forward error correction for 8 bit data

[fun2design83]

Hi all,
Its been a while I am reading about forward error correction (FEC) and I am not sure what scheme works the best. My application requires to send a byte (signal). this signal after passing the channel should be decoded in receiver and it should be corrected if it has error. In this system, we do have just one way communication, and re-sending the signal in case of error is so expensive. What FEC scheme do you recommend and why?

 

[ads-ee]

What you use is based on the requirements, things like BER, the symbol rate requirements, etc. none of which you've touched on.

As you have no requirements, use Hamming (7:4) on each nibble of your bytes. Reason: it's easy to implement. ;-)

 

[fun2design83]

What you use is based on the requirements, things like BER, the symbol rate requirements, etc. none of which you've touched on.

As you have no requirements, use Hamming (7:4) on each nibble of your bytes. Reason: it's easy to implement. ;-)

just bare with me, I have limited background in communication, so you might here some none-sense from me. sorry about that in advance.

I am thinking to use instead of NRZ use 4 different voltage level which are going to be translated into 4 symbols. so if everything is done right, each symbol should represent 2 bits, so in order to send this byte I need to send 4 symbols. these symbols are far apart from each other about 1dB. I don't know how much noise would be introduced by the channel ( I don't have the communication channel ready yet to do test). So to answer your question I dont know whats my BER.

 

[ads-ee]

What you just described isn't FEC. It seems like what you are describing is your modulation scheme, which doesn't sound very robust. Why don't you just use something well known like BPSK or QPSK, both of these are still used for channels with lots of noise and fading issues.

You probably should read some introductory stuff on modulation first. https://www.pitt.edu/~dtipper/DM.pdf

The FEC is supposed to correct the small number of errors you do receive and correct them so you don't need to re-transmit. Most communication channels use something like Reed-Solomon, Turbo codes, and LDPC codes. The current "art" is LDPC codes even thought they've existed for decades the hardware requirements were prohibitive, until recently. Depending on whether the channel has burst errors you might need to interleave/de-interleave the symbols before/after transmit/receive.

Given your limited communication systems background, you probably need to do a lot of reading before making any decisions, or you can find someone with the proper credentials to design the system for you.