On 14 Jun 2005 00:39:42 -0700, porterboy76@yahoo.com wrote:

>Thanks Eric...
>
>> If the bit rate is high
>> enough then the delay relative to the voice channel is often small
>> enough.
>
>Oh, I hadn't thought of that. So, let me get this right...
>The decoding delay might be of the order of a few thousand bits, but if
>we have a high information rate, say 3Mb/s, then a few thousand bits is
>not such a long time, and can be less than the tolerable delay for
>voice comms... is that it? I was going to avoid iterative decoding
>altogether, but maybe I should consider it if that is the case...

That's pretty much it assuming the decoder has to decode the entire
stream, i.e., run at the aggregate rate rather than the voice rate.

Even with a low-rate system it is still possible to design the decoder
such that it runs much faster than the data rate in order to reduce
the latency.    Since iteratively decoded codes are always block codes
the latency is associated with the decode time after reception of the
last bit in the block.   If you just clock the bejeezus out of the
thing the latency will be low, but the complexity and power
consumption may be higher than you'd like to live with.


Eric Jacobsen
Minister of Algorithms, Intel Corp.
My opinions may not be Intel's opinions.
http://www.ericjacobsen.org

Thanks Eric...

> If the bit rate is high
> enough then the delay relative to the voice channel is often small
> enough.

Oh, I hadn't thought of that. So, let me get this right...
The decoding delay might be of the order of a few thousand bits, but if
we have a high information rate, say 3Mb/s, then a few thousand bits is
not such a long time, and can be less than the tolerable delay for
voice comms... is that it? I was going to avoid iterative decoding
altogether, but maybe I should consider it if that is the case...

PB

On 13 Jun 2005 06:40:34 -0700, porterboy76@yahoo.com wrote:

>
>QUESTION
>Do you know where I might find tabulated coding gains for different
>coding schemes? For example BICM, LDPC, MLC, Turbo Codes etc.

It's difficult to find a single source for such things since the gains
depend on the specific code designs, which vary widely, as well as the
channel models used to evaluate them.

>CONTEXT
>I have been working for a while on different coding schemes, so far I
>have studied block codes, convolutional codes and trellis coded
>modulation. I am interested in coding for the bandwidth limited
>channel, so coded modulation is of more use than standard binary codes,
>which typically have spectral efficiency less than 1 bit/s/Hz. The
>industry standard TCM seems to be "Pragmatic-TCM" introduced by Viterbi
>et al in 1989, however the coding gains for this scheme (which uses a
>convolutional code of constraint length K=7, together with
>set-partioning constellation mapping) are typically less than 3dB at
>BER = 10^-5. I understand 3dB is a significant gain in communications,
>but I was hoping to do better with some of the more modern coding
>schemes, mentioned above. Another question, do all of these modern
>schemes use iterative decoding?

Capacity-approaching codes are all iteratively decoded.   If you can
get by with gain less than that, then there are lots of other
alternatives, like concatenated CC-RS, etc.

> Does the delay incurred make these
>codes impractical for delay sensitive applications like voice
>communication?

It can, but it depends on the application.   If the bit rate is high
enough then the delay relative to the voice channel is often small
enough.   Alternatively, you may be able to use small block sizes, but
that often degrades the gain of the code.


Eric Jacobsen
Minister of Algorithms, Intel Corp.
My opinions may not be Intel's opinions.
http://www.ericjacobsen.org

QUESTION
Do you know where I might find tabulated coding gains for different
coding schemes? For example BICM, LDPC, MLC, Turbo Codes etc.

CONTEXT
I have been working for a while on different coding schemes, so far I
have studied block codes, convolutional codes and trellis coded
modulation. I am interested in coding for the bandwidth limited
channel, so coded modulation is of more use than standard binary codes,
which typically have spectral efficiency less than 1 bit/s/Hz. The
industry standard TCM seems to be "Pragmatic-TCM" introduced by Viterbi
et al in 1989, however the coding gains for this scheme (which uses a
convolutional code of constraint length K=7, together with
set-partioning constellation mapping) are typically less than 3dB at
BER = 10^-5. I understand 3dB is a significant gain in communications,
but I was hoping to do better with some of the more modern coding
schemes, mentioned above. Another question, do all of these modern
schemes use iterative decoding? Does the delay incurred make these
codes impractical for delay sensitive applications like voice
communication?