Hi Ben I think I can help you out on question 1:The scaling parameter affects where the PFB's windows overlap; for a width of 1 they overlap at the 3dB point, meaning that an input signal that lies exactly between two bins will appear in both bins at -3dB (correct me if I'm wrong list!). The spacing between bins, and the total number of bins, will not change -- just the width of the window applied.
I've attached a plot contrasting three different scaling parameters for a PFB + FFT (real) running on the ROACH. Hope it sheds some light!
Cheers Danny On Thursday, 27 January 2011 at 12:23, [email protected] wrote:
Two quick PFB/FFT questions. 1) The bin width scaling parameter in PFB_FIR. Can someone explain how this works? If you increase the bin width, doesn't the output rate of the PFB need to increase? Does a value of 2 here make the bins twice as wide as a PFB with bins that are "normally" spaced? 2) We are feeding the CASPER FFT complex data with 2^N points. The output is complex data with 2^(N-1) points. Where did my negative frequencies go? Since the input is complex the negative frequencies are independent of the positive freqs, and I need them! Thanks, Ben
iADC Calibration.pdf
Description: Adobe PDF document
