[hpsdr] spectrum increased frequency resolution?

[Tayloe Dan-P26412]

>If I have a block of samples (say N = 4096, fs = 122.88 Msps), 
>a straightforward FFT produces data with frequency resolution of 30kHz.

>Is there a technique that will allow me to obtain information on a
finer 
>frequency grid (say 1kHz or 300Hz or ...)--other than "simply"
acquiring 
>additional data, extending the length of the time record?

It seems to me that you could take the time series of the output of one
of the 30 KHz FFT results and do yet another FFT on one of those
selected 4096 streams, which would be a two stage FFT.  

A 98.3 Mbps sample rate and front end, high speed N=512 FFT could break
down the frequency spectrum into 192 KHz outputs streams, one of which
could be selected as the input for a final, much lower speed N=4096 FFT.

The problem with this is that is that if you wanted to tune
continuously, it would be hard to handle tuning up against the 192 KHz
band edges.  An alternative to solve this problem might be to use a
N=1024 FFT, get 96 KHz streams, then use a two point inverse FFT to
create a single "merged" 192 Kbps data stream from two selected 96 KHz
streams.  This would allow a receiver to pick the two 96 Kbps streams
that place the receiver center closest to the 192 Kbps bandwidth and
solves the band edge problem.  Heck, you could even go down to 48 Kbps
streams and pick the four 48 Kbps streams that would place the
recombined 192 Kbps window closest to the desired center frequency.  

A 2x2 or 4x4 inverse FFT to recombine streams ought to be trivial.
However, that would get rid of the need to create an asic digital mixer
with all its inherent DDS like LO spur problems and mixer loss and would
also fix the band edge tuning problem.

It seems doable. Has anyone done anything like that before?

- Dan, N7VE


 1244586657.0