Partitioned interleaver memory for map decoder
AssigneeQUALCOMM Incorporated
InventorInyup Kang
Life of this patent
Estimating the term…
Abstract
The present invention is a novel and improved technique for decoding with particular application to turbo, or iterative, coding techniques. In accordance with one embodiment of the invention a system for decoding includes a channel deinterleaver RAM for storing a block of symbol estimates and for reading out at least three different portions of said block in substantially simultaneous fashion. A set of 3 state metric calculators receives the three portion in a substantially simultanous fashion and generate corresponding sets of state metric calculations from said three portions of the channel deinterleaver block.
Claims8
A system for decoding comprising:
- a) channel deinterleaver RAM for storing a block of symbol estimates and for reading out at least three different portions of said block before any of the below-recited state metric calculations have been generated;
- b) set of 3 state metric calculators, each state metric calculator for generating a set of state metric calculations;
- c) multiplexer bank for coupling said set of 3 state metric calculators to said at least three different portions of said block.
A system for performing turbo decoding comprising:
- a) channel deinterleaver RAM for storing a block of symbol estimates and for reading out at least three portions of said block, hereafter referred to as a first portion, a second portion, and a third portion, before any of the below-recited state metric calculations have been generated;
- b) set of 3 state metric calculators, each state metric calculator for generating a set of state metric calculations;
- c) multiplexer bank for coupling said set of 3 state metric calculators to said at least three different portions of said block;
- d) APP memory for storing a priori values and for reading out at least three different portions of said a priori values.
A method for performing turbo decoding comprising:
- a) reading out at least three different portions of a channel estimate block at least three different times, with all of the portions being read out before any of the below-recited state metric calculations have been generated for that time;
- b) each time, generating state metrics for each different portion during each different read out;
- c) storing a first set of said state metrics;
- d) calculating symbol estimates using a first set of estimates and a second set of state metrics, where said second set of state metrics is calculated using an initialization value calculated using a third set state metrics.
Dependencies are inferred from the claim text.