FM Demod Questions

Hi Folks,

I have a telecom application coming up and would appreciate advice on the
best choice of DSP for the job.

The DSP would be required to perform bandpass filtering and demodulation of
an FM carrier containing voice modulation.  The BPF must have good adjacent
channel rejection at +/- 20 kHz from wanted channel center frequency, and
good in-band group delay.

It also must detect carrier ON and OFF times to a resolution of better than
100 nsec. in the presence of noise down to about C/N = 10 dB.

I'd prefer to place the DSP at the 140 MHz receiver input.  Otherwise, it
would go after the first mixer at around 20 MHz.  Choice depends partly upon
how much more expensive the higher speed DSPs would be.

Suggestions on choice of DSP for this application and guesstimate of 140 MHz
vs. 20 MHz cost would be appreciated.

George

George W wrote:
> Hi Folks,
> 
> I have a telecom application coming up and would appreciate advice on the
> best choice of DSP for the job.
> 
> The DSP would be required to perform bandpass filtering and demodulation of
> an FM carrier containing voice modulation.  The BPF must have good adjacent
> channel rejection at +/- 20 kHz from wanted channel center frequency, and
> good in-band group delay.
> 
> It also must detect carrier ON and OFF times to a resolution of better than
> 100 nsec. in the presence of noise down to about C/N = 10 dB.
> 
> I'd prefer to place the DSP at the 140 MHz receiver input.  Otherwise, it
> would go after the first mixer at around 20 MHz.  Choice depends partly upon
> how much more expensive the higher speed DSPs would be.
> 
> Suggestions on choice of DSP for this application and guesstimate of 140 MHz
> vs. 20 MHz cost would be appreciated.

The following is based on not much experience, and some purely 
theoretical knowledge:

Certainly having to deal with a 40 MHz sample rate rather than a 300 MHz 
sample rate gives you a cost advantage.  Probably one of the newer SHARC 
incarnations would do.

Further, I'd say doing as much pre-selection in analog is to your 
advantage as far as saturation / intermod products / image-signal 
generation goes.  If you can make a voltage-controlled filter to 
preselect the mixer input, so much the better.  Then after the mixer, I 
recommend at least 1 analog IF stage to tighten down the bandwidth somewhat.

Once you establish what your IF bandwidth is, probably in the 100-200 
KHz range after 1 stage, you might be able to get away with 
undersampling at that rate and using the aliasing to your advantage, 
though the precision of the sampling frequency must be as tight as if it 
were at 40 MHz.  Jitter in the sample rate is your enemy in this approach.

Other folks around here have even published papers on DSP FM demod, so 
you might look in Google Groups or similar search engines for where they 
mention those.

Good luck!