Typos in wiki

[HenkPoley]

https://github.com/bmc0/dsp/wiki/System-Wide-DSP-Guide

"defualt.pa" -> "default.pa"

below that remove "pacmd ", as it stops PulseAudio from loading.

[bmc0]

Thanks! Fixed in fc6c959. Not sure how I missed the pacmd thing...

[HenkPoley]

FYI. I've made a little writeup about wiring it all together with speaker calibration measurements: http://mythtv-multikabel.blogspot.nl/2015/11/room-correction-using-drc-fir-and.html

The part about your code is after the heading "Applying the correction". If you have time, let me know if there's anything wrong. For example the stuff with gain and clipping seems odd to me. But maybe it's a side effect of how DRC builds the speaker calibration. I also need to limit it to 2048 taps or so, or else the sound breaks down a lot when using LADSPA in PulseAudio.

[bmc0]

A few comments:

• If you're using the fir effect, you also need libsndfile and/or ffmpeg in order read the impulse files.
• dsp will fail to play an mp3 if you don't have libmad and/or ffmpeg support compiled in.
• There is currently no global tilde expansion done within the ladspa_dsp configuration file(s), so you need to give a full path to the fir effect. I'll probably add tilde expansion at some point.
• The sample rate for the impulse needs to match the sample rate of the audio stream you're running through fir. It does not currently resample the impulses. I'll probably add impulse resampling at some point since it would be quite useful.
• You may want to play around with the minimum partition length for the fir effect. It defaults to 2048, but you can try larger or smaller values: fir 512 <impluse_file> or fir 4096 <impulse_file>, for example. I've run a 720000 tap stereo filter (a reverb impulse) before without any problems.

Regarding the clipping, look through the information regarding normalization in the drc manual. There are multiple normalization options.

[HenkPoley]

I've added libsndfile-dev and libmad0-dev here and there in the document. btw, when I installed ffmpeg, I got a build failure on ubuntu 14.04.3, I'll submit a separate bug report for that.

I've changed the fir lines to include the full path.

I've adapted the part that talks about the bitrate for dsp. But PulseAudio by default uses 44.1kHz, and if you change that (at least you used to) you run into audio problems anyways.

When I played with the partition length on the ladspa_dsp plugin in PulseAudio, it didn't exactly go well. Might have been because I didn't add the channels=2 part of the invocation.

[bmc0]

I tried running a long filter with pulseaudio and got some cracking. Increasing the default-fragment-size-msec setting seemed to fix it... Kinda weird because I have zero issues when I load the plugin with ALSA (which is what I always use), even when running filters with over 500000 taps.

The configure script doesn't check library versions, and the ffmpeg code probably doesn't work with old versions of libav (which Ubuntu probably uses instead of actual ffmpeg, since it's Debian based).

[bmc0]

Another thing that may help is setting the maximum partition size as well for the fir effect: fir 1024 16384 <impulse_file>.

You could also try using Jconvolver. It's multi-threaded and likely higher performance. My convolution code is pretty simple and unoptimized.

[HenkPoley]

I've had trouble going the BruteFIT / Jconvolver road before. I believe you still need to have one of those running and then make everything else use that as audio device, isn't it? Wasn't really stable for me. Jack makes it even more complicated since it basically requires another (usually GUI) program to set up the wiring, instead of having some config file. I just want to turn on my MythTV system and have it work mostly as an appliance. Then have MythTV start Chrome for example, and audio should still work, and not have something hog the (virtual) device.

Might try your LADSPA plugin inside ALSA next.