|
|
 | | From: | Gui Xie | | Subject: | about forward error correction codes? | | Date: | Sat, 8 Jan 2005 23:40:46 +0900 |
|
|
 | Hi:
Are there any forward error correction codes which can correct 30 to 40
percentage bit errors in a bit sequence?
Thanks
Gui Xie
|
|
| | From: | Mark Adler | | Subject: | Re: about forward error correction codes? | | Date: | 12 Jan 2005 21:03:06 -0800 |
|
|
| Gui Xie wrote:
> Could you please tell me
> what code achieves the most t/n in the coding literature?
With all due respect, you will learn more by perusing the literature
yourself.
mark
|
|
| | From: | Mark Adler | | Subject: | Re: about forward error correction codes? | | Date: | 8 Jan 2005 10:46:43 -0800 |
|
|
| Gui Xie wrote:
> Are there any forward error correction codes which can correct
> 30 to 40 percentage bit errors in a bit sequence?
Yes. Reed-Solomon codes are nicely configurable to define a message
length and codeword length to get the desired performance. However 30%
to 40% is definitely pushing it, depending on what your noise source
looks like and what error rate you can accept after decoding. Since
Reed-Solomon corrects m-tuples of bits (often m = 8), it works best
when the bit errors come in bursts so they mostly end up in the same
m-tuples. (By the way, if you can identify erasures, i.e. missing data
such as a lost packet or a failed hard drive, the R-S code can correct
twice as many m-tuple erasures as it can m-tuples of unknown location
in error.) If the bit errors at your quoted rates are randomly
scattered, you'll need a lot of redundancy in the coding, using any
method, to get significant improvement in the error rate.
Note that at exactly a 50% bit error rate, all information is lost and
correction is not possible. Though at 60%, it's possible again since
that's just like a 40% rate with the bits inverted. In any case,
you're in a very difficult regime.
mark
|
|
| | From: | Gui Xie | | Subject: | Re: about forward error correction codes? | | Date: | Sun, 9 Jan 2005 21:23:06 +0900 |
|
|
| Thank you very much for your help.
Denote a code by (n,k,t), where n, k, t are the codeword length, message
length, and error correction capability respectively. This code can correct
at most t bit errors in the codeword of the n bits. Could you please tell me
what code achieves the most t/n in the coding literature? I know, the BCH
code (255,9,63) get 24% for t/n.
Thanks.
"Mark Adler" ????
news:1105210003.435502.190530@c13g2000cwb.googlegroups.com...
> Gui Xie wrote:
> > Are there any forward error correction codes which can correct
> > 30 to 40 percentage bit errors in a bit sequence?
>
> Yes. Reed-Solomon codes are nicely configurable to define a message
> length and codeword length to get the desired performance. However 30%
> to 40% is definitely pushing it, depending on what your noise source
> looks like and what error rate you can accept after decoding. Since
> Reed-Solomon corrects m-tuples of bits (often m = 8), it works best
> when the bit errors come in bursts so they mostly end up in the same
> m-tuples. (By the way, if you can identify erasures, i.e. missing data
> such as a lost packet or a failed hard drive, the R-S code can correct
> twice as many m-tuple erasures as it can m-tuples of unknown location
> in error.) If the bit errors at your quoted rates are randomly
> scattered, you'll need a lot of redundancy in the coding, using any
> method, to get significant improvement in the error rate.
>
> Note that at exactly a 50% bit error rate, all information is lost and
> correction is not possible. Though at 60%, it's possible again since
> that's just like a 40% rate with the bits inverted. In any case,
> you're in a very difficult regime.
>
> mark
>
|
|
|
