FPGA Central

Hi all ,

I am trying to implement BER calculation while actual data
communication is in progress. Currently I have a BER mode for the
system where the system generates pseudo random numbers from the
transmitter and is recalculated on the receiver. What is the usual
pattern for doing BER calculation while data communication is on
going?

BR
Rate

On Nov 2, 9:17*am, ratemonotonic <niladri1...@gmail.com> wrote:
> Hi all ,
>
> I am trying to implement BER calculation while actual data
> communication is in progress. Currently I have a BER mode for the
> system where the system generates pseudo random numbers from the
> transmitter and is recalculated on the receiver. What is the usual
> pattern for doing BER calculation while data communication is on
> going?

A couple things come to mind:

* Compute BER on a portion of the data which is predictable and ignore
the rest. If the errors are uniformly distributed then the long term
average BER on the subset should be the same as for the entire data
set. Some packet data formats (like DVB transport streams) contain
known bytes, or if you're defining your own system you can insert
known data.

* Some error correction systems provide other metrics on errors. For
example:
- A Reed-Solomon decoder operating within it's error correction
capability can tell you how many errors it has corrected in a block.
- Some Viterbi decoders provide information on internal states which
can be used to infer rough error rates. Usually this is not very
accurate, but it can be useful for gross estimation of system
performance.

Eric

On Sun, 2 Nov 2008 08:17:57 -0800 (PST), ratemonotonic
<[email protected]> wrote:

>Hi all ,
>
>I am trying to implement BER calculation while actual data
>communication is in progress. Currently I have a BER mode for the
>system where the system generates pseudo random numbers from the
>transmitter and is recalculated on the receiver. What is the usual
>pattern for doing BER calculation while data communication is on
>going?
>
>BR
>Rate

In addition to what the other Eric said, there are some other commonly
used methods. Some are system dependent in that the data
format/framing and/or the FEC used can make a difference on which ways
make the most sense:

1. Re-encode and compare. With convolutional code systems the raw
BER can be estimated by re-encoding the corrected data and comparing
it to the received data. This gives a good estimate of the raw
(i.e., pre-FEC decoder) error rate, and if the channel is reasonably
flat at the slicer and the noise reasonably white, then a map from raw
BER to corrected BER can be pretty reliable within reason. Many
systems estimate Eb/No this way, too, since mapping raw BER to Eb/No
is pretty reliable in a flat channel.

2. Compute block or packet error rate using CRCs or other means in
the data formatting. As mentioned, RS decoders can provide
reasonably reliable codeword error metrics, and sometimes
decompressors (if they're in the stream) can tell when the input is
wonky as well. If the system can map CRC or RS codeword error rate
or some other block error rate reliably to BER, then that's a
possibility.

Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.ericjacobsen.org

Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

On 2 Nov, 20:17, Eric Jacobsen <eric.jacob...@ieee.org> wrote:
> On Sun, 2 Nov 2008 08:17:57 -0800 (PST), ratemonotonic
>
> <niladri1...@gmail.com> wrote:
> >Hi all ,
>
> >I am trying to implement BER calculation while actual data
> >communication is in progress. Currently I have a BER mode for the
> >system where the system generates pseudo random numbers from the
> >transmitter and is recalculated on the receiver. What is the usual
> >pattern for doing BER calculation while data communication is on
> >going?
>
> >BR
> >Rate
>
> In addition to what the other Eric said, there are some other commonly
> used methods. Some are system dependent in that the data
> format/framing and/or the FEC used can make a difference on which ways
> make the most sense:
>
> 1. Re-encode and compare. With convolutional code systems the raw
> BER can be estimated by re-encoding the corrected data and comparing
> it to the received data. This gives a good estimate of the raw
> (i.e., pre-FEC decoder) error rate, and if the channel is reasonably
> flat at the slicer and the noise reasonably white, then a map from raw
> BER to corrected BER can be pretty reliable within reason. Many
> systems estimate Eb/No this way, too, since mapping raw BER to Eb/No
> is pretty reliable in a flat channel.
>
> 2. Compute block or packet error rate using CRCs or other means in
> the data formatting. As mentioned, RS decoders can provide
> reasonably reliable codeword error metrics, and sometimes
> decompressors (if they're in the stream) can tell when the input is
> wonky as well. If the system can map CRC or RS codeword error rate
> or some other block error rate reliably to BER, then that's a
> possibility.
>
> Eric Jacobsen
> Minister of Algorithms
> Abineau Communicationshttp://www.ericjacobsen.org
>
> Blog:http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php

Thanks for the input. I needed this pointing in the right direction.
Our radio has not got FEC implemented , but this is to be implemented
very soon. The RS decoder method will fit into our system. CRC is
implemented and is used for packet error rate indication.