Technical Field

This invention relates to a method and appar- atus for correcting an error in a multi-bit data word.

Background Art

In modern day processing systems, semiconduc¬ tor memory devices such as dynamic random access memory devices have been utilized. Due to the inherent capa- citance structure of such memory devices, errors are introduced into the data stored therein. In order to overcome this deficiency in the use of RAM memory de¬ vices, error coding techniques have been employed. Some of these techniques include the use of error correcting codes contained in each data word stored in the memory device or systems .which implement the well-known Hamming code (see "Error Detection and Error Correcting Codes" by R. W. Hamming, Bell Systems Technical Journal, Vol. 26, No. 2, April 1950, pp. 147-160) or other similar codes. Such systems employ redundant data which involves complex networks and large memory storage areas.

Disclosure of the Invention

It is the object of the present invention to provide a method and apparatus for correcting an error in a multi-bit data word, which is simple in construc¬ tion and hence low in cost.

Therefore, according to the present invention, there is provided a method for correcting an error in a multi-bit data word, characterized by the step of storing a multi-bit data word in a memory unit; gener¬ ating a first error correction code word from the data word being stored in said memory unit by dividing said data word by a polynomial; generating a second error correction code from said data word when reading said data word from said memory unit; comparing said first

-2-

and second erro ^' r correction code words; determining an erroneous bit in the data word read from said memory unit if said first and second error correction code words do not match; and correcting the erroneous bit. According to another aspect of the present invention, there is provided apparatus for correcting an error in a multi-bit data word, characterized by a memory unit; division means adapted to divide by a poly¬ nomial a data word applied for storage in said memory unit thereby generating a first error correction code word from said data word, and further adapted to gen¬ erate a second error correction code word from said data word when reading said data word from said memory unit; and processing means adapted to compare the first and second error correction code words, determine any erroneous bit in the data word read from said memory unit in dependence on the result of the comparison and correct the erroneous bit.

Preferably, the memory unit in which the ulti- bit data word is stored is a dynamic RAM (random access memory) and the error correction code word in the form of the remainder generated by the division of the data ^' word by the polynomial is stored in a static RAM. This has the advantage that the storage area available for data word storage in the dynamic RAM is not diminished by the storage of error correction code words.

In accordance with a preferred embodiment of the invention division by the polynomial is effected in a multi-stage shift register constructed to effect such division. When the data word is read out of the dynamic RAM memory unit the word is stored in the static RAM memory and again divided by the polynomial. The remain¬ der generated as a result of the second division is compared with the remainder stored in the static RAM memory unit to determine if an error has occurred in storing the data word in the dynamic RAM memory unit. If the remainders do not match, indicating an error in

the data word read out of the dynamic RAM memory device, the remainder stored in the static RAM memory unit is loaded into the shift register followed by a predeter¬ mined number of zero bits. Each time a zero bit is loaded into the shift register, the output of each stage of the shift register is examined and a count of the zero bits being loaded in the shift register is generated. The error bit in the data word is detected when the output of the last stage of the shift register is a binary bit one and the remaining stages each have an output of a binary bit zero. The location of the error bit in the data word stored in the static RAM is pointed to by the count occurring at this time, which is then complemented correcting the error bit.

Brief Description of the Drawings

One embodiment of the invention will now be described by way of example with reference to the accompanying drawings, in which:

Figs. 1A and IB taken together define a block diagram of a portion of a data processing system;

Figs. 2A and 2B taken together define a block diagram of the remaining portion of the data processing system;

Fig. 3 is a block diagram of the dynamic RAM controller showing its relationship to the dynamic RAM main memory unit;

Figs. 4A and 4B taken together disclose the logic circuit of the error correction polynomial block of the dynamic RAM controller of Fig . 3 ; Fig. 5 is a block diagram of the circuits which control access to the dynamic RAM memory unit by the dynamic RAM controller;

Fig. 6 is a block diagram of the timing gen¬ eration logic circuit block of the dynamic RAM control- ler;

Fig. 7 is a schematic representation of the timing signals used in transferring data between the

OMP

dynamic RAM memory unit and ^' the DMA controller;

Fig. 8 is a flowchart of a write operations- Fig. 9 is a flowchart of a read operation; Fig. 10 is a flowchart of an error correction operation.

Best Mode for Carrying Out the Invention

Referring now to Figs. 1A, IB, 2A and 2B, there is disclosed a block diagram of the data proces¬ sing system which incorporates the present invention. The data processing system includes a main processor board and an auxiliary memory board which includes a dynamic random access memory (DRAM) unit for increasing the memory capacity of the main processor board. The main processor board includes a main processor 20 (Fig. 1A) which communicates with the auxiliary memory board over a number of bus lines and control lines in a manner that is well-known in the art. The main processor 20 may consist of an Intel 8085 microprocessor which in¬ cludes internal registers, counters, pointers and asso- ciated logic circuits well-known in the art. The pro¬ cessor 20 outputs and receives over a bi-directional 8- bit bus 22 the low order Address-Data bits AD _n -AD ₇ inclusive; outputs over line 24 the active low Hold

Acknowledge control signal HLDA; outputs over the 8-bit bus 26 the Read and Write control signals RD, WR re¬ spectively and the high order Address signals A„ , A-, ₂ - ₁₅ inclusive; receives over the input line 28 the Restart signal RST5.5; outputs over line 30 the Master

Reset signal MRST, outputs over line 32 the Low Power Fail Detect signal LPFD; outputs over line 34 the In¬ ternal Reset signal IRS and outputs over the 2-bit bus 36 the 5 volt Power Supply signal and the ground con¬ nection GRD. The data bits appearing on the bus 22 are stored in a 74LS373 buffer 38 prior to transfer to either the main processor 20 or to an Intel 8155 I/O controller unit 40 which controls the orderly flow of

the data bits between the main processor 20 and an Intel 8085 memory processor unit 62 (Fig. IB). The control signals appearing on the bus 26 are inputted into a 74LS138 decoder 44 which in response to receiving the signals stated above outputs the Status Strobe signal

STB over line 46 to a 74LS367 bus driver 48 functioning as a status port unit which also receives over the 8-bit bus 50 the high order Address-Data signals AD^-AD-, inclusive from the I/O controller 40, which signals are then transmitted over the 4-bit bus 52 to the 8-bit bi¬ directional bus 22 connected between the buffer 38 and the main processor 20 under the control of the strobe signal STB. The decoder 44 also outputs over line 56 the Receive Status Strobe signal ASTB to the controller 40 and over line 58 the Write Status Strobe signal

DSTB, which' signals strobe the data signals out of the controller 40.

The I/O controller 40 outputs the Address-Data signals AD _Q -AD_ inclusive over the 8-bit bi-directional bus 60 (Figs. 1A and IB) to the memory processor 62 which controls the transfer of data between the main processor 20 (Fig. 1A) and an auxiliary dynamic random access memory (DRAM) unit 64 (Fig. 2B) . Other signals transmitted between the I/O controller 40 and the memory processor 62 include the Clock Out signal CLK OUT, the Read and Write signals RD, WR, the Address Latch Enable signal ALE, and the I/O Memory Write signal i/o/M. These signals are transmitted between the I/O controller 40 and the processor unit 62 over the 5-bit bi-direc- tional bus 66. Appearing on the 3-bit bus 68 is the

Interrupt signal INTR while the Timing signals TIMER OUT appears on line 70 and the Reset signal RESET OUT appears on line 72. All of these signals control the transfer of the data bits between the I/O controller 40 (Fig. 1A) and the memory processor 62 in a manner that is well- known in the art.

As shown in Fig. IB, the memory processor unit

62 will transmit/receive the data bits D _Ω ~D ₇ inclusive

>

-6-

over the bi-directional bus 74 and the low order address bits A _Q -A ₇ inclusive over the 8-bit bidirectional bus 76 to a latch member 78 which, upon the Latch Enable signal ALE appearing on line 80 becoming active, will output the address signals A--A- inclusive over the 8-bit bus 82. These signals are transmitted to a ROM memory unit 84 (Fig. IB) and a static RAM (SRAM) memory unit 86 in which the address bits are used in addressing data locations in the memory units 84 and 86. The SRAM memory unit 86 includes a first storage portion for storing the data words being transferred between the processor 62 and the memory 64, a second storage portion for storing an error correction code for each of the data words stored in the memory unit 84 in a manner that will be described more fully hereinafter and a third storage portion for storing a bit count, this latter storage portion functioning as a zero bit counter. The data bits D _Q -D ₇ are stored in the first storage portion prior to a write operation of the memory unit 64 (Fig. 2B) . During a read operation, the data bits read from the memory unit 64 are stored in the first storage portion of the memory unit 86 prior to transfer over bus 74 to the processor unit 62.

Appearing on the output line 88 of the memory processor unit 62 is the Hold Acknowledge signal HLDA while the high order address bits A_-A, ,. inclusive appear on the 8-bit bi-directional bus 90 for trans¬ mission to a SN74LS138 three to eight decoder unit 92 and also to the ROM memory unit 84 and the SRAM memory unit 86 for addressing data locations in the memory.

The decoder 92 which receives the low order address bits A--A-, inclusive over the bus 82 will also output memory enabling signals over lines 94 and 96 to the memory units 84 and 86 respectively. The decoder 92 will also output an enabling signal over line 98 to the I/O con¬ troller 40 (Fig. 1A) and over line 100 to a flip-flop 102 (Fig. 1A) which outputs an enabling signal over line

OMPI

104 to the status port unit 48 enabling its operation. The flip-flop 102 is reset by the Master Reset signal

MRST appearing on line 30. The decoder 92 (Fig. IB) further outputs the Chip Select signals CS _Q and CS, on lines 106 and 108 respectively which are used in ad¬ dressing the DRAM memory unit 64 (Fig. 2B) in a manner that is well-known in the art. The decoder 92 further outputs Error Correction Clock signals ECC ₁ -ECC _a inclu¬ sive over line 109 and ECC ₈ -ECC, _g inclusive over line 111 which are used in enabling the reading of signals representing the remainder of a division operation generated during the transfer of data bits between the DRAM memory unit 64 and the SRAM memory unit 86 as will be described more fully hereinafter. Further included in the data processing system is a battery back-up unit 110 (Fig. IB) which supplies power over line 112 to the SRAM memory unit 86 in case of a power failure in the system.

Referring now to Figs. 2A and 2B, which il- lustrate in block form the remaining portion of the data processing system incorporating the present invention, there is shown a direct memory access (DMA) controller 114 used in transferring data between the memory pro¬ cessor unit 62 (Fig. IB) and the DRAM memory unit 64 (Fig. 2B) . The DMA controller 114 receives the Hold

Acknowledge signal HLDA over line 88 (Fig. IB) enabling the controller to transfer data bits from the DRAM memory unit 64 to the SRAM memory unit 86 and then to the memory processor unit 62 (Fig. IB) which then in turn transfers the data bits to the main processor 20.

The controller 114 further outputs the low order Address bits A _Q -A ₇ inclusive ^" over the 8-bit bi-directional bus 82 and receives the Chip Select signal CS over line 106, the clock signals CLK OUT which are transmitted over the 4-bit bus 66 and inverted by the inverters 116 before being received by the controller 114 over bus 117 and the reset signal RESET OUT over line 72. In response

OMPI

-b-

to receiving the signal HOLD, the 8085 memory processor 62 will output the signal HLDA over line 88. to the DMA controller 114 (Fig. 13) notifying the DMA controller of the completion of the transfer of the data bits from the DRAM memory unit 64. The raising of the Interrupt signal TRAP over line 120 notifies the memory processor 62 of the starting or stopping of a data transfer oper¬ ation by the DMA controller 114.

The DMA controller 114 further outputs in a multiplex arrangement the data and address signals required in accessing the DRAM memory unit 64 over an 8- bit bi-directional bus 122 in which the high order address bits A _g -A, ₅ inclusive are stored in a latch 124 (Fig. 2A) under the control of the Address Strobe signal ADSTB appearing on line 126. The address bits are then outputted over the 8-bit bus 90 for use in addres¬ sing the storage locations in the DRAM memory unit 64 (Fig. 2B) and the SRAM memory unit 86 upon the gener¬ ation of the Address Enable signal AEN which appears on line 127. The data signals D _Q -D ₇ inclusive appearing on the 8-bit bi-directional bus 74 are transmitted during a write operation of the DRAM memory unit 64 from the SRAM memory unit 86 (Fig. IB) to a DRAM controller 128 (Fig. 2A) whose construction will be described more fully hereinafter.

After the data bits D _Q -D ₇ inclusive appear on the bi-directional bus 74 during a read or write oper¬ ation, the DMA controller 114 outputs the Acknowledge signal DACK over line 130 to the DRAM controller 128. As will be explained more fully hereinafter, the DRAM controller 128 during a write operation generates an error correction code word for each data word being written into the DRAM memory unit 64. This error cor¬ rection code word is then stored in the second storage portion of the SRAM memory unit 86 (Fig. IB). When the same data word is again read from the DRAM memory unit 64, a second error correction code word is generated

OMPI *Λr _ WIPO ^~

which, if no error is present in the data word being read from the DRAM memory unit 64, will be the same as the first error correction code word. This construction allows the data words which do not contain error cor- rection code bytes to be stored in the DRAM memory unit 64, thereby allowing more data words to be stored in the DRAM memory unit.

During a data transfer operation, the memory processor 62 (Fig. IB) transmits the sector address signals over the data bus 74 and strobes these signals into the DRAM controller 128 with the Chip Select sig¬ nals CS. appearing on line 108 (Fig. 2A) . The DRAM controller 128 outputs the Memory Read control signals MEMR over line 132 and the Memory Write control signal MEMW over line 134 for transferring data .between the

SRAM memory unit 86 (Fig. IB) and the 'DRAM memory unit 64 (Fig. 2B) . The controller 128 further outputs the Column Refresh signals CAS0-CAS2 inclusive over lines 136-140 inclusive to a SN74LS138 three to eight decoder 142 (Fig. 2B) and the Row Refresh control signals RAS0- RAS2 over lines 144-148 inclusive to a second SN74LS138 three to eight decoder 152. The decoders 142 and 152 output the appropriate column and row refresh signals over lines 154 and 156 to the DRAM memory unit 64 for refreshing the cells in the dynamic memory unit in a manner that is well-known in the art.

The data signals D , D _ouτ stored in the memory unit 64 are serially transmitted over the 2-bit bus 180 to the DRAM controller 128. The address signals RA--RA _g inclusive, used in addressing the storage sec¬ tors of the memory unit 64 during a read or write oper¬ ation, are transmitted in parallel over the 9-bit bus 160 to a buffer unit 162 for storage therein which in turn is connected to the memory unit 64 over the 9-bit buses 164 (Fig. 2B) . The DRAM memory unit 64 is pro¬ vided with up to 128 64K dynamic RAM chips or 128 256K x 1 dynamic RAM chips providing a memory capacity of 4

megabytes. When addressing the memory unit 64, the signal WE is transmitted from the controller 128 over line 166 for enabling the access operation of the memory unit 64 to occur. During the initialization of the system, the switches 168a and 168b (Fig. 2A) notify the controller 128 of the system configuration.

Referring now to Fig. 3, there is shown in dotted outline a block diagram of the DRAM controller 128 (Fig. 2A) which includes a DRAM address generator 170 which receives over the 8-bit bi-directional bus 74 the address bits used for addressing the storage sectors of the DRAM memory unit 64 during a read or write oper¬ ation. The address bits appearing on the bus 74 are multiplexed by an address multiplexing circuit 172 under the timing control of thirteen timing signals generated by a timing- generation logic unit 174 and transmitted over the 13-bit bus 175. The logic unit 174 also out¬ puts timing signals over a 3-bit bus 176 to a refresh multiplexer 179, comprising the three to eight decoder 142 (Fig. 2B) , and a Clock signal CLOCK over line 177 to an error correction polynomial unit 182. The multi¬ plexed address signals are transmitted by the multi¬ plexing circuit 172 over a 9-bit bus 173 to the DRAM memory unit 64. The data signals appearing on the 8-bit bus 74 are also transmitted to a serial-to-parallel/ parallel-to-serial converter 178 which converts the data from parallel to serial when the data is being written into the memory 64 and from serial-to-parallel when the data is read from the memory 64. The serial data sig- nals are outputted from the converter 178 over line 180 to an error correction polynomial unit 182 which divides the serial data signals appearing on line 180 by a polynomial to generate a remainder comprising the error correction code word in a manner to be described more fully hereinafter. The serial data signals D _IN , ^D _0U T are also transmitted over line 180 to the memory unit 64 during a write operation and are outputted over line 180 during a read operation.

-li-

Further included in the DRAM controller unit 128 is a control unit 184 which receives over bus 186 the control signals WR, RD, HOLD and will output the Acknowledge signal DACK over the same bus. In response to receiving these control signals, the control unit 184 will output the appropriate control signals over line 188 for controlling the transfer of the address signals between the read/write multiplexer 189, comprising the three to eight decoder 152 (Fig. 2B) , and the memory unit 64. The decoder 152 receives over a 4-bit bus 190 a portion of the address signals for use in addressing the memory unit 64. During the time the memory unit 64 is being accessed, the decoders 142 and 152 output the column and row refresh signals CAS, RAS over the 8-bit buses 154 and 156 to the memory unit 64 enabling the cells in the memory unit to be refreshed.

Referring now to Figs. 4A and 4B, there is shown details of the error correction polynomial unit 182 (Fig. 3) which includes a plurality of 74LS74 flip- flops 192a-192o inclusive forming a dynamic shift regis¬ ter generally indicated by the numeral 193 (Fig. 4A) . As disclosed on page 360 of the publication "Error Correction ^* Codes" by W. W. Wesley Peterson and E. J. Weldon, Jr., MIT Press, 1972, the method for deriving t e polynomial for a 256 byte word length is described in corollary 11.2. Utilizing the irreducible poly¬ nomial of degree 9 from the Tables found on page 476 of the reference, we derive the following equation:

(1) g(x)=(x ⁹ +x ⁴ +l) (x ⁶ +l) which gives g(x)=x ¹⁵ +x ¹⁰ +x ⁹ +x ^δ +x ⁴ +l.

To find the maximum number of binary bits that can pass through the polynomial and still correct a one bit error in a bit length series of 9 is found by the following equation. ⁶ ^ ^9"15

If x is equal to 2, then n max is eq ^ual to 3066.

Subtracting the term 15 which is equal to the number of flip-flop stages in the polynomial logic unit 182 from the above figure, we find that n is equal to ' 3051.

Since the actual bit length word of the memory unit 64 is 256 bytes or 2048 bits, the error correction code is shortened by the insertion of a number of zeros into the polynomial unit 182 which is equal to the difference between the actual bit length of the word stored in the memory unit 64 and the maximum number of bits that can be corrected. In the present example, this difference is 1003.

The data being inputted into the polynomial unit 182 appears on line.180 (Figs. 3 and 4A) and is inputted into one input of an Exclusive OR gate 194 which also receives at its other input the output signal of the last 74LS74 flip-flop 192o (Fig. 4B) of the shift register 193 appearing on the feedback line 196. The Exclusive OR gate 194 will complement the binary input signal appearing on line 180 if the binary signal on line 196 is a 1. The output signal of the Exclusive OR gate 194 will output a signal over line 198 to the D input of the first flip-flop 192a of the shift register 193 and also over line 200 to the input of each of a number of Exclusive OR gates 202-208 inclusive. By examining Figs. 4A and 4B, it will be seen that the Exclusive OR gate 194 receiving the output of the flip- flop 192o will correspond to the term in the poly- nomial equation (1) disclosed previously. In a similar manner, the Exclusive OR gate 202 corresponds to the term x , the gate 204 to the term x , the gate 206 to the term x 9 and the gate 208 to the term x10. As the binary data bits appearing on line 180 are shifted through the flip-flops 190a-192o inclusive under the control of the clock signals CLOCK appearing on line 177

-1 J-

(Figs. 3 and 4A) , the Exclusive OR gates 194 and 202-208 inclusive will perform successive additions-in modulo 2 arithmetic of the binary bits resulting in the binary bit word being divided by the polynomial, thereby producing a multi-bit remainder appearing on the Q output lines 210 of the flip-flops 192a-192o inclusive and at the input of a pair of 74LS244 tri-state buffer units 212 (Fig. 4A) and 214 (Fig. 4B) . At the conclu¬ sion of the operation of the polynomial unit 182, the error correction code clock signals ECC, „ inclusive appearing on line 109 and the clock signals ECC _{Q *} . _ inclusive appearing on line 111 will enable the data bits appearing on the input lines 210 of the buffer units 212 and 214 to be transmitted over the 8-bit bus 74 for storage in 'the SRAM memory unit 86 (Fig. IB) for use in an error detecting procedure as will be described more fully hereinafter.

Referring now to Fig. 5, there is disclosed the logic circuit of the serial-to-parallel/parallel-to- serial converter unit 178 located in the DRAM controller 128 (Fig. 3) which includes a 74LS199 converter 222 receiving the data bits D _Q -D ₇ appearing on bus 74 and the serial data bits appearing on lines 180a and 180b of the bus 180 (Fig. 2B) . The converter 222 will convert the binary bits to the proper sequence between the parallel bus 74 and the serial lines 180a and 180b under the control of signals appearing on line 224 and gener¬ ated by a Start/Stop control unit 226 located in the DMA/CPU control unit 184 (Fig. 3) in response to the control unit 226 receiving the clock signals 4096,

4096 over the lines 228b and 228a of the bus 228 from the timing generation logic unit 174 (Fig. 3) and the read and write control signals RD, WR appearing on lines 186a and 186b of the bus 186. The control unit 226 controls the transfer of data bits between the DRAM memory unit 64 (Fig. 2B) and the DMA controller unit 114 (Fig. 2A) . As disclosed in Fig. 7, upon the 4096 clock

OMPI

signals 230 going high, the DMA request (DREQ) signal 232 will go high enabling data to be transferred over bus 74 between the SRAM memory unit 86 and the converter 178 (Fig. 3). For every data byte transferred, the DMA controller 114 will raise the DMA acknowledge (DACK) signal 234 (Fig. 7) over line 130 (Fig. 2A) of the bus 186 to a DMA handshake 74LS74 flip-flop 236 (Fig. 5) which resets the flip-flop resulting in the lowering of the DREQ signal 232 appearing on line 238. When the 4096 clock signal 230 goes low (Fig. 7), the control unit 226 will lower the transfer/refresh enable signal 240 over line 188 resulting in the completion of a DRAM memory access operation and the start of a refresh operation of the DRAM memory unit 64. Upon the raising of the next 4096 clock signal by the timing generation logic unit 174 (Fig. 3), the DREQ signal 232 is again raised allowing for the transfer of data from the next sector to occur. The read and write signals RD and WR appearing on lines 186a and 186b respectively iden- tify the type of memory access operation that is occur¬ ring in a manner that is well-known in the art.

Referring now to Fig. 6, there is shown a block diagram of the timing generation logic unit 174 located in the DRAM controller 128 (Fig. 3), which includes an oscillator 242 outputting the system clock signals of 4.068 MHz. to a timing sequence unit 244 comprising a plurality of flip-flops wired in series for outputting the strobe signals RAS and CAS and the 4.068 MHz. system clock signals over the bus lines 176 to the multiplexer 189 (Fig. 3) enabling the high address signals Ag-A, ₅ to be strobed into the DRAM memory unit 64. The sequence unit 244 also outputs the 7-bit clock signal 246 (Fig. 7) over line 248 to the converter 222 (Fig. 5) controlling the transfer of an 8-bit data word appearing on the bus 74 into the converter 222. The 4.068 MHz. system clock signals are also transmitted over line 250 to a 13-bit ripple counter 252 which

outputs a count of 0-4096 over the 12-bit bus 175 to the address multiplexing circuit 172 (Fig. 3) which outputs the signals over bus 173 as the row address signals in the DRAM memory unit 64. The counter 252 will also output the 4096, 4096 clock signals (Fig. 7) over lines 228b and 228a, respectively, to the start/ stop control unit 226 (Fig. 5) for initiating either an access operation or a refresh operation, the latter occurring at the time the clock signal 4096 occurs. Referring now to Fig. 8, there is shown a flowchart of a write operation of the data processing system. The processor 62 (Fig. IB) will initiate a write operation (block 254) by loading (block 256) the address bits A _n -A, _ς into the DRAM controller 128 (Fig. 2A) and set up the DMA controller 114 for a write oper¬ ation by loading the addresses of the data bits stored in the first storage portion of the SRAM memory unit 86 (Fig. IB) which are to be transferred to the DRAM memory unit, reset (block 258) the flip-flops 192a-192o inclu- sive (Figs. 4A and 4B) of the polynomial logic unit 182 (Fig. 3) and start (block 260) transferring the data bits from the SRAM memory unit 86 (Fig. IB) to the serial-to-parallel converter unit 17-3 (Fig. 3) upon the appearance of the high 4096 clock signals 230 (Fig. 7). The data bits are then stored in the DRAM memory unit 64 (block 262). The data is also transferred (block 264) from the converter unit 178 to the error correction polynomial unit 182 (Fig. 3) where a 15 bit remainder representing an error correction code is generated. After all the data bits have been stored in the memory unit 64 and have been transferred to the polynomial logic unit 182 (block 268), the processor will enable the buffers 212, 214 (Figs. 4A and 4B) to output the remainder of the polynomial unit operation for storage in the second storage portion of the SRAM memory unit 86 at the same address as the data is stored in the DRAM memory unit 64, thereby completing a valid write oper¬ ation (block 270).

In reading the same data from the DRAM memory unit 64 (block 272) (Fig. 9), the processor _. 62 (Fig. IB) will again load the address bits of the data to be read (block 274) into the DRAM controller unit 128 (Fig. 2A) and set up the DMA controller 114 for a read operation, reset (block 276) the flip-flops 192a-192o inclusive (Figs. 4A and 4B) of the polynomial unit 182 (Fig. 3) and wait (block 278) for the appearance of a high 4096 clock signal 230 (Fig. 7) before transferring the. serial data bits from the DRAM memory unit 64 to the serial-to- parallel converter unit 178 (Fig. 3) which transfers the parallel data bits to the first storage portion of the SRAM memory unit 86 (block 280). The data bits are also transferred from the memory unit 64 to the polynomial unit 182 (block 282) where a remainder is generated by the operation of the polynomial unit. At the completion of the transfer of the data bits (block 284), the pro¬ cessor will read the remainder present at the output of the buffer units 212, 214, (Figs. 4A and 4B) and the remainder stored in the second storage portion of the

SRAM memory unit 86 at the same address as the data bits read from the DRAM memory unit 64 (block 286) and com¬ pare the two remainders (block 288) . If the two re¬ mainders are the same, the memory processor unit 62 interrupts the main processor unit 20 indicating the presence of data stored in the buffer 38 (Fig. 1A) (block 292) which is to be transferred to the main processor unit. If the remainders do not match, indi¬ cating an error in the data read from the memory unit 64, an error correction operation occurs (block 290) correcting the error if it is limited to a single bit error.

Referring now to Fig. 10, there is shown a flowchart of an error correction operation. After the processor 62 (Fig. IB) senses that the error correction codes generated by the error correction polynomial unit 182 (Fig. 3) do not match, the processor 62 will shift

(block 292) the original 15 bit error correction code stored in the SRAM memory unit 86 (Fig. IB), into the polynomial unit 182. After shifting the error correc¬ tion code into the polynomial unit 182, the processor 62 will shorten (block 294) the cyclic code operation of the polynomial unit 182 by shifting 1,003 binary zero bits into the polynomial unit after which the processor will start shifting a second string of zero bits into the polynomial unit 182. Prior to this operation, the processor 62 resets the count stored on the third storage portion of the memory unit 86 representing the bit counter portion of the SRAM memory unit 86 (Fig. IB) . Upon shifting the first zero bit of the second string into the polynomial unit 182, the processor will store (block 296) a count in the third portion of the SRAM memory unit 86. After shifting in the first zero bit, the processor 62 will read (block 298) the output signals of the flip-flops 192a-192o (Figs. 4A and 4B) by enabling the buffers 212, 214 to output the signals over bus 74 to the processor 62. If the output signal of the flip-flop 192o (Fig. 4B) is not a binary bit one (block 300), it will check (block 302) the count stored in the bit counter portion of the SRAM memory unit. 86. If the count is not equal to 2,048 representing the maximum bit length of a data word that can be stored in the memory unit 64, the processor will increment (block 304) the count in the bit counter portion of the memory unit 86 and shift (block 306) another zero bit into the polynomial unit 182. This process will be repeated until the output signals of the polynomial unit 182 comprise a sequence of binary bit zeros in the first 14 flip-flops 192a-192n inclusive and a binary bit one in the 15th flip-flop 192o, indicating the detection (block 308) of an error bit whose bit location in the data word stored in the SRAM memory unit 86, is pointed to by the count residing in the bit counter portion of the SRAM memory unit. The processor will then complement (block

OMPI

310) the error bit of the data word stored in the SRAM memory unit 86 correcting the error bit and. generating a signal indicating a good read operation (block 312) . If after inserting 2,048 zero bits in the polynomial unit 182 the processor has not found the required sequence of signals in the polynomial unit (block 300) and finding the bit count to be equal to 2,048 (block 302), the processor 62 will recognize (block 314) the presence of a multi-bit uncorrectable error condition and generate a signal indicating a bad read operation (block 316).

It will be seen from this construction that a single error bit is corrected as part of a read oper¬ ation of the dynamic RAM memory unit 64 which utilizes a simple dynamic shift register constructed to divide the word being read from the memory by a polynomial. This arrangement requires a small area of the SRAM memory unit 86 to be utilized for storage of a count repre¬ senting the number of zero bits shifted into the poly- nomial unit 182.

The Intel integrated circuits referenced in this application are commercially available from the Intel Corporation of Santa Clara, California while the remaining circuit references are commercially available from the Texas Instruments Corporation of Dallas, Texas.