[Pafgbt] save PAF cross-correlations rather than formed beam outputs

[Brian Jeffs]

Thanks D.J., that gives us some perspective.  

My note could have been a bit misleading since in the CASPER F engine,  for a 2048 point FFT we get 1024 frequency channels (real input data, so we throw away the symmetric mirror image frequency bins).  I was quoting N as FFT size, so to get the number of channels you mention requires an FFT that is double the size, e.g. N=8192 for the 20 MHz BW Galactic HI case would give 4096 channels.  This is more than we can do with our existing system.  Perhaps with more ROACHs we could get this big an FFT working on fewer inputs, but I have not seen it done.

Brian

On May 21, 2013, at 2:25 PM, D.J. Pisano wrote:

> Hey all,
> 
> Just a quick note on the frequency resolution in response to Brian's
> email.  For extragalactic HI applications, 2048 channels across 20 MHz
> will be about right.  For Galactic HI, either a 10 MHz band (with 2048
> channels) or a 20 MHz band with 4096 would be best.
> 
> --D.J.
> 
> ------------------------------------------------------------------------------------
> D.J. Pisano, Assistant Professor
> WVU Department of Physics      phone:  304-293-4886
> P.O. Box 6315                               fax:    304-293-5732
> Morgantown, WV 26506 USA    djpisano at mail.wvu.edu
> ------------------------------------------------------------------------------------