Dear Sir,

As far as I know, Reed-Solomon codes and BCH codes are usually hard codes, meaning no soft decoding. The soft decoding is so difficult because these codes operate over a finite field (a Galois field), and this mathematical structure is not compatible with the domain of the real numbers R, which you need for soft decoding.

Usually, if you want soft decoding, you need to resort to either Viterbi coding or LPDC coding.

The advantage of the BCH and Reed-Solomon codes, however, is the math over the finite field, which results in the complexity of the decoder being proportional to the number of check bytes, instead of being a function of the block length.

In practical application of error correction, to have both the advantage of the finite field, and the advantage of the soft decoding (which results in better SNR), tandem codes are used, eg:

1. A first LPDC code offering soft decoding followed by

2. A BCH code offering very low outgoing error rate.

or

1. A first Viterbi code, offering soft decoding. This code need not protect all

    symbol bits. It can eg work protecting a few LSB, with the rest  unprotected

2. A Reed-Solomon code offering very low outgoing error rate.

Whether you use a Viterbi or LPDC code first, is up to you. Both have advantages

and disadvantages. If you use a Viterbi code, please note it has a tendency to

produce burst errors (if it fails), and then the second layer should be a Reed-Solomon code (and not BCH), as its performance on burst errors is better.

Regards,

Henri.

http://ieeexplore.ieee.org/abstract/document/6012528/