hi danny, (and laura)
if i understand your email correctly,
you wants to process four real streams,
where each real stream is 400 Msps.
so you need four real to complex transforms,
where each of the four FFT's has two parallel
real inputs at 200 MHz,
and one complex output at 200 MHz.
if we have a block for that, great (i'm not sure if we do).
if we don't have a block for that:
you could use a pair of the the original casper biplex FFT,
each biplex processes two complex FFT's at once, both at
200 Mcomplex samples per second. and then you'd have
to apply one of the two standard tricks to turn complex FFT's
into real FFT's - you can do two reals in one complex,
or you can do a real of length 2N in a complex of length N.
i think best is do two reals in one complex, and
you'll also need a corner turner in front of the pair of biplex
to arrange your data to turn the four streams at 400 Msps
into eight real streams at 200 Msps.
dan
On 03/18/2010 11:19 AM, Laura Spitler wrote:
Hi Danny,
I think what you want is the regular 'fft' with the "number of
simultaneous inputs" set to 0. This gives you two independent streams
labels as pol0 and pol1.
Cheers,
Laura
On Thu, Mar 18, 2010 at 11:12 AM, Danny Price
<[email protected]> wrote:
G'day all
Does anyone have experience with using a 2^1 simultaneous input FFT? The
fft_wideband_real block has a minimum of 4 inputs; has anyone ever modified
this or built a 2 input version?
The reason I ask is that we would like to make a 0-200MHz spectrometer; the
initial plan was clocking the ROACH at 200MHz => ADC at 800MHz, then
dropping two of the four ADC samples. We're after high spectral resolution
and would like to pack four spectrometers onto one board. Any advice?
Thanks
Danny
--
Dan Werthimer
Space Sciences Lab and Berkeley Wireless Research Center
University of Calfornia, Berkeley
