Gene Heskett wrote: > On Thursday 11 October 2007, Jonatan Liljedahl wrote: >> Jonatan Liljedahl wrote: >>> I'm having problems with distorted sound with ALSA. It's not >>> clipping distortion but more like a dirty spectra. The test I am >>> doing is playing a pure 440Hz sawtooth wave, and the sound is >>> very unclean, there's inharmonic partitials that shouldn't be >>> there, actually below the fundamental frequency I think. Compared >>> to an analog sawtooth oscillator it sounds really awful. >> Here's a picture of the resulting waveform when playing a pure >> sawtooth and recording it straight in again on my Terratec Phase >> 26. As you can see, there is some kind of ringing happening... >> >> But, this is interesting, when recording a pure sawtooth from my >> analogue modular into my soundcard (through a mixing desk) there >> was some ringing too, but it didn't sound at all as dirty as the >> previous computer-generated sawtooth. When playing back this >> recorded analogue sawtooth, I had a hard time hearing the >> difference. >> >> So, I guess it's not the ringing this is about. And now when I >> listen again with different frequencies I definitely hear that the >> polution of the spectra is subtones, frequencies below the >> fundamental. Perhaps the problem is some cyclic jitter? But it >> seems dependent on the high overtones of the computer-generated >> waveforms since I can't hear it when recording a sawtooth from the >> outside... >> >> Any ideas what this is and what I can do about it? > > The fall time of that computer generated sawtooth is probably 100x > the fundamental frequency of the sawtooth. I suspect what your ears > are hearing is aliasing because some portions of that exceed the > sampling frequency, which would be non-harmonically related, and > translated to a tone that is the difference between the sampling > frequency and the upper harmonics of that 400hz tone. This is why > good digitizers will have a truly brick wall analog filter, cutting > off by 60+ db, anything that approaches the 'nyquist frequency' of > the sampler itself. And it's both expensive in terms of component > count and cost, board real estate and design time to do that without > audible group delay effects, which are an entirely different horse, > and the filtering _must_ be done before digitization. Aliasing, once > in the digital data stream, cannot be removed.
This was indeed it. Though I didn't know that a soundfile must not contain any sharp discontinuities, I thought that the hardware would filter it out. But you say this is not possible once it's digital, how come? Why can't you just run a soundfile through a filter to take out everything above nyquist? (upsampling it first, I guess) > Recalculate the sawtooth so that the fall time stays within the > systems bandwidth, and I'd bet a bottle of suds most of the effect > you are hearing will go away. The fall time should not be less than > 1/frequency, answer is in seconds. Yes, I found code for bandlimited sawtooth oscillator in the Synthesis ToolKit which worked very well! -- /Jonatan [ http://kymatica.com ] ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/ _______________________________________________ Alsa-user mailing list Alsa-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/alsa-user
