRITE low 18 15 ns

34 t

su(WRH-CK2L)

Setup time, HWRITE high or HCS high to LCLK2 no longer high

†

39 t

d(CK1L-RYH)

Delay time from LCLK1↑ to HRDY high tQ+20 tQ+18 ns

46 t

d(WRL-RYH)

Delay time from later of HCS or HWRITE low to HRDY high (TMS34020 ready)

25 20 ns

47 t

d(WRH-RYL)

Delay time from earlier of HCS or HWRITE high to HRDY low (end of write)

25 20 ns

48 t

d(CK2L-OEL)

Delay time from LCLK2↓ to HOE low tQ+15+

tQ+13.5+

49 t

d(CK1H-OEH)

Delay time from LCLK1↓ to HOE high tQ+15 tQ+13.5 ns

51 t

d(RYH-OEH)

Delay time from HRDY↑ to HOE high 2tQ+15 2tQ+13.5 ns

†

Setup time to ensure recognition of input on this clock edge.

NOTE 4:

= tQ if using the clock stretch;

= 0 otherwise

HOE

HRDY

HCS

HWRITE

LCLK2

LCLK1

Q4Q3Q2Q1Q4Q3Q2Q1Q4Q3Q2Q1Q4

‡

Q3Q2Q1Q4

‡

See clock stretch, page 20.

Figure 39. Host-Interface-Cycle Timing Responses (Write-Cycle)

Page 76

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

local-bus timing: bus control inputs (see Note 4 and Figure 40)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

52 t

a(CMV-LAV)

†

Access time, CAMD valid after address valid on LAD 3tQ–45 3tQ–37 ns

53 t

h(LA-CMV)

†

Hold time, CAMD valid after address no longer valid on LAD

0 0 ns

54 t

a(BCV-ALL)

†

Access time, control valid (LRDY, PGMD, SIZE16, BUSFLT) after ALTCH

low

3tQ–35+

3tQ–27+

55 t

h(CK2H-BCV)

†

Hold time, control (LRDY, PGMD, SIZE16, BUSFLT) valid after LCLK2 high

0 0 ns

56 t

su(BCV-CK2H)

†

Setup time, SIZE16 valid before LCLK2 no longer low

20 15 ns

†

CAMD, LRDY , PGMD, SIZE16, and BUSFL T are synchronous inputs. The specified setup, access and hold times must be met for proper device operation.

NOTE 4:

= tQ if using the clock stretch;

= 0 otherwise

Address

Valid

LAD

BUSFLT

SIZE16

PGMD

LRDY

ALTCH

CAMD

LCLK2

LCLK1

Q1Q4Q3Q2Q1Q4

‡

Q3Q2Q1Q4Q3Q2Q1

Valid

‡

See clock stretch, page 20.

Figure 40. Local-Bus Timing: Bus Control Inputs

Page 77

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

local-bus timing: bus control inputs (see Note 4 and Figure 41)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

57 t

d(CK2H-ALL)

Delay time, ALTCH low after LCLK2 no longer low

tQ+15 tQ+13.5 ns

58 t

d(CK1L-ALH)

Delay time, ALTCH high after LCLK1 no longer high

tQ+15 tQ+13.5 ns

59 t

d(CK1H-LAV)

Delay time, LAD0–LAD31 address valid after LCLK1 no longer low

tQ+22 tQ+20 ns

60 t

h(LAV-CK2L)

Hold time, LAD0–LAD31 address valid after LCLK2 low

tQ–15+

tQ–12+

61 t

d(CT-LAD)

Delay time, LAD0–LAD31 driven after earlier of DDIN no longer high or CAS

longer low or TR

/QE no longer low

tQ–5+

†

tQ–5+

†

62 t

LAV-CTV

)

Hold time, LAD0–LAD31 read data valid after earlier of DDIN low or RAS

, CAS, or

/QE low

3.5 3.5 ns

63 t

d(CK2L-LAV)

Delay time, LAD0 – LAD31 data valid after LCLK2 no longer high (write)

tQ+22+

tQ+20+

64 t

h(CK2L-LAV)

Hold time, LAD0–LAD31 data valid after LCLK2 low (write)

tQ–15 tQ–13.5 ns

65 t

d(CK1H-RCV)

Delay time, RCA0–RCA12 row address valid after LCLK1 no longer low

tQ+22 tQ+22 ns

66 t

d(CK2L-RCV)

Delay time, LAD0–LAD31 column address valid after LCLK2 no longer high

tQ+22+

tQ+20+

67 t

h(RCV-CK2L)

Hold time, RCA0-RCA12 address valid after LCLK2 low

tQ–15 tQ–13.5 ns

68 t

d(CK1H-DIH)

Delay time, DDIN high after LCLK1 no longer low

tQ+15 tQ+13.5 ns

69 t

d(CK1L-DIL)

Delay time, DDIN low after LCLK1 no longer high

tQ+15 tQ+13.5 ns

70 t

d(CK1H-DOL)

Delay time, DDOUT low after LCLK1 no longer low

tQ+15 tQ+13.5 ns

71 t

d(CK1L-DOH)

Delay time, DDOUT high after LCLK1 no longer high

tQ+15 tQ+13.5 ns

72 t

d(CK2L-DOL)

Delay time, DDOUT low after LCLK2 no longer high

tQ+15+

tQ+13.5+

73 t

su(LAV-ALL)

Setup time, LAD0–LAD31 data valid before ALTCH

no longer high

tQ–16 tQ–13.5 ns

74 t

en(DAV-DIH)

Enable time, data valid after DDIN high (see Note 7)

2tQ–20 2tQ–17 ns

75 t

dis(DAV-DIL)

Disable time, data in the high-impedance state after DDIN low (see Note 7)

tQ–12+

†

tQ–10+

†

These values are derived from characterization data and are not tested.

NOTES: 4.

= tQ if using the clock stretch;

= 0 otherwise

7. DDIN is used to control LAD bus buffers between the SMJ34020A and local memory. Parameter 74 references the time for these data buffers to go from the high-impedance state to an active level. Parameter 75 references the time for the buffers to go from an active level to the high-impedance state.

Page 78

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

2nd Column1st ColumnRow

Data OutData InAddress/Status

Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

DDOUT

LAD0–LAD31

DDIN

RCA0–RCA12

ALTCH

LCLK2

LCLK1

/QE

CAS

Figure 41. Local-Bus Timing: Bus Control Inputs (Continued)

Page 79

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

local-bus timing: RAS, CAS0–CAS3, WE, TR/QE, and SF (see Notes 4 and 8 and Figure 42)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

62 t

h(LAV-CTV)

Hold time, LAD0–LAD31 read data valid after earlier of DDIN, low after RAS

CAS

, or TR/QE high

3.5 3.5 ns

76 t

d(CK1L-REL)

Delay time, RAS low after LCLK1 no longer high

tQ+12+

tQ+10+

77 t

d(CK1L-REH)

Delay time, RAS high after LCLK1 no longer high

tQ+12 tQ+10 ns

78 t

d(CK1H-CEL)

Delay time, CAS low after LCLK1 no longer low

tQ+12 tQ+10 ns

79 t

d(CK1L-CEH)

Delay time, CAS high after LCLK1 no longer high

tQ+12 tQ+10 ns

80 t

d(CK2L-WEL)

Delay time, WE low after LCLK2 no longer high

tQ+15+

tQ+13.5+

81 t

d(CK1L-WEH)

Delay time, WE high after LCLK1 no longer high

tQ+15 tQ + 15 ns

82 t

d(CK2L-TRL)

Delay time, TR/QE low after LCLK2 no longer high

tQ+15+

tQ+13.5+

83 t

d(CK1L-TRH)

Delay time, TR/QE high after LCLK1 no longer high

tQ+15 tQ+13.5 ns

84 t

d(CK1H-SFV)

Delay time, SF valid after LCLK1 no longer low

tQ+22 tQ+20 ns

85 t

d(CK2L-SFV)

Delay time, SF valid after LCLK2 no longer high

tQ+22+

tQ+20+

86 t

d(CK2L-SFZ)

Delay time, SF in the high-impedance state after LCLK2 no longer high

tQ+22

†

tQ+20

†

87 t

su(ADV-REL)

‡

Setup time, row address valid before RAS

no longer high

2tQ–22 2tQ–20 ns

88 t

h(ADV-REL)

‡

Hold time, row address valid after RAS low

tQ–5+

89 t

su(RCV-CEL)

Setup time, column address valid before CAS

no longer high

tQ–22 tQ–20 ns

90 t

h(RCV-CEH)

Hold time, column address valid after CAS

high

tQ–15 tQ– 13.5 ns

91 t

su(CAV-CEL)

Setup time, write data valid before CAS no longer high

tQ–22 tQ–20 ns

92 t

h(CAV-CEH)

Hold time, write data valid after CAS no longer low

tQ–15 tQ– 13.5 ns

93 t

a(LAV-REL)

Access time, data-in valid after RAS low (assuming maximum transition time)

4tQ–8+

94 t

a(LAV-CEL)

Access time, data-in valid after CASL no longer high

2tQ–8+

95 t

a(LAV-RCV)

Access time, data-in valid after column address valid

3tQ–20+

3tQ – 12 ns

†

These values are derived from characterization data and are not tested.

‡

Parameters 87 and 88 also apply to WE

, TR/QE, and SF relative to RAS.

NOTES: 4.

= tQ if using the clock stretch;

= 0 otherwise

8. Parameter 96 has been eliminated.

Page 80

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

local-bus timing: RAS, CAS0–CAS3, WE, TR/QE, and SF (see Notes 4 and 8 and Figure 42) (continued)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

97 t

su(WEL-CEL)

Setup time, write low before CAS no longer high (on write cycles)

tQ–15 tQ– 13.5 ns

98 t

w(REH)

Pulse duration, RAS high 4tQ–12+

4tQ–10+

99 t

w(REL)

Pulse duration, RAS low 4ntQ–12+

4ntQ–4+

100 t

w(CEH)

Pulse duration, CAS high 2tQ–12 2tQ–10 ns

101 t

w(CEL)

Pulse duration, CAS low 2tQ–12 2tQ–8 ns

102 t

d(REL-CEH)

Delay time, RAS low to CAS no longer low

4tQ–12+

4tQ–4+

NOTES: 4.s = tQ if using the clock stretch;

= 0 otherwise

8. Parameter 96 has been eliminated.

Page 81

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

2nd Data Out1st Data OutAddress/Status

Data In 2Data In 1Address/Status

2nd Column1st ColumnRow

ALTCH

TR/QE

CAS0–CAS3

LAD (WRITE)

LAD (READ)

RCA

RAS

LCLK2

LCLK1

Q1 Q2 Q3 Q4†Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

102

101

100

†

See clock stretch, page 20.

Figure 42. Local-Bus Timing: RAS, CAS0–CAS3, WE, TR/QE, and SF

Page 82

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

The refresh pseudo-address present on LAD0–LAD31 is the output from the 16-bit refresh address register([IO] register located at C000 01F0h) on LAD16–LAD31. LAD0–LAD3 have the refresh status code (status code = 0011), and LAD4–LAD15 are held low.

CBR refresh: RAS and CAS0–CAS3 (see Note 4 and Figure 42)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

76 t

d(CK1L-REL)

Delay time, RAS low after LCLK1 no longer high

tQ+12 +

tQ+10 +

77 t

d(CK1L-REH)

Delay time, RAS high after LCLK1 no longer high

tQ+12 tQ+10 ns

78 t

d(CK1H-CEL)

Delay time, CAS low after LCLK1 no longer low

tQ+12 tQ+10 ns

79 t

d(CK1L-CEH)

Delay time, CAS high after LCLK1 no longer high

tQ+12 tQ+10 ns

102 t

d(REL-CEH)

Delay time, RAS low to CAS no longer low

4tQ–12 +

4tQ – 4+

103 t

d(CEL-REL)

Delay time, CAS low to RAS no longer high

2tQ–15 2tQ – 13.5 ns

104 t

d(REH-CEL)

Delay time, RAS high to CAS no longer high

2tQ–15 +

2tQ –13.5+

NOTE 4:s = tQ if using the clock stretch;

= 0 otherwise

Page 83

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

DDOUT

(see Note A)

RCA

(see Note A)

LAD

(see Note A)

ALTCH

(see Note A)

CAS

RAS

LCLK2

LCLK1

Refresh Pseudo-Address

Q1 Q2 Q3 Q4†Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

102

103

104

†

See clock stretch, page 20.

NOTE A: ALTCH

, LAD, RCA, and DDOUT are shown for reference only.

Figure 43. CBR Refresh: RAS and CAS0–CAS3

Page 84

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

multiprocessor-interface timing: GI, ALTCH, RAS, R0 and R1 (see Figure 44)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

105 t

a(GIV-RQV)

Access time, GI valid after R0 and R1 valid (see Note 9)

2tQ–40 2tQ–30 ns

105.1 t

su(GIV-CK1H)

Setup time, GI valid before LCLK1 no longer low (see Note 9)

40 35 ns

106 t

h(CK1H-GIV)

Hold time, GI valid after LCLK1 no longer low 0 0 ns

107 t

d(CK2H-RQV)

Delay time, LCLK2 no longer low to R0 or R1 valid tQ+15 tQ+13.5 ns

108 t

d(CK2H-RQNV)

Delay time, LCLK2 high to R0 or R1 no longer valid tQ–15 tQ–13.5 ns

NOTE 9: These timings must be met to ensure that GI is recognized on this clock cycle.

For a SMJ34020A to gain control of the local bus during a given cycle, GI must be low at the start of Q1 (indicating that the bus arbitration logic is granting the bus to this processor).

R0–R1

LCLK2

LCLK1

106

105

Q4 Q1 Q2 Q3 Q4†Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1

Valid

108

107

105.1

†

See clock stretch, page 20.

Figure 44. Multiprocessor-Interface Timing: GI, ALTCH, RAS, R0 and R1

Page 85

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

multiprocessor-interface timing: high-impedance signals (see Note 4 and Figure 45)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

84 t

d(CK1H-SFV)

Delay time, SF valid after LCLK1 no longer low tQ+22 tQ+20 ns

86 t

d(CK2L-SFZ)

Delay time, SF in the high-impedance state after LCLK2 no longer high

tQ+22+

†

tQ+20+

†

109 t

d(CK2L-ADZ)

Delay time, LAD and RCA in the high-impedance state after LCLK2 no longer high

tQ+22+

†

tQ+20 +

†

110 t

d(CK1H-ADV)

Delay time, LAD and RCA valid after LCLK1 no longer low

tQ+22 tQ+20 ns

111 t

d(CK1H-CTZ)

Delay time, ALTCH, RAS, CAS, WE, TR/QE, HOE, and HDST in the high-impedance state after LCLK1 no longer low

tQ+15

†

tQ+13.5

†

112 t

d(CK2L-CTH)

Delay time, ALTCH, RAS, CAS, WE, TR/QE, HOE, and HDST in the high-impedance state after LCLK2 no longer high

tQ+15+

tQ+13.5+

113 t

d(CK1H-DIZ)

Delay time, DDIN in the high-impedance state after LCLK1 no longer low

tQ+15

†

tQ+13.5

†

114 t

d(CK2L-DIL)

Delay time, DDIN low after LCLK2 no longer high tQ+15 +

tQ+13.5+

115 t

d(CK2L-DOZ)

Delay time, DDOUT in the high-impedance state after LCLK2 no longer high

tQ+15 +

†

tQ+13.5 +

†

116 t

d(CK2L-DOH)

Delay time, DDOUT high after LCLK2 no longer high tQ+15+

tQ+13.5+

†

These values are derived from characterization data and are not tested.

NOTE 4:

= tQ if using the clock stretch;

= 0 otherwise

Page 86

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

110

DDOUT

DDIN

ALTCH

, RAS,

CAS

, WE,

/QE, HOE, HDST

LAD0–LAD31, RCA0–RCA12

–R1

LCLK2

LCLK1

Q2Q1Q4Q3Q2Q1Q4

†

Q3Q2Q1Q4

Hi-Z

112

109

115

111

113

114

116

†

See clock stretch, page 20.

Figure 45. Multiprocessor-Interface Timing: High-Impedance Signals

Page 87

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

video-shift-clock timing: SCLK (see Figure 46)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

117 t

c(SCK)

Cycle time, period of video serial clock SCLK 35 50 25 50 ns

118 t

w(SCKH)

Pulse duration, SCLK high 12 10 ns

119 t

w(SCKL)

Pulse duration, SCLK low 12 10 ns

120 t

t(SCK)

Transition time, (rise and fall) of SCLK 2

†

This value is determined through computer simulation and is not tested.

SCLK

120

117

119

118

Figure 46. Video-Shift-Clock Timing: SCLK

video-interface timing: VCLK and video outputs (see Figure 47)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

123 t

c(VCK)

Cycle time, period of video input clock VCLK 62.5 100 62.5 100 ns

124 t

w(VCKH)

Pulse duration, VCLK high 28 28 ns

125 t

w(VCKL)

Pulse duration, VCLK low 28 28 ns

126 t

t(VCK)

Transition time, (rise and fall) of VCLK 2

†

127 t

d(VCKL-HSL)

Delay time, VCLK low to HSYNC, VSYNC, CSYNC/VBLNK or CBLNK

/VBLNK low

40 40 ns

128 t

d(VCKL-HSH)

Delay time, VCLK low to HSYNC, VSYNC, CSYNC/HBLNK, or CBLNK

/VBLNK high

40 40 ns

129 t

h(VCKL-HSL)

Hold time, VCLK no longer high to HSYNC, VSYNC, CSYNC

/HBLNK, or CBLNK/VBLNK no longer high

†

130 t

h(VCKL-HSH)

Hold time, VCLK no longer high to HSYNC, VSYNC, CSYNC

/HBLNK, or CBLNK/VBLNK no longer low

†

This value is determined through computer simulation and is not tested.

VCLK

HSYNC

VSYNC CSYNC/HBLNK CBLNK/VBLNK

(outputs)

130

127

129

128

126

123

125

124

Figure 47. Video-Interface Timing: VCLK and Video Outputs

Page 88

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

video-interface timing: external sync inputs (see Note 10 and Figure 48)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

131 t

su(SL-VCKH)

Setup time, HSYNC, VSYNC, CSYNC low to VCLK no longer low 20 20 ns

132 t

su(SH-VCKH)

Setup time, HSYNC, VSYNC, CSYNC high to VCLK no longer low 20 20 ns

133 t

h(VCKH-SV)

Hold time, HSYNC, VSYNC, CSYNC valid after VCLK high 20 20 ns

NOTE 10: Setup and hold times on asynchronous inputs are required only to assure recognition at indicated clock edges.

See Note BSee Note A

DCBA

HSYNC

VSVNC CSYNC (inputs)

VCLK

132

133

131

133

NOTES: A. If the falling edge of the sync signal occurs more than t

h(VCKH-SV)

after VCLK edge A and at least t

su(SL-VCKH)

before

edge B, the transition is detected at edge B instead of edge A.

B. If the rising edge of the sync signal occurs more than t

h(VCKH-SV)

after VCLK edge C and at least t

su(SH-VCKH)

before

edge D, the transition is detected at edge D instead of edge C.

Figure 48. Video-Interface Timing: External Sync Inputs

interrupt timing: LINT1 and LINT2 (see Figure 49)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

134 t

su(LINTL-CK2H)

Setup time, LINT1 or LINT2 low before LCLK2 no longer low

tQ+45

†

tQ+40

†

135 t

w(LINTL)

Pulse duration, LINT1 or LINT2 low 8t

‡

†

Although LINT1

and LINT2 can be asynchronous to the SMJ34020A, this setup ensures recognition of the interrupt on this clock edge.

‡

This pulse duration minimum ensures that the interrupt is recognized by internal logic; however, the level must be maintained until it has been acknowledged by the interrupt service routine.

LINT

LCLK2

LCLK1

135

134

Figure 49. Interrupt Timing: LINT1 and LINT2

Page 89

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

host-interrupt timing: HINT (see Figure 50)

’34020A-32 ’34020A-40

NO.UNIT

136 t

d(CK1H-HINTV)

Delay time, LCLK1 no longer low to HINT valid 30 25 ns

HINT

LCLK2

LCLK1

136

Figure 50. Host-Interrupt Timing: HINT

emulator-interface timing (see Figure 51)

’34020A-32 ’34020A-40

MIN MAX MIN MAX

UNIT

137 t

su(EMV-CK1H)

Setup time, EMU0–EMU2 valid to LCLK1 no longer low 30 25 ns

138 t

h(EMV-CK1H)

Hold time, EMU0–EMU2 valid after LCLK1 no longer low 0 0 ns

139 t

d(CK1L-SCV)

Delay time, EMU3 valid after LCLK1 low 25 20 ns

140 t

h(CK2H-SCNV)

Hold time, LCLK2 high before EMU3 not valid tQ–15 tQ–13.5 ns

EMU3

EMU0–EMU2

LCLK2

LCLK1

139

137

138

137

138

140

Figure 51. Emulator-Interface Timing

Page 90

SMJ34020A GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

MECHANICAL DATA

GA-GB (S-CPGA-P15 X 15) CERAMIC PIN GRID ARRAY PACKAGE

4040114-8/B 10/94

Down

Cavity

Down

Cavity

Cavity Up

Outline

Small

Outline

Large

MAXIMUM PINS WITHIN MATRIX – 225

C1 0.025 (0,63) 0.060 (1,52)

0.060 (1,52)0.040 (1,02)C

B1 0.095 (2,41) 0.205 (5,21)

0.205 (5,21)0.110 (2,79)B

A1 1.480 (37,59) 1.535 (38,99)

1.590 (40,38)

DIM MIN MAX Notes

1.540 (39,12)A

A or A1 SQ

4 Places

DIA TYP

1.400 (35,56) TYP

0.050 (1,27) DIA

0.120 (3,05)

0.016 (0,41)

0.022 (0,55)

0.140 (3,56)

B or B1

C or C1

J H G

F E D C

123456789K10L11 12M13N14

0.100 (2,54)

NOTES: A. All linear dimensions are in inches (millimeters).

B. This drawing is subject to change without notice. C. Index mark may appear on top or bottom depending on package vendor. D. Pins are located within 0.005 (0,13) radius of true position relative to each other at maximum material condition and within

0.015 (0,38) radius relative to the center of the ceramic.

E. This package can be hermetically sealed with metal lids or with ceramic lids using glass frit.

F. The pins can be gold plated or solder dipped.

G. Falls within MIL-STD-1835 CMGA7-PN and CMGA19-PN and JEDEC MO-067AG and MO-066AG, respectively

Page 91

SMJ34020A

GRAPHICS SYSTEM PROCESSOR

SGUS011B – APRIL 1991 – REVISED AUGUST 1995

POST OFFICE BOX 1443 • HOUSTON, TEXAS 77251–1443

MECHANICAL DATA

HT (S-CQFP-F132) CERAMIC QUAD FLATPACK

4040169-3/B 10/94

0.800 (20,32) TYP

0.018 (0,46) MAX

0.105 (2,67) MAX

0.002 (0,05)

0.130 (3,30) MAX

0.014 (0,36) At Braze Pads

1.525 (38,74)

0.960 (24,38)

0.935 (23,75)

1.480 (37,59)

132

0.013 (0,330)

0.006 (0,152)

100

132



0.004 (0,10)

0.008 (0,20)

0.025 (0,64)

NOTES: A. All linear dimensions are in inches (millimeters).

B. This drawing is subject to change without notice. C. This package is hermetically sealed with a metal lid. D. The terminals are gold plated. E. Falls within JEDEC MO-090 AB

Page 92

IMPORTANT NOTICE

T exas 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.

CERT AIN 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 APPLICA TIONS IS UNDERST OOD 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.

Datasheet SM34020AGBM32, SM34020AGBM40, SM34020AHTM32, SM34020AHTM40, SMJ34020AGBM40 Datasheet (Texas Instruments)

Specifications and Main Features

Frequently Asked Questions

User Manual