Below, are some command files containing examples of the use of the LDPC programs, together with the output I obtained for these examples. Output on other machines might conceivably be slightly different, due to different round-off errors. The run-examples script runs all the example scripts and compares their output with the outputs that I obtained (on a Pentium machine).
ex-ham7b, output in ex-ham7b-out
A (7,4) Hamming code used with a BSC. Demonstrates encoding of random messages and decoding to minimize bit error rate by exhaustive enumeration.
ex-ham7a, output in ex-ham7a-out
A (7,4) Hamming code used with an AWGN channel. Tested using zero messages. Decoded by exhaustive enumeration to minimize either block or bit error rate, and by probability propagation.
ex-dep, output in ex-dep-out
Examples of how parity check matrices with linearly dependent rows (ie, redundant parity checks) are handled. This is probably not of great interest to most users.
ex-ldpc-encode, output in ex-ldpc-encode-out
Encodes messages with an LDPC code using sparse, dense, and mixed representations of the generator matrix.
ex-ldpc36-1000a, output in ex-ldpc36-1000a-out
A (2000,1000) LDPC code with 3 checks per bit and 6 bits per check. Three encoding methods are tried out, and the code is tested on an AWGN channel at various noise levels, using random messages.
ex-ldpc36-5000a, output in ex-ldpc36-5000a-out
A (10000,5000) LDPC code with 3 checks per bit and 6 bits per check. Tested on an AWGN channel at various noise levels, using random messages. Pipes are used to avoid creating lots of files.
ex-ldpcvar-5000a, output in ex-ldpcvar-5000a-out
A (10000,5000) LDPC code with the number of checks per bit varying from 2 to 7. Tested on an AWGN channel at various noise levels, using random messages. Pipes are used to avoid creating lots of files. Performance is better than for the code above in which the number of checks is the same for all bits.
ex-wrong-model, output in ex-wrong-model-out
Tests what happens when messages are decoded using the wrong noise model, including using the right type of model but with the wrong noise level, and using the wrong type of model (ie, using an AWLN model for messages transmitted through an AWGN channel, or vice versa).