Abstract: A method and apparatus for decoding a linear block encoded string of information bits comprising: converting the string into a plurality of codewords. Performing hard and soft decisions on each codeword to generate a hard and soft decision vector. Computing the syndrome and finding the location of the two minimum values by Galois Field Arithmetic. Designating these values LOW1 and LOW2 and xoring with a Nc1, thus generating Nc2. Swapping Nc1 with Nc2 and determining the lowest soft decision value, Min1 and a next lowest value, Min2. The two bit locations creating Min1 are designated as MinA and MinB. MinA being replaced with Min2 minus the value MinA. MinB being replaced with Min2 minus the value at MinB. Generating an output codeword by subtracting Min1 from all other bit locations values and 2's complementing all soft values with 0 in their location. Creating the new soft value vector.

Abstract: A memory block structure for use within a viterbi decoder includes multiple dual port RAMs configured as multiport RAMs. The memory block structure is configured to allow a one-word write operation and an N-word read operation during a single clock cycle in order to achieve one decoded output symbol per clock period using the viterbi algorithm. By using dual port RAMs, a more densely packed and less expensive memory block structure is achieved. An encoded stream of input symbols are input to the viterbi decoder and written to the memory block structure one word at a time. Once X+Y bits have been written to the memory block structure, the decoder will then read N words from the memory block structure, simultaneously, reading back through X+Y words and outputting Y bits N at a time at the end of the trace back through memory.

Abstract: A single clock cycle adaptive data compressor/decompressor with a string reversal mechanism is described which can perform data compression and decompression at the rate of one uncompressed symbol per clock cycle. The LZW data compression algorithm has been improved for use in this data compressor. The compressor builds a string table as the data is received. Each string within the table is made up of the address within the table of the longest previously seen matching string and the one character that makes this string different. This data compressor/decompressor utilizes a content addressable memory to store the string table. This content addressable memory allows the compressor to store the current symbol string in a table while that same string table is simultaneously searched for the current string. During decompression the characters within a symbol string are output in reverse of the order in which they were input.

Abstract: A direct memory access controller controls many direct memory access ports using a token passing scheme. The system multiplexes the port's accesses to external random access memory by daisy-chaining a loop of direct access memory ports and passing the token around to each port. Once a port receives the token it may request as many random access memory accesses as it requires. These accesses may be either read operations or write operations with both using the same multiplexed data bus. The latency inherent in reading an external RAM causes no loss in the access efficiency. When the port has completed its data transfer or if the port does not require a data transfer, the token is passed to the next direct memory access port for its data transfer. The token is passed around to all connected ports until all have had an opportunity to complete any memory transfers which they required. Each port is identical except for a binary identification code that is used to represent each port.

Abstract: An adaptive lossless data compression system with systolic string matching logic performs compression and decompression at the maximum rate of one symbol per clock cycle. The adaptive data compression system uses an improvement of the LZ1 algorithm. A content addressable memory (CAM) is used to store the last n input symbols. The CAM is stationary, stored data is not shifted throughout the CAM, but rather the CAM is used as a circular queue controlled by a Write Address Pointer Counter (WREN). During a compression operation, a new input symbol may be written to the CAM on each clock cycle, while simultaneously the rest of the CAM is searched for the input symbol. Associated with each word of the CAM array is a String Match State Machine (SMSM) and, an address logic module (ALM). These modules detect the occurrence of strings stored in the CAM array that match the current input string and report the address of the longest matching string nearest to the Write Address Pointer.

Abstract: A content addressable memory cell includes six transistors connected together to perform memory read, memory write, and matching operations. This cell has the ability to perform typical memory write and memory read operations as well as the capability of signalling whether or not its stored data matches data that is being searched for. A cross-coupling scheme is used in the memory cell so that a high potential will always be stored on the gate of a transistor whose source is at ground. This cross-coupling scheme increases the amount of charge stored on the storage transistor and decreases the required frequency of refresh operations. In addition to the transistors configured to store data, an additional transistor configured as a diode is used as a rapid discharge path to maximize the efficiency of the cell during a read operation. During a match operation another transistor is utilized to discharge the Match line quickly in the event the stored data does not match the data that is being searched for.

Abstract: A content addressable memory cell includes six transistors connected together to perform memory read, memory write, and matching operations. This cell has the ability to perform typical memory write and memory read operations as well as the capability of signalling whether or not its stored data matches data that is being searched for. A cross-coupling scheme is used in the memory cell so that a high potential will always be stored on the gate of a transistor whose source is at ground. This cross-coupling scheme increases the amount of charge stored on the storage transistor and decreases the required frequency of refresh operations. In addition to the transistors configured to store data, an additional transistor configured as a diode is used as a rapid discharge path to maximize the efficiency of the cell during a read operation. During a match operation another transistor is utilized to discharge the Match line quickly in the event the stored data does not match the data that is being searched for.

Abstract: The present invention is for a Error Correction Unit (ECU) that uses a single stack architecture for the generation, reduction and evaluation of the polynomials involved in the correction of a Reed-Solomon code. The circuit uses the same hardware to generate the syndromes, reduce the .OMEGA.(x) and .LAMBDA.(x) polynomials and evaluate the .OMEGA.(x) and .LAMBDA.(x) polynomials. Some of the specifics involved in calculating and reducing the polynomials mentioned above are novel as well. First, the implementation of the general Galois field multiplier is new and faster than previous implementations. Second, the circuit for implementing the Galois field inverse function has not appeared in prior art designs. Third, a novel method of generating the .OMEGA.(x) and .LAMBDA.(x) polynomials (including alignment of these polynomials prior to evaluation) is utilized. Fourth, corrections are performed in the same order as they are received using a premultiplication step prior to evaluation.