Forums

Zero flushing in viterbi decoder

Started by san_jack July 17, 2009
Hi all,

In zero termination method of convolution encoder, we pad zeros in the end
of each packet.

Should we remove the padded zeros in the viterbi decoder. 
>Hi all, > >In zero termination method of convolution encoder, we pad zeros in the
end
>of each packet. > >Should we remove the padded zeros in the viterbi decoder.
The answer to that seems obvious, so maybe you mean something deeper I am missing. The obvious answer is you keep the entire received stream until it has served its purpose in the decode process. That is, it has allowed you to decode the last of the actual data bits with the proper level of error protection, and then you dump them as you know a priori they contain to actual data. Steve
>>Hi all, >> >>In zero termination method of convolution encoder, we pad zeros in the >end >>of each packet. >> >>Should we remove the padded zeros in the viterbi decoder. > >The answer to that seems obvious, so maybe you mean something deeper I
am
>missing. > >The obvious answer is you keep the entire received stream until it has >served its purpose in the decode process. That is, it has allowed you to >decode the last of the actual data bits with the proper level of error >protection, and then you dump them as you know a priori they contain no >actual data.
That was poorly worded. The temporal smearing which the trellis produces does of course mean those bits do contain a percentage of the information about the data you need to decode. If they didn't, there would certainly be no point in keeping them. I should have said they contain no *further* actual data. Steve
steveu <steveu@coppice.org> wrote:

> (Attribution lost) wrote,
>> In zero termination method of convolution encoder, we pad >> zeros in the end of each packet.
>>Should we remove the padded zeros in the viterbi decoder.
> The answer to that seems obvious, so maybe you mean something > deeper I am missing.
It almost sounds like a protocol question, e.g. does the next layer of processing past the decoder expect to see the tail bits or not. The answer is usually no.
> The obvious answer is you keep the entire received stream > until it has served its purpose in the decode process. That is, > it has allowed you to decode the last of the actual data bits > with the proper level of error protection, and then you dump > them as you know a priori they contain to actual data.
S.