Re: [Discuss-gnuradio] QAM receiving and frequency locking.

[Tom Rondeau]

Mason wrote:
> Hi list,
>    So I'm still working on trying to get QAM transmissions between two 
> USRPs using the following patch:
> http://www.cae.wisc.edu/~masond/gr-qam-patch-rev3.diff
>
> The QAM receiver works in simulation but the constellation just spins 
> and spins when sent over the air.
> It apparently is unable to cope well with offsets of frequency between 
> sender and receiver. Does anyone know either a) better values for the 
> 'magic numbers' of the receiver to make it track better, (or even a 
> source for the selection of the original values) or b) a better 
> receiver algorithm to use than the M&M?
>
> --Mason

You'll want to make another receiver. It looks like the qam_receiver is 
just the mpsk_receiver. This one calculates the frequency offset by the 
instantaneous phase offset in each symbol. For anything not BPSK or 
QPSK, it uses the constellation to calculate this by first finding which 
point the symbol most likely represents. With QAM, you have both a 
sensitivity problem doing this under low SNR. Specifically, all of the 
inner points (i.e., the ones with the lowest SNR) are going to bias the 
reciever and prevent it from getting lock. You really want to make this 
into a QPSK problem where you are only looking at the outter-most 
corners and making decisions based on the high SNR symbols.

Tom