soft decision decoder for binary BCH codes

Hi,
Is there a good algorithms to perform soft decision decoding (maybe
iterative) for binary BCH codes ?

Thanks.

edim <edim0@walla.com> wrote:

>Is there a good algorithms to perform soft decision decoding (maybe
>iterative) for binary BCH codes ?

If the number r of check bits is reasonably small you can
use a Viterbi, BCJR, or generalized Viterbi algorithm with
2^r states.  Constructing the appropriate trellis is 
quite simple.   The BCJR algorithm will give you the
same exact result as a brute-force, maximum-likelyhood
decoder that searches all possible codewords.

"Small" might mean something like r < 10 or 12, depending
on your implementation constraints.

For larger r iterative decoding is not a bad idea, but
I don't think there's an accepted best way to construct
such a decoder.

Steve

edim wrote:

> Hi,
> Is there a good algorithms to perform soft decision decoding (maybe
> iterative) for binary BCH codes ?

There are two general approaches, but they are not so good.

1) You can use any trellis search algorithm such as Viterbi, M-algorithm 
or Fano sequential decoder.

2) You can find the nearest valid BCH codeword by hard decision, then 
construct all codewords within some distance from that word, then check 
those codewords to find which makes the best match with the signal.

Binary BCH codes are designed for the simple algebraic decoding. If the 
soft decision is meant, then there are *lot* better codes. Why looking 
for the difficulties?


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

Vladimir Vassilevsky  <antispam_bogus@hotmail.com> wrote:

>Binary BCH codes are designed for the simple algebraic decoding. If the 
>soft decision is meant, then there are *lot* better codes. Why looking 
>for the difficulties?

I can speculate that sometimes, one is handed a BCH code and
needs (or think one needs) to decode it past its hard-decision
limit.

Furthermore some BCH codes are close to optimal for their length.
(Optimal meaning approaching the sphere-packing limit described
by Pollara et.al.)  A (23,12) Golay code is also a BCH code
and there is no more optimal code of the same rate and size.

Steve

On Jan 20, 9:36&#4294967295;am, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:
> edim wrote:
> > Hi,
> > Is there a good algorithms to perform soft decision decoding (maybe
> > iterative) for binary BCH codes ?
>
> There are two general approaches, but they are not so good.
>
> 1) You can use any trellis search algorithm such as Viterbi, M-algorithm
> or Fano sequential decoder.
>
> 2) You can find the nearest valid BCH codeword by hard decision, then
> construct all codewords within some distance from that word, then check
> those codewords to find which makes the best match with the signal.
>
> Binary BCH codes are designed for the simple algebraic decoding. If the
> soft decision is meant, then there are *lot* better codes. Why looking
> for the difficulties?
>
> Vladimir Vassilevsky
> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com

I tried soft decoding POCSAG once, and it improved performance
significantly. The signal you are decoding is usually a given you need
to do your best with.

Regards,
Steve

<steveu@coppice.org> wrote:

>I tried soft decoding POCSAG once, and it improved performance
>significantly. 

I would expect maybe 2 dB improvement in AWGN for that code.

Steve

"Steve Pope" <spope33@speedymail.org> wrote in message
news:gl3jvo$hq4$1@blue.rahul.net...
> Vladimir Vassilevsky  <antispam_bogus@hotmail.com> wrote:
>
> >Binary BCH codes are designed for the simple algebraic decoding. If the
> >soft decision is meant, then there are *lot* better codes. Why looking
> >for the difficulties?
>
> I can speculate that sometimes, one is handed a BCH code and
> needs (or think one needs) to decode it past its hard-decision
> limit.
>
> Furthermore some BCH codes are close to optimal for their length.
> (Optimal meaning approaching the sphere-packing limit described
> by Pollara et.al.)  A (23,12) Golay code is also a BCH code
> and there is no more optimal code of the same rate and size.

IIRC there is a tailbiting convolutional code with the constraint length of
only 4, which is equvalent to the binary Golay code (Zigangirov's book).
If the soft decision is meant, there is no point in using BCH codes.

Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com

"Steve Pope" <spope33@speedymail.org> wrote in message
news:gl3ma8$kcr$1@blue.rahul.net...
> <steveu@coppice.org> wrote:
>
> >I tried soft decoding POCSAG once, and it improved performance
> >significantly.
>
> I would expect maybe 2 dB improvement in AWGN for that code.
>

That's about right for AWGN, however the POCSAG decoder is located after the
FM discriminator. So I doubt if there will be much difference of soft
decision vs hard decision.

Vladimir Vassilevsky
DSP and Mixed Signal Consultant
www.abvolt.com

On Jan 20, 5:58�am, "Vladimir Vassilevsky"
<antispam_bo...@hotmail.com> wrote:
> "Steve Pope" <spop...@speedymail.org> wrote in message
>
> news:gl3ma8$kcr$1@blue.rahul.net...
>
> > <ste...@coppice.org> wrote:
>
> > >I tried soft decoding POCSAG once, and it improved performance
> > >significantly.
>
> > I would expect maybe 2 dB improvement in AWGN for that code.
>
> That's about right for AWGN, however the POCSAG decoder is located after the
> FM discriminator. So I doubt if there will be much difference of soft
> decision vs hard decision.
>
> Vladimir Vassilevsky
> DSP and Mixed Signal Consultantwww.abvolt.com

Why can't soft values after FM discriminator provide some gain in a
soft decoder?

John

Hi,
The reason I want soft decoding for BCH code is because in my system I
have a concatenation of LDPC and BCH codes, so the output of LDPC decoder
goes to BCH decoder, so I want to use the soft output of the LDPC message
passing decoder as input to BCH decoder.