leban (giancarlo) wrote on Wed, 14 December 2005 09:18 |
zetterstroem wrote on Wed, 30 November 2005 19:31 | "The sound of a compressor is a simple gain envelope and amplifier distortion."
|
This is terribly true. A lot of plugins simply use an envelope follower (and this is easy to implement, just a rms measure using a lowpass filter for a compressor, or a peak measure for a limiter) BUT there is a change in amplitude in real-time made using this value. Every sample has a different gain -> this could be read as distortion.
But a la2 is not so linear.... it distorts too... otherwise it would be easily modelled using convolution... My question Do we need the la2 'colour' or 'compression' itself?
|
i think it's all percieved as distortion, because the ear is not linear. i think one cannot have compression without distortion since we are trying to induce the psychoacoustic effect that something is louder than it is, or has more dynamics than it does.
i think we need the color and the compression. but an analog chain has more color already, so the proportions to add color vs. amplitude envelope control would be different for digital (assuming digital artifacts exist only below the threshold of perception, which i think is entirely possible and a reasonable assumption).
and the technique would also be different. not necessarily according to the analog tradition, but emulating it, like a magician would, like artists do. i dream and experiment and learn from experience and have lots of failures. the rules of analog seem to me to lead to a good sounding record. the rules of digital are only solid guideposts to avoiding disaster. so i think to be successful in digital, you must think completely outside the box. the irony of it all.
[not to mention that a lump of ferrite just intuitively seems more "natural" than a string of 1s and 0s, and of course sounds natural right out of the box... even a cheap lump of ferrite could perhaps work if we are only interested in making distortion. by comparison, i think cheap dsp always sucks.. and you can't usually identify cheap dsp by physical attributes.. so imo, not easy to do digital. i can't always predict the outcome as easily with digital, whereas i think it is easier to conjure an analog chain and predict the results without ever hearing it]
convolution? i think with any dsp process there is a going to be a tradeoff between accuracy in the time domain vs. the frequency domain. this is the biggest hurdle for any digital processor, imo. but I am sure electrical engineers have their own set of constraints, and 100 years of time for the most clever amongst them to work these out.
second order harmonics. i don't think it's rocket science to make these.. but you know.. there is magic in electrons running through tubes, like frosty the snowman's hatand mysteriously filled stockings on christmas morning [imho!]. i've worked with digital since 1985, and the first thing i learned was how to make it sound "warmer"; by now i think am pretty good at it.
jeff dinces