Re: [Discuss-gnuradio] OFDM Updates

[Tom Rondeau]

Dan Halperin wrote:
> Shravan Rayanchu wrote:
>   
> > Basically, I seem to completely lose some of the packets in the air.
> > Of the packets I receive, almost all the packets are received
> > correctly. Initially, the error rate was too high (The packets were
> > getting lost and also among the packets received, lot of them were in
> > errors), so I increased tx-amplitude to ~3000.
> >
> > Am I using the right version of the code ? Is the tarball release
> > better to use ? Can you please let me know if there are any parameters
> > which I need to change ?
> >     
>
> Tom mentioned an existing problem in the email you replied to, which you
> didn't address in your response. Could that be the problem or have you
> ruled it out?
>   
Yes, thanks for pointing that out, Dan. The problem I discussed in my 
original email is certainly the problem; it's exactly what I was seeing.

> For debugging these types of errors, I really do suggest (from
> experience!) that you start saving the outputs of the intermediate
> stages to disk and seeing what they look like. It might require some
> understanding of the receiver, but then again you probably want that
> knowledge anyway...
>   
Excellent point. Visualization tools are a key to understanding and 
debugging this stuff.

By using the --log option, the receiver will dump output data files for 
every important (and even some not-so-important) block in the chain. The 
gr_plot_XXX.py scripts are useful for getting a quick look at the 
output. There is a local gr_plot_ofdm.py script distributed as part of 
the ofdm example directory that provides a specific way of looking at 
the output of the OFDM system.

> It's likely (as I found with older versions of the DBPSK code, for
> instance) that some of the synchronization and/or timing algorithms
> aren't working in your setup. But maybe there's lots of cochannel
> interference. Maybe the RSSI is low. Maybe the frequency offset of your
> daughterboards is too large to be handled by the PLLs...
>   
Always the case, really. These methods are as straight-forward as we can 
make them to do the basic receivers for different modulations. Most 
standard/commercial digital radios do a whole lot more to make sure the 
signal is properly received. Cochannel interference will quickly kill 
these implementations. And in the case of the M-PSK code, my first 
version was textbook while the second version was the right way; that 
helps, too :)

Tom