Re: [Discuss-gnuradio] FFT Baselines curved?

[Lamar Owen]

On Tuesday 19 April 2005 10:54, n4hy wrote:
> Cascaded Integrator Comb filters are pretty simple filters that are very
> fast to compute
> but produce those slow roll offs. There are more complex things that
> could be done and
> hopefully now that the USRP is really flowing out the door,
> daughtercards and all, and
> as we come up to speed, we can work on these subtleties in the core code.

Yeah, we're trying to apply these as spectrometers at PARI, and getting a flat 
baseline is pretty much required.  I personally am working on spec-an stuff 
for AM broadcast (to do 47CFR73.44 occupied bandwidth measurements, where 12 
bit resolution is fine).

For the spectrometer project, we don't actually want an FFT.  We want a swept 
filter analyzer instead, for integration purposes.  FFT is acceptable, but 
most radio astronomy spectrometers are swept filter.  Getting swept filter in 
software isn't too hard, either, but pre-filtering with the FPGA would be 
nice, then set up the software filters as fixed-frequency.

The skirts next to signals are quite curved, too, and make the box not as 
useful as it could be in my application.  Straight A/D at a low enough sample 
rate to get over the USB with no filtering at all I could deal with, but I've 
not dug into the code deep enough as yet to see how that works.

I can choose basically any IF I want, and 10.7MHz is fairly standard in the RA 
world, which falls in the passband of the USRP nicely.

My personal USRP I ordered with a TVRX (plus a BasicTX/RX pair), and I must 
say the WFM demod is nice for UHF NTSC TV audio.  I see the BTSC stereo 
subcarriers properly, and can even see some hsync leakage in the audio.  

I may also write an NTSC block;  I do enough video work that having a 
vectorscope and a waveform monitor is required.  Currently I drag around an 
SGI O2 with the analog video card; the SGI diagnostic software for that card 
has vectorscope and waveform monitor capabilities.

Does anyone (Matt?) have a good characterization of the frequency response 
curves of the BasicRX?  If not, I'll take our calibrated noise sources and 
generators here and produce one; but it will be offset by the curvature 
already noted.
-- 
Lamar Owen
Director of Information Technology
Pisgah Astronomical Research Institute
1 PARI Drive
Rosman, NC  28772
(828)862-5554
www.pari.edu