This sound file contains a mono and stereo version of a single compressive EQ filter. What's that?
A compressive EQ filter applies a dynamically controlled amount of EQ cut or boost in a user specified bandwidth. The filter is a bell shaped filter centered at Fc and its width is determined by the filter Q parameter.
The compression takes place in 3 stages, described by two thresholds:
ThreshHi? and
ThreshLow?. The boost or cut applied in the bandpass of the EQ filter is described by the
GainDB? parameter. This gain is the amount of gain applied at the lower threshold. For signal levels from 0 dBFS down to the upper threshold, the system is unity gain, and applies zero equalization. Between the upper and lower thresholds, the amount of boost or cut is dynamically controlled, as with compression, so that at the lower threshold, the user specified amount of boost or cut is applied. Below the lower threshold, the system has unit slope again, but with that maximum boost or cut.
How is this useful? I'm still working on that one... BSS sells a hardware unit that performs this job with 4 channels. [
http://www.bss.co.uk/products/dynamics/dpr-901ii/index.html ]. I have provided one channel here, but you can string them together in series to perform multiple jobs. I haven't provided the hi or low shelving version, but using the shelving filters in Kyma or those provided by me from other postings, you can snip and paste one of these together if you need it.
There are two toggle buttons: Side allows you to listen to the sidechain as it is being boosted or cut. Cmpr enables the compressive EQ with Side is deselected. Deselecting Cmpr and Side serves as a system bypass.
My impression as I start to use this, is that it is a bit more flexible than the Dolby style spectral processor. The Dolby system employs 3 bands across the spectrum, and applies its magic in each of those adjoining bands. This compressive EQ allows a more surgical approach to brightening or dimming certain frequencies in the sound. The example given by BSS is one of enhancing the sibilance of a singer as she fades during quiet portions of a song. That gives added life to the song, and keeps the listener alert to her singing. I'm sure there are many other uses for this thing...
So how does this work? I start by peeling off the EQ filter portion of the sound. This filtered component with unity gain is sent into a dual slope compressor (actually 3 slopes, 2 thresholds...) and also into a compensating delay line. At the output of the compressor the difference is formed between the compressed filtered signal and the uncompressed filtered signal. That uncompressed filtered signal is the portion that is present in that filter passband from the original signal. Once that difference is formed, it is simply added back into the delay compensated original signal. This subtracts out the portion of the original signal in that passband, and adds back in the compressed or expanded portion. Voila!
[ BTW... the stereo version is not simply two channels pasted together... that wouldn't compress a stereo signal properly. Instead, the stereo version uses some
StereoCompressor? Class objects that I constructed to take the max of the absolute value between the left and right channels, and uses that max value as the sidechain signal to the compressors. That way if one channel gets very loud, it pulls the other along with it to maintain the proper stereo balance. ]
--
DavidMcClain - 17 Dec 2003
Sorry but I really don't get this one! Here is my (bad) understanding:
-- If VCS is set as:
--TreshHi -20db
-- TreshLow -40db
-- CenterFreq 2500hz
-- GainDB +12
,when the input is (after filtering) -60db up to -40db, it should be boosted by 12db at 2500hz;
when input is in the -40 to -20db range it should be boosted by 12db (from -40db)
to 0db (when input signal reaches -20db); finally when input goes beyond -20db
no EQ should apply. What did I miss?
-- User.KarlMousseau - 07 Jan 2004
Hi Karl,
It sounds to me like you have a proper understanding. I don't have it running in front of me right now. But the compression is a 3-segment compressor. Full gain below the lower threshold, diminishing gain between the thresholds, and unity gain above the upper threshold.
So, why do you think you missed anything? This is modeled after the BSS DPR-944 Parametric EQ, the BSS DPR-901, and combined with some features of the Dolby Spectral Processor. The 3 segment compression is quite useful after all. It permits you to emphasize mid-level loudness sounds without disrupting the gain of the louder portions. Having it run through a bandpass filter first allows you to emphasize things like sibilance that has faded as the singer backs away from the mic.
The BSS DPR-944 actually manages to get some kind of variable bandwidth filter in their analog circuitry. The bandpass offered here is a simple 2-pole BPF which corresponds to the BSS most surgical setting.
Cheers,
-- User.DavidMcClain - 22 Jan 2004
