On Thu, 8 Jan 1998, Juhana K Kouhia wrote:
>look at them. One major difference between the compressors in most of the
>programs and the one used at Lucasfilm (Moorer) is that Moorer's program
>divides the audio to several parts with respect to frequency and process each
>of them separately with the "compressor".
That's a multiband compressor - basically, it's a whole different kind of
beast from the rather simplistic tool usually dubbed as a 'compressor'.
The multiband devices are the ones most often used to achieve high
compression without noticeable audible effects. A couple of examples would
be the compressors used at radio stations and recording studios to
try and overcome the low dynamic range of the media used (FM radio and tape,
>them. I don't understand these Attack and Release things of compressor at
>all but would like to have an explanation on them, please.
Basically, when using a compressor, you're dealing with causal systems. That
means you do not know what kind of data is coming in until it actually does.
If you want to flatten the dynamics of such a signal, you will want to track
the envelope of the signal, and recreate it in such a way that overall
dynamic range is reduced. When you have the current amplitude of the signal
(from the envelope tracker), you need to know how much to attenuate the
signal (well, you could amplify the silent parts, but this is the usual way
to do things...). You do this by maintaining a current attenuation value and
periodically updating it. The idea is to predict the incoming (mean)
amplitude and to attenuate only strong signals. So what you do is, you
increase the amount of attenuation when your envelope tracker tells the
incoming signal is loud and reduce it when the signal is quieter. This all
works fine, except what it amounts to is amplitude modulation. And as soon
as you amplitude modulate by a quickly changing signal (read: lots of high
frequency content), you begin to create clicks, pops and various other side
effects. Also, you need to take care of the fact that when a sudden loud
passage appears, you cannot react immediately (that sounds quite strange),
but gradually, and during that time your output will overshoot/clip. So you
need separate variables to control the speed at which you increase the
attenuation when the signal suddenly grows louder and, on the other hand,
when it keeps itself to low amplitudes. These are called attack and release
times, respectively. In the analog implementation they are simply knobs to
control two filters that do the work of tracking attack and release state
operation, in the digital one, well, you guess... ;)
By the way, at the extreme, when no filtering is done on the attenuation
changes and the gain is updated per sample, the compressor reduces to a
waveshaper. So it is rather easy to see why the filtering is of use...
Sampo Syreeni <email@example.com>, Decoy/Dawn, Student (Math, Helsinki University)
PS. Harmillista, että lista on englanninkielinen. Tuollaisen tekstin
vääntäminen on yhtä tuskaa pidemmän päälle... Kuinka pitkään olet leikkinyt
SAOLin kanssa? Itse olen kys. ohjelmistosta miltei täysin pihalla (eli en
siis omista kopiota...). Lähettele postia, jos siltä tuntuu - uudet DSP- ja
tietokonemusiikkiviitteet ovat aina kiinnostavia!
This archive was generated by hypermail 2b29 : Mon Jan 28 2002 - 11:46:31 EST