National Instruments Network Card DP8400 User Manual

National Semiconductor  
Application Note 302  
Charles Carinalli  
Mike Evans  
The DP8400 Family of  
Memory Interface Circuits  
February 1986  
INTRODUCTION  
The DP8409A multi-mode dynamic RAM controller/driver  
was the first controller to resolve all of these problems. This  
Schottky bipolar device provides the flexibility of external  
access control, along with automatic access timing genera-  
tion, without the need for an external timing generator clock.  
In addition, on-board capacitive drivers allow direct drive for  
over 88 DRAMs. With the simple addition of refresh clocks,  
the circuit can perform hidden refresh automatically. It is the  
DP8409A design that has been used as the spring board for  
a whole family of controllers with faster speed performance  
while maintaining maximum pin upgrade compatibility.  
The rapid development in dynamic random access memory  
(DRAM) chip storage capability, coupled with significant  
component cost reductions, has allowed designers to build  
large memory arrays with high performance specifications.  
However, the development of memory arrays continues to  
have a common set of problems generated by the complex  
timing and refresh requirements of DRAMs. These include:  
how to quickly drive the memories to take advantage of their  
speed, minimization of board space required by the support  
circuitry and the need for error detection and correction.  
Unfortunately, these problems must be addressed with each  
new system design. Full system solutions will vary greatly,  
depending on the DRAM array size, memory speed, and the  
processor.  
All Control On-Chip  
Figure 1 is a block diagram of the DP8409A. the ADS input  
strobes the parallel memory address into the row latches  
R08, the column latches C08, and bank select B0 and  
B1. The nine output drivers may be multiplexed between the  
row or column input latches, or the 9-bit on-chip refresh  
counter. One of four RAS outputs is selected during an ac-  
cess cycle by setting the bank select inputs B0 or B1. All  
four RAS outputs are active during refresh. Either external  
or automatic control is available on-chip for the CAS output,  
while an on-chip buffer is provided to minimize skew associ-  
ated with WE output generation.  
This application note introduces a complete family of DRAM  
support circuits that provides a straightforward solution to  
the above problems while allowing a high degree of flexibili-  
ty in application with little or no performance penalty. The  
DP8400 family (Table I) includes DRAM controllers, error  
detection/correction circuits, octal address buffers and sys-  
tem control circuits. The LSI blocks are designed with flex-  
ible interfaces, making application possible with all existing  
DRAMs including the recently announced 1 Mbit devices.  
Additionally, interface is easy to all popular microprocessors  
with memory word widths possible from 8 to 80 bits.  
All DRAM address and control outputs on the DP8409A can  
directly drive in excess of 500 pF, or the equivalent of 88  
DRAMs (4 banks of 22 DRAMs). All output drivers are  
closely matched, significantly reducing output skew. Each  
output stage has symmetrical high and low logic level drive  
capability, insuring matched rise and fall time characteris-  
tics.  
TABLE I. DP8400 Family Members  
DP8400-2,  
DP8402A  
16 and 32 Bit Error  
Checker/Correctors  
Flexibility and Upgradability to 256k or 1 Mbit DRAMs  
DP8408A, DP8409A,  
DP8417, DP8418,  
DRAM Controller/Drivers  
The 9 multiplexed address outputs and 9-bit internal refresh  
counter of the DP8409A direct addressing capability for  
256k DRAMs. Careful design of memory boards, using 64k  
DRAMs with the DP8409A, insures direct upgradability to  
256k DRAMs. This can be done by simply allowing for board  
address extension by two bits and designing the ninth ad-  
dress trace (Q8) of the DP8409A to connect to pin 1 of the  
DRAMs (A8). This is, in general, a non-connected pin in  
64ks and the ninth address in 256ks. All that need be done  
is to remove the 64ks and replace them with 256ks, thereby  
increasing the memory on the same board by a 4 to 1 ratio.  
The resulting development cost saving can be significant.  
DP8419, DP8428, DP8429  
DP8420, DP84244  
DP84XX2  
DRAM Buffer Drivers  
Microprocessor  
Interface Circuits  
FULL FUNCTION DRAM CONTROLLER  
The heart of any DRAM array design is the controller func-  
tion. Previous LSI controllers supplied a minimum function  
of address multiplexing with an on-board refresh counter.  
This required external delay line timing and logic to control  
memory access, additional logic to perform memory refresh,  
and external drivers to drive the capacitive memory array.  
The complete solution results in significant access delay in  
relation to DRAM speeds and skews in output sequencing,  
as well as a large component count.  
Although the new 1 Mbit DRAMs require the larger 18 pin  
package, which will require a memory board redesign, up-  
grading the controller portion of the board may need no  
redesign when converting from the DP8409A or DP8419 to  
the new DP8429 1 Mbit DRAM controller driver.  
Three mode pins (M0, M1 and M2) offer externally select-  
able modes of operation, a key reason for the DP8409A’s  
application flexibility (Table II). The operational modes are  
divided between external and automatic memory control.  
A previous LSI solution brought much of this logic on-chip.  
However, it is limited in application to certain microproces-  
sors and has the disadvantage of all access timing originat-  
ing from an external clock, whose phase uncertainty gener-  
ates a delay in actually knowing when an access has start-  
ed.  
C
1995 National Semiconductor Corporation  
TL/F/5012  
RRD-B30M115/Printed in U. S. A.  
Download from Www.Somanuals.com. All Manuals Search And Download.  
TL/F/5012–2  
FIGURE 2. Typical Application of DP8409A Using External Control and Refresh in Modes 0 and 4  
TL/F/5012–3  
FIGURE 3. This figure demonstrates the automatic accessing capability of the DP8409A. Only one strobing edge,  
RASIN, is required for generation of all DRAM access timing signals. This is accomplished with on-chip  
delay generators, eliminating the need for external delay lines. No access timing clock is necessary.  
3
Download from Www.Somanuals.com. All Manuals Search And Download.  
Refreshing  
curred, via the refresh request output (RF I/O pin). The sys-  
tem acknowledges the request for a forced refresh by set-  
ting M2 (refresh) low on the DP8409A and preventing fur-  
ther access to the DP8409A. The DP8409A then uses  
RGCK to generate an automatic forced refresh. The refresh  
request pin then returns to the inactive state, and the  
DP8409A allows the processor to take full system control  
after the forced refresh has been completed.  
The DP8409A also provdes hidden refresh capability while  
in one of the automatic access modes (Figure 4). In this  
mode, it will automatically perform a refresh without the sys-  
tem being interrupted. To do this, the DP8409A requires two  
clock signals, refresh clock (RFCK) which defines the re-  
fresh period (usually 16 ms), and RAS generator clock  
(RGCK), which is typically the microprocessor clock.  
Highest priority is given to hidden refreshing through use of  
level sensing of RFCK. A refresh cycle begins when RFCK  
transitions to a high level. If during the time RFCK is high the  
DP8409A is deselected (CS in the high state) and the proc-  
essor is accessing another portion of the system such as  
another memory segment, or ROM, or a peripheral, then a  
hidden refresh is performed. When a read or write cycle is  
initiated by the processor, the RASIN input on the DP8409A  
transitions low. With CS high, this causes the present state  
of the internal refresh counter to be placed on the address  
outputs, followed by the four RAS outputs transitioning low,  
strobing the refresh address into the DRAM array. When the  
cycle ends, RASIN will terminate, thus forcing the RAS out-  
puts back to their inactive state and ending the hidden re-  
fresh. The refresh counter is then incremented and another  
microprocssor cycle can begin immediately. However, to  
save power, the DP8409A will allow only one hidden refresh  
to occur during a given RFCK cycle.  
OCTAL MEMORY DRIVERS  
For those applications where the memory array is extremely  
large or the controller design is unique to a particular appli-  
cation requirement, specialized high capacitive load ad-  
dress and control buffers are required. However, like any  
other element in a DRAM system, selection of the improper  
driver can have significant impact on system performance.  
In the past, this function has been performed using Schottky  
logic family circuits such as the DM74S240 octal inverter or  
the DM74S244 octal buffer. The output stages of these de-  
vices have good drive capability, but their performance with  
heavy capacitive loads is not ideal for DRAM arrays. The  
key disadvantage of these devices is their non-symmetrical  
rise and fall time characteristics and their long propagation  
delays with heavy load capacitance. The former is a result  
of impedance mismatch in the upper and lower output  
stages. The latter stems from process capability and circuit  
design techniques not tailored to the DRAM application.  
The combined result of all these factors is increased output  
skew in address and control lines when these devices are  
used as buffers.  
In the event that a hidden refresh does not occur, the  
DP8409A must force a refresh before the RFCK’s next  
positive-goingtransition. Thesystemisnotifiedafterthenega-  
tive-going RFCK transition that a hidden refresh has not oc-  
TL/F/5012–4  
FIGURE 4. Hidden and Forced Refresh Timing of the DP8409A  
4
Download from Www.Somanuals.com. All Manuals Search And Download.  
Two new devices are now available for this application. The  
DP84240 is pin and function compatible with the  
DM74S240. The DP84244 is likewise compatible with the  
DM74S244. However, this is where the similarity between  
the devices ends. Both the DP84240 and the DP84244  
have been designed specifically to drive DRAM arrays. Fig-  
ure 5 shows a typical application of the DP84244, used in  
conjunction with the DP8409A, to drive a very large memory  
array.  
with this high speed, chip power dissipation is still main-  
tained at a reasonable level as demonstrated by the graphs  
shown in Figures 7a, 7b of power versus frequency.  
The DP84240 and the DP84244 are fabricated on a high  
performance oxide-isolated Schottky bipolar process. Spe-  
cial circuit techiques have been used to minimize internal  
delays and skews. Additionally, both rise and fall time char-  
acteristics track closely as a function of load capacitance.  
This has been accomplished through impedance matching  
of the upper and lower output stages. The result of these  
characteristics is a substantial reduction of skew in both the  
address and control lines to the DRAM array.  
Figures 6a, 6b show some typical performance curves for  
these circuits. Note that, at over 500 pF, the propagation  
delay through these drivers is on the order of 15 ns. This  
delay includes propagation delay and rise or fall time. Even  
TL/F/5012–5  
FIGURE 5. The DP84244 Used as a Buffer in a Large Memory Array (greater than 88 DRAMs) Controlled by the DP8409A  
TL/F/5012–6  
TL/F/5012–7  
FIGURE 6a. t  
PLH  
Measured to 2.7V on Output vs. C  
FIGURE 6b. t Measured to 0.8V on Output vs. C  
PHL L  
L
5
Download from Www.Somanuals.com. All Manuals Search And Download.  
through these driversÐa delay not shown by the data sheet  
specifications. Additionally, the problem becomes increas-  
ingly severe as multiple driver inputs are used in parallel for  
bus expansion applications.  
Both the DP84240 and the DP84244 are designed to signifi-  
cantly reduce both static and dynamic input capacitance.  
When these devices are driven with standard logic circuits,  
no appreciable overhead delay need be added to the basic  
device delay specifications due to input pulse distortion.  
ERROR CORRECTION  
The determination of whether a DRAM system requires er-  
ror correction must be resolved early in the system design.  
A positive answer to this question may have far-reaching  
impact on board development time and component cost. It  
is clear, however, that such a decision cannot be taken  
lightly.  
TL/F/5012–8  
FIGURE 7a. Typical Power Dissipation for DP84240 at  
e
V
CC  
5.5V (All 8 drivers switching simultaneously)  
The type and origin of errors in DRAM systems are many  
and can result from a number of sources (Table III). Current  
estimates of soft error rates due to alpha particles in 64k  
RAMs indicate some hope that these error rates will be simi-  
lar or possibly better than those found in 16k DRAMsÐbut  
the facts are still somewhat unclear. However, it is clear that  
the use of 256k DRAMs and the introduction in the near  
future of 1 Mbit DRAMs with even smaller memory cells and  
greater chip densities will place a significant challenge on  
DRAM chip designers to keep these rates down. It is be-  
lieved by some that error correction may become mandato-  
ry in future DRAM system designs. Currently, the decision to  
add error correction is not so straightforward. It depends on  
many factors, not the least of which is the end user’s per-  
ception of its value to system uptime and reliability.  
TL/F/5012–9  
FIGURE 7b. Typical Power Dissipation for DP84244 at  
e
V
CC  
5.5V (All 8 drivers switching simultaneously)  
TABLE III. The Sources and Types of Memory Errors  
The output stages of the DP84240 and the DP84244, al-  
though well matched, are relatively low impedance. Output  
impedance is under 10X. Some DRAM arrays will require  
the addition of damping resistors in series with the outputs  
of the drivers. These damping resistors are used to minimize  
undershoot which may have a harmful effect on the DRAMs  
if allowed to become large. This undershoot is caused by  
the high transient currents from the drivers necessary to  
drive the capacitive loads. These high currents pass through  
Error  
Sources  
System Action  
Type  
Alpha Particles  
System Noise  
Chip Patterns  
Power Glitches  
Temporary system errorÐ  
may be overwritten with a  
low probability of repetition  
#
#
#
#
Soft  
Stuck Memory Bit  
Permanent failureÐmay  
act as logic 1 or 0  
#
#
#
a distributed inductive/capacitive circuit created by the  
board traces and the DRAM load, causing the undershoot.  
Hard Memory Chip Interface  
Interface Circuit Failure  
The damping resistor has specifically not been placed on-  
chip because its value is dependent on the DRAM array size  
and board layout. In fact, address lines will quite often re-  
quire a different resistor value from the DRAM control lines.  
The resistor must be tuned for a particular board layout  
since too high a resistor will produce an excessively slow  
edge and too low a resistor will not remove the udershoot.  
Values for damping resistors may vary from 15X to 150X,  
depending on the application. Placing any value of damping  
resistor on-chip, other than a value less than the minimum,  
severely restricts the application of these high performance  
circuits.  
Generally, error correction will always be found in highly reli-  
able systems during DRAMs, such as process control equip-  
ment, banking terminals, and military systems where high  
data integrity and minimum downtime are priorities. Howev-  
er, the importance of error correction has grown substantial-  
ly, to the point that it is now used as selling feature in the  
vast majority of large memory-based systems. In fact, some  
major computer houses have adopted quidelines for use by  
their designers in the development of DRAM arrays. A  
somewhat common set has been foundÐif the memory ar-  
ray is on the order of (/4 million bytes, then word parity  
should be used. This permits the detection of single bit er-  
rors but does not allow error correction. When the total  
memory approaches (/2 million bytes, then double bit error  
detection and single bit error correction should be added.  
Another key advantage of both the DP84240 and the  
DP84244 is their low input capacitance. Previous address  
buffer/drivers (such as the DM74S240/244) have high input  
capacitance. Fast edges at the inputs of these drivers be-  
come slower and distorted due to this dynamic input capaci-  
tance. This problem must be factored as an additional delay  
The decision to add error correction to a system is costly,  
both in memory overhead and control hardware. Table IV  
6
Download from Www.Somanuals.com. All Manuals Search And Download.  
TABLE IV. Check Bit Overhead for Multiple Bit Error  
Detection and Single Bit Error Correction  
to 80-bit data words. It is a 16-bit chip that is easily expand-  
able with the simple addition of more DP8400s for each 16-  
bit word increment.  
Number of Bits  
in Memory  
Number of  
Check Bits  
Required  
Percentage  
of Excess  
Memory  
Figures 9a, 9b and 9c demonstrate its basic operation in the  
write and read memory access cycles. Figure 9a shows the  
normal write cycle, where system data is used by the  
DP8400 to generate parity bits, called check bits, based on  
certain combinations of the data bits. This combination is  
defined by the DP8400’s matrix shown in Figure 10. When-  
ever a ‘‘1’’ occurs in any row, the corresponding input data  
bit at the top of the column helps determine the parity for  
that check bit labeled at the end of the row. These check  
bits are written along with the data at the same memory  
address. Also, during a memory write cycle the DP8400  
checks system byte parity. This is parity associated with the  
data bytes transmitted between the processor and the  
memory card. This is an optional feature that may prove  
very valuable in multiple board memory systems.  
Data Word  
8
5
6
63%  
38%  
16  
24  
32  
48  
64  
6 (7)  
7
25% (29%)  
22%  
7 (8)  
8
15% (17%)  
13%  
Note: The number stated assumes the use of the DP8400; the number in  
parentheses is required by other error correction circuits.  
lists the number of additional memory chips required to sup-  
port single bit error correction and double bit error detection  
as a function of the memory data word width.  
Sometime later a read will occur at this same memory ad-  
dress. The reading of memory data may be performed in  
two ways, as shown in Figures 9b and 9c. In the read cycle,  
the DP8400 uses the data read from memory and internally  
regenerates check bits using the same matrix. These newly  
generated check bits are then compared (using X-OR  
gates) with the check bits read from memory to detect er-  
rors. The result of this comparison is called a syndrome  
word. Any differences in the generated versus read check  
bits will result in at least one syndrome bit true. This indi-  
cates an error in either the read data or check bit field or  
both.  
This table also shows the percentage of DRAM overhead  
required to implement this function. Adding error correction  
also increases the memory access delay, since the informa-  
tion contained in the overhead chips must be analyzed in  
each read and generated in each write operation.  
DP8400 16-Bit Expandable Error Correction Chip  
The DP8400 expandable error checker/corrector is shown  
in block diagram form in Figure 8. This circuit offers a high  
degree of flexibility in applications which range from 8-bit  
TL/F/501210  
FIGURE 8. DP8400 Simplified Block Diagram  
7
Download from Www.Somanuals.com. All Manuals Search And Download.  
TL/F/501211  
FIGURE 9a. Normal Write Mode with DP8400  
TL/F/501212  
FIGURE 9b. Normal Read Mode Using the Error Monitoring Method with the DP8400  
TL/F/501213  
FIGURE 9c. Normal Read Mode Using the Always Correct Method with the DP8400  
*C2, C3 generate odd parity  
TL/F/501214  
FIGURE 10. DP8400 Matrix  
8
Download from Www.Somanuals.com. All Manuals Search And Download.  
A key advantage of the DP8400 is that it has three error  
flags detailing the type of error occurrence. These are gen-  
erated using the syndrome word in the manner shown in  
Figure 11. The resulting error type identifications are shown  
in Table V. The three error flags allow complete error type  
identification, plus the unique determination of double bit  
errors, which will be key during the discussion of double bit  
error correction. Also, on a memory read, the DP8400 gen-  
erates byte parity bits for transmission to the processor  
along with the data.  
Double Bit Error Correct  
The probability of double bit errors in DRAM systems is rela-  
tively low, but as memory array sizes grow, the occurrence  
of these error types must be considered. Adopting certain  
practices, such as rewriting a memory location whenever an  
error is detected, or using ‘‘memory scrubbing’’ techniques,  
can significantly reduce the probability of a double soft error  
occurrence. Memory scrubbing is when the system, during  
low usage, actually accesses memory solely for the purpose  
of identifying and correcting single soft errors. This is an  
important technique if there are segments of the memory  
that are not always being accessed so that soft error occur-  
rences would not be quickly found.  
The occurrence of a double error comprising one soft and  
one hard must now be considered. This type of error has a  
higher probability than two soft errors. The hard error may  
be due to a catastrophic chip failure, and a subsequent soft  
error will create two errors. This can be a source of concern  
since most error correction chips cannot handle double er-  
rors of this type. Therefore, most systems will ‘‘crash’’ when  
a catastrophic chip failure is coupled with a soft error in the  
same memory address.  
The DP8400 has been designed to handle just such an oc-  
currence. It can correct any double bit error, as long as at  
least one of the errors is a hard error. The DP8400 does this  
without the need for extra hardware required for the basic  
double bit detect/single bit correct system implementation.  
This method is called the double complement correct tech-  
nique and is demonstrated in Figure 12 using a 4-bit data  
word for simplicity. In this example, a single hard error is  
located in the most significant bit of a particular memory  
location and a soft error occurs at the next bit. The position  
of the errors is not important since the errors may be distrib-  
uted in either the data or check bit field or both. First, the  
data word and corresponding check bits are written to this  
memory location. When a later read of this location occurs,  
step A, two errors are directly reported by the DP8400 error  
flags. The system detects this, disables memory; and places  
the DP8400 in the complement write mode. This causes the  
previously read data and check bits to be complemented in  
the DP8400 and written back to the same memory address,  
step B, writing over the previous soft error. Obviously this  
does not modify the cell where the hard error exits. The  
system then reads from the same address again, but this  
time it places the DP8400 in the complement read mode,  
step C. The DP8400 again complements the memory data  
and check bits and generates new check bits based on the  
new data word. At this point, the chip detects a single bit  
error in the bit position where the soft error occurred, and  
using the conventional single error correction procedure, re-  
turns corrected data to the system, step D.  
TL/F/501215  
TABLE V. Error Flags after Normal Read  
AE  
0
E1  
E0  
0
Error Type  
No Error  
0
1
1
0
Single Check Bit Error  
Single Data Error  
Double-Bit Error  
1
1
0
1
1
0
All Others  
Invalid Conditions  
There are two basic memory read methods that may be  
used with the DP8400. The first is shown in Figure 9b and is  
called the error monitoring method. Here, the read data is  
assumed to be correct and the processor immediately acts  
on the data. If the DP8400 detects an error, the processor is  
interrupted using the any error flag (AE). Using this method,  
there is no detection delay in most memory reads since  
errors seldom occur, but when an error does occur, the  
processor must be capable of accepting an interrupt and a  
read cycle extension to obtain the corrected data from the  
DP8400.  
In the second read, the complement read, the hard error  
repeats since this bit location again receives a bit which is  
complemented with respect to itself. But the soft error has  
been overwritten and does not repeat. Effectively, the mem-  
ory has complemented the hard bit error position twice and  
the soft bit error position only once, while the DP8400 com-  
plements both positions twice. Therefore, after the second  
read, there is only one error left, the soft error. Since this is  
now a single error it can be directly corrected.  
A second approach is called the always correct method,  
Figure 9c. In this case, the data is always assumed to be in  
error and the processor always waits for the DP8400 to ana-  
lyze whether an error exits. Then the corrected or un-  
changed data is read from the DP8400. Although this meth-  
od results in longer memory read time, every memory read  
will always be of the same delay except when a double error  
occurs. The selection of which method to use depends on  
many factors, including the processor, system structure,  
and performance.  
9
Download from Www.Somanuals.com. All Manuals Search And Download.  
After the complement correct cycle, the memory must be  
rewritten with the corrected data since the address now  
contains data that is complemented. Full error reporting is  
available from the DP8400 after the second read, the com-  
plement read, of memory. This is shown in Table VI.  
ice. Using this technique, software may be provided in the  
system to warn the operator that the system is in a degrad-  
ed operational mode and that field service should occur  
shortly. In the meantime, the system will continue to operate  
properly. The key to the effectiveness of the DP8400 in this  
application is its three error flags which allow complete error  
reportingÐincluding a unique double error indication.  
This method is a very effective tool to avoid system crash  
due to memory chip failure, and can do much to reduce  
unscheduled field service calls. The only time the system  
will see a double error that is not directly correctable is  
when a double soft error occurs. The probability of this is  
very low if the previously discussed techniques are used.  
The extra time taken to do an additional read and write of  
memory is insignificant when the alternative is a system that  
has a catastrophic failure that requires immediate field serv-  
DP8402A, 3, 4, 5 32-Bit Error Detector and Corrector  
(EDAC)  
In addition to the popular DP8400-2 16-bit error checker/  
corrector, National offers a family of 32-bit Error Detector  
and Correctors (EDACs). With  
DP8402A, 3, 4, function in  
a few exceptions, the  
similar manner to the  
5
a
DP8400-2. One major exception is that the DP8402A, 3, 4, 5  
are not expandable beyond 32 bits.  
TL/F/501216  
FIGURE 12. Double Error Correct Complement Hard Error  
MethodÐ1 Hard Error and 1 Soft Error in Data Bits  
TABLE VI. DP8400 Error Flags after a Complement Read  
AE E1 E0  
Error Type  
Two Hard Errors  
0
1
1
1
0
1
1
0
0
0
1
0
One Hard Error, One Soft Check Bit Error  
One Hard Error, One Soft Data Bit Error  
Two Soft Errors, Not Corrected  
10  
Download from Www.Somanuals.com. All Manuals Search And Download.  
MICROPROCESSOR INTERFACE CIRCUITS  
This system structure requires the insertion of few or no wait  
states during memory access cycle, thus maximizing  
a
The major 8-bit, 16-bit and 32-bit microprocessors have dif-  
ferent control signal timing. There are also a number of  
speed options. The DP8400 family was designed, not for a  
specific microprocessor, but rather, significant control flexi-  
bility has been provided on both the DP84XX DRAM control-  
ler/drivers and the DP84XX error correction devices for  
easy interface to any microprocessor. However, a certain  
amount of ‘‘glue’’ is necessary to interface to these LSI cir-  
cuits, usually in the form of a number of MSI/SSI logic cir-  
cuits. Not only can this be costly in board space utilization,  
but it is usually the one place where the most design related  
problems occur in system development.  
throughput. The DP84XX2 circuits have been designed to  
work with all of National’s DRAM controller/drivers to con-  
trol refreshing so that system throughput is affected only  
when absolutely necessary. First, in any refresh clock peri-  
od of 16 ms, hidden refreshing is given maximum opportuni-  
ty. This can be helped with the optional DP84300 refresh  
interval generator which offers maximum high-to-low ratio-  
ing of RFCK. Second, when a hidden refresh does not occur  
in a particular RFCK cycle, a forced refresh may still not  
affect a slow access cycle. The worst-case is when an ac-  
cess is pending during a forced refresh, in which case a  
three wait state delay is usually the maximum penalty.  
Figures 13 and 14 show the DP8400 family solution to this  
problemÐthe DP84XX2 series of microprocessor interface  
circuits. Figure 13 shows how the DP84300 refresh timer  
and the DP84XX2 microprocessor interface circuit connect  
to the DP8409A and various microprocessors for a typical  
application. Figure 14 shows the DP8409A and the DP8400  
together in a microprocessor-based memory system using  
DRAMs, with double bit error detect and single bit error cor-  
rect capability. In addition, it shows that with the simple ad-  
dition of some standard data buffers, how the system can  
implement byte writing to the DRAM array.  
Usually two DP84XX2 type chips would be required to inter-  
face between any microprocessor and the DP8400/  
DP8409A combined system. These chips would handle the  
read/write control as well as error detection and correction  
control. Table VII shows the individual DP84XX2 circuits  
that would be used in systems with no error correction, thus  
requiring only the DP84XX DRAM controller/driver function.  
TL/F/501217  
FIGURE 13. Connecting the DP8409A between 16-Bit Microprocessor and Memory  
11  
Download from Www.Somanuals.com. All Manuals Search And Download.  
The DP8400 DRAM interface family provides complete solu-  
tions to memory support. This begins with the LSI functions  
such as the DP8400 expandable error checker/corrector  
and the DP8409A DRAM controller/driver. It continues with  
the DP84240 and the DP84244 high performance buffer/  
drivers. Finally, it concludes with easy interface to popular  
microprocessors with the use of the DP84XX2 series. It is  
the first family of DRAM support circuits designed  
for universal applications with multiple microprocessors,  
with no manufacturers CPU enjoying a favorite role.  
Data sheets and more detailed application information are  
available for all the members of the DP8400 family. Contact  
your local National Semiconductor representative or Nation-  
al Semiconductor directly.  
TL/F/501218  
FIGURE 14. Flexible Application of the DP8409A and DP8400.  
This Figure Shows an Application with a 16-Bit Microprocessor.  
TABLE VII. The DP84300 Series of Interface  
Circuits for Various 16-Bit Microprocessors  
System Using  
Microprocessor  
DP84XX DRAM Controller/Driver  
National & TI  
Series 32000  
DP84412  
DP84512  
National & TI  
Series 32332  
Motorola  
DP84322 or  
DP84422  
68000/08/10  
Motorola  
68020  
DP84522  
DP84532  
DP84432  
Intel  
80286  
Intel  
8086/186/88/188  
Zilog  
8000  
(2) 74S64  
(1) 74S04  
12  
Download from Www.Somanuals.com. All Manuals Search And Download.  
13  
Download from Www.Somanuals.com. All Manuals Search And Download.  
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.  
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  
Corporation  
National Semiconductor  
Europe  
National Semiconductor  
Hong Kong Ltd.  
National Semiconductor  
Japan Ltd.  
a
1111 West Bardin Road  
Arlington, TX 76017  
Tel: 1(800) 272-9959  
Fax: 1(800) 737-7018  
Fax:  
(
49) 0-180-530 85 86  
@
13th Floor, Straight Block,  
Ocean Centre, 5 Canton Rd.  
Tsimshatsui, Kowloon  
Hong Kong  
Tel: (852) 2737-1600  
Fax: (852) 2736-9960  
Tel: 81-043-299-2309  
Fax: 81-043-299-2408  
Email: cnjwge tevm2.nsc.com  
a
a
a
a
Deutsch Tel:  
English Tel:  
Fran3ais Tel:  
Italiano Tel:  
(
(
(
(
49) 0-180-530 85 85  
49) 0-180-532 78 32  
49) 0-180-532 93 58  
49) 0-180-534 16 80  
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.  
Download from Www.Somanuals.com. All Manuals Search And Download.  

Miele Vacuum Cleaner S 246i User Manual
Murray Lawn Mower 405005x99A User Manual
NAD Cassette Player 602 User Manual
NEC Flat Panel Television P701AVT User Manual
NordicTrack Home Gym NTL078060 User Manual
Olivetti Printer 16W User Manual
Optima Batteries Projector TWR1693 User Manual
Panasonic Computer Drive LF 1004AB User Manual
Panasonic Intercom System KX TD161 User Manual
Patton electronic Modem 1088 C User Manual