Audacity and JACK : I need some terms defined.

[Brendo613]

I have, on Ubuntu 8.10, Audacity recording through JACK. At 44100 khz, 16 bit, I recorded for 1 1/2 hours last night via a USB mixer sending a stereo mix, and amazingly only had two xruns (I believe these are skips in the audio). I realize, however, that while I know a bit about my digital recording setup, I have some areas in which I am completely clueless. To that point, I need to know what these things are:

*In Audacity, the sample rate converter - real time, high quality, dither ... I don't know what any of this is
*In Audacity, the entire spectrogram category ... FFT, what are these things?
*In JACK, I basically just configure different numbers at random. I don't know any of these settings' meanings: frames per period, periods per buffer, port maximum.

Obviously, some intelligent minds went to work writing these programs, and I'd quite benefit knowing how to tweak the settings offered in these applications

[steve]

I have, on Ubuntu 8.10, Audacity recording through JACK.

How? I've never got it to work - I just get a constant stream of xruns as soon as Audacity opens. (perhaps my computer is just too old).

*In Audacity, the sample rate converter - real time, high quality, dither

The "real time" settings are when you are listening to Audacity playing.
When you have audio that does not have the same sample rate as the "project rate" (the overall sample rate for the project), Audacity must resample the audio to make it the same sample rate as the project (so that all tracks play at the same speed). The most important thing here is that playback is smooth and not interrupted, so a fast method of resampling is used.

When you mix tracks or export the final mixdown, sound quality is more important than speed, so a slower, higher quality method is used.

"Dither" is a method of dealing with "rounding errors" - it is a bit slower than simple truncating, but produces better sound quality. By default, some method of dither is used for the high quality resampling, but not for the real-time setting. The different dither settings are different algorithms for exactly how the dither is accomplished. There are advantages / disadvantages to each method - "shaped" tends to give good overall results.

Most of these terms can be found on wikipedia if you want more specific detail.

There are a number of guides to achieving the optimum settings with Jack. Lowering latency can produce more xruns, so the general procedure is to get latency as low as possible without xruns. However, as Audacity is not a real time processing program, latency is not really that important, so going for "safer" (higher latency) settings than you would use for other programs is probably the most desirable.

In the "Spectrum" category, higher FFT will give finer resolution in the display of frequencies when displaying tracks in the spectrum view, but drawing the tracks will take longer.

[Brendo613]

steve, thanks much for that reply. I remember trying realtime and low latency kernels before, but upon retrying a realtime kernel, I'm not able to boot into it ... but I'm not going to b!tch about that on this forum.

As far as helping you out (it's only fair after your informative post), I had to tweak settings a bit to get anything usable. I have a latency rating now of 232ms, with a 256 buffer size and 20 periods/buffer, if memory serves correct (I'm away from the home computer presently). I plan on starting another thread, to address the details of how I've both been configuring JACK, and on how to better configure it.

Experiment with those settings, or post back what you have, and I'll do my best to help you.

[steve]

or post back what you have,

I think the defining issue about what I have, is a 500MHz Pentium III.

When I eventually get round to installing Linux on my Pentium IV I'll give it another go. I did manage to get it working (sort of) with Audacity 1.3.5 (or was it 1.3.4?) but no joy at all since then. (That was using the real time kernel).