May 1992 DP8520A/DP8521A/DP8522A microCMOS Programmable 256k/1M/4M Video RAM Controller/Drivers General Description Features The DP8520A/21A/22A video RAM controllers provide a low cost, single chip interface between video RAM and all 8-, 16- and 32-bit systems. The DP8520A/21A/22A generate all the required access control signal timing for VRAMs. An on-chip refresh request clock is used to automatically refresh the VRAM array. Refreshes and accesses are arbitrated on chip. If necessary, a WAIT or DTACK output inserts wait states into system access cycles, including burst mode accesses. RAS low time during refreshes and RAS precharge time after refreshes and back to back accesses are guaranteed through the insertion of wait states. Separate on-chip precharge counters for each RAS output can be used for memory interleaving to avoid delayed back to back accesses because of precharge. An additional feature of the DP8522A is two access ports to simplify dual accessing. Arbitration among these ports and refresh is done on chip. Y Control Ý of Pins Ý of Address (PLCC) Outputs Y Y Y Y Y Y Y Y On chip high precision delay line to guarantee critical VRAM access timing parameters microCMOS process for low power High capacitance drivers for RAS, CAS, DT/OE and VRAM address on chip On chip support for nibble, page and static column VRAMs Byte enable signals on chip allow byte writing in a word size up to 16 bits with no external logic Selection of controller speeds: 20 MHz and 25 MHz On board Port A/Port B (DP8522A only)/refresh arbitration logic Direct interface to all major microprocessors (application notes available) 4 RAS and 4 CAS drivers (the RAS and CAS configuration is programmable) Largest VRAM Possible Direct Drive Memory Capacity Access Ports Available DP8520A 68 9 256 kbit 4 Mbytes Single Access Port DP8521A 68 10 1 Mbit 16 Mbytes Single Access Port DP8522A 84 11 4 Mbit 64 Mbytes Dual Access Ports (A and B) Block Diagram DP8520A/21A/22A VRAM Controller TL/F/9338 – 5 FIGURE 1 TRI-STATEÉ is a registered trademark of National Semiconductor Corporation. PALÉ is a registered trademark of and is used under license from Monolithic Memories, Inc. C1995 National Semiconductor Corporation TL/F/9338 RRD-B30M105/Printed in U. S. A. DP8520A/DP8521A/DP8522A microCMOS Programmable 256k/1M/4M Video RAM Controller/Drivers PRELIMINARY Table of Contents 1.0 INTRODUCTION 7.0 ADDITIONAL ACCESS SUPPORT FEATURES 7.1 Address Latches and Column Increment 2.0 SIGNAL DESCRIPTIONS 7.2 Address Pipelining 2.1 Address, R/W and Programming Signals 7.3 Delay CAS During Write Accesses 2.2 VRAM Control Signals 2.3 Refresh Signals 8.0 RAS AND CAS CONFIGURATION MODES 2.4 Port A Access Signals 8.1 Byte Writing 2.5 Port B Access Signals (DP8522A) 8.2 Memory Interleaving 2.6 Common Dual Port Signals (DP8522A) 8.3 Address Pipelining 2.7 Power Signals and Capacitor Input 8.4 Error Scrubbing 2.8 Clock Inputs 8.5 Page/Burst Mode 3.0 PORT A ACCESS MODES 9.0 PROGRAMMING AND RESETTING 3.1 Access Mode 0 9.1 Mode Load Only Programming 3.2 Access Mode 1 9.2 Chip Selected Access Programming 9.3 External Reset 4.0 REFRESH OPTIONS 9.4 Definition of Programming Bits 4.1 Refresh Control Modes 4.1.1 Automatic Internal Refresh 4.1.2 Externally Controlled/Burst Refresh 4.1.3 Refresh Request/Acknowledge 10.0 TEST MODE 11.0 VRAM CRITICAL TIMING OPTIONS 11.1 Programming Values of tRAH and tASC 11.2 Calculation of tRAH and tASC 4.2 Refresh Cycle Types 4.2.1 Conventional Refresh 4.2.2 Staggered Refresh 4.2.3 Error Scrubbing Refresh 12.0 DUAL ACCESSING (DP8522A) 12.1 Port B Access Mode 4.3 Extending Refresh 12.2 Port B Wait State Support 4.4 Clearing the Refresh Address Counter 12.3 Common Port A and Port B Dual Port Functions 4.5 Clearing the Refresh Request Clock 12.3.1 GRANTB Output 12.3.2 LOCK Input 5.0 PORT A WAIT STATE SUPPORT 5.1 WAIT Type Output 13.0 ABSOLUTE MAXIMUM RATINGS 5.2 DTACK Type Output 14.0 DC ELECTRICAL CHARACTERISTICS 5.3 Wait State Support for VRAM Transfer Cycles 15.0 AC TIMING PARAMETERS 5.4 Dynamically Increasing the Number of Wait States 16.0 FUNCTIONAL DIFFERENCES BETWEEN THE DP8520A/21A/22A AND THE DP8520/21/22 5.5 Guaranteeing RAS Low Time and RAS Precharge Time 17.0 DP8520A/21A/22A USER HINTS 6.0 DP8520A/21A/22A VIDEO RAM SUPPORT 18.0 DESCRIPTION OF A DP8522A/DP8500 SYSTEM INTERFACE 6.1 Support for VRAM Transfer Cycles 6.2 Support for VRAM Access Cycles through Port A 2 1.0 Introduction asserted and must precede the input VSRL, Video Shift Register Load, asserting by enough CLK periods to guarantee any access in progress or pending refresh can finish. VSRL asserting causes DT/OE to transition low immediately. Both VSRL and DT/OE assert before RAS and CAS assert for the transfer. The cycle is ended by DT/OE negating. This is caused by either VSRL negating or by four rising edges of CLK from VSRL asserting, whichever comes first. The DP8520A/21A/22A have greatly expanded refresh capabilities compared to other VRAM controllers. There are three modes of refreshing available. These modes are internal automatic refreshing, externally controlled/burst refreshing, and refresh request/acknowledge refreshing. Any of these modes can be used together or separately to achieve the desired results. In any combination of these modes, the programming of ECAS0 determines the use of the RFIP (RFRQ) pin. ECAS0 asserted during programming causes this pin to function as RFIP which will assert just prior to a refresh cycle and will negate when the refresh is completed. ECAS0 negated during programming causes this pin to function as RFRQ which indicates an internal refresh request when asserted. When using internal automatic refreshing, the DP8520A/ 21A/22A will generate an internal refresh request from the refresh request clock. The DP8520A/21A/22A will arbitrate between the refresh requests and accesses. Assuming an access is not currently in progress, the DP8520A/21A/22A will grant a refresh, assert RFIP if programmed, and on the next positive clock edge, refreshing will begin. If an access had been in progress, the refresh will begin after the access has terminated. To use externally controlled/burst refresh, the user disables the internal refresh request by asserting the input DISRFRSH. A refresh can now be externally requested by asserting the input RFSH. The DP8520A/21A/22A will arbitrate between the external refresh request and accesses. Assuming an access is not currently in progress, the DP8520A/21A/22A will grant a refresh, assert RFIP if programmed, and on the next positive clock edge, refreshing will begin. If an access had been in progress, the refresh would take place after the access has terminated. With refresh request/acknowledge mode, the DP8520A/ 21A/22A broadcasts the internal refresh request to the system through the RFRQ output pin. External circuitry can determine when to refresh the VRAM through the RFSH input. The controllers have three types of refreshing available: conventional, staggered and error scrubbing. Any refresh control mode can be used with any type of refresh. In a conventional refresh, all of the RAS outputs will be asserted and negated at once. In a staggered refresh, the RAS outputs will be asserted one positive clock edge apart. Error scrubbing is the same as conventional refresh except that a CAS will be asserted during a refresh allowing the system to run that data through an EDAC chip and write it back to memory, if a single bit error has occurred. The refreshes can be extended with the EXTEND REFRESH input, EXTNDRF. The DP8520A/21A/22A have wait support available as DTACK or WAIT. Both are programmable. DTACK, Data Transfer ACKnowledge, is useful for processors whose wait signal is active high. WAIT is useful for processors whose The DP8520A/21A/22A are CMOS Video RAM controllers that incorporate many advanced features including the capabilities of address latches, refresh counter, refresh clock, row, column and refresh address multiplexor, delay line, refresh/access/VRAM transfer cycle arbitration logic and high capacitive drivers. The programmable system interface allows any manufacturer’s microprocessor or bus to directly interface via the DP8520A/21A/22A to VRAM arrays up to 64 Mbytes in size. After power up, the DP8520A/21A/22A must first be programmed before accessing the VRAM. The chip is programmed through the address bus. There are two methods of programming the chip. The first method, mode load only, is accomplished by asserting the signal mode load, ML. A valid programming selection is presented on the row, column, bank and ECAS inputs, then ML is negated. When ML is negated, the chip is programmed with the valid programming bits on the address bus. The second method, chip selected access, is accomplished by asserting ML and performing a chip selected access. When CS and AREQ are asserted for the access, the chip is programmed. During this programming access, the programming bits affecting the wait logic become effective immediately, allowing the access to terminate. After the access, ML is negated and the rest of the programming bits take effect. Once the DP8520A/21A/22A has been programmed, a 60 ms initialization period is entered. During this time, the DP8520A/21A/22A controllers perform refreshes to the VRAM array so further VRAM warm up cycles are unnecessary. The DP8520A/21A/22A can now be used to access the VRAM. There are two modes of accessing with the controller. The two modes are Mode 0, which initiates RAS synchronously, and Mode 1, which initiates RAS asynchronously. To access the VRAM using Mode 0, the signal ALE is asserted along with CS to ensure a valid VRAM access. ALE asserting sets an internal latch and only needs to be pulsed and not held throughout the entire access. On the next rising clock edge, after the latch is set, RAS will be asserted for that access. The DP8520A/21A/22A will place the row address on the VRAM address bus, guarantee the programmed value of row address hold time of the VRAM, place the column address on the VRAM address bus, guarantee the programmed value of column address setup time and assert CAS. AREQ can be asserted anytime after the clock edge which starts the access RAS. RAS and CAS will extend until AREQ is negated. The other access mode, Mode 1, is asynchronous to the clock. When ADS is asserted, RAS is asserted. The DP8520A/21A/22A will place the row address on the VRAM address bus, guarantee the programmed value of row address hold time, place the column address on the VRAM address bus, guarantee the programmed value of column address setup time and assert CAS. AREQ can be tied to ADS or can be asserted after ADS is asserted. AREQ negated will terminate the access. The DP8520A/21A/22A also provides full support for VRAM transfer cycles. To begin the cycle, the input AVSRLRQ, Advanced Video Shift Register Load Request, is 3 1.0 Introduction (Continued) The DP8522A has all the features previously mentioned. Unlike the DP8520A/21A, the DP8522A has a second port to allow a second CPU to access the memory array. This port, Port B, has two control signals to allow a CPU to access the VRAM array. These signals are access request for Port B, AREQB, and Advanced Transfer ACKnowledge for Port B, ATACKB. Two other signals are used by both Port A and Port B for dual accessing purposes. The signals are lock, LOCK and grant Port B, GRANTB. All arbitration for the two ports and refresh is done on-chip by the DP8522A through the insertion of wait states. Since the DP8522A has only one input address bus, the address lines have to be multiplexed externally. The signal GRANTB can be used for this purpose since it is asserted when Port B has access to the VRAM array and negated when Port A has access to the VRAM array. Once a port has access to the array, the other port can be ‘‘locked out’’ by asserting the input LOCK. AREQB, when asserted, is used by Port B to request an access. ATACKB, when asserted, signifies that access RAS has been asserted for the requested Port B access. By using ATACKB, the user can generate an appropriate WAIT or DTACK like signal for the Port B CPU. The following explains the terminology used in this data sheet. The terms negated and asserted are used. Asserted refers to a ‘‘true’’ signal. Thus, ‘‘ECAS0 asserted’’ means the ECAS0 input is at a logic 0. The term ‘‘COLINC asserted’’ means the COLINC input is at a logic 1. The term negated refers to a ‘‘false’’ signal. Thus, ‘‘ECAS0 negated’’ means the ECAS0 input is at a logic 1. The term ‘‘COLINC negated’’ means the input COLINC is at a logic 0. The table shown below clarifies this terminology. wait signal is active low. The user can choose either at programming. These signals are used by the on-chip arbitor to insert wait states to guarantee the arbitration between accesses and refreshes or precharge. Both signals are independent of the access mode chosen. DTACK will assert a programmed number of clock edges from the event that starts the access RAS. DTACK will be negated, when the access is terminated, by AREQ being negated. DTACK can also be programmed to toggle with the ECAS inputs during burst/page mode accesses. WAIT is asserted during the start of the access (ALE and CS, or ADS and CS) and will negate a number of clock edges from the event that starts the access RAS. After WAIT is negated, it will stay negated until the next access. WAIT can also be programmed to toggle with ECAS inputs during a burst/page mode access. Both signals can be dynamically delayed further through the WAITIN signal to the DP8520A/21A/22A. The DP8520A/21A/22A have address latches, used to latch the bank, row and column address inputs. Once the address is latched, a column increment feature can be used to increment the column address. The address latches can also be programmed to be fall through. The RAS and CAS drivers can be configured to drive a one, two or four bank memory array up to 32 bits in width. The two ECAS signals can then be used to select one pair of CAS drivers for byte writing with no external logic for systems with a word length of up to 16 bits. When configuring the DP8520A/21A/22A for more than one bank, memory interleaving can be used. By tying the low order address bits to the bank select lines, B0 and B1, sequential back to back accesses will not be delayed since the DP8520A/21A/22A have separate precharge counters per bank. The DP8520A/21A/22A are capable of performing address pipelining. In address pipelining, the DP8520A/ 21A/22A guarantee the column address hold time and switch the internal multiplexor to place the row address on the address bus. At this time, another memory access to another bank can be initiated. 4 Signal Action Logic Level Active High Asserted High Active High Negated Low Active Low Asserted Low Active Low Negated High Connection Diagrams TL/F/9338 – 2 Top View FIGURE 2 Order Number DP8520AV-20 or DP8520AV-25 See NS Package Number V68A 5 Connection Diagrams (Continued) TL/F/9338 – 3 Top View FIGURE 3 Order Number DP8521AV-20 or DP8521AV-25 See NS Package Number V68A TL/F/9338 – 4 Top View FIGURE 4 Order Number DP8522AV-20 or DP8522AV-25 See NS Package Number V84A 6 2.0 Signal Descriptions Pin Name Device (If not applicable to all) Input/ Output Description 2.1 ADDRESS, R/W AND PROGRAMMING SIGNALS R0 – 10 R0 – 9 DP8522A DP8520A/21A I I ROW ADDRESS: These inputs are used to specify the row address during an access or refresh to the VRAM or for a VRAM transfer cycle. They are also used to program the chip when ML is asserted (except R10). C0 – 10 C0 – 9 DP8522A DP8520A/21A I I COLUMN ADDRESS: These inputs are used to specify the column address during an access to the VRAM or for a VRAM transfer cycle. They are also used to program the chip when ML is asserted (except C10). B0, B1 I BANK SELECT: Depending on programming, these inputs are used to select a group of RAS and CAS outputs to assert during an access. They are also used to program the chip when ML is asserted. ECAS0 – 1 I ENABLE CAS: These inputs are used to enable a single or group of CAS outputs when asserted. In combination with the B0, B1 and the programming bits, these inputs select which CAS output or CAS outputs will assert during an access. ECAS0 must be asserted for either CAS0 or CAS1 to assert during an access. ECAS1 must be asserted for either CAS2 or CAS3 to assert during an access. The ECAS signals can also be used to toggle a group of CAS outputs for page/nibble mode accesses. They also can be used for byte write operations. If ECAS0 is negated during programming, continuing to assert the ECAS0 while negating AREQ or AREQB during an access, will cause the CAS outputs to be extended while the RAS outputs are negated (the ECASn inputs have no effect during scrubbing refreshes). WIN I WRITE ENABLE IN: This input is used to signify a write operation to the VRAM. This input asserted will also cause CAS to delay to the next positive clock edge if address bit C9 is asserted during programming. COLINC (EXTNDRF) I I COLUMN INCREMENT: When the address latches are used, and a refresh is not in progress, this input functions as COLINC. Asserting this signal causes the column address to be incremented by one. When a refresh is in progress, this signal, when asserted, is used to extend the refresh cycle by any number of periods of CLK until it is negated. ML I MODE LOAD: This input signal, when low, enables the internal programming register that stores the programming information. O O O VRAM ADDRESS: These outputs are the multiplexed output of the R0 – 9, 10 and C0 – 9, 10 and form the VRAM address bus. These outputs contain the refresh address whenever a refresh is in progress. They contain high capacitive drivers with 20X series damping resistors. RAS0 – 3 O ROW ADDRESS STROBES: These outputs are asserted to latch the row address contained on the outputs Q0 – 8, 9, 10 into the VRAM. When a refresh is in progress, the RAS outputs are used to latch the refresh row address contained on the Q0 – 8, 9, 10 outputs in the VRAM. These outputs contain high capacitive drivers with 20X series damping resistors. CAS0 – 3 O COLUMN ADDRESS STROBES: These outputs are asserted to latch the column address contained on the outputs Q0 – 8, 9, 10 into the VRAM. These outputs have high capacitive drivers with 20X series damping resistors. DT/OE O DATA TRANSFER/OUTPUT ENABLE: This output transitions low before RAS goes low and transitions high before RAS goes high during a video RAM shift register load operation (see VSRL pin description). During normal write accesses this output is held high, and for read accesses this output is asserted after CAS is asserted, and is negated after CAS negates. 2.2 VRAM CONTROL SIGNALS Q0 – 10 Q0 – 9 Q0 – 8 DP8522A DP8521A DP8521A 7 2.0 Signal Descriptions (Continued) Pin Name Device (If not applicable to all) Input/ Output Description 2.3 REFRESH SIGNALS RFIP(RFRQ) O REFRESH IN PROGRESS or REFRESH REQUEST: When ECAS0 is asserted during programming, this output functions as RFIP, and is asserted prior to a refresh cycle and is negated when all the RAS outputs are negated for that refresh. When ECAS0 is negated during programming, this output functions as RFRQ. When asserted, this pin specifies that 13 ms or 15 ms have passed. If DISRFSH is negated, the DP8520A/21A/22A will perform an internal refresh as soon as possible. If DISRFRSH is asserted, RFRQ can be used to externally request a refresh through the input RFSH. This output has a high capacitive driver and a 20X series damping resistor. RFSH I REFRESH: This input asserted with DISRFRSH already asserted will request a refresh. If this input is continually asserted, the DP8520A/21A/22A will perform refresh cycles in a burst refresh fashion until the input is negated. If RFSH is asserted with DISRFSH negated, the internal refresh address counter is cleared (useful for burst refreshes). DISRFSH I DISABLE REFRESH: This input is used to disable internal refreshes and must be asserted when using RFSH for externally requested refreshes. ADS (ALE) I I ADDRESS STROBE or ADDRESS LATCH ENABLE: Depending on programming, this input can function as ADS or ALE. In mode 0, the input functions as ALE and when asserted along with CS causes an internal latch to be set. Once this latch is set an access will start from the positive clock edge of CLK as soon as possible. In Mode 1, the input functions as ADS and when asserted along with CS, causes the access RAS to assert if no other event is taking place. If an event is taking place, RAS will be asserted from the positive edge of CLK as soon as possible. In both cases, the low going edge of this signal latches the bank, row and column address if programmed to do so. CS I CHIP SELECT: This input signal must be asserted to enable a Port A access. AREQ I ACCESS REQUEST: This input signal in Mode 0 must be asserted some time after the first positive clock edge after ALE has been asserted. When this signal is negated, RAS is negated for the access. In Mode 1, this signal must be asserted before ADS can be negated. When this signal is negated, RAS is negated for the access. WAIT (DTACK) O O WAIT or DTACK: This output can be programmed to insert wait states into a CPU access cycle. With R7 negated during programming, the output will function as a WAIT type output. In this case, the output will be active low to signal a wait condition. With R7 asserted during programming, the output will function as DTACK. In this case, the output will be negated to signify a wait condition and will be asserted to signify the access has taken place. Each of these signals can be delayed by a number of positive clock edges or negative clock levels of CLK to increase the microprocessor’s access cycle through the insertion of wait states. WAITIN I WAIT INCREASE: This input can be used to dynamically increase the number of positive clock edges of CLK until DTACK will be asserted or WAIT will be negated during a VRAM access. 2.4 PORT A ACCESS 2.5 PORT B ACCESS SIGNALS AREQB DP8522A only I PORT B ACCESS REQUEST: This input asserted will latch the row, column and bank address if programmed, and requests an access to take place for Port B. If the access can take place, RAS will assert immediately. If the access has to be delayed, RAS will assert as soon as possible from a positive edge of CLK. ATACKB DP8522A only O ADVANCED TRANSFER ACKNOWLEDGE PORT B: This output is asserted when the access RAS is asserted for a Port B access. This signal can be used to generate the appropriate DTACK or WAIT type signal for Port B’s CPU or bus. 8 2.0 Signal Descriptions (Continued) Pin Name Device (If not applicable to all) Input/ Output Description 2.6 COMMON DUAL PORT SIGNALS GRANTB DP8522A only O GRANT B: This output indicates which port is currently granted access to the VRAM array. When GRANTB is asserted, Port B has access to the array. When GRANTB is negated, Port A has access to the VRAM array. This signal is used to multiplex the signals R0 – 8, 9, 10; C0 – 8, 9, 10; B0 – 1; WIN; LOCK and ECAS0 –1 to the DP8522A when using dual accessing. LOCK DP8522A only I LOCK: This input can be used by the currently granted port to ‘‘lock out’’ the other port from the VRAM array by inserting wait states into the locked out port’s access cycle until LOCK is negated. AVSRLRQ I ADVANCED VIDEO SHIFT REGISTER LOAD REQUEST: This must precede the VSRL input going low by the amount of time necessary to guarantee that any currently executing access and pending refresh can finish. This input disables Port B and refresh requests until four CLK periods after VSRL has transitioned low. This input may be held low until the video RAM transfer cycle is completed or may be momentarily pulsed low. VSRL I VIDEO SHIFT REGISTER LOAD: This input causes the DT/OE output to transition low immediately. Therefore, when executing a video RAM shift register load, VSRL transitions low before RAS goes low. The DT/OE output will transition high from VSRL going high or four CLK periods (rising clock edges) from VSRL going low, whichever occurs first. VSRL low also disables the WIN input from affecting the DT/OE logic, until the video shift register load access is over. 2.7 VRAM TRANSFER CYCLE SIGNALS 2.8 POWER SIGNALS AND CAPACITOR INPUT VCC I POWER: Supply Voltage. GND I GROUND: Supply Voltage Reference. CAP I CAPACITOR: This input is used by the internal PLL for stabilization. The value of the ceramic capacitor should be 0.1 mF and should be connected between this input and ground. 2.9 CLOCK INPUTS There are two clock inputs to the DP8520A/21A/22A, CLK and DELCLK. These two clocks may both be tied to the same clock input, or they may be two separate clocks, running at different frequencies, asynchronous to each other. CLK I SYSTEM CLOCK: This input may be in the range of 0 Hz up to 25 MHz. This input is generally a constant frequency but it may be controlled externally to change frequencies or perhaps be stopped for some arbitrary period of time. This input provides the clock to the internal state machine that arbitrates between accesses and refreshes. This clock’s positive edges and negative levels are used to extend the WAIT (DTACK) signals. Ths clock is also used as the reference for the RAS precharge time and RAS low time during refresh. All Port A and Port B accesses are assumed to be synchronous to the system clock CLK. DELCLK I DELAY LINE CLOCK: The clock input DELCLK, may be in the range of 6 MHz to 20 MHz and should be a multiple of 2 (i.e., 6, 8, 10, 12, 14, 16, 18, 20 MHz) to have the DP8520A/21A/22A switching characteristics hold. If DELCLK is not one of the above frequencies the accuracy of the internal delay line will suffer. This is because the phase locked loop that generates the delay line assumes an input clock frequency of a multiple of 2 MHz. For example, if the DELCLK input is at 7 MHz and we choose a divide by 3 (program bits C0 – 2) this will produce 2.333 MHz which is 16.667% off of 2 MHz. Therefore, the DP8520A/21A/22A delay line would produce delays that are shorter (faster delays) than what is intended. If divide by 4 was chosen the delay line would be longer (slower delays) than intended (1.75 MHz instead of 2 MHz). (See Section 10 for more information.) This clock is also divided to create the internal refresh clock. 9 3.0 Port A Access Modes Sometime after the first positive edge of CLK after ALE and CS have been asserted, the input AREQ must be asserted. In single port applications, once AREQ has been asserted, CS can be negated. Once AREQ is negated, RAS and DTACK, if programmed, will be negated. If ECAS0 is asserted during programming, CAS will be negated with AREQ. If ECAS0 was negated during programming, a single CAS or group of CASs will continue to be asserted after RAS has been negated given that the appropriate ECASs inputs were asserted as shown in Figure 5b . This allows the VRAM to have data present on the data out bus while gaining RAS precharge time. ALE can stay asserted several periods of CLK. However, ALE must be negated before or during the period of CLK in which AREQ is negated. When performing address pipelining, the ALE input cannot be asserted to start another access until AREQ has been asserted for at least one clock period of CLK for the present access. The DP8520A/21A/22A have two general purpose access modes. With one of these modes, any microprocessor can be interfaced to VRAM. A Port A access to VRAM is initiated by two input signals: ADS (ALE) and CS. The access is always terminated by one signal: AREQ. These input signals should be synchronous to the input clock, CLK. One of these access modes is selected at programming through the B1 input signal. In both modes, once an access has been requested by CS and ADS (ALE), the DP8522A will guarantee the following: The DP8520A/21A/22A will have the row address valid to the VRAMs’ address bus, Q0–8, 9, 10 given that the row address setup time to the DP8520A/21A/22A was met; The DP8520A/21A/22A will bring the appropriate RAS or RASs low; The DP8520A/21A/22A will guarantee the minimum row address hold time, before switching the internal multiplexor to place the column address on the VRAM address bus, Q0 – 8, 9, 10; The DP8520A/21A/22A will guarantee the minimum column address setup time before asserting the appropriate CAS or CASs; The DP8520A/21A/22A will hold the column address valid the minimum specified column address hold time in address pipelining mode and will hold the column address valid the remainder of the access in non-pipelining mode. The chip includes a WIN pin to signify a write operation to the DP8520A/21A/22A. When asserted, WIN will cause CAS to delay to the next positive clock edge if address bit C9 is asserted during programming. When negated, WIN will cause the DT/OE output to follow the CAS outputs for a read access, if ECAS0 is negated during programming. WE, write enable, must be externally gated from the processor to the VRAM as there is no output pin from the WIN input pin available on chip. 3.2 ACCESS MODE 1 Access Mode 1, shown in Figure 6a , is selected by asserting the input B1 during programming. This mode allows accesses, which are not delayed by precharge, Port B access, VRAM transfer cycle or refresh, to start immediately from the access request input, ADS. To initiate a Mode 1 access, CS is asserted followed by ADS asserted. If the programmed precharge time from the last access or VRAM refresh had been met and a refresh of the VRAM, a Port B access to the VRAM, or a VRAM transfer cycle was not in progress, the RAS or group of RASs selected by programming and the bank select inputs would be asserted from ADS being asserted. If a VRAM refresh, a Port B access, or a VRAM transfer cycle is in progress or precharge time is required, the controller will wait until these events have taken place and assert RAS or the group of RASs from the next positive edge of CLK. When ADS is asserted or sometime after, AREQ must be asserted. At this time, ADS can be negated and AREQ will continue the access. Once AREQ is negated, RAS and DTACK, if programmed, will be negated. If ECAS0 was asserted during programming, CAS will be negated with AREQ. If ECAS0 was negated during programming, a single CAS or group of CASs will continue to be asserted after RAS has been negated given that the appropriate ECAS inputs were asserted as shown in Figure 6b . This allows a VRAM to have data present on the data out bus while gaining RAS precharge time. ADS can continue to be asserted after AREQ has been asserted and negated, however a new access would not be started until ADS is negated and asserted again. ADS and AREQ can be tied together in applications not using address pipelining. If address pipelining is programmed, it is possible for ADS to be negated after AREQ is asserted. Once AREQ is asserted, ADS can be asserted again to initiate a new access. 3.1 ACCESS MODE 0 Access Mode 0, shown in Figure 5a, is selected by negating the input B1 during programming. This access mode allows accesses to VRAM to always be initiated from the positive edge of the system input clock, CLK. To initiate a Mode 0 access, ALE is pulsed high and CS is asserted. Pulsing ALE high and asserting CS, sets an internal latch which requests an access. If the precharge time from the last access or VRAM refresh had been met and a refresh of VRAM, a Port B access, or a VRAM transfer cycle was not in progress, the RAS or group of RASs would be initiated from the first positive edge of CLK. If a VRAM refresh is in progress or precharge time is required, the controller will wait until these events have taken place and assert RAS on the next positive edge of CLK. 10 3.0 Port A Access Modes (Continued) TL/F/9338 – 6 FIGURE 5a. Access Mode 0 TL/F/9338 – 7 FIGURE 5b. Access Mode 0 Extending CAS 11 3.0 Port A Access Modes (Continued) TL/F/9338 – 8 FIGURE 6a. Access Mode 1 TL/F/9338 – 9 FIGURE 6b. Access Mode 1 Extending CAS 12 4.0 Refresh Options Pulsing RFSH low, sets an internal latch, that is used to produce the internal refresh request. The refresh cycle will take place on the next positive edge of CLK as shown in Figure 7b . If an access to VRAM is in progress or precharge time for the last access has not been met, the refresh will be delayed. Since pulsing RFSH low sets a latch, the user does not have to keep RFSH low until the refresh starts. When the last refresh RAS negates, the internal refresh request latch is cleared. By keeping RFSH asserted past the positive edge of CLK which ends the refresh cycle as shown in Figure 8 , the user will perform another refresh cycle. Using this technique, the user can perform a burst refresh consisting of any number of refresh cycles. Each refresh cycle during a burst refresh will meet the refresh RAS low time and the RAS precharge time (programming bits R0 – 1). If the user desires to burst refresh the entire VRAM (all row addresses) he could generate an end of count signal (burst refresh finished) by looking at one of the DP8520A/21A/ 22A high address outputs (Q7, Q8, Q9 or Q10) and the RFIP output. The Qn outputs function as a decode of how many row addresses have been refreshed (Q7 e 128 refreshes, Q8 e 256 refreshes, Q9 e 512 refreshes, Q10 e 1024 refreshes). The DP8520A/21A/22A support a wide variety of refresh control mode options including automatic internally controlled refresh, externally controlled/burst refresh, refresh request/acknowledge and any combination of the above. With each of the control modes above, different types of refreshes can be performed. These different types include all RAS refresh, staggered refresh and error scrubbing during all RAS refresh. There are three inputs, EXTNDRF, RFSH and DISRFSH, and one output, RFIP (RFRQ), associated with refresh. There are also ten programming bits; R0–1, R9, C0 – 6 and ECAS0 used to program the various types of refreshing. The two inputs, RFSH and DISRFSH, are used in the externally controlled/burst refresh mode and the refresh request/acknowledge mode. The output RFRQ is used in the refresh request/acknowledge mode. The input EXTNDRF is used in all refresh modes and the output RFIP is used in all refresh modes except the refresh request/acknowledge mode. Asserting the input EXTNDRF, extends the refresh cycle single or multiple integral clock periods of CLK. The output RFIP is asserted one period of CLK before the first refresh RAS is asserted. If an access is currently in progress, RFIP will be asserted up to one period of CLK before the first refresh RAS, once AREQ or AREQB is negated for the access (see Figure 7a ). The DP8520A/21A/22A will increment the refresh address counter automatically, independent of the refresh mode used. The refresh address counter will be incremented once all the refresh RASs have been negated. In every combination of refresh control mode and refresh type, the DP8520A/21A/22A is programmed to keep RAS asserted a number of CLK periods. The values of RAS low time during refresh are programmed with the programming bits R0 and R1. 4.1.3 Refresh Request/Acknowledge The DP8520A/21A/22A can be programmed to output internal refresh requests. When the user programs ECAS0 negated during programming, the RFIP output functions as RFRQ. RFRQ will be asserted from a positive edge of CLK as shown in Figure 9a. Once RFRQ is asserted, it will stay asserted until the RFSH is pulsed low with DISRFSH asserted. This will cause an externally requested/burst refresh to take place. If DISRFSH is negated, an automatic internal refresh will take place as shown in Figure 9b . RFRQ will go high and then assert if additional periods of the internal refresh clock have expired and neither an externally controlled refresh nor an automatically controlled internal refresh have taken place as shown in Figure 9c . If a time critical event, or long access like page/static column mode access can not be interrupted, RFRQ pulsing high can be used to increment a counter. The counter can be used to perform a burst refresh of the number of refreshes missed (through the RFSH input). 4.1 REFRESH CONTROL MODES There are three different modes of refresh control. Any of these modes can be used in combination or singularly to produce the desired refresh results. The three different modes of control are: automatic internal refresh, external/ burst refresh and refresh request/acknowledge. 4.1.1 Automatic Internal Refresh The DP8520A/21A/22A have an internal refresh clock. The period of the refresh clock is generated from the programming bits C0–3. Every period of the refresh clock, an internal refresh request is generated. As long as a VRAM access is not currently in progress and precharge time has been met, the internal refresh request will generate an automatic internal refresh. If a VRAM access is in progress, the DP8520A/21A/22A on-chip arbitration logic will wait until the access is finished before performing the refresh. The refresh/access arbitration logic can insert a refresh cycle between two address pipelined accesses. However, the refresh arbitration logic can not interrupt an access cycle to perform a refresh. To enable automatic internally controlled refreshes, the input DISRFSH must be negated. 4.2 REFRESH CYCLE TYPES Three different types of refresh cycles are available for use. The three different types are mutually exclusive and can be used with any of the three modes of refresh control. The three different refresh cycle types are: all RAS refresh, staggered RAS refresh and error scrubbing during all RAS refresh. In all refresh cycle types, the RAS precharge time is guaranteed: between the previous access RAS ending and the refresh RAS0 starting; between refresh RAS3 ending and access RAS beginning; between burst refresh RASs. 4.2.1 Conventional RAS Refresh A conventional refresh cycle causes RAS0 – 3 to all assert from the first positive edge of CLK after RFIP is asserted as shown in Figure 10 . RAS0 – 3 will stay asserted until the number of positive edges of CLK programmed have passed. On the last positive edge, RAS0 – 3, and RFIP will be negated. This type of refresh cycle is programmed by negating address bit R9 during programming. 4.1.2 Externally Controlled/Burst Refresh To use externally controlled/burst refresh, the user must disable the automatic internally controlled refreshes by asserting the input DISRFSH. The user is responsible for generating the refresh request by asserting the input RFSH. 13 4.0 Refresh Options (Continued) TL/F/9338 – 10 Explanation of Terms RFRQ e ReFresh ReQuest internal to the DP8520A/21A/22A. RFRQ has the ability to hold off a pending access. RFSH e Externally requested ReFreSH RFIP e ReFresh In Progress ACIP e Port A or Port B (DP8522A only) ACcess In Progress. This means that either RAS is low for an access or is in the process of transitioning low for an access. FIGURE 7a. DP8520A/21A/22A Access/Refresh Arbitration State Program TL/F/9338 – 11 FIGURE 7b. Single External Refresh (2 Periods of RAS Low during Refresh Programmed) 14 4.0 Refresh Options (Continued) TL/F/9338 – 12 FIGURE 8. External Burst Refresh (2 Periods of RAS Precharge, 2 Periods of Refresh RAS Low during Refresh Programmed) TL/F/9338 – 13 FIGURE 9a. Externally Controlled Single and Burst Refresh with Refresh Request (RFRQ) Output (2 Periods of RAS Low during Refresh Programmed) TL/F/9338 – 14 FIGURE 9b. Automatic Internal Refresh with Refresh Request (3T of RAS low during refresh programmed) *In Figures 9a and 9b , where the DP8520A/21A/22A operate in the refresh request/acknowledge mode, the RFIP output pin functions as RFRQ. An RFIP timing waveform is included in the figure solely for the purpose of simplifying the diagrams. 15 16 FIGURE 11. Staggered RAS Refresh FIGURE 10. Conventional RAS Refresh FIGURE 9c. Refresh Request Timing TL/F/9338 – 17 TL/F/9338 – 16 TL/F/9338 – 15 4.0 Refresh Options (Continued) 4.0 Refresh Options (Continued) lows a CAS or group of CASs to assert during the all RAS refresh as shown in Figure 12 . This allows data to be read from the VRAM array and passed through an Error Detection And Correction Chip, EDAC. It is important to note that while an error scrubbing during refresh access is being peformed, it is the system designer’s responsibility to properly control the WE input of the VRAM. WE should be high during the initial access of the VRAM, which could be accomplished by gating RFIP, if programmed, with the processor access circuitry that creates WE. If the EDAC determines that the data contains a single bit error and corrects that error, the refresh cycle can be extended with the input extend refresh, EXTNDRF, and a read-modify-write operation can be performed, and the corrected data can be written back to the VRAM by bringing WE low. The DP8522A has a 24-bit internal refresh address counter that contains the 11 row, 11 column and 2 bank addresses. The DP8520A/21A have a 22-bit internal refresh address counter that contains the 10 row, 10 column and 2 bank addresses. These counters are configured as bank, column, row with the row address as the least significant bits. The bank counter bits are then used with the programming selection to determine which CAS or group of CASs will assert during a refresh. 4.2.2 Staggered RAS Refresh A staggered refresh staggers each RAS or group of RASs by a positive edge of CLK as shown in Figure 11 . The number of RASs, which will be asserted on each positive edge of CLK, is determined by the RAS, CAS configuration mode programming bits C4–C6. If single RAS outputs are selected during programming, then each RAS will assert on successive positive edges of CLK. If two RAS outputs are selected during programming then RAS0 and RAS1 will assert on the first positive edge of CLK after RFIP is asserted. RAS2 and RAS3 will assert on the second positive edge of CLK after RFIP is asserted. If all RAS outputs were selected during programming, all RAS outputs would assert on the first positive edge of CLK after RFIP is asserted. Each RAS or group of RASs will meet the programmed RAS low time and then negate. 4.2.3 Error Scrubbing during Refresh The DP8520A/21A/22A support error scrubbing during all RAS VRAM refreshes. Error scrubbing during refresh is selected through bits C4–C6 with bit R9 negated during programming. Error scrubbing can not be used with staggered refresh (see Section 9.0). Error scrubbing during refresh al- TL/F/9338 – 18 FIGURE 12. Error Scrubbing during Refresh 17 4.0 Refresh Options (Continued) ry array. By asserting RFSH one period of CLK before DISRFSH is asserted and then keeping both inputs asserted, the DP8520A/21A/22A will clear the refresh address counter and then perform refresh cycles separated by the programmed value of precharge as shown in Figure 14b . An end-of-count signal can be generated from the Q VRAM address outputs of the DP8520A/21A/22A and used to negate RFSH. 4.3 EXTENDING REFRESH The programmed number of periods of CLK that refresh RASs are asserted can be extended by one or multiple periods of CLK. Only the all RAS (with or without error scrubbing) type of refresh can be extended. To extend a refresh cycle, the input extend refresh, EXTNDRF, must be asserted before the positive edge of CLK that would have negated all the RAS outputs during the refresh cycle and after the positive edge of CLK which starts all RAS outputs during the refresh as shown in Figure 13 . This will extend the refresh to the next positive edge of CLK and EXTNDRF will be sampled again. The refresh cycle will continue until EXTNDRF is sampled low on a positive edge of CLK. 4.5 CLEARING THE REFRESH REQUEST CLOCK The refresh request clock can be cleared by negating DISRFSH and asserting RFSH for 500 ns, one period of the internal 2 MHz clock as shown in Figure 15 . By clearing the refresh request clock, the user is guaranteed that an internal refresh request will not be generated for approximately 15 ms, one refresh clock period, from the time RFSH is negated. This action will also clear the refresh address counter. 4.4 CLEARING THE REFRESH ADDRESS COUNTER The refresh address counter can be cleared by asserting RFSH while DISRFSH is negated as shown in Figure 14a . This can be used prior to a burst refresh of the entire memo- TL/F/9338 – 19 FIGURE 13. Extending Refresh with the Extend Refresh (EXTNDRF) Input TL/F/9338 – 20 FIGURE 14a. Clearing the Refresh Address Counter TL/F/9338 – 21 FIGURE 14b. Clearing the Refresh Counter during Burst TL/F/9338 – 22 FIGURE 15. Clearing the Refresh Request Clock Counter 18 5.0 Port A Wait State Support asserted by the CPU, wait states (extra clock periods) are inserted into the current access cycle as shown in Figure 16 . Once WAIT is sampled negated, the access cycle is completed by the CPU. WAIT is asserted at the beginning of a chip selected access and is programmed to negate a number of positive edges and/or negative levels of CLK from the event that starts the access. WAIT can also be programmed to function in page/burst mode applications. Once WAIT is negated during an access, and the ECAS inputs are negated with AREQ asserted, WAIT can be programmed to toggle, following the ECAS inputs. Once AREQ is negated, ending the access, WAIT will stay negated until the next chip selected access. For more details about WAIT Type Output, see Application Note AN-773. Wait states allow a CPU’s access cycle to be increased by one or multiple CPU clock periods. The wait or ready input is named differently by CPU manufacturers. However, any CPU’s wait or ready input is compatible with either the WAIT or DTACK output of the DP8520A/21A/22A. The user determines whether to program WAIT or DTACK (R7) and which value to select for WAIT or DTACK (R2, R3) depending upon the CPU used and where the CPU samples its wait input during an access cycle. The decision to terminate the CPU access cycle is directly affected by the speed of the VRAMs used. The system designer must ensure that the data from the VRAMs will be present for the CPU to sample or that the data has been written to the VRAM before allowing the CPU access cycle to terminate. The insertion of wait states also allows a CPU’s access cycle to be extended until the VRAM access has taken place. The DP8520A/21A/22A insert wait states into CPU access cycles due to; guaranteeing precharge time, refresh currently in progress, user programmed wait states, the WAITIN signal being asserted and GRANTB not being valid (DP8522A only). If one of these events is taking place and the CPU starts an access, the DP8520A/21A/22A will insert wait states into the access cycle, thereby increasing the length of the CPU’s access. Once the event has been completed, the DP8520A/21A/22A will allow the access to take place and stop inserting wait states. There are six programming bits, R2–R7; an input, WAITIN; and an output that functions as WAIT or DTACK. 5.2 DTACK TYPE OUTPUT With the R7 address bit asserted during programming, the user selects the DTACK type output. As long as DTACK is sampled negated by the CPU, wait states are inserted into the current access cycle as shown in Figure 17. Once DTACK is sampled asserted, the access cycle is completed by the CPU. DTACK, which is normally negated, is programmed to assert a number of positive edges and/or negative levels from the event that starts RAS for the access. DTACK can also be programmed to function during page/ burst mode accesses. Once DTACK is asserted and the ECAS inputs are negated with AREQ asserted, DTACK can be programmed to negate and assert from the ECAS inputs toggling to perform a page/burst mode operation. Once AREQ is negated, ending the access, DTACK will be negated and stays negated until the next chip selected access. For more details about DTACK Type Output, see Application Note AN-773. 5.1 WAIT TYPE OUTPUT With the R7 address bit negated during programming, the user selects the WAIT output. As long as WAIT is sampled TL/F/9338 – 23 FIGURE 16. WAIT Type Output TL/F/9338 – 44 FIGURE 17. DTACK Type Output 19 5.0 Port A Wait State Support (Continued) will then insert wait states into the transfer cycle. The transfer cycle is completed from either VSRL negating or four clocks from VSRL asserting. The first event of these two to take place causes WAIT to negate (DTACK to assert) immediately or one half system clock period later, depending on how the user had programmed WAIT to end (DTACK to start) during a non-burst type of access. The wait logic is intimately connected to the DP8520A/21A/22A graphics functions and the WAIT output functions the same as the DT/OE output (see Figure 18 ). 5.3 WAIT STATE SUPPORT FOR VIDEO RAM SHIFT REGISTER LOAD OPERATIONS FOR PORT A If using the DP8520A/21A/22A in a system using video VRAMs, the CPU that controls loading the Video RAM shift register must be connected to Port A. The input AVSRLRQ asserts, signaling an advanced request for a Video RAM shift register load operation. Sometime later, the input VSRL asserts, signifying that the transfer cycle has started, and this action causes the DT/OE output to transfer low. VSRL asserting also asserts WAIT (keeps DTACK negated) and TL/F/9338 – 74 FIGURE 18. Wait State Timing during a VRAM Transfer Cycle (WAIT Programmed as OT, WAIT Sampled at the ‘‘T3’’ Rising Clock Edge) 20 5.0 Port A Wait State Support (Continued) 5.4 DYNAMICALLY INCREASING THE NUMBER OF WAIT STATES 5.5 GUARANTEEING RAS LOW TIME AND RAS PRECHARGE TIME The user can increase the number of positive edges of CLK before DTACK is asserted or WAIT is negated. With the input WAITIN asserted, the user can delay DTACK asserting or WAIT negating either one or two more positive edges of CLK. The number of edges is programmed through address bit R6. If the user is increasing the number of positive edges in a delay that contains a negative level, the positive edges will be met before the negative level. For example if the user programmed DTACK of (/2T, asserting WAITIN, programmed as 2T, would increase the number of positive edges resulting in DTACK of 2(/2T as shown in Figure 19. Similarly, WAITIN can increase the number of positive edges in a page/burst access. WAITIN can be permanently asserted in systems requiring an increased number of wait states. WAITIN can also be asserted and negated, depending on the type of access. As an example, a user could invert the WRITE line from the CPU and connect the output to WAITIN. This could be used to perform write accesses with 1 wait state and read accesses with 2 wait states as shown in Figure 20 . The DP8520A/21A/22A will guarantee RAS precharge time between accesses; between refreshes; and between access and refreshes. The programming bits R0 and R1 are used to program combinations of RAS precharge time and RAS low time referenced by positive edges of CLK. RAS low time is programmed for refreshes only. During an access, the system designer guarantees the time RAS is asserted through the DP8520A/21A/22A wait logic. Since inserting wait states into an access increases the length of the CPU signals which are used to create ADS or ALE and AREQ, the time that RAS is asserted can be guaranteed. Precharge time is also guaranteed by the DP8520A/21A/ 22A. Each RAS output has a separate positive edge of CLK counter. AREQ is negated setup to a positive edge of CLK to terminate the access. That positive edge is 1T. The next positive edge is 2T. RAS will not be asserted until the programmed number of positive edges of CLK have passed as shown in Figure 21. Once the programmed precharge time has been met, RAS will be asserted from the positive edge of CLK. However, since there is a precharge counter per RAS, an access using another RAS will not be delayed. Precharge time before a refresh is always referenced from the access RAS negating before RAS0 for the refresh asserting. After a refresh, precharge time is referenced from RAS3 negating, for the refresh, to the access RAS asserting. TL/F/9338 – 75 FIGURE 19. WAITIN Example (DTACK is Sampled at the ‘‘T3’’ Falling Clock Edge) 21 5.0 Port A Wait State Support (Continued) TL/F/9338 – 76 FIGURE 20. WAITIN Example (WAIT is Sampled at the End of ‘‘T2’’) TL/F/9338 – 77 FIGURE 21. Guaranteeing RAS Precharge (DTACK is Sampled at the ‘‘T2’’ Falling Clock Edge) 22 6.0 DP8520A/21A/22A Video RAM Support The DP8520A/21A/22A provides full support for all access modes of video RAMs through the addition of three pins (AVSRLRQ, VSRL, and DT/OE) to the standard DP8420A/21A/22A. The access modes of video RAMs can be split up into two groups; video RAM transfer cycles (read with the serial port in active or in standby mode, write, and pseudo write transfer cycles), and non-transfer cycles. The DP8520A/21A/22A support of video RAMs allows the full capabilities of the National Semiconductor Advanced Graphics chip set (DP8500 Series) to be realized. See Figures 22, 23, and 58a. TL/F/9338 – 52 FIGURE 22. The Video RAM (A Dual Ported Memory) Ideal solution for graphics frame buffer. Screen refresh can occur at the same time as random access to the frame buffer for screen update and manipulation. TL/F/9338 – 56 FIGURE 23. The DP8500 Raster Graphics Processor Interfaced to the DP8520A/21A/22A Video RAM Controller 23 6.0 DP8520A/21A/22A Video RAM Support (Continued) completed. Figure 23 shows the case of an externally requested refresh being disabled, because of a previous AVSRLRQ, until the VRAM shift register load has been completed. The VSRL input causes the DT/OE output to assert immediately, regardless of what else may be happening in the DP8520A/21A/22A. Therefore, it is the system designer’s responsibility to guarantee that all pending accesses have been completed by the time the VSRL input asserts. The system designer can guarantee this by issuing AVSRLRQ far enough in advance to guarantee that all pending accesses have been completed by the time VSRL asserts. The AVSRLRQ input does not override the LOCK input (see Section 12.0) for dual port systems, and as a result, the designer must also guarantee that Port A can be accessed by assuring that GRANTB and LOCK are both not asserted when AVSRLRQ is asserted. Generally, the VSRL is the status of the upcoming access cycle (of the graphics processor). Therefore, this input precedes the inputs ADS and AREQ that execute the VRAM shift register load transfer cycle. This sequence of events guarantees the correct relationship of DT/OE, RAS and CAS (DT preceding RAS and CAS when asserting and negating). The wait logic is also intimately connected to the graphics functions on the DP8520A/21A/22A. The DT/OE (and WAIT/DTACK) relationship to VSRL during a VRAM transfer cycle depends upon how the DP8520A/21A/22A was programmed with respect to the ECAS0 input. If ECAS0 was negated during programming, the DT/OE output will follow the VSRL input. If ECAS0 was asserted during programming, the DT/OE output will follow VSRL asserting. DT/OE will then negate either when VSRL negates or from the fourth rising clock edge after VSRL asserted, whichever event takes place first. This allows DT to negate before RAS and CAS negate, thus guaranteeing the correct timing relationship during the transfer cycle (see Figure 23 ). The WE input of the VRAM determines whether the access is a read or write transfer cycle (see Figures 24 and 25 respectively). 6.1 SUPPORT FOR VRAM TRANSFER CYCLES (TO THE SERIAL PORT OF THE VRAM) The DP8520A/21A/22A supports VRAM transfer cycles with the serial port in the active or standby mode. Active or standby refers to whether data is or is not currently being shifted in or out of the VRAM serial port (i.e., whether the shift clock (SCLK) is currently active). The DP8520A/ 21A/22A support for data transfer cycles with the serial port in the active mode includes the ability to support transfer cycles with the serial port in the standby mode. Hereafter, the term VRAM transfer cycle means VRAM transfer cycle with the serial port in the active mode. In order to support VRAM transfer cycles, the DP8520A/ 21A/22A must be able to guarantee timing with respect to its input CLK (which must be synchronous to VRAM shift clock), RAS, CAS, and DT/OE. Figure 23 shows the timing of a graphics memory system where the DP8520A/21A/22A is being used with the National Semiconductor DP8500 Raster Graphics Processor (RGP). If the DP8520A/21A/22A is being used in a graphics frame buffer application, it has the ability to support a VRAM transfer cycle during active video time (ex. mid scan line). This is one of the very attractive features supported by the National Semiconductor Advanced Graphics chip set. Most of the commercial graphics controller chip sets available will only support VRAM transfer cycles during blanking periods (while the VRAM is in standby mode). The DP8520A/21A/22A supports VRAM transfer cycles during active video time by being able to guarantee an exact instant during which the transfer of VRAM data to the VRAM shift register will occur. This exact instant can be guaranteed through the AVSRLRQ and VSRL inputs. The input AVSRLRQ disables any further internally or externally requested refreshes or Port B access requests from being executed. The AVSRLRQ input does this by making the VRAM controller arbitration logic think that a Port A access is in progress from the point where the AVSRLRQ input asserts until the VRAM shift register load operation is 24 6.0 DP8520A/21A/22A Video RAM Support (Continued) TL/F/9338 – 78 FIGURE 24. Video RAM Timing READ Transfer Cycle, B Port Active (Transfer VRAM Row Data into Shift Register) TL/F/9338 – 55 FIGURE 25. Video RAM Timing WRITE Transfer Cycle, B Port Active (Transfer Shift Register Data into VRAM Row) 25 6.0 DP8520A/21A/22A Video RAM Support (Continued) to take place from the serial port. In addition to both read and write transfer cycles, the DP8520A/21A/22A also supports pseudo write transfer cycles (see Figure 26 ). A pseudo write transfer cycle must be performed after a read transfer cycle if the subsequent operation is a write transfer cycle. The DP8520A/21A/22A VRAM controller is operated as if it is doing a write transfer cycle, but since the SOE input to the VRAM is negated (disabling the serial port), a transfer doesn’t take place. The purpose of this pseudo write transfer cycle is to switch the SI/O (0 – 3) lines of the VRAM’s serial port from output mode to input mode. During a transfer cycle (VSRL asserted during the access) WIN is disabled from affecting the DT/OE logic until the transfer cycle is completed as shown by CAS negating. During a transfer cycle, the SOE (Serial Output Enable) input to the VRAM is asserted and is used as an output control for a read transfer cycle and is used as a write enable control during a write transfer cycle. When SOE is negated, serial access is disabled, and a transfer cycle cannot take place. SOE asserted during a read enables the serial input/output bus SI/O (0 – 3) while the VRAM data bus (DQ0–3) is put into a high impedance state, thus allowing the transfer cycle TL/F/9338 – 79 FIGURE 26. Video RAM Timing Pseudo WRITE Transfer Cycle, B Port Active (Transfer Shift Register Data into VRAM Row) 26 6.0 DP8520A/21A/22A Video RAM Support (Continued) be negated for a read access once CASn is negated (see Figure 28 ), causing the VRAM outputs to be enabled. If CASn toggles during a page mode read access, then the DT/OE logic will also toggle following the CASn input. 6.2 SUPPORT FOR VRAM ACCESS CYCLES THROUGH PORT A USING THE DP8520A/21A/22A With the DP8520A/21A/22A, the output DT/OE will remain negated during write accesses (see Figure 27 ), but during read accesses it will assert after CASn asserts. DT/OE will TL/F/9338 – 54 FIGURE 27. Video RAM Random Access Write Cycle TL/F/9338 – 53 FIGURE 28. Video RAM Random Access Read Cycle 27 7.0 Additional Access Support Features For Port B, if GRANTB is asserted, the address will be latched with AREQB asserted. If GRANTB is negated, the address will latch on the first or second positive edge of CLK after GRANTB is asserted depending on the programming bits R0, R1. To support the different modes of accessing, the DP8520A/ 21A/22A have multiple access features. These features allow the user to take advantage of CPU or VRAM functions. These additional features include: address latches and column increment for page/burst mode support; address pipelining to allow a new access to start to a different bank of VRAM after CAS has been asserted and the column address hold time has been met; and delay CAS, to allow the user with a multiplexed bus to ensure valid data is present before CAS is asserted. 7.2 ADDRESS PIPELINING Address pipelining is the overlapping of accesses to different banks of VRAM. If the majority of successive accesses are to a different bank, the accesses can be overlapped. Because of this overlapping, the cycle time of the VRAM accesses are greatly reduced. The DP8520A/21A/22A can be programmed to allow a new row address to be placed on the VRAM address bus after the column address hold time has been met. At this time, a new access can be initiated with ADS or ALE, depending on the access mode, while AREQ is used to sustain the current access. The DP8522A supports address pipelining for Port A only. This mode can not be used with page, static column or nibble modes of operations because the VRAM column address is switched back to the row address after CAS is asserted. This mode is programmed through address bit R8 (see Figures 30 and 31 ). During address pipelining in Mode 0, shown in Figure 32 , ALE cannot be pulsed high to start another access until AREQ has been asserted for the previous access for at least one period of CLK. DTACK, if programmed, will be negated once AREQ is negated. WAIT, if programmed to insert wait states, will be asserted once ALE and CS are asserted. In Mode 1, shown in Figure 33, ADS can be negated once AREQ is asserted. After meeting the minimum negated pulse width for ADS, ADS can again be asserted to start a new access. DTACK, if programmed, will be negated once AREQ is negated. WAIT, if programmed, will be asserted once ADS is asserted. In either mode with either type of wait programmed, the DP8520A/21A/22A will still delay the access for precharge if sequential accesses are to the same bank or if a refresh takes place. 7.1 ADDRESS LATCHES AND COLUMN INCREMENT The address latches can be programmed, through programming bit B0, to either latch the address or remain permanently in fall-through mode. If the address latches are used to latch the address, the rising edge of ALE in Mode 0 places the latches in fall-through. Once ALE is negated, the address present on the row, column and bank inputs is latched. In Mode 1, the address latches are in fall-through mode until ADS is asserted. ADS asserted latches the address. Once the address is latched, the column address can be incremented with the input COLINC. With COLINC asserted, the column address is incremented. If COLINC is asserted with all of the bits of the column address asserted, the column address will return to zero. COLINC can be used for sequential accesses of static column VRAMs. COLINC can also be used with the ECAS inputs to support sequential accesses to page mode VRAMs as shown in Figure 29. COLINC should only be asserted when a refresh is not in progress as indicated by RFIP, if programmed, being negated during an access since this input functions as an extend refresh when a refresh is in progress. The address latches function differently with the DP8522A. The DP8522A will latch the address of the currently granted port. If Port A is currently granted, the address will be latched as described in Section 7.1. If Port A is not granted, and requests an access, the address will be latched on the first or second positive edge of CLK after GRANTB has been negated depending on the programming bits R0, R1. 28 7.0 Additional Access Support Features (Continued) TL/F/9338 – 80 FIGURE 29. Column Increment TL/F/9338 – 81 FIGURE 30. Non-Address Pipelined Mode TL/F/9338 – 82 FIGURE 31. Address Pipelined Mode 29 FIGURE 33. Mode 1 Address Pipelining (DTACK 1(/2T Programmed, DTACK is Sampled at the ‘‘T3’’ Falling Clock Edge) FIGURE 32. Mode 0 Address Pipelining (WAIT of 0, (/2T has Been Programmed. WAIT is Sampled at the ‘‘T3’’ Falling Clock Edge) 30 TL/F/9338 – 84 TL/F/9338 – 83 7.0 Additional Access Support Features 7.0 Additional Access Support Features VRAM will be setup to CAS asserting as shown in Figures 34 and 35. If the possibility exists that data still may not be present after the first positive edge of CLK, CAS can be delayed further with the ECAS inputs. If address bit C9 is negated during programming, read and write accesses will be treated the same (with regard to CAS). 7.3 DELAY CAS DURING WRITE ACCESSES Address bit C9 asserted during programming will cause CAS to be delayed until the first positive edge of CLK after RAS is asserted when the input WIN is asserted. Delaying CAS during write accesses ensures that the data to be written to TL/F/9338 – 85 FIGURE 34. Mode 0 Delay CAS TL/F/9338 – 86 FIGURE 35. Mode 1 Delay CAS 31 8.0 RAS and CAS Configuration Modes The DP8520A/21A/22A allow the user to configure the VRAM array to contain one, two or four banks of VRAM. Depending on the functions used, certain considerations must be used when determining how to set up the VRAM array. Programming address bits C4, C5 and C6 along with bank selects, B0 –1, and CAS enables, ECAS0–1, determine which RAS or group of RASs and which CAS or group of CASs will be asserted during an access. Different memory schemes are described. The DP8520A/21A/22A is specified driving a heavy load of 72 VRAMs, representing four banks of VRAM with 16-bit words and 2 parity bits. The DP8520A/21A/22A can drive more than 72 VRAMs, but the AC timing must be increased. Since the RAS and CAS outputs are configurable, all RAS and CAS outputs should be used for the maximum amount of drive. 8.1 BYTE WRITING By selecting a configuration in which all CAS outputs are selected during an access, each ECAS input enables a pair of CAS outputs to select a byte in a word size of up to 16 bits. In this case, the RAS outputs are used to select which of up to 4 banks is to be used as shown in Figures 36 and 37. The user can also configure the VRAM array into an 8 bank system as shown in Figure 38. This setup can be used along with byte writing for an 8-bit system if the LOW BYTE and HIGH BYTE are connected to ECAS0 and ECAS1 respectively. The user can connect upper address bits to ECAS0,1 for use in an 8 bank – 16-bit system, but cannot use byte writing in this case. TL/F/9338 – 88 FIGURE 36. VRAM Array Setup for 16-Bit System (C6, C5, C4 e 1, 1, 0 during Programming) 32 8.0 RAS and CAS Configuration Modes (Continued) TL/F/9338 – 87 FIGURE 37. VRAM Array Setup for 16-Bit, 1 Bank System (C6, C5, C4 e 0, 0, 0 Allowing Error Scrubbing or C6, C5, C4 e 0, 1, 1 No Error Scrubbing during Programming) TL/F/9338 – 89 FIGURE 38. 8 Bank VRAM Array (C6, C5, C4 e 1, 1, 0 during Programming) 33 8.0 RAS and CAS Configuration Modes (Continued) 8.2 MEMORY INTERLEAVING 8.3 ADDRESS PIPELINING Memory interleaving allows the cycle time of VRAMs to be reduced by having sequential accesses to different memory banks. Since the DP8520A/21A/22A have separate precharge counters per bank, sequential accesses will not be delayed if the accessed banks use different RAS outputs. To ensure different RAS outputs will be used, a mode is selected where either one or two RAS outputs will be asserted during an access. The bank select or selects, B0 and B1, are then tied to the least significant address bits, causing a different group of RASs to assert during each sequential access as shown in Figure 39. In this figure there should be at least one clock period of all RAS’s negated between different RAS’s being asserted to avoid the condition of a CAS before RAS refresh cycle. Address pipelining allows several access RASs to be asserted at once. Because RASs can overlap, each bank requires either a mode where one RAS and one CAS are used per bank as shown in Figure 40 or where two RASs and two CASs are used per bank as shown in Figure 41. In order to perform byte writing while using address pipelining, external gating on the CAS outputs must be used. If the array is not layed out this way, a CAS to a bank can be low before RAS, which will cause a refresh of the VRAM, not an access. To take full advantage of address pipelining, memory interleaving is used. To memory interleave, the least significant address bits should be tied to the bank select inputs to ensure that all ‘‘back to back’’ sequential accesses are not delayed, since different memory banks are accessed. TL/F/9338 – 90 FIGURE 39. Memory Interleaving with Byte Writing Capability (C6, C5, C4 e 1, 1, 0 during Programming) 34 8.0 RAS and CAS Configuration Modes (Continued) TL/F/9338 – 91 FIGURE 40. VRAM Array Setup for 4 Banks Using Address Pipelining (C6, C5, C4 e 1, 1, 1 or C6, C5, C4 e 0, 1, 0 (Also Allowing Error Scrubbing) during Programming) TL/F/9338 – 92 FIGURE 41. VRAM Array Setup for Address Pipelining with 2 Banks (C6, C5, C4 e 1, 0, 1 or C6, C5, C4 e 0, 0, 1 (Also Allowing Error Scrubbing) during Programming) 8.4 ERROR SCRUBBING In error scrubbing during refresh, the user selects one, two or four RAS and CAS outputs per bank. When performing error detection and correction, memory is always accessed as words. Since the CAS signals are not used to select individual bytes, their corresponding ECAS inputs can be tied low as shown in Figures 42 and 43 . TL/F/9338 – 93 FIGURE 42. VRAM Array Setup for 4 Banks Using Error Scrubbing (C6, C5, C4 e 0, 1, 0 during Programming) TL/F/9338 – 94 FIGURE 43. VRAM Array Setup for Error Scrubbing with 2 Banks (C6, C5, C4 e 0, 0, 1 during Programming) 35 8.0 RAS and CAS Configuration Modes (Continued) The ECAS inputs may then be toggled with the DP8520A/ 21A/22A’s address latches in fall-through mode, while AREQ is asserted. The ECAS inputs can also be used to select individual bytes. When using nibble mode VRAMS, the third and fourth address bits can be tied to the bank select inputs to perform memory interleaving. In page or static column modes, the two address bits after the page size can be tied to the bank select inputs to select a new bank if the page size is exceeded. 8.5 PAGE/BURST MODE In a static column, page or burst mode system, the least significant bits must be tied to the column address in order to ensure that the page/burst accesses are to sequential memory addresses, as shown in Figure 44. In a nibble mode system, the two least significant address bits (A2, A3) must be tied to the highest row and column address inputs (depends on VRAM size) to ensure that the toggling bits of nibble mode VRAMs are to sequential memory addresses. TL/F/9338 – 95 *See table below for row, column & bank address bit map. A0,A1 are used for byte addressing in this example. Addresses Column Address Page Mode/Static Column Mode Page Size Nibble Mode* R9, C9 e A2, A3 C0 – 8 e X 256 Bits/Page 512 Bits/Page 1024 Bits/Page 2048 Bits/Page C0 – 7 e A2 – 9 C8 – 10 e X C0– 8 e A2– 10 C9,10 e X C0– 9 e A2– 11 C10 e X C0– 10 e A2– 12 Row Address X X X X X B0 B1 A4 A5 A10 A11 A11 A12 A12 A13 A13 A14 *Assuming 1 M-bit Vrams are being used. Assume that the least significant address bits are used for byte addressing. Given a 32-bit system A0,A1 would be used for byte addressing. X e DON’T CARE, the user can do as he pleases. FIGURE 44. Page, Static Column, Nibble Mode System 36 9.0 Programming and Resetting The DP8520A/21A/22A must be programmed by one of two possible programming sequences before it can be used. After power up, the DP8520A/21A/22A must be externally reset (see External Reset) before programming. After programming, the DP8520A/21A/22A enters a 60 ms initialization period. During this initialization period, the DP8520A/ 21A/22A performs refreshes about every 15 ms; this makes further VRAM warmup cycles unnecessary. The chip can be programmed as many times as the user wishes. After the first programming, the 60 ms initialization period will not be entered into unless the chip is reset. Refreshes occur during the 60 ms initialization period. If ECAS0 was asserted during programming, the RFIP (RFRQ) pin will act as RFIP and will be asserted throughout the initialization period, otherwise the pin will act like RFRQ and toggle every 13 ms – 15 ms in conjunction with internal refresh requests. If the user attempts an access during the initialization period, wait states will be inserted into the access cycle until the initialization period is complete and RAS precharge time has been met. The actual initialization time period is given by the following formula: T e 4096*(Clock Divisor Select) *(Refresh Clock Fine Tune) /(DELCK Frequency) 9.1 MODE LOAD ONLY PROGRAMMING MODE LOAD, ML, asserted enables an internal 23-bit programmable register. To use this method, the user asserts ML, enabling the internal programming register. After ML is asserted, a valid programming selection is placed on the address bus (and ECAS0), then ML is negated. When ML is negated, the value on the address bus (and ECAS0) is latched into the internal programming register and the DP8520A/21A/22A is programmed, as shown in Figure 45. After ML is negated, the DP8520A/21A/22A will enter the 60 ms initialization period only if this is the first programming after power up or reset. Using this method, a set of transceivers on the address bus can be put at TRI-STATEÉ by the system reset signal. A combination of pull-up and pull-down resistors can be used on the address inputs of the DP8520A/21A/22A to select the programming values, as shown in FIgure 46 . 9.2 CHIP SELECTED ACCESS PROGRAMMING The chip can also be programmed by asserting ML and performing a chip selected access. ADS (or ALE) is disabled internally until after programming. To program the chip using this method, ML is asserted. After ML is asserted, CS is asserted and a valid programming selection is placed on the address bus. When AREQ is asserted, the chip is programmed with the programming selection on the address bus. After AREQ is negated, ML can be negated as shown in Figure 47 . TL/F/9338 – 96 FIGURE 45. Mode Load Only Programming TL/F/9338 – 97 *Pull-Up or Pull-Down Resistors on Each Address Input FIGURE 46. Programming during System Reset 37 9.0 Programming and Resetting (Continued) TL/F/9338 – 98 FIGURE 47. CS Access Programming Using this method, various programming schemes can be used. For example if extra upper address bits are available, an unused high order address bit can be tied to the signal ML. Using this method, one need only write to a page of memory, thus asserting the high order bit and in turn programming the chip as shown in Figure 48 . TL/F/9338 – A0 FIGURE 49. Programming the DP8520A/21A/22A through the Address Bus and an I/O Port TL/F/9338–99 Another simple way the chip can be programmed is the first write after system reset. This method requires only a flipflop and an OR gate as shown in Figure 50. At reset, the flipflop is preset, which pulls the Q output low. Since WR is negated, ML is not enabled. The first write access is used to program the chip. When WR is asserted, ML is asserted. WR negated clocks the flip-flop, negates ML, and programs the DP8520A/21A/22A with the address and ECAS0 available at that time. CS does not need to be asserted using this method. FIGURE 48. Programming the DP8520A/21A/22A through the Address Bus Only An I/O port can also be used to assert ML. After ML is asserted, a chip selected access can be performed to program the chip. After the chip selected access, ML can be negated through the I/O port as shown in Figure 49 . TL/F/9338 – A1 FIGURE 50. Programming the DP8520A/21A/22A on the First CPU Write after Power Up 38 9.0 Programming and Resetting (Continued) will program the chip. If ML is negated before or at the same time as DISRFSH as shown in Figure 52, the chip will not be programmed. After the chip is programmed, the 60 ms initialization period will be entered into if this is the first programming after power up or reset. The user must perform an external reset before programming the DP8520A/21A/ 22A. 9.3 EXTERNAL RESET At power up, if the internal power up reset worked, all internal latches and flip-flops are cleared. The power up state is entered by asserting ML and DISRFSH for 16 positive edges of CLK. After 16 clocks if the user negates DISRFSH before negating ML as shown in Figure 51, ML negated TL/F/9338 – A2 FIGURE 51. Chip Reset and Programmed TL/F/9338 – A3 FIGURE 52. Chip Reset but Not Programmed 39 9.0 Programming and Resetting (Continued) 9.4 PROGRAMMING BIT DEFINITIONS Symbol Description ECAS0 Extend CAS/Refresh Request Select 0 The CASn outputs will be negated with the RASn outputs when AREQ (or AREQB, DP8522A only) is negated. The RFIP pin will function as refresh in progress. During a video shift register load operation, the DT/OE output will be negated by either the 4th rising clock edge after the input VSRL asserts, or by the VSRL input negating, whichever occurs first, when this mode is programmed. 1 The CASn outputs will be negated, during an access (Port A (or Port B, DP8522A only)) when their corresponding ECASn inputs are negated. This feature allows the CAS outputs to be extended beyond the RAS outputs negating. Scrubbing refreshes are NOT affected. During scrubbing refreshes the CAS outputs will negate along with the RAS outputs regardless of the state of the ECAS inputs. The RFIP output will function as ReFresh ReQuest (RFRQ) when this mode is programmed. The DT/OE output will be negated by the input VSRL negating when this mode is programmed. B1 Access Mode Select 0 1 ACCESS MODE 0: ALE pulsing high sets an internal latch. On the next positive edge of CLK, the access (RAS) will start. AREQ will terminate the access. ACCESS MODE 1: ADS asserted starts the access (RAS) immediately. AREQ will terminate the access. B0 Address Latch Mode 0 1 ADS or ALE asserted for Port A or AREQB asserted for Port B with the appropriate GRANT latch the input row, column and bank address. The row, column and bank latches are fall through. C9 Delay CAS during WRITE Accesses 0 1 CAS is treated the same for both READ and WRITE accesses. During WRITE accesses, CAS will be asserted by the event that occurs last: CAS asserted by the internal delay line or CAS asserted on the positive edge of CLK after RAS is asserted. C8 Row Address Hold Time 0 1 Row Address Hold Time e 25 ns minimum Row Address Hold Time e 15 ns minimum C7 Column Address Setup Time 0 1 Column Address Setup Time e 10 ns minimum Column Address Setup Time e 0 ns minimum C6, C5, C4 RAS and CAS Configuration Modes/Error Scrubbing during Refresh 0, 0, 0 RAS0–3 and CAS0–3 are all selected during an access. For a particular CAS to be asserted, its corresponding ECAS input must be asserted. B0 and B1 are not used during an access. Error scrubbing during refresh. 0, 0, 1 RAS and CAS pairs are selected during an access by B1. For a particular CAS to be asserted, its corresponding ECAS input must be asserted. B1 e 0 during an access selects RAS0–1 and CAS0 –1. B1 e 1 during an access selects RAS2–3 and CAS2 –3. B0 is not used during an Access. Error scrubbing during refresh. 0, 1, 0 RAS and CAS singles are selected during an access by B0 – 1. For a particular CAS to be asserted, its corresponding ECAS input must be asserted. B1 e 0, B0 e 0 during an access selects RAS0 and CAS0. B1 e 0, B0 e 1 during an access selects RAS1 and CAS1. B1 e 1, B0 e 0 during an access selects RAS2 and CAS2. B1 e 1, B0 e 1 during an access selects RAS3 and CAS3. Error scrubbing during refresh. 0, 1, 1 RAS0–3 and CAS0–3 are all selected during an access. For a particular CAS to be asserted, its corresponding ECAS input must be asserted. B1, B0 are not used during an access. No error scrubbing. (RAS only refreshing) 40 9.0 Programming and Resetting (Continued) 9.4 PROGRAMMING BIT DEFINITIONS (Continued) Symbol Description C6, C5, C4 RAS and CAS Configuration Modes (Continued) 1, 0, 0 RAS pairs are selected by B1. CAS0 – 3 are all selected. For a particular CAS to be asserted, its corresponding ECAS input must be asserted. B1 e 0 during an access selects RAS0 – 1 and CAS0 –3. B1 e 1 during an access selects RAS2 – 3 and CAS0 –3. B0 is not used during an access. No error scrubbing. RAS and CAS pairs are selected by B1. For a particular CAS to be asserted, its corresponding ECAS input must be asserted. B1 e 0 during an access selects RAS0 – 1 and CAS0 –1. B1 e 1 during an access selects RAS2 – 3 and CAS2 –3. B0 is not used during an access. No error scrubbing. 1, 0, 1 1, 1, 0 RAS singles are selected by B0– 1. CAS0 – 3 are all selected. For a particular CAS to be asserted, its corresponding ECAS input must be asserted. B1 e 0, B0 e 0 during an access selects RAS0 and CAS0 –3. B1 e 0, B0 e 1 during an access selects RAS1 and CAS0 –3. B1 e 1, B0 e 0 during an access selects RAS2 and CAS0 –3. B1 e 1, B0 e 1 during an access selects RAS3 and CAS0 –3. No error scrubbing. 1, 1, 1 RAS and CAS singles are selected by B0, 1. For a particular CAS to be asserted, its corresponding ECAS input must be asserted. B1 e 0, B0 e 0 during an access selects RAS0 and CAS0. B1 e 0, B0 e 1 during an access selects RAS1 and CAS1. B1 e 1, B0 e 0 during an access selects RAS2 and CAS2. B1 e 1, B0 e 1 during an access selects RAS3 and CAS3. No error scrubbing. C3 Refresh Clock Fine Tune Divisor 0 Divide delay line/refresh clock further by 30 (If DELCLK/Refresh Clock Clock Divisor e 2 MHz e 15 ms refresh period). Divide delay line/refresh clock further by 26 (If DELCLK/Refresh Clock Clock Divisor e 2 MHz e 13 ms refresh period). 1 C2, C1, C0 Delay Line/Refresh Clock Divisor Select 0, 0, 0 0, 0, 1 0, 1, 0 0, 1, 1 1, 0, 0 1, 0, 1 1, 1, 0 1, 1, 1 Divide DELCLK by 10 to get as close to 2 MHz as possible. Divide DELCLK by 9 to get as close to 2 MHz as possible. Divide DELCLK by 8 to get as close to 2 MHz as possible. Divide DELCLK by 7 to get as close to 2 MHz as possible. Divide DELCLK by 6 to get as close to 2 MHz as possible. Divide DELCLK by 5 to get as close to 2 MHz as possible. Divide DELCLK by 4 to get as close to 2 MHz as possible. Divide DELCLK by 3 to get as close to 2 MHz as possible. R9 Refresh Mode Select 0 1 RAS0–3 will all assert and negate at the same time during a refresh. Staggered Refresh. RAS outputs during refresh are separated by one positive clock edge. Depending on the configuration mode chosen, either one or two RASs will be asserted. R8 Address Pipelining Select 0 Address pipelining is selected. The VRAM controller will switch the VRAM column address back to the row address after guaranteeing the column address hold time. Non-address pipelining is selected. The VRAM controller will hold the column address on the VRAM address bus until the access RASs are negated. 1 41 9.0 Programming and Resetting (Continued) 9.4 PROGRAMMING BIT DEFINITIONS (Continued) Symbol Description R7 WAIT or DTACK Select 0 1 WAIT type output is selected. DTACK (Data Transfer ACKnowledge) type output is selected. R6 Add Wait States to the Current Access if WAITIN is Low 0 1 WAIT or DTACK will be delayed by one additional positive edge of CLK. WAIT or DTACK will be delayed by two additional positive edges of CLK. R5, R4 WAIT/DTACK during Burst (See Section 5.1.2 or 5.2.2) 0, 0 NO WAIT STATES; If R7 e 0 during programming, WAIT will remain negated during burst portion of access. If R7 e 1 programming, DTACK will remain asserted during burst portion of access. 0, 1 1T; If R7 e 0 during programming, WAIT will assert when the ECAS inputs are negated with AREQ asserted. WAIT will negate from the positive edge of CLK after the ECASs have been asserted. If R7 e 1 during programming, DTACK will negate when the ECAS inputs are negated with AREQ asserted. DTACK will assert from the positive edge of CLK after the ECASs have been asserted. 1, 0 (/2T; If R7 e 0 during programming, WAIT will assert when the ECAS inputs are negated with AREQ asserted. WAIT will negate on the negative level of CLK after the ECASs have been asserted. If R7 e 1 during programming, DTACK will negate when the ECAS inputs are negated with AREQ asserted. DTACK will assert from the negative level of CLK after the ECASs have been asserted. 1, 1 0T; If R7 e 0 during programming, WAIT will assert when the ECAS inputs are negated. WAIT will negate when the ECAS inputs are asserted. If R7 e 1 during programming, DTACK will negate when the ECAS inputs are negated. DTACK will assert when the ECAS inputs are asserted. R3, R2 WAIT/DTACK Delay Times (See Section 5.1.1 or 5.2.1) 0, 0 NO WAIT STATES; If R7 e 0 during programming, WAIT will remain high during non-delayed accesses. WAIT will negate when RAS is negated during delayed accesses. NO WAIT STATES; If R7 e 1 during programming, DTACK will be asserted when RAS is asserted. 0, 1 (/2T; If R7 e 0 during programming, WAIT will negate on the negative level of CLK, after the access RAS. 1T; If R7 e 1 during programming, DTACK will be asserted on the positive edge of CLK after the access RAS. 1, 0 NO WAIT STATES, (/2T; If R7 e 0 during programming, WAIT will remain high during non-delayed accesses. WAIT will negate on the negative level of CLK, after the access RAS, during delayed accesses. (/2T; If R7 e 1 during programming, DTACK will be asserted on the negative level of CLK after the access RAS. 1, 1 1T; If R7 e 0 during programming, WAIT will negate on the positive edge of CLK after the access RAS. 1(/2T; If R7 e 1 during programming, DTACK will be asserted on the negative level of CLK after the positive edge of CLK after the access RAS. R1, R0 RAS Low and RAS Precharge Time 0, 0 RAS asserted during refresh e 2 positive edges of CLK. RAS precharge time e 1 positive edge of CLK. RAS will start from the first positive edge of CLK after GRANTB transitions (DP8522A). 0, 1 RAS asserted during refresh e 3 positive edges of CLK. RAS precharge time e 2 positive edges of CLK. RAS will start from the second positive edge of CLK after GRANTB transitions (DP8522A). 1, 0 RAS asserted during refresh e 2 positive edges of CLK. RAS precharge time e 2 positive edges of CLK. RAS will start from the first positive edge of CLK after GRANTB transitions (DP8522A). 1, 1 RAS asserted during refresh e 4 positive edges of CLK. RAS precharge time e 3 positive edges of CLK. RAS will start from the second positive edge of CLK after GRANTB transitions (DP8522A). Note 1: The configuration modes allow RASs and CASs to be grouped such that each RAS and CAS will drive one-fourth of the total VRAM array whether the array is organized as 1, 2, or 4 banks. Note 2: In order for a CAS output to go low during an access, it must be both selected and enabled. ECAS0–1 are used to enable the CAS outputs (ECAS0 enables CAS0,1; ECAS1 enables CAS2,3). Selection is determined by the configuration mode and B1, B0. 42 11.2 CALCULATION OF tRAH AND tASC 10.0 Test Mode There are two clock inputs to the DP8520A/21A/22A. These two clocks, DELCLK and CLK can either be tied together to the same clock or be tied to different clocks running asynchronously at different frequencies. The clock input, DELCLK, controls the internal delay line and refresh request clock. DELCLK should be a multiple of 2 MHz. If DELCLK is not a multiple of 2 MHz, tRAH and tASC will change. The new values of tRAH and tASC can be calculated by the following formulas: If tRAH was programmed to equal 15 ns then tRAH e 30*(((DELCLK Divisor)* 2 MHz/(DELCLK Frequency))b1) a 15 ns. If tRAH was programmed to equal 25 ns then tRAH e 30*(((DELCLK Divisor)* 2 MHz/(DELCLK Frequency))b1) a 25 ns. If tASC was programmed to equal 0 ns then tASC e 15* ((DELCLK Divisor)* 2 MHz/(DELCLK Frequency)) b 15 ns. If tASC was programmed to equal 10 ns then tASC e 25* ((DELCLK Divisor)* 2 MHz/(DELCLK Frequency)) b 15 ns. Since the values of tRAH and tASC are increased or decreased, the time to CAS asserted will also increase or decrease. These parameters can be adjusted by the following formula: Delay to CAS e Actual Spec. a Actual tRAH b Programmed tRAH a Actual tASC b Programmed tASC. Staggered refresh in combination with the error scrubbing mode places the DP8520A/21A/22A in test mode. In this mode, the 24-bit refresh counter is divided into a 13-bit and 11-bit counter. During refreshes both counters are incremented to reduce test time. 11.0 VRAM Critical Timing Parameters The two critical timing parameters, shown in Figure 53 , that must be met when controlling the access timing to a VRAM are the row address hold time, tRAH, and the column address setup time, tASC. Since the DP8520A/21A/22A contain a precise internal delay line, the values of these parameters can be selected at programming time. These values will also increase and decrease if DELCLK varies from 2 MHz. 11.1 PROGRAMMABLE VALUES OF tRAH AND tASC The DP8520A/21A/22A allow the values of tRAH and tASC to be selected at programming time. For each parameter, two choices can be selected. tRAH, the row address hold time, is measured from RAS asserted to the row address starting to change to the column address. The two choices for tRAH are 15 ns and 25 ns, programmable through address bit C8. tASC, the column address setup time, is measured from the column address valid to CAS asserted. The two choices for tASC are 0 ns and 10 ns, programmable through address bit C7. TL/F/9338 – A4 FIGURE 53. tRAH and tASC 43 12.0 Dual Accessing Functions (DP8522A) suming that Port A is not accessing the VRAM (CS, ADS/ ALE and AREQ) and RAS precharge for the particular bank has completed. It is important to note that for GRANTB to transition to Port B, Port A must not be requesting an access at a rising clock edge (or locked) and Port B must be requesting an access at that rising clock edge. Port A can request an access through CS and ADS/ALE or CS and AREQ. Therefore during an interleaved access where CS and ADS/ALE become asserted before AREQ from the previous access is negated, Port A will retain GRANTB e 0 whether AREQB is asserted or not. Since there is no chip select for Port B, AREQB must incorporate this signal. This mode of accessing is similar to Mode 1 accessing for Port A. The DP8522A has all the functions previously described. In addition to those features, the DP8522A also has the capabilities to arbitrate among refresh, Port A and a second port, Port B. This allows two CPUs to access a common VRAM array. VRAM refresh has the highest priority followed by the currently granted port. The ungranted port has the lowest priority. The last granted port will continue to stay granted even after the access has terminated, until an access request is received from the ungranted port (see Figure 54a ). The dual access configuration assumes that both Port A and Port B are synchronous to the system clock. If they are not synchronous to the system clock they should be externally synchronized (Ex. By running the access requests through several Flip-Flops, see Figure 58a ). 12.1 PORT B ACCESS MODES (DP8522A) Port B accesses are initiated from a single input, AREQB. When AREQB is asserted, an access request is generated. If GRANTB is asserted and a refresh is not taking place or precharge time is not required, RAS will be asserted when AREQB is asserted. Once AREQB is asserted, it must stay asserted until the access is over. AREQB negated, negates RAS as shown in Figure 54b. Note that if ECAS0 e 1 during programming the CAS outputs may be held asserted (beyond RASn negating) by continuing to assert the appropriate ECAS input (the same as Port A accesses). If Port B is not granted, the access will begin on the first or second positive edge of CLK after GRANTB is asserted (See R0, R1 programming bit definitions) as shown in Figure 54c , as- TL/F/9338 – A5 Explanation of Terms AREQA e Chip Selected access request from Port A AREQB e Chip Selected access request from Port B LOCK e Externally controlled LOCKing of the Port that is currently GRANTed. FIGURE 54a. DP8522A PORT A/PORT B ARBITRATION STATE DIAGRAM. This arbitration may take place during the ‘‘ACCESS’’ or ‘‘REFRESH’’ state (see Figure 7a ). TL/F/9338 – A6 FIGURE 54b. Access Request for Port B TL/F/9338 – A7 FIGURE 54c. Delayed Port B Access 44 12.0 Dual Accessing Functions (DP8522A) (Continued) 12.2 PORT B WAIT STATE SUPPORT (DP8522A) Advanced transfer acknowledge for Port B, ATACKB, is used for wait state support for Port B. This output will be asserted when RAS for the Port B access is asserted, as shown in Figures 55 and 56. Once asserted, this output will stay asserted until AREQB is negated. With external logic, ATACKB can be made to interface to any CPU’s wait input as shown in Figure 57 . TL/F/9338 – B0 C) Synchronize ATACKB to CPU B Clock. This is useful if CPU B runs asynchronous to the DP8522. FIGURE 57. Modifying Wait Logic for Port B 12.3 COMMON PORT A AND PORT B DUAL PORT FUNCTIONS An input, LOCK, and an output, GRANTB, add additional functionality to the dual port arbitration logic. LOCK allows Port A or Port B to lock out the other port from the VRAM. When a Port is locked out of the VRAM, wait states will be inserted into its access cycle until it is allowed to access memory. GRANTB is used to multiplex the input control signals and addresses to the DP8522A. TL/F/9338 – A8 A) Extend ATACK to (/2T ((/2 Clock) after RAS goes low. 12.3.1 GRANTB Output The output GRANTB determines which port has current access to the VRAM array. GRANTB asserted signifies Port B has access. GRANTB negated signifies Port A has access to the VRAM array. TL/F/9338 – A9 B) Extend ATACK to 1T after RAS goes low. TL/F/9338 – B1 FIGURE 55. Non-Delayed Port B Access TL/F/9338 – B2 FIGURE 56. Delayed Port B Access 45 12.0 Dual Accessing Functions (DP8522A) (Continued) the VRAM array, the GRANTB output will transition from a rising clock edge from AREQ or AREQB negating and will preceed the RAS for the access by one or two clock periods. GRANTB will then stay in this state until the other port requests an access and the currently granted port is not accessing the VRAM as shown in Figure 58b . Since the DP8522A has only one set of address inputs, the signal is used, with the addition of buffers, to allow the currently granted port’s addresses to reach the DP8522A. The signals which need to be bufferred are R0–10, C0–10, B0 – 1, ECAS0 – 1, and LOCK. All other inputs are not common and do not have to be buffered as shown in Figure 58a. If a Port, which is not currently granted, tries to access Dual Accessing Using the DP8522A TL/F/9338 – 1 *If Port B is synchronous the Request Synchronizing logic will not be required. FIGURE 58a. Dual Accessing Using the DP8522A to interface a CPU and a Graphics Processor to a Video RAM System 46 12.0 Dual Accessing Functions (DP8522A) (Continued) TL/F/9338 – B3 FIGURE 58b. Wait States during a Port B Access 12.3.2 LOCK Input 12.4 DUAL PORTING AND VRAM TRANSFER CYCLES The input AVSRLRQ disables any further internally or externally requested refreshes or Port B access requests from being executed. The AVSRLRQ input does this by making the VRAM controller arbitration logic think that a Port A access is in progress from the point where the AVSRLRQ input goes low until the VRAM shift register load operation is completed. When the LOCK input is asserted, the currently granted port can ‘‘lock out’’ the other port through the insertion of wait states to that port’s access cycle. LOCK does not disable refreshes, it only keeps GRANTB in the same state even if the other port requests an access, as shown in Figure 59 . LOCK can be used by either port. The user must guarantee that LOCK is not asserted with Port B granted when AVSRLRQ is asserted to request a VRAM transfer cycle. TL/F/9338 – B4 FIGURE 59. LOCK Function 47 13.0 Absolute Maximum Ratings (Note 1) All Input or Output Voltage with Respect to GNDÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀb0.5V to a 7V Power Dissipation @ 20 MHzÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀ0.5W If Military/Aerospace specified devices are required, please contact the National Semiconductor Sales Office/Distributors for availability and specifications. Temperature under Bias ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀ0§ C to a 70§ C Storage Temperature ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀb65§ C to a 150§ C 14.0 DC Electrical Characteristics TA e 0§ C to a 70§ C, VCC e 5V g 10%, GND e 0V Symbol Parameter Conditions Min VIH Logical 1 Input Voltage Tested with a Limited Functional Pattern VIL Logical 0 Input Voltage Tested with a Limited Functional Pattern VOH1 Q and WE Outputs IOH e b10 mA VOL1 Q and WE Outputs IOL e 10 mA VOH2 All Outputs except Qs, WE IOH e b3 mA VOL2 All Outputs except Qs, WE IOL e 3 mA IIN Input Leakage Current VIN e VCC or GND IIL ML ML Input Current (Low) VIN e GND ICC1 Standby Current CLK at 8 MHz (VIN e VCC or GND) ICC1 Standby Current ICC1 ICC2 Typ Max Units 2.0 VCC a 0.5 V b 0.5 0.8 V VCC b 1.0 V 0.5 VCC b 1.0 V V b 10 0.5 V 10 mA 200 mA 6 15 mA CLK at 20 MHz (VIN e VCC or GND) 8 17 mA Standby Current CLK at 25 MHz (VIN e VCC or GND) 10 20 mA Supply Current CLK at 8 MHz (Inputs Active) (ILOAD e 0) (VIN e VCC or GND) 20 40 mA ICC2 Supply Current CLK at 20 MHz (Inputs Active) (ILOAD e 0) (VIN e VCC or GND) 40 75 mA ICC2 Supply Current CLK at 25 MHz (Inputs Active) (ILOAD e 0) (VIN e VCC or GND) 50 95 mA CIN* Input Capacitance fIN at 1 MHz 10 pF *Note: CIN is not 100% tested. 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A The AC timing parameters are grouped into sectional numbers as shown below. These numbers also refer to the timing diagrams. 1 – 38 Common parameters to all modes of operation 50 – 56 Difference parameters used to calculate; RAS low time, RAS precharge time, CAS high time and CAS low time 100 – 121 Common dual access parameters used for Port B accesses and inputs and outputs used only in dual accessing 200 – 212 Refresh parameters 300 – 315 Mode 0 access parameters used in both single and dual access applications 400 – 416 Mode 1 access parameters used in both single and dual access applications 450 – 455 Special Mode 1 access parameters which supersede the 400 – 416 parameters when dual accessing 500 – 506 Programming parameters 600 – 605 Graphics parameters for VRAM transfer cycles Unless otherwise stated VCC e 5.0V g 10%, 0 k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0 – 8, 9, and 10; or CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9, and 10. 48 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) 8520A/21A/22A-25 Number Symbol Common Parameter Description CL CH Min Max Min Max 25 0 25 1 fCLK CLK Frequency 0 2 tCLKP CLK Period 40 40 3, 4 tCLKPW CLK Pulse Width 12 5 fDCLK DELCLK Frequency 5 20 5 20 6 tDCLKP DELCLK Period 50 200 50 200 7, 8 tDCLKPW DELCLK Pulse Width 12 12 9a tPRASCAS0 RAS Asserted to CAS Asserted (tRAH e 15 ns, tASC e 0 ns) 30 30 9b tPRASCAS1 RAS Asserted to CAS Asserted (tRAH e 15 ns, tASC e 10 ns) 40 40 9c tPRASCAS2 (RAS Asserted to CAS Asserted (tRAH e 25 ns, tASC e 0 ns) 40 40 9d tPRASCAS3 (RAS Asserted to CAS Asserted (tRAH e 25 ns, tASC e 10 ns) 50 50 10a tRAH Row Address Hold Time (tRAH e 15) 15 15 10b tRAH Row Address Hold Time (tRAH e 25) 25 25 11a tASC Column Address Setup Time (tASC e 0) 0 0 11b tASC Column Address Setup Time (tASC e 10) 10 10 12 tPCKRAS CLK High to RAS Asserted following Precharge 22 26 13 tPARQRAS AREQ Negated to RAS Negated 31 35 14 tPENCL ECAS0–1 Asserted to CAS Asserted 20 27 15 tPENCH ECAS0–1 Negated to CAS Negated 20 27 16 tPARQCAS AREQ Negated to CAS Negated 47 54 17 tPCLKWH CLK to WAIT Negated 31 31 18 tPCLKDL0 CLK to DTACK Asserted (Programmed as DTACK of 1/2, 1, 1(/2 or if WAITIN is Asserted) 28 28 36 36 19 tPEWL ECAS Negated to WAIT Asserted during a Burst Access 20 tSECK ECAS Asserted Setup to CLK High to Recognize the Rising Edge of CLK during a Burst Access 49 19 12 19 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) Unless otherwise stated VCC e 5.0V g 10%, 0§ C k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0–8, 9 and 10; or CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9 and 10. 8520A/21A/22A-25 Number Symbol Common Parameter Description CL Min 21 tPEDL CH Max Min Max ECAS Asserted to DTACK Asserted during a Burst Access (Programmed as DTACK0) 38 38 ECAS Negated to DTACK Negated during a Burst Access 38 38 22 tPEDH 23 tSWCK WAITIN Asserted Setup to CLK 26 tPAQ Row, Column Address Valid to Q0–8, 9, 10 Valid 26 35 27 tPCINCQ COLINC Asserted to Q0– 8, 9, 10 Incremented 30 39 28 tSCINEN COLINC Asserted Setup to ECAS Asserted to Ensure tASC e 0 ns 17 19 29a tSARQCK1 AREQ Negated Setup to CLK High with 1 Period of Precharge 37 37 29b tSARQCK2 AREQ Negated Setup to CLK High with l 1 Period of Precharge Programmed 15 15 30 tPAREQDH AREQ Negated to DTACK Negated 27 27 31 tPCKCAS CLK High to CAS Asserted when Delayed by WIN 25 32 32 tSCADEN Column Address Setup to ECAS Asserted to Guarantee tASC e 0 14 16 33 tWCINC COLINC Pulse Width 20 20 34a tPCKCL0 CLK High to CAS Asserted following Precharge (tRAH e 15 ns, tASC e 0 ns) 72 79 34b tPCKCL1 CLK High to CAS Asserted following Precharge (tRAH e 25 ns, tASC e 0 ns) 82 89 34c tPCKCL2 CLK High to CAS Asserted following Precharge (tRAH e 25 ns, tASC e 0 ns) 82 89 34d tPCKCL3 CLK High to CAS Asserted following Precharge (tRAH e 25 ns, tASC e 10 ns) 92 99 35 tCAH Column Address Hold Time (Interleave Mode Only) 36 tPCQR CAS Asserted to Row Address Valid (Interleave Mode Only) 90 90 37 tPCASDTH CAS Asserted to DT/OE Asserted for a VRAM Read Access 14 24 38 tPCASDTL CAS Negated to DT/OE Negated for a VRAM Read Access 14 24 5 5 32 50 32 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) Unless otherwise stated VCC e 5.0V g 10%, 0§ C k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0–8, 9 and 10; or CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9 and 10. 8520A/21A/22A-25 Number Symbol Difference Parameter Description CL Min 50 tD1 CH Max Min Max (AREQ or AREQB Negated to RAS Negated) Minus (CLK High to RAS Asserted) 14 14 51 tD2 (CLK High to Refresh RAS Negated) Minus (CLK High to RAS Asserted) 11 11 52 tD3a (ADS Asserted to RAS Asserted (Mode 1)) Minus (AREQ Negated to RAS Negated) 4 4 4 4 53 tD3b (CLK High to RAS Asserted (Mode 0)) Minus (AREQ Negated to RAS Negated) 54 tD4 (ECAS Asserted to CAS Asserted) Minus (ECAS Negated to CAS Negated) 55 tD5 (CLK to Refresh RAS Asserted) Minus (CLK to Refresh RAS Negated) 6 6 56 tD6 (AREQ Negated to RAS Negated) Minus (ADS Asserted to RAS Asserted ((Mode 1)) 10 10 51 b7 7 b7 7 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) Unless otherwise stated VCC e 5.0V g 10%, 0§ C k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0–8, 9 and 10; or CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9 and 10. 8520A/21A/22A-25 Number Symbol Common Dual Access Parameter Description CL Min 100 tHCKARQB 101 102 CH Max Min Max AREQB Negated Held from CLK High 3 3 tSARQBCK AREQB Asserted Setup to CLK High 7 tPAQBRASL AREQB Asserted to RAS Asserted 37 41 103 tPAQBRASH AREQB Negated to RAS Negated 32 36 105 tPCKRASG CLK High to RAS Asserted for Pending Port B Access 44 48 106 tPAQBATKBL AREQB Asserted to ATACKB Asserted 45 45 107 tPCKATKB CLK High to ATACKB Asserted for Pending Access 51 51 108 tPCKGH CLK High to GRANTB Asserted 32 32 109 tPCKGL CLK High to GRANTB Negated 29 29 110 tSADDCKG Row Address Setup to CLK High That Asserts RAS following a GRANTB Change to Ensure tASR e 0 ns for Port B 11 16 LOCK Asserted Setup to CLK Low to Lock Current Port 5 5 7 111 tSLOCKCK 112 tPAQATKBH AREQ Negated to ATACKB Negated 21 21 113 tPAQBCASH AREQB Negated to CAS Negated 47 54 114 tSADAQB Address Valid Setup to AREQB Asserted 7 12 116 tHCKARQG AREQ Negated Held from CLK High 5 5 117 tWAQB AREQB High Pulse Width to Guarantee tASR e 0 ns 26 31 118a tPAQBCAS0 AREQB Asserted to CAS Asserted (tRAH e 15 ns, tASC e 0 ns) 87 94 118b tPAQBCAS1 AREQB Asserted to CAS Asserted (tRAH e 15 ns, tASC e 10 ns) 97 104 118c tPAQBCAS2 AREQB Asserted to CAS Asserted (tRAH e 25 ns, tASC e 0 ns) 97 104 118d tPAQBCAS3 AREQB Asserted to CAS Asserted (tRAH e 25 ns, tASC e 10 ns) 107 114 120a tPCKCASG0 CLK High to CAS Asserted for Pending Port B Access (tRAH e 15 ns, tASC e 0 ns) 96 103 CLK High to CAS Asserted for Pending Port B Access (tRAH e 15 ns, tASC e 0 ns) 106 113 CLK High to CAS Asserted for Pending Port B Access (tRAH e 25 ns, tASC e 0 ns) 106 113 CLK High to CAS Asserted for Pending Port B Access (tRAH e 25 ns, tASC e 10 ns) 116 123 120b 120c 120d 121 tPCKCASG1 tPCKCASG2 tPCKCASG3 tSBADDCKG Bank Address Valid Setup to CLK High That Starts RAS for Pending Port B Access 52 10 10 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) Unless otherwise stated VCC e 5.0V g 10%, 0§ C k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9 and 10. Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0–8, 9 and 10; or 8520A/21A/22A-25 Number Symbol Refresh Parameter Description CL Min CH Max Min Max 200 tSRFCK RFSH Asserted Setup to CLK High 22 22 201 tSDRFCK DISRFSH Asserted Setup to CLK High 22 22 202 tSXRFCK EXTENDRF Setup to CLK High 12 12 204 tPCKRFL CLK High to RFIP Asserted 31 205 tPARQRF AREQ Negated to RFIP Asserted 50 50 206 tPCKRFH CLK High to RFIP Negated 51 51 207 tPCKRFRASH CLK High to Refresh RAS Negated 29 33 208 tPCKRFRASL CLK High to Refresh RAS Asserted 23 27 209a tPCKCL0 CLK High to CAS Asserted during Error Scrubbing (tRAH e 15 ns, tASC e 0 ns) 73 80 CLK High to CAS Asserted during Error Scrubbing (tRAH e 15 ns, tASC e 10 ns) 83 90 CLK High to CAS Asserted during Error Scrubbing (tRAH e 25 ns, tASC e 0 ns) 83 90 CLK High to CAS Asserted during Error Scrubbing (tRAH e 25 ns, tASC e 10 ns) 93 100 209b 209c 209d tPCKCL1 tPCKCL2 tPCKCL3 15 31 210 tWRFSH RFSH Pulse Width 211 tPCKRQL CLK High to RFRQ Asserted 40 40 212 tPCKRQH CLK High to RFRQ Negated 40 40 53 15 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) Unless otherwise stated VCC e 5.0V g 10%, 0§ C k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9 and 10. Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0–8, 9 and 10; or 8520A/21A/22A-25 Number Symbol Mode 0 Access Parameter Description CL Min CH Max Min 300 tSCSCK CS Asserted to CLK High 13 13 301a tSALECKNL ALE Asserted Setup to CLK High Not Using On-Chip Latches or if Using On-Chip Latches and B0, B1, Are Constant, Only 1 Bank 15 15 ALE Asserted Setup to CLK High, if Using On-Chip Latches if B0, B1 Can Change, More Than One Bank 29 29 301b tSALECKL 302 tWALE ALE Pulse Width 13 13 303 tSBADDCK Bank Address Valid Setup to CLK High 18 18 304 tSADDCK Row, Column Valid Setup to CLK High to Guarantee tASR e 0 ns 11 16 Row, Column, Bank Address Held from ALE Negated (Using On-Chip Latches) 8 8 Row, Column, Bank Address Setup to ALE Negated (Using On-Chip Latches) 2 2 305 306 tHASRCB tSRCBAS Max 307 tPCKRL CLK High to RAS Asserted 22 26 308a tPCKCL0 CLK High to CAS Asserted (tRAH e 15 ns, tASC e 0 ns) 72 79 308b tPCKCL1 CLK High to CAS Asserted (tRAH e 15 ns, tASC e 10 ns) 82 89 308c tPCKCL2 CLK High to CAS Asserted (tRAH e 25 ns, tASC e 0 ns) 82 89 308d tPCKCL3 CLK High to CAS Asserted (tRAH e 25 ns, tASC e 10 ns) 92 99 309 tHCKALE ALE Negated Hold from CLK High 0 0 310 tSWINCK WIN Asserted Setup to CLK High to Guarantee CAS is Delayed b 16 b 16 311 tPCSWL CS Asserted to WAIT Asserted 22 22 312 tPCSWH CS Negated to WAIT Negated 25 25 313 tPCLKDL1 CLK High to DTACK Asserted (Programmed as DTACK0) 32 32 314 tPALEWL ALE Asserted to WAIT Asserted (CS is Already Asserted) 29 29 315 AREQ Negated to CLK High That Starts Access RAS to Guarantee tASR e 0 ns (Non-Interleaved Mode Only) 54 34 39 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) Unless otherwise stated VCC e 5.0V g 10%, 0§ C k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9 and 10. Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0–8, 9 and 10; or 8520A/21A/22A-25 Number Symbol Mode 1 Access Parameter Description CL Min CH Max Min 400a tSADSCK1 ADS Asserted Setup to CLK High 13 13 400b tSADSCKW ADS Asserted Setup to CLK (to Guarantee Correct WAIT or DTACK Output; Doesn’t Apply for DTACK0) 25 25 5 Max 401 tSCSADS CS Setup to ADS Asserted 402 tPADSRL ADS Asserted to RAS Asserted 25 5 29 403a tPADSCL0 ADS Asserted to CAS Asserted (tRAH e 15 ns, tASC e 0 ns) 75 82 403b tPADSCL1 ADS Asserted to CAS Asserted (tRAH e 15 ns, tASC e 10 ns) 85 92 403c tPADSCL2 ADS Asserted to CAS Asserted (tRAH e 25 ns, tASC e 10 ns) 85 92 403d tPADSCL3 ADS Asserted to CAS Asserted (tRAH e 25 ns, tASC e 10 ns) 95 102 404 tSADDADS Row Address Valid Setup to ADS Asserted to Guarantee tASR e 0 ns 9 14 405 tHCKADS ADS Negated Held from CLK High 0 0 406 tSWADS WAITIN Asserted Setup to ADS Asserted to Guarantee DTACK0 Is Delayed 0 0 407 tSBADAS Bank Address Setup to ADS Asserted 11 11 408 tHASRCB Row, Column, Bank Address Held from ADS Asserted (Using On-Chip Latches) 10 10 409 tSRCBAS Row, Column, Bank Address Setup to ADS Asserted (Using On-Chip Latches) 2 2 410 tWADSH ADS Negated Pulse Width 12 17 411 tPADSD ADS Asserted to DTACK Asserted (Programmed as DTACK0) 412 tSWINADS WIN Asserted Setup to ADS Asserted (to Guarantee CAS Delayed during Writes Accesses) 35 b 10 35 b 10 413 tPADSWL0 ADS Asserted to WAIT Asserted (Programmed as WAIT0, Delayed Access) 29 29 414 tPADSWL1 ADS Asserted to WAIT Asserted (Programmed WAIT 1/2 or 1) 29 29 415 tPCLKDL1 CLK High to DTACK Asserted (Programmed as DTACK0, Delayed Access) 32 32 416 AREQ Negated to ADS Asserted to Guarantee tASR e 0 ns (Non Interleaved Mode Only) 55 31 36 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) Unless otherwise stated VCC e 5.0V g 10%, 0§ C k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9 and 10. Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0–8, 9 and 10; or 8520A/21A/22A-25 Number Symbol Mode 1 Dual Access Parameter Description CL Min 450 tSADDCKG Row Address Setup to CLK High That Asserts RAS following a GRANTB Port Change to Ensure tASR e 0 ns CH Max 11 Min Max 16 451 tPCKRASG CLK High to RAS Asserted for Pending Access 38 42 452 tPCLKDL2 CLK to DTACK Asserted for Delayed Accesses (Programmed as DTACK0) 43 43 453a tPCKCASG0 CLK High to CAS Asserted for Pending Access (tRAH e 15 ns, tASC e 0 ns) 86 93 CLK High to CAS Asserted for Pending Access (tRAH e 15 ns, tASC e 10 ns) 96 103 CLK High to CAS Asserted for Pending Access (tRAH e 25 ns, tASC e 0 ns) 96 103 CLK High to CAS Asserted for Pending Access (tRAH e 25 ns, tASC e 10 ns) 106 113 453b 453c 453d tPCKCASG1 tPCKCASG2 tPCKCASG3 454 tSBADDCKG Bank Address Valid Setp to CLK High That Asserts RAS for Pending Access 4 4 455 tSADSCK0 ADS Asserted Setup to CLK High 11 11 56 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) Unless otherwise stated VCC e 5.0V g 10%, 0§ C k TA k 70§ C, the output load capacitance is typical for 4 banks of 18 VRAMs per bank, including trace capacitance (see Note 2). CH e 50 pF loads on all outputs except CH e 125 pF loads on RAS0 – 3 and CAS0 –3 and CH e 380 pF loads on Q0 – 8, 9 and 10. Two different loads are specified: CL e 50 pF loads on all outputs except CL e 150 pF loads on Q0–8, 9 and 10; or 8520A/21A/22A-25 Number Programing Parameter Description Symbol CL Min CH Max tHMLADD Mode Address Held from ML Negated 501 tSADDML Mode Address Setup to ML Negated 6 6 502 tWML ML Pulse Width 15 15 503 tSADAQML Mode Address Setup to AREQ Asserted 0 0 504 tHADAQML Mode Address Held from AREQ Asserted 38 38 505 tSCSARQ CS Asserted Setup to AREQ Asserted 6 6 506 tSMLARQ ML Asserted Setup to AREQ Asserted 10 10 600 tSCKVSRL VSRL Low Setup to CLK Rising Edge to guarantee counting VSRL as being Low (used to determine when to end Graphics Shift load access) 7 7 VSRL Low from CLK High (to guarantee VSRL is not counted as being Low until the next rising clock edge) 3 3 AVSRLRQ Low before CLK Rising Edge to guarantee locking the VRAM to only Port A accesses 12 12 601 602 tHVSRLCK tSCKAVSRL 7 Min 500 Max 7 603 tPVSRLDTL VSRL Low to DT/OE Low during Graphics Shift Register Load access 23 33 604 tPVSRLDTH VSRL Negated to DT/OE Negated 22 32 605 tPCKDTL CLK to DT/OE Negated 27 37 Note 1: ‘‘Absolute Maximum Ratings’’ are the values beyond which the safety of the device cannot be guaranteed. They are not meant to imply that the device should be operated at these limits. The table of ‘‘Electrical Characteristics’’ provides conditions for actual device operation. Note 2: Input pulse 0V to 3V; tR e tF e 2.5 ns. Input reference point on AC measurements is 1.5V. Output reference points are 2.4V for High and 0.8V for Low. Note 3: AC Production testing is done at 50 pF. 57 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) TL/F/9338 – B5 FIGURE 60. Clock, DELCLK Timing TL/F/9338 – B6 FIGURE 61. 100: Dual Access Port B 58 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) TL/F/9338 – B7 FIGURE 62. 100: Port A and Port B Dual Access TL/F/9338 – B8 FIGURE 63. 200: Refresh Timing 59 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) TL/F/9338 – B9 FIGURE 64. 300: Mode 0 Timing 60 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) TL/F/9338 – C0 (Programmed as C4 e 1, C5 e 1, C6 e 1) FIGURE 65. 300: Mode 0 Interleaving 61 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) TL/F/9338 – C1 FIGURE 66. 400: Mode 1 Timing 62 15.0 AC Timing Parameters: DP8520A/DP8521A/DP8522A (Continued) TL/F/9338 – C2 FIGURE 67. 400: COLINC Page/Static Column Access Timing TL/F/9338 – C3 FIGURE 68. 500: Programming 63 15.0 AC Timing Parameters: DP8520A/21A/22A (Continued) TL/F/9338 – 57 FIGURE 69. Graphics Timing Diagram 64 16.0 Functional Differences between the DP8520A/21A/22A and the DP8520/21/22 17.0 DP8520A/21A/22A User Hints 1. All inputs to the DP8520A/21A/22A should be tied high, low or the output of some other device. Note: One signal is active high. COLINC (EXTNDRF) should be tied low to disable. 1. Extending the Column Address Strobe (CAS) CAS can be extended indefinitely after AREQ transitions high in non-interleaved mode only, providing that the user program the DP8520A/21A/22A with the ECAS0 (negated) during programming. To extend CAS, the user continues to assert any or multiple ECASs after negating AREQ. Extending CAS with RAS negated can be used to gain RAS precharge time. By negating AREQ, RAS will be negated. The user can then continue to assert a one or both of the ECASs, which will keep CAS asserted. By keeping CAS asserted with RAS negated, the VRAM will keep the data valid until CAS is negated. Even though CAS will be extended, DTACK output will always end from AREQ negated. 2. Extending DT/OE Functionality The DT/OE output will follow the CAS output during a VRAM read access, and will remain negated during a VRAM write access. For the DP8520/21/22, the DT/OE output remained negated for all VRAM access cycles. This will allow the VRAM to drive the data bus. There are 2 options for the function of the DT/OE output during a video shift register load operation. With ECAS0 negated during programming, the DT/OE output will follow the VSRL input during video shift register load operations. With the ECAS0 asserted during programming, VSRL will assert DT/OE. VSRL negated before four rising clock edges will cause DT/OE to be negated. VSRL asserted more than four rising clock edges will cause DT/OE to be negated from the fourth rising clock edge. 3. Dual Accessing RAS will be asserted either one or two clock periods after GRANTB has been asserted. The amount of RAS low and high time, programmed by bits R0 and R1, determines the number of clock periods after GRANTB changes before RAS will start. This is shown in the table below. 2. Each ground on the DP8520A/21A/22A must be decoupled to the closest on-chip supply (VCC) with 0.1 mF ceramic capacitor. This is necessary because these grounds are kept separate inside the DP8520A/21A/ 22A. The decoupling capacitors should be placed as close as possible with short leads to the ground and supply pins of the DP8520A/21A/22A. 3. The output called ‘‘CAP’’ should have a 0.1 mF capacitor to ground. 4. The DP8520A/21A/22A has 20X series damping resistors built into the output drivers of RAS, CAS, address and DT/OE. Space should be provided for external damping resistors on the printed circuit board (or wirewrap board) because they may be needed. The value of these damping resistors (if needed) will vary depending upon the output, the capacitance of the load, and the characteristics of the trace as well as the routing of the trace. The value of the damping resistor also may vary between the wire-wrap board and the printed circuit board. To determine the value of the series damping resistor it is recommended to use an oscilloscope and look at the furthest VRAM from the DP8520A/21A/22A. The undershoot of RAS, CAS, DT/OE and the addresses should be kept to less than 0.5V below ground by varying the value of the damping resistor. The damping resistors should be placed as close as possible with short leads to the driver outputs of the DP8520A/21A/22A. 5. The circuit board must have a good VCC and ground plane connection. If the board is wire-wrapped, the VCC and ground pins of the DP8520A/21A/22A, the VRAM associated logic and buffer circuitry must be soldered to the VCC and ground planes. 6. The traces from the DP8520A/21A/22A to the VRAM should be as short as possible. 7. ECAS0 should be held low during programming if the user wishes that the DP8520A/21A/22A be compatible with a DP8520/21/22 design. RAS RAS Asserted RAS Asserted from R0, R1 Precharge During GRANTB Change Time Refresh 0, 0 0, 1 1, 0 1, 1 1T 2T 2T 3T 2T 2T 3T 4T 18.0 Description of a DP8522A/ DP8500 System Interface 1 Rising Clock Edge 1 Rising Clock Edge 2 Rising Clock Edges 2 Rising Clock Edges Several simple block and timing diagrams are inserted to help the user design a system interface between the DP8520A/21A/22A VRAM controller and the Raster Graphics Processor DP8500 (as shown in Figure 70 ). For accessing the VRAM, the DP8520A/21A/22A uses the RGP’s PHI 2 clock as an input clock and it runs in Mode 1 (asynchronous mode). This allows the user to guarantee row, column and bank address set up times to a rising clock edge (as shown in timing calculations provided). This system design uses a PALÉ to interface the access request logic and the wait logic between the DP8522A and the RGP. External logic is also needed for plane control. 4. Refresh Clock Counter The refresh clock counter will count and assert RFRQ externally when it is time to do a refresh. This will occur even when internal refreshes are disabled. This allows the user to run the chip in a request/acknowledge mode for refreshing. ECAS0 is used to program the RFIP output to act as either refresh request (RFRQ) or RFIP. ECAS0 asserted during programming causes the RFIP output to function as RFIP. ECAS0 negated during programming causes the RFIP output to function as RFRQ. 5. Clearing the Refresh Clock The refresh clock counter is cleared by negating DISRFSH and asserting RFSH for at least 500 ns. 65 18.0 Description of a DP8522A/DP8500 System Interface (Continued) TL/F/9338 – C4 FIGURE 70. DP8422A/DP8500 (RGP) Interface Block Diagram The main idea of the block diagram in Figure 73 is to cause the video DRAM shift register load operation to happen correctly. Once the DP8500 (RGP) issues the Display Refresh REQuest signal (DRREQ) the system knows that the video shift register load operation should occur at a certain defined time later. In the block diagram this time is controlled by the counter device. This counter determines when VSRL transitions high, thereby causing the video shift register load operation. In the upper part of the block diagram is the ‘‘REFRESH’’ output that is used as the ‘‘VSRL’’ input of the DP8522A and is also used to create ‘‘REFRESH NOT DONE’’. To creat the ‘‘REFRESH’’ output a NAND latch is used. This latch is set when a screen refresh is in progress, shown by the RGP outputs ALE, B0, and B1 all being high. If the status of the RGP is anything other than screen refresh the latch is reset. The latch is also reset during a screen refresh when the load shift register counter times out. This counter determines when the ‘‘VSRL’’ input of the DP8522A transitions high, causing the video DRAMs to load a row of data into their shift registers. The ‘‘REFRESH’’ signal along with the load shift register counter output ‘‘NOT DONE’’ are used to create the ‘‘REFRESH NOT DONE’’ signal. This output is used for two purposes. One of which is to hold the ‘‘WAIT’’ output low, thereby inserting WAIT states into the RGP video shift register load access. The other purpose is to hold ‘‘DT/OE’’ low until ‘‘VSRL’’ transitions high. Important setup timing parameters which must be met for a DP8500(RGP) – DP8522A system (assuming RGP is running at 20 MHz (TCP e 50 ns)). 1. Address Setup to ADS Asserted e 1 TCP b ÝtALV a Derating the DP8500 Spec for Light Load b tPF373 a Min PAL Delay to Produce ADS & AREQ e 50 ns b 38 ns a 5 ns b 8 ns a 2 ns e 11 ns (Using Light Load Timing Specs, the DP8520A/21A/22A Needs) 9 ns Setup for Row Address to ADS Asserted 11 ns Setup for Bank Address to ADS Asserted 2. ADS Setup to Clock Rising Edge e 1 TCP b ÝtALEV b Max PAL Delay e 50 ns b 26 ns b 10 ns e 14 ns (DP8520A/21A/22A Needs 7 ns) 3. WAIT Negated Setup to Clock e 1 TCP b $18 b Max PAL Delay e 50 ns b 28 ns b 10 ns e 12 ns (The DP8500 Needs 5 ns Setup Time) Note 1: ‘‘$’’ symbol refers to a DP8520A/21A/22A timing parameter. Note 2: ‘‘Ý’’ symbol refers to a DP8500 timing parameter. Note 3: ALE asserted by the RGP (DP8500) should use the system PAL to assert WAIT in order to guarantee proper setup time. DTACK low should then be used to negate the WAIT signal through PAL equations. 66 18.0 Description of a DP8522A/DP8500 System Interface (Continued) TL/F/9338 – C5 FIGURE 71. DP8522A/DP8500 (RGP) Instruction Read Cycle Timing 67 18.0 Description of a DP8522A/DP8500 System Interface (Continued) TL/F/9338 – C6 FIGURE 72. DP8500 (20 MHz)/DP8522A Write Operand Cycle Timing TL/F/9338 – C7 FIGURE 73. Mid-Scan Line Load Application Example 68 FIGURE 74. Mid-Scan Line Load Functional Timing TL/F/9338 – C8 18.0 Description of a DP8522A/DP8500 System Interface (Continued) 69 DP8520A/DP8521A/DP8522A microCMOS Programmable 256k/1M/4M Video RAM Controller/Drivers Physical Dimensions inches (millimeters) Plastic Chip Carrier (V) Order Number DP8520AV-20, DP8520AV-25, DP8521AV-20 or DP8521AV-25 NS Package Number V68A Plastic Chip Carrier (V) Order Number DP8522AV-20 or DP8522AV-25 NS Package Number V84A LIFE SUPPORT POLICY NATIONAL’S PRODUCTS ARE NOT AUTHORIZED FOR USE AS CRITICAL COMPONENTS IN LIFE SUPPORT DEVICES OR SYSTEMS WITHOUT THE EXPRESS WRITTEN APPROVAL OF THE PRESIDENT OF NATIONAL SEMICONDUCTOR CORPORATION. As used herein: 1. Life support devices or systems are devices or systems which, (a) are intended for surgical implant into the body, or (b) support or sustain life, and whose failure to perform, when properly used in accordance with instructions for use provided in the labeling, can be reasonably expected to result in a significant injury to the user. National Semiconductor Corporation 1111 West Bardin Road Arlington, TX 76017 Tel: 1(800) 272-9959 Fax: 1(800) 737-7018 2. A critical component is any component of a life support device or system whose failure to perform can be reasonably expected to cause the failure of the life support device or system, or to affect its safety or effectiveness. National Semiconductor Europe Fax: (a49) 0-180-530 85 86 Email: cnjwge @ tevm2.nsc.com Deutsch Tel: (a49) 0-180-530 85 85 English Tel: (a49) 0-180-532 78 32 Fran3ais Tel: (a49) 0-180-532 93 58 Italiano Tel: (a49) 0-180-534 16 80 National Semiconductor Hong Kong Ltd. 13th Floor, Straight Block, Ocean Centre, 5 Canton Rd. Tsimshatsui, Kowloon Hong Kong Tel: (852) 2737-1600 Fax: (852) 2736-9960 National Semiconductor Japan Ltd. Tel: 81-043-299-2309 Fax: 81-043-299-2408 National does not assume any responsibility for use of any circuitry described, no circuit patent licenses are implied and National reserves the right at any time without notice to change said circuitry and specifications.