John Ackermann N8UR wrote: > Thanks, Chuck -- sounds like useful new features. I will download and play. > > I meant to mention earlier in reference to the filter and the tradeoff > between skirts and CPU usage -- have you looked at the filtering scheme > used in the Flex-Radio SDR-1000 software? I don't recall the exact > details right now, but it works something along the lines of doing an > FFT to convert the baseband signal back into the frequency domain, and > implementing the filter as a mask against the FFT bins, and then an > inverse FFT back to the time domain. (I'm an idiot about this stuff, so > probably got some component of that wrong). The end result is almost > "brick wall" filter skirts without using up too much CPU.
We do that already. Instead of using gr.fir_filter_ccc use gr.fft_filter_ccc. We should have the fff version in there soon, too. Probably when Eric gets back. Even without that, you should be able to do multi-thousand tap filters on a reasonable machine at the rates you are talking about here. On the other hand, if you are using a filter with more than a few hundred taps, you're doing something wrong, IMHO. In the case of a decimating filter, you can split the decimation into multiple stages, thus greatly easing the requirements. Matt _______________________________________________ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org http://lists.gnu.org/mailman/listinfo/discuss-gnuradio
