View EZ030AN_1064834.PDF datasheet online --- IC-ON-LINE

Datasheet File OCR Text:

publication# 17580 rev. a amendment /0 issue date: january 1994 ez-030 demonstration board theory of operation application note by david stoenner advanced micro devices most microprocessor evaluation and demonstration circuits are designed to illustrate the highest per- formance and flexibility the microprocessor can deliver, but without ease-of-design considerations. advanced micro devices has developed second generation risc microprocessors with attention to the design ease as well as the execution performance. introduction the am29030 ? microprocessor provides higher perfor- mance, lower system cost, and easier design than its predecessors in the 29k ? family. the standalone ez-030 demonstration board displays the ease of design and low-cost system features of the am29030 micropro- cessor, producing an elegant high-performance low- complexity design. this application note explores the theory of operations and design considerations of the ez-030 demonstration board. the ez-030 demonstration board circuit is designed to demonstrate the simplicity of a page mode dram memory system to support the am29030 microproces- sor. although the circuit is simple and flexible, it does not sacrifice execution performance. the ez-030 demon- stration board performance is not restricted by its dram controller. the ez-030 demonstration board has four functional sections (see figure 1): the am29030 microprocessor and supporting parts the memory and memory control components the uart and rs232 support the power-on reset chip and push button all of the above circuitry fits into 14 square inches of board space (16 in 2 including connectors and stand- offs). am29030 microprocessor the am29030 microprocessor is a 32-bit risc micropro- cessor executing the 29k family instruction set. the predecessor of the am29030 microprocessor, the am29000 ? microprocessor has a modified harvard three-bus architecture with a single address bus multi- plexed for its instruction and data buses. the am29030 microprocessor has a simplified two-bus architecture (address bus and instruction/data bus) and a large on- chip i-cache (8 kbyte). mux serial interface (1 pack) scc 85c30 (1 pack) am29030 microprocessor (1 pack) 27c010 (1 pack) memory control (3 packs) 1 mbit 32 zip dram (8 packs) max232 (1 pack) 8 32 22 8 17 11 (3 packs) row/column control address oe host 17580a001 figure 1. ez-030 demonstration board block diagram ras and cas
amd ez-030 demonstration board theory of operation 2 the first generation of amd signal-compatible embedded risc microprocessors are the am29000, am29005 ? , and am29050 ? microprocessors. each of these microprocessors has an external modified har- vard architecture to expose the internal instruction and data buses. the system designer can create two inde- pendent memory systems to provide necessary band- width for a design. the am29000 microprocessor provides the highest price/performance value of any risc microprocessor. the second generation of the 29k family extends perfor- mance while reducing design complexity and design-in effort. the am29030 and am29035 ? microprocessors extend the range of price/performance beyond the am29000 processor. the am29030 microprocessor bus has reduced the signal complement, relaxed the timing requirements, and added special memory support to re- duce system cost, making it a designer-friendly part. clock the am29030 microprocessor clock timing is forgiving and flexible. for example, the cpu can tolerate an input clock duty cycle of 40 percent to 60 percent. the am29030 microprocessor also has scalable clocking ? technology that runs the internal processor core at twice the bus frequency. the ez-030 demonstration board uses this scalable clocking feature. using the configura- tion jumpers on the ez-030 demonstration board, the de- signer can allow for a double frequency or 32-mhz input clock (inclk) while mai ntaining a 16-mhz external memory clock (memclk). because only the pga pack- ages of the am29030 can dissipate the heat reliably at 25 and 33 mhz, their timing is used for analysis in this ap- plication note. an addendum at the end of this applica- tion note shows how the cerquad 16-mhz am29035 can be used in this memory environment. running the pro- cessor at 2x frequency provides nearly 1.5 times the per- formance of a 1x inclk, even though the external memory bus runs at 1x. this feature allows the ez-030 demonstration board to get the most performance for its dram bandwidth. a side benefit of using the scalable clocking feature with the internally derived memclk is relaxed interface timing restrictions (see table 1). processor outputs are valid closer to the rising edge of memclk and input data has less required setup time. interface the ez-030 demonstration board exploits features of the am29030 microprocessor to derive the maximum performance for the minimum cost and complexity. the first feature the ez-030 demonstration board uses is the am29030 microprocessor's narrow read feature for eprom. by asserting the read-narrow pin (rdn ) during an eprom access of the am29030 microprocessor, the am29030 microprocessor reads instructions one byte at a time and generates an address for each byte. the instruction data is read from the least significant byte (bits 31 to 24) of the bus, because the ez-030 demon- stration board is a big-endian design. on the ez-030 demonstration board, an 8-bit eprom is attached with the read narrow interface. this eprom contains minimon29k ? code, used for bootstrap code to download programs into dram. the eprom also contains debugger support and host interface (hif) support. hif allows the host operating-system standard i/o services, such as open, close, read and write file, and others, to be passed through the ez-030 demon- stration board to the host. table 1. pga interface timing restrictions 1 preliminary 25/33 mhz no. parameter description test conditions min max unit 7 synchronous output valid delay memclk output memclk input 1 2 13 17 ns 7a synchronous output valid delay for id31id0 memclk output memclk input 1 2 13 17 ns 9 synchronous input setup time memclk output memclk input memclk=inclk 17 17 12 ns 9a synchronous input setup time for id31id0 memclk output memclk input memclk=inclk 9 9 6 ns notes: 1 subject to change without notice. refer to the aam29030 and am29035 microprocessors data sheeto for a complete listing of ac/dc parameters (order# 18150).
amd ez-030 demonstration board theory of operation 3 the am29030 microprocessor is very well suited for page-mode or static column dram. the microprocessor takes advantage of page-mode memory systems to pro- vide single-cycle accesses to same page requests. the am29030 microprocessor has a special signal to notify the memory system that accesses are sequential. these sequential accesses occur for cache line fills, and multiple loads and stores of registers. the ez-030 card utilizes only the sequential burst signal for its page mode implementation. sequential accesses are marked by the burst signal of the am29030 microprocessor. the burst signal never spans dram page boundaries; in fact, the burst access does not cross 1-kbyte memory boundaries. the 1k boundary is known as a processor page boundary, be- cause the processor checks the memory management unit (mmu) at this boundary for page faults. if the proces- sor mmu is not activated, this boundary is still respected. thus, page-mode and static column drams snap in with no need for special address comparators. the am29030 microprocessor furnishes a new address during each burst access cycle, which can be astrobedo into the page-mode dram by the cas strobe pulsing at an appropriate time. to generate this pulsing cas strobe, the ez-030 demonstration board uses a delay line. the delay line produces a delayed version of memclk. the real memclk and the delayed version are aandedo to- gether to produce the cas pulse train used during the burst signal. (the am29030 microprocessor has on- chip page comparators and interleave support features that the ez-030 demonstration board does not use.) jtag connector the ez-030 demonstration board has a standard 10-pin jtag connector to access the am29030 microproces- sor part on the board. the jtag ieee scan standard provides low-cost emulation and simpler in-circuit testing. using this connector, it is possible to access ev- ery pin of the am29030 microprocessor through the five jtag pins. because the jtag reset pin must be tied to the reset of the processor when the jtag port is not in use, a jumper must be disconnected when a jtag de- vice is attached to the ez-030 demonstration board. see the ez-030 demonstration board reference manu- al for details. memory description the memory system of the ez-030 demonstration board is composed of three pal ? devices, three 74ls157 muxs, one eprom (27c010), and eight zip drams (256k x 470 or 1 mbyte x 470). the three pal ics that form the memory controller of the ez-030 demonstration board are the master control pal device (mstr_con), which is a palace16v815, the ras/cas decode pal device (cas_dec), which is a pal16v87, and the re- fresh timer pal device (refresh), which is a pal- ace16v825 (see figure 2). state machine description figure 3 is a state machine diagram of the ez-030 memory system. the ready signal, rdy , is the termina- tion of any access. since the microprocessor samples the rdy signal to indicate the completion of an access, rdy can be supplied by the serial communications con- trol pal ics or the memory control pal ics. below is a description of the rdy derivation. the state machine represents the primary states and state transitions. these are idle, row address (ras ), burst read, burst write, refresh cas, and refresh ras. the state machine trajectories create the following: (1) three-cycle random accesses to dram, (2) one-cycle sequential (burst) accesses to dram, and (3) three- cycle refresh. the following sections describe the pal equation source files. memory master timing generator 16v815 mstr_con memory strobe decoder 16v87 cas_dec cas ras mux select delayed memclk from am29030 processor memory refresh timer 16v825 refresh rom oe 17580a002 figure 2. memory system block diagram
amd ez-030 demonstration board theory of operation 4 dram access ras cas write cas read write idle refresh cas refresh ras ref access burst burst & 17580a003 refresh req burst ref_rcq write burst refresh access figure 3. dram state machine diagram ready signal generation mstr_con is the timing generator for the memory sys- tem. when the am29030 microprocessor asserts the access request signal req , the mstr_con pal device produces a ready signal (rdy ) [1.2a]. the rdy signal is a combinatorial signal. that is, it is a combina- tion of the ready terms for both memory mapped de- vices, the serial port, and the memory system. the memory ready signal, mem_rdy, is a registered signal as shown by the a:=o equate symbol [1.1a]. mem_rdy : = req*a31*a30*mem_rdy [1.1a] + mem_access*mem_rdy [1.1b] + mem_access*mem_rdy*burst [1.1c] rdy = mem_rdy + serial_rdy [1.2a] equation 1. mstr_con pal equations the mstr_con pal device has memclk as its clocked input. if there is a microprocessor access re- quest while the most significant address lines, a31 and a30, are zero (eprom address space), the mem_rdy [1.1a] signal activates in the next cycle. the am29030 microprocessor does not sample the rdy signal until the next cycle. this results in a two-cycle access, one cycle for the mstr_con pal device to sample the request and one cycle for the am29030 microprocessor to sample the rdy signal. for subsequent eprom accesses, the ready signal goes active for one cycle and then inactive for the next cycle, continuing two-cycle timing. during a dram access, mem_access must be active before mem_rdy activates [1.1b]. mem_rdy is the memory system component of the rdy signal. the next section shows that mem_access is a registered signal that indicates a dram access was requested. since mem_access is a registered signal, the rdy for dram is delayed one more cycle than the eprom, or three cycles total. these three cycles are: one cycle for mem_access to go active, one cycle for mem_rdy to go active (which triggers the processor rdy signal in the same cycle), and one cycle for the rdy signal to be sampled by the microprocessor. during the burst access of the am29030 microproces- sor, the rdy signal is held active as long as burst is active [1.1c]. this provides a single-cycle access ac- knowledge because rdy is sampled every cycle. the eprom does not support burst . dram timing the idle signal represents the time when the dram system is between accesses. the 70-ns dram has a ras precharge time of 50 ns [2.1a]. then, the mstr_con pal device requires a visit of the idle state before it can go to the ras state and then onto the cas state. this visit takes at least one cycle which is 63 ns at 16 mhz. while idle is asserted, ras is deas- serted. therefore, the ras strobe remains inactive for 63 ns, which is long enough to meet the precharge duration requirement. mem_access is a signal that indicates a dram ac- cess is in progress. idle is a signal that indicates dram is inactive. idle asserts when no dram activity is de- tected, specifically, when there is no dram access and no refresh access (mem_access *ref_access ) [2.1a]. idle is reinforced if: no access is requested to idle := idle *mem_access*ref_access [2.1a] + reset [2.1b] + idle *mem_access*st1*mem_rdy* burst [2.1c] + idle *ref_access*ref_req [2.1d] + idle*(req*a31*/a30 )*ref_req * emul [2.1e] ref_access := idle*ref_req*ref_access [2.2a] + ref_access*ref_req [2.2b] mem_access := idle*req*a31*a30*ref_req * mem_access *emul [2.3a] + mem_access*st1 [2.3b] + mem_access*burst [2.3c] st1 := mem_access*mem_rdy [2.4a] + st1*burst [2.4b] equation 2. mstr_con pal equations
amd ez-030 demonstration board theory of operation 5 ras0 = mem_access [3.1a] + ref_access*memclk [3.1b] + ras0*ref_access [3.1c] mux = mem_access*memclk [3.2a] + mux*mem_access [3.2b] cas0 = ref_access [3.3a] + mux*mem_access*st1*write *memclk* delay_in [3.3b] + mux*mem_access*st1*write *cas0* delay_in [3.3c] + mux*mem_access*st1*write *cas0* memclk [3.3d] + mux*mem_access*st1*write*we0* memclk [3.3e] equation 3. cas_dec pal equations dram (2 msbs a31, a30 = 10 2 ) [2.1e], no refresh re- quest is pending (ref_req ), and no emulator ac- cesses are attempted (emul ). idle is asserted if a non-burst (simple) access is complete: (mem_ac- cess*st1*mem_rdy*burst ) [2.1c]. idle is also asserted if a refresh has completed: (ref_ac- cess*ref_req ) [2.1d]. ref_access is a signal that acknowledges the ref_req when the dram is idle [2.2a]. the ref_ac- cess releases after the ref_req is released, ac- knowledging the completion of a refresh [2.2b]. the refresh operation can occur while eprom or serial port accesses are also occurring. mem_access is initiated when the am29030 micro- processor accesses dram space, where a31=1 and a30=0 [2.3a]. this is qualified by no refresh request and no emulator access. mem_access persists for one more cycle until st1 is asserted. mem_access is rein- forced by burst which might take from 1 cycle to 255 cycles. then mem_access waits for an idle cycle. since mem_access creates ras, the idle cycle is the ras precharge delay. mem_access creates ras fairly directly through the cas_dec pal device. the st1 signal is used to flag the last cycle of a memory access [2.3b]. the cas_dec pal device decodes the st1 signal to form the cas memory strobes for the proper memory devices [2.4a]. during a write, only the bytes with we strobes result in cas strobes. this is how partial words (bytes and half-words) are stored. the st1 signal persists during burst accesses [2.4b]. the cas_dec pal device the cas_dec pal device is a pal device 16v87. the row and column address strobes are decoded by the cas_dec pal equations. additionally, the strobe signal, mux, which chooses column address instead of row ad- dress, is decoded out of this pal device. see equation 3 for the equations for ras, cas, and the row and column address strobes. the ras0 strobe is used by all eight zip dram chips to clock in the row address. ras0 is derived from mem_access of the mstr_con pal device. mem_access indicates the beginning of a memory cycle and ras initiates the cycle [3.1a]. the additional two terms of ras0 create the pulse for cas before ras refresh. the ras signal is delayed half a memclk from the onset of ref_access [3.1b]. the mux [3.2a and 3.2b] signal connects to the a/b con- trol input of the three 74ls157 two-input muxs. this sig- nal is used to choose between the row addresses and the column addresses. the mux signal chooses the column address half a memclk cycle after the ras strobe activates. this first-half clock cycle gives the row addresses 30 ns of hold time. the required time is 15 ns for 70-ns drams. the second-half memclk cycle al- lows the column address to propagate through the 74ls157 mux. this delay signal provides the mux switching time. the switching delay of the 74ls157 mux is 27 ns, and the half-cycle plus the delay signal is 50 ns. both the mstr_con and the cas_dec pal devices allow additional setup time tolerance. the setup time for the column address on a 70-ns dram is 0. only one of the cas equations is shown here [3.3a3.3e] because the rest of the cas signals cas1cas3 are symmetrical. the first term of the cas equation is the cas before ras refresh. cas becomes active as soon as a refresh cycle begins. the ref_access signals the beginning of the refresh cycles. the second, third and fourth terms are for normal memory accesses. the cas strobe for the first read or simple read cycle to dram oc- curs in the last of two cycles. the delay si gnal (delay_in), is memclk passed through a 20-ns delay line. the cas strobe continues until memclk becomes active in the last equation. the delay signal is necessary for the burst read access that might follow the simple access. a burst read access starts at the end of a simple access. the second through fourth terms of the cas equation shape the cas pulse to provide the appropriate setup time for the column ad- dress [3.3b3.3d]. the fourth term [3.3d] allows the assertion of memclk to deassert any previous cas. the first term [3.3a] takes ef- fect, allowing a delay of delay_in (20 ns) before cas becomes active. these two terms have now provided a 20-ns pulse that meets the 10-ns cas precharge specifi- cation of a 70-ns dram. the 20-ns pulse of cas inactive allows the 14-ns address propagation delay through the 74ls157 muxs. the second term carries the cas signal into the third term. delay_in overlaps memclk by 20 ns, so that cas is reinforced by the second term until delay_in goes inactive (see figure 4 and figure 5). the page-mode access time from cas active to data ac- cessed is 20 ns, which allows the microprocessor 15 ns of setup time (9 ns is the requirement).
amd ez-030 demonstration board theory of operation 6 st1 ras cas rdy mem_acc req figure 4. dram read timing diagram memclk delay ram data 17580a004 rdy cas ras st1 mem_acc req figure 5. dram write timing diagram memclk delay ram data 17580a005
amd ez-030 demonstration board theory of operation 7 the single-cycle write cas is easier than the single-cycle read cas, because the data is available near the begin- ning of the cycle. all that is necessary is a cas strobe at the back half of the cycle after the column address is set up. the equation term: mux*mem_access*st1* write *we0*memclk [3.3e] activates cas signal to become active as memclk (i.e., it is active during the back half of memclk). the remain- ing signals in this term are conditions for the cas pulse(s) to occur. note the we0 signal would be we1, we2, or we3 for cas1, cas2, or cas3. this we signal is a byte-write strobe signal from the am29030 microproces- sor. only the cas lines of the microprocessor-requested bytes are written. refresh timer the refresh timer logic is a simple counter with two exposed signals, ref_req and ref_access. ref_req is the request generated every 129 clocks. ref_access is the acknowledge signal that a refresh access is in progress. refresh access only clears the ref_req signal. therefore, refreshes can be pending for up to 129 cycles before they are lost (see equation 4). q0 := q0 q1:=q1:+: q0 [4.1] q2:=q2:+: (q1*q0) [4.2] q3:=q3:+: (q2*q1*q0) [4.3] ... q6:=q6:+: (q5*q4 *q3*q2*q1*q0) [4.4] ref_req:= q6*q5*q4*q3*q2*q1*q0 [4.5a] +ref_req*ref_access [4.5b] equation 4. refresh pal equations eprom the rom_cs signal is routed to both the rdn and to the eprom (27c010) chip-select. because of the read- narrow feature of the am29030 microprocessor, the eprom accesses are easy to facilitate. there is one trick thoughethe read-narrow signal (rdn ) of the am29030 microprocessor must be driven during a pro- cessor reset to indicate whether an 8-bit or a 16-bit eprom is used. for an 8-bit eprom, the rdn signal must be driven inactive (high) [5.1]. this is the default state of the eprom chip-select (rom_cs) signal from the mstr_con pal device. the rom_cs equation only produces a chip-select and rdn during a request to the eprom address space. the eprom address space is any address with the msbs, a31 and a30, both equal to zero [5.1]. note that when a processor is reset, execution starts from location zero, which is in the eprom address space. rom_cs = emul *req*a31 *a30 [5.1] equation 5. mstr_con pal equation serial port support the serial communications controller (scc) chip is at- tached to the most significant byte of the am29030 micro- processor. to provide 5 v-only operation of rs232 drives, a max232 chip is used to create 10 v level signals for the receive data, the transmit data, the data terminal ready, and data set ready signals. the con- nector is configured so that the ez-030 demonstration board is data communications equipment (dce) and at- taches to a host, such as a pc-at that is data term inal equipment (dte). the ez-030 debug monitor serial pro- tocol requires an intelligent host. the serial port section consists of the am85c30 chip, the serial controller pal device (serial), the max232 chip, a baud rate oscillator, and the subminiature 9-pin d connector. the am85c30 chip the am85c30 chip is known as the serial communica- tions controller (scc). it is a dual-channel, multi-protocol data communications peripheral chip. the scc can be configured to provide asynchronous sdlc and hdlc in- terface. this device can check and generate crc codes. the ez-030 circuit with standard target universal de- bugger interface (udi) target monitor software utilizes a very small subset of the functionality of this scc part. the ez-030 board has a connector for only one of the ports on the am85c30 chip. the scc is connected to an independent 2.4576-mhz oscillator for baud rate precision. the scc interface pal device the serial pal device creates the signals to interface the am29030 microprocessor to the am85c30 chip. serial is clocked by mem_clk. the 8530_cs is used for the chip enable (ce) to the scc [6.1]. one timing specification on the am85c30 chip is enforced by soft- ware. because the scc is run by an independent oscil- lator, a circuit inside the part prevents metastability by multiply-sampling source data. because of this resam- pling delay, accesses to the chip must be spaced by six clocks of the scc oscillator plus 300 ns. to accommodate this extended timing spec, the ez-030 software accessing the scc directly must delay be- tween accesses. when computing this delay, the cache and double internal frequency considerations must be accounted for. for the 2.43567-mhz oscillator attached to a 16-mhz internal frequency part, the number of delay cycles between accesses must be 45 cycles.
amd ez-030 demonstration board theory of operation 8 8530_cs = req*a31*a30 [6.1] serial_rdy := nc1*nc2*serial_rdy [6.2] nc1 := (8530_rd + 8530_wr)*nc1*serial_rdy [6.3] nc2 := (8530_rd + 8530_wr)*(nc1:+:nc2) *serial_rdy [6.4] equation 6. scc interface pal equations 8530_rd:=8530_cs*write *8530_rd *serial_rdy [7.1a] + 8530_rd*serial_rdy [7.1b] + reset [7.1c] 8530_wr := 8530_cs*write*8530_wr *serial_rdy [7.2a] + 8530_wr*nc1 *serial_rdy [7.2b] + 8530_wr*nc2 *serial_rdy + reset [7.2c] equation 7. scc interface pal equations the serial_rdy signal is a term used by the mstr_con pal device [1.2a] to provide the rdy sig- nal to the am29030 microprocessor [6.2]. the nc1 and nc2 (no connect 1 and 2) terms form a counter to pro- vide the proper delay for reads and writes to allow trans- fer [6.3] and [6.4]. the resulting access is active five cycles: one cycle for nc1, one cycle for nc2, one cycle for nc1 to be active again, one cycle for serial_rdy to sample nc1 and nc2, and one cycle for the rdy to be sampled by the am29030 microprocessor. note that the software timing restriction allowing six scc oscilla- tor cycles between accesses is over and above the nc1 and nc2. the reset signal causes a simultaneous read and write signal to the am85c30 chip, which indicates a reset to the scc [7.1c] and [7.2c]. this can be done by software as well. the serial_rdy signal is initiated one cycle after the 8530_rd signal, which begins one propaga- tion delay after the am29030 microprocessor access to the serial port. the 8530_rd signal lasts until the serial_rdy is done [7.1b]. the 8530_wr write pulse lasts for three clocks, bracketed by one clock cycle on either side [7.2a]. (this is for the data hold time require- ment of the am85c30 chip. note that a slower than 8-mhz part requires an extra wait state here.) finishing touches the ez-030 demonstration board uses a tl7705 chip to provide a clean, adebounced,o power-on reset signal and push-button reset signal. it is possible to attach a jumper between the dtr signal and the push button to allow host-controlled reset serial port. the reset signal configures the processor for 8-bit operation and resets the scc and the am29030 microprocessor. the 74ls157 muxs have long propagating delays and switching times, and therefore do not require series ter- mination resistors for the highly capacitive eight fan-out dram address bus. if high-speed muxs or pal devices are used in lieu of the 74ls157 muxs, it is recom- mended that series 33-ohm termination resistors be used to moderate pulse switching times and dampen reflections.
amd ez-030 demonstration board theory of operation 9 addendum for 16-mhz am29035 cerquad this application note presented how to make a simple memory interface for the am29030, but the timing was for the pga package parts that, because of the superior power distribution, can have the better timings shown in table 1. the cerquad parts, though, have different tim- ings than their pga counterparts, as shown in table 2. this needs to be taken into account in the worst-case analysis of the design. this addendum shows how to produce a design for the cerquad parts that can still maintain the performance of the design shown previously. the problem comes in worst-casing the address to the cas control path to meet the id31id0 setup. the cas term cannot be lowered until the address is valid at the output of the multiplexers (74ls157's). the equation is: t 7 + 74ls157 (16v87) pal fastest delay 22 + 14 2 = 34 ns making the delay line 34 ns instead of 20 ns. this, then, adds to the cas an access of 20 ns plus the worst-case delay time of the pal device, plus the setup of the am29035. this equation becomes: 34 ns + 20 ns + 7 ns + 17 ns => 78 ns this 78 ns significantly exceeds the 63 ns allowed so modifications must be made to the design. there are two paths to fix the design. either fix the ad- dress clock to q delay, or fix the data setup side. to fix the data setup, a register with a better setup, say a 74f374 with a setup of 4 ns, can be inserted. this solves the problem, but adds an extra clock to each access. this also adds four parts to the design for the full 32-bit data bus. there is a less expensive solution. since the only problem is on cas and not ras, only eight address lines need to be dealt with instead of 32 data lines. a 20v810-ns pal device can be used as a clock-enabled register with a control-enabled count. this then reduces the equivalent of t 7 from 22 ns to 7 ns. along with using a 5-ns pal device for cas decode, this results in the set- up time required with 1 ns of margin. to fix the 1 ns of margin, the delay line should be reduced to 15 ns and the 74ls157s changed to 74f157s. this results in an access equation of: delay line + pal delay + cas access + setup 63 15 + 5 + 20 + 19 = 59 63 ns the equation for this pal device (add_cntr.pds) is given at the end of the pal equation section. this pal device should fit between the intersection of the address bus for the prom and the multiplexers, as shown in figure 1, and is only needed on the lower address bits a9a2. there are two control terms on this pal device load, which cause the register to be updated with the ad- dress and inc, which in turn cause the register to count upward. in this design, the term idle can be used in the master control pal device to drive load. the address bus is valid 1 clock before idle becomes false, so the register is updated with the correct address and held. holding this address until the first cas pulse enables the st1 to be used as the needed inc term to produce counts in the burst sequence. these modifications to the basic memory system make the significantly cheaper cerquad part function at or near its top speed. only one small change is needed to the eprom. the speed of the eprom must be de- creased to 70 ns instead of the 90 ns as used in the 25- and 35-mhz designs. table 2. 16-mhz am29035 cerquad interface timing restrictions 1 preliminary 16 mhz no. parameter description test conditions min max unit 7 synchronous output valid delay memclk output memclk input 1 2 20 24 ns 7a synchronous output valid delay for id31id0 memclk output memclk input 1 2 22 26 ns 9 synchronous input setup time memclk output memclk input memclk=inclk 21 21 17 ns 9a synchronous input setup time for id31id0 memclk output memclk input memclk=inclk 17 17 13 ns notes: 1 subject to change without notice. refer to the aam29030 and am29035 microprocessors data sheeto for a complete listing of ac/dc parameters (order# 18150).
amd ez-030 demonstration board theory of operation 10 appendix a. pal equations this appendix contains pal equations for the ez-030 demonstration board. ;palasm ? design description ; declaration segment title ras and cas generator for ez-030 pattern cas_dec.pds revision a company amd date 03/03/92 chip u3 pal16l8 (or palce16v8) ; pin declarations pin 1 memclk pin 2 /mem_access ; pin 3 /ref_access ; pin 4 delay_in ; pin 5 /we0 pin 6 /we1 pin 7 /we2 ; pin 8 /we3 ; pin 9 /write pin 10 gnd pin 11 /st1 ; pin 13 /cas0 pin 14 /cas1 pin 15 /cas2 ; pin 16 /cas3 ; pin 17 /mux ; pin 18 /ras0 pin 20 vcc ; ; boolean equation segment equations minimize_off ras0 = mem_access + ref_access*/memclk + ras0*ref_access mux = mem_access*/memclk + mux*mem_access cas0 = ref_access + mux*mem_access*st1*/write*memclk*delay_in + mux*mem_access*st1*/write*cas0*delay_in + mux*mem_access*st1*/write*cas0*/memclk + mux*mem_access*st1*write*we0*/memclk
amd ez-030 demonstration board theory of operation 11 cas1 = ref_access + mux*mem_access*st1*/write*memclk*delay_in + mux*mem_access*st1*/write*cas1*delay_in + mux*mem_access*st1*/write*cas1*/memclk + mux*mem_access*st1*write*we1*/memclk cas2 = ref_access + mux*mem_access*st1*/write*memclk*delay_in + mux*mem_access*st1*/write*cas2*delay_in + mux*mem_access*st1*/write*cas2*/memclk + mux*mem_access*st1*write*we2*/memclk cas3 = ref_access + mux*mem_access*st1*/write*memclk*delay_in + mux*mem_access*st1*/write*cas3*delay_in + mux*mem_access*st1*/write*cas3*/memclk + mux*mem_access*st1*write*we3*/memclk ; ;palasm design description ; declaration segment title memory cycle generator for ez-030 pattern mstr_con.pds revision a company amd date 03/03/92 chip u4 palce16v8 ; correction 4/25/92 ; cs which is used for read-narrow initialization of the am29030 at reset ; was set to 0 should be 1 correction below ; ; pin declarations pin 1 memclk ; pin 2 /reset ; pin 3 /req pin 4 a31 pin 5 a30 pin 6 /burst ; pin 7 /emul ; pin 8 /serial_rdy ; pin 9 /ref_req ; pin 10 gnd ; pin 11 /oe ; pin 12 /rdy pin 13 /mem_access registered ; pin 14 /ref_access registered ; pin 15 /rom_cs pin 16 /idle registered ; pin 18 /mem_rdy ; pin 19 /st1 ; pin 20 vcc ;
amd ez-030 demonstration board theory of operation 12 ; boolean equation segment equations idle := /idle*/mem_access*/ref_access + reset + /idle*mem_access*st1*mem_rdy*/burst + /idle*ref_access*/ref_req + idle*/(req*a31*/a30)*/ref_req*/emul ref_access := idle*ref_req*/ref_access + ref_access*ref_req mem_access := idle*req*a31*/a30*/ref_req*/mem_access*emul + mem_access*/st1 + mem_access*burst st1 := mem_access*/mem_rdy + st1*burst mem_rdy := req*/a31*/a30*/mem_rdy + mem_access*/mem_rdy + mem_access*mem_rdy*burst rom_cs = /emul*req*/a31*/a30 rdy = mem_rdy + serial_rdy ; ;palasm design description ; declaration segment title refresh timer for ez-030 pattern refresh.pds revision a company amd date 03/03/92 chip u2 palce16v8 ; pin declarations pin 1 memclk ; pin 2 /ref_access ; pin 11 /oe ; pin 12 /ref_req registered ; pin 13 /q0 registered ; pin 14 /q1 registered ; pin 15 /q2 registered ; pin 16 /q3 registered ; pin 17 /q4 registered ; pin 18 /q5 registered ; pin 19 /q6 registered ;
amd ez-030 demonstration board theory of operation 13 ; boolean equation segment equations q0 := /q0 q1 := q1 :+: q0 q2 := q2 :+: (q1*q0) q3 := q3 :+: (q2*q1*q0) q4 := q4 :+: (q3*q2*q1*q0) q5 := q5 :+: (q4*q3*q2*q1*q0) q6 := q6 :+: (q5*q4*q3*q2*q1*q0) ref_req := q6*q5*q4*q3*q2*q1*q0 + ref_req*/ref_access ; title 85c30 controller pattern serial.pds revision a company amd date 04/20/1992 chip u18 pal16v8 clk /reset /req a31 a30 /write /int nc nc gnd /oe /8530_wr /8530_rd /8530_cs /nc1 /nc2 /nc3 /intr*(0) /serial_rdy vcc equations 8530_cs = req*a31*a30 8530_rd := 8530_cs*/write*/8530_rd*/serial_rdy + 8530_rd*/serial_rdy + reset 8530_wr := 8530_cs*write*/8530_wr*/serial_rdy + 8530_wr*/nc1*/serial_rdy + 8530_wr*/nc2*/serial_rdy + reset nc1 := (8530_rd + 8530_wr)*/nc1*/serial_rdy nc2 := (8530_rd + 8530_wr)*(nc1:+:nc2)*/serial_rdy serial_rdy := nc1*nc2*/serial_rdy intr*(0) = int
amd ez-030 demonstration board theory of operation 14 ;palasm design description ; declaration segment title address counter and incrementer for the ez-030 speed fix pattern add_cntr.pds revision a author david stoenner company amd date 10/19/93 chip u10 palce20v8 ; pin declarations pin 1 memclk pin 2 /inc pin 3 a2 pin 4 a3 pin 5 a4 pin 6 a5 pin 7 a6 pin 8 a7 pin 9 a8 pin 10 a9 pin 12 gnd pin 13 /oe pin 14 /load pin 22 q0 registered ; pin 21 q1 registered ; pin 20 q2 registered ; pin 19 q3 registered ; pin 18 q4 registered ; pin 17 q5 registered ; pin 16 q6 registered ; pin 15 q7 pin 24 vcc ; boolean equation segment equations q0 := inc*/q0 + load*a2 q1 := inc*(q1 :+: q0) + load*a3 q2 := inc*(q2 :+: (q1*q0)) + load*a4
amd ez-030 demonstration board theory of operation 15 q3 := inc*(q3 :+: (q2*q1*q0)) + load*a5 q4 := inc*(q4 :+: (q3*q2*q1*q0)) + load*a6 q5 := inc*(q5 :+: (q4*q3*q2*q1*q0)) + load*a7 q6 := inc*(q6 :+: (q5*q4*q3*q2*q1*q0)) + load*a8 q7 := inc*(q7 :+: (q6*q5*q4*q3*q2*q1*q0)) + load*a9 ; copyright ? 1993 advanced micro devices. all rights reserved. am29000, amd, pal, and palasm are registered trademarks of advanced micro devices, inc. 29k, am29005, am29030, am29035, am29050, minimon29k, and scalable clocking are trademarks of advanced micro devices, inc. product names used in this publication are for identification purposes only and may be trademarks of their respective companies.

▲Up To Search▲

Price & Availability of EZ030AN

	To Download EZ030AN Datasheet File
If you can't view the Datasheet, Please click here to try to view without PDF Reader .