Systems and methods for a turbo decoder in a universal mobile telecommunication system (UMTS)

A turbo decoder comprising:

• a plurality of maximum a posteriori (MAP) engines;
• a first plurality of extrinsic memory banks and a second plurality of extrinsic memory banks;
• and
• wherein each of the first and second pluralities of extrinsic memory banks is accessible by at least one of the plurality of MAP engines, the turbo decoder further comprising:
• a plurality of systematic memory banks;
• a plurality of parity memory banks;
• and
• wherein the turbo decoder is configured such that during decoding:
• in a first half of a decoding iteration, a MAP engine of the plurality of MAP engines:
• (i) receives as input systematic bits si from a systematic memory bank of the systematic memory banks, a priori data ei from the first plurality of extrinsic memory banks, and parity-1 bits pi from a first parity memory bank of the plurality of parity memory banks,
• (ii) generates v′i=si+e′i, and
• (iii) stores v′i in the second plurality of extrinsic memory banks, where e′i is the a priori information generated by the MAP engine during the first half of the decoding iteration, and
• in a second half of a decoding iteration, the MAP engine:
• (iv) receives as input interleaved data vj from the second plurality of extrinsic memory banks, and parity-2 bits pj from a second parity memory bank of the plurality of parity memory banks,
• (v) generates extrinsic data e′j, and
• (vi) stores e′j in the first plurality of extrinsic memory banks.

A method of decoding using a turbo decoder comprising N MAP engines,

• a first plurality of extrinsic memory banks, and a second plurality of extrinsic memory banks, wherein each of the first and second pluralities of extrinsic memory banks is configured to store and retrieve data according to a R×C matrix having a format similar to that of an interleaver table, the method comprising:
• in a first half of a decoding iteration, the N MAP engines reading data from the first plurality of extrinsic memory banks by row according to a first predetermined sequence;
• in the first half of the decoding iteration, the N MAP engines writing data to the second plurality of extrinsic memory banks by row according to the first predetermined sequence;
• in a second half of the decoding iteration, the N MAP engines reading data from the second plurality of extrinsic memory banks by column according to a second predetermined sequence.