High speed turbo decoder

A method for decoding a segment of a received code by an instance of one or more instances of a Maximum-A-Posteriori (MAP) turbo decoder, the segment being subdivided into windows, the method comprising:

• for each window, concurrently computing, by forward and backward gamma engines respectively, gamma branch metrics in a forward and a backward direction, wherein the gamma branch metrics are based on received bits, a combination of transmitted bits, and a-priori information for the received bits;
• for each window, sequentially computing, by forward and backward state metric engines comprising respective lambda engines and coupled to the respective forward and backward gamma engines, a forward and a backward state metric, respectively, and determining, by the lambda engines, Log Likelihood Ratios (LLRs) for received and decoded bits, and soft decisions for the received bits, wherein the forward and the backward state metrics are based on a respective gamma branch metric and a respective initial value;
• and
• computing, by a post-processor, extrinsic data based on the forward and backward state metrics, and providing a decoded segment.

A device, comprising:

• a processor;
• and
• a memory configured to store a program instructions to be executed by the processor, where executing the program instructions causes the processor to perform operations for receiving a signal over a wireless channel and decoding a segment of a received code by an instance of one or more instances of a Maximum-A-Posteriori (MAP) turbo decoder, the segment being subdivided into windows, the operations comprising:
• concurrently computing, for each window, gamma branch metrics in a forward and a backward direction, wherein the gamma branch metrics are based on received bits, a combination of transmitted bits, and a-priori information for the received bits;
• sequentially computing, for each window, a forward and a backward state metric and determining Log Likelihood Ratios (LLRs) for received and decoded bits, and soft decisions for the received bits, wherein the forward and the backward state metrics are based on a respective gamma branch metric and a respective initial value;
• and
• computing extrinsic data based on the forward and backward state metrics, and providing a decoded segment.