LDPC Decoder Message Formatting Based on Activity Factor Minimization Using Differential Density Evolution

The power consumption of digital implementations of low-density parity-check decoders is studied using a density evolution approach which allows the capture of the switching statistics of the messages through the code communications network. Messages are crafted specifically to reduce the switching activity during a decoding cycle, which forms the major cause of power consumption. Low-power message formats are designed, particularly differential signaling with various bit sizes. The performance and potential of these novel message formats are evaluated and compared to hardware measurements and simulations. It is shown that the power consumption of an LDPC decoder can be as low as a few tens of pico joules per decoded information bit, even with standard state-of-the art silicon technology.