Yes, yet again! On JC's Blowtorch thread I mentioned trying to create some interesting test files, sound and music tracks, for use in the subjective evaluating of how easy it is to hear various types of distortion or abberation artifacts, at various levels of inclusion.
This is just to say the ball's rolling -- first up I'm playing with a clearly recorded playing on triangles, plenty of high harmonics and nice long decay trails. I can distort the waveform with an abitrarily complex mathematical function, mix an attenuated copy of that with original, and see how audible the end result is ...
This is being done at the moment on a PC with ordinary, single fullrange, monitor speakers -- absolutely nothing special is being done as yet to improve the "transparency" of the playback.
What's clear straight away, in the first, rough tryouts, is the power of the masking effect -- at any point where the average intensity of the clean signal is of the order of 60dB louder than the distortion component then it's close to impossible to pick something going on ... but, at 40dB difference then even the low end nature of the reproduction chain allows me to hear the distortion.
And herein lies the first complexity: depending on precisely the algorithm used for generating distortion, determining in the time domain, looking at the waveform itself, what the maximum levels of particular distortion peaks will be at any particular moment is not obvious. So, how does one specify that the distortion is so many dB below the waveform in a complex signal? ... I've got a hunch this is going to critical ...
Interesting times ahead ...
This is just to say the ball's rolling -- first up I'm playing with a clearly recorded playing on triangles, plenty of high harmonics and nice long decay trails. I can distort the waveform with an abitrarily complex mathematical function, mix an attenuated copy of that with original, and see how audible the end result is ...
This is being done at the moment on a PC with ordinary, single fullrange, monitor speakers -- absolutely nothing special is being done as yet to improve the "transparency" of the playback.
What's clear straight away, in the first, rough tryouts, is the power of the masking effect -- at any point where the average intensity of the clean signal is of the order of 60dB louder than the distortion component then it's close to impossible to pick something going on ... but, at 40dB difference then even the low end nature of the reproduction chain allows me to hear the distortion.
And herein lies the first complexity: depending on precisely the algorithm used for generating distortion, determining in the time domain, looking at the waveform itself, what the maximum levels of particular distortion peaks will be at any particular moment is not obvious. So, how does one specify that the distortion is so many dB below the waveform in a complex signal? ... I've got a hunch this is going to critical ...
Interesting times ahead ...
Last edited:
I recall this site, if the archive has saved everything, it has several interesting "pre-distorted" soundfiles:
http://web.archive.org/web/20070303180526/http://www.pcavtech.com/
If something has only second or third harmonic distortion, the THD level of the distortion needs to be much higher to be audible than a signal with higher harmonics.
http://web.archive.org/web/20070303180526/http://www.pcavtech.com/
If something has only second or third harmonic distortion, the THD level of the distortion needs to be much higher to be audible than a signal with higher harmonics.
Thanks, benb - this is going to be an exploration, and the software I'm using allows me rectify, change quantisation behaviour, simulate tube characteristics, clipping and saturation, weird and wonderful forms of fuzz - and that's only some of the presets.
I aim to see which distortion type cuts through best, makes itself most noticeable, as at least one objective ...
I aim to see which distortion type cuts through best, makes itself most noticeable, as at least one objective ...
Just an observation of probable results for simulation of actual amplifier sins: crossover distortion (as in a no-bias push-pull BJT amp) would surely be one of the most audible and annoying, and at the other extreme, a slight curve of a vacuum tube response (so that one polarity has slightly more gain than the other) does mostly second harmonic, and at moderate levels can actually make something sound "better." I recall Goldwave and Cool Edit (later bought by Adobe and renamed Audition) were two audio file editors with such features from the late 1990s.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.