[pulseaudio-discuss] Updates to filtering work. Review requested.

[Matthew Robbetts]

On 20 Apr 2013, at 00:27, Justin Chudgar <justin at justinzane.com> wrote:

> https://github.com/justinzane/pulseaudio
> 
> Gentlemen, Ladies and Bots: 
> 
> Thanks again for all your help getting me to this point. Based on various 
> input, I have done a major refactor of what was module-lfe-lp. The actual 
> filtering code now lives in pulsecore/biquad-filter.c/h so that it may be used 
> by other modules or by pulse itself. Module-lfe-lp has been renamed module-
> lfe-lowpass-sink.c and now relies on biquad-filter to do the heavy lifting. 
> 
> I have tried to cleanup most of the fprint and comment cruft and have moved 
> all doc comments to biquad-filter.h. You will note that I wrote the biquad 
code 
> with the assumption that it will get extended to handle other filter types, 
> like notch and bandpass, as well as 2nd, 4th, 6th and 8th order filtering. 
This 
> is not used by my module, but it seemed logical to enable easy extension of 
> functionality as needed.
> 
> I would appreciate your continued review and guidance. For your convenience, 
I 
> attached a graph of the power spectral density comparison for a standard 
> surround-51 setup with lpfreq=200.0Hz. To my uneducated eye, it looks like 
> things are working.
> 
> Justin<test_200.png>_______________________________________________
> pulseaudio-discuss mailing list
> pulseaudio-discuss at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/pulseaudio-discuss

Hi Justin,

Can you comment on exactly what you measured to make that graph? It looks 
awfully clean to be an acoustic measurement, but is an odd shape for a signal-
level measurement.
_______________________________________________
pulseaudio-discuss mailing list
pulseaudio-discuss at lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/pulseaudio-discuss


The source audio and python test code are in pulseaudio/pytests in the repo. 
Basically, the source is a simple chirp-like sound generated in audacity and 
saved as a stereo wav. The  test just plays the stereo source through null-
sink with lfe-lowpass-sink filtering at a cutoff of 100Hz. As far as the PSD 
graph goes, check out "Spectrum" on pypi. It is a spectrum.pdanniell PSD.