k y m a • t w e a k y — the kyma collective || view the current website || February 2015 Archive

/ WebHome / Sounds / Share.DiscussStereoFrequencyBandCompressor

Search


Share Section


Sounds
Samples
Timelines
Tools
Microsounds

Home 
Topics 
More... 

All Sections


Products
Order
Company
Community
Share
Learn

Login / Register 
Change password
Forgot password?

Symbolic Sound


Home
Kyma Forum
Eighth Nerve

TWiki Links


twiki.org
The attached Sound performs frequency selective compression of the incoming spectrum. It uses two ganged 2-pole bandpass filters to select out the band of interest. Then compression is applied in such a way that, at the threshold level, we have unity gain plus or minus any additional gain you specify in the VCS.

The motivation for this Sound is some recordings that have waaaay too much energy in the region 1.5-2.5 KHz during lead parts, but otherwise the backing effects are just fine. I don't want to remove those backing contributions, but I do want to chop down the intensity of the lead synth worms and other wierd pads in the music. (I almost suspect that the sound engineer had some hearing loss around 2 KHz and he ended up overcompensating so it sounded right to him.)

So it isn't so much a matter of compressive EQ needed here as it is straightforward compression of signals above the threshold level. Everything below threshold is just fine.

Using two ganged filters provides much steeper skirts to the bandpass than would be had with only one filter. Each filter provides about -6 dB/octave rolloff on either side of the center frequency. (These filters use my DSP code).

An interesting feature of these filters is that a band reject from the same family of filters is mathematically identical to subtracting the bandpass filter output from the incoming signal. Hence, doing it twice in a row is fine too. This is not the case had I used ganged Butterworth LPF and HPF to create a "bandpass" filter. Those Butterworth filters have complex phase properties that produce notches and or bumps in the spectrum when you add or subtract filtered signals from the incoming signal.

The stereo compressor uses a pair of compressors, one for each channel, fed by the maximum of the absolute value of the left and right signals. This is one way to run a stereo system and keep the balance during compression. But another way would be to feed the SqrtMagnitude? of the left and right channels. The difference is slight, but there is a difference. The SqrtMagnitude? will always give the same or higher apparent level than the maximum of either channel. Hence a SqrtMagnitude? driven compressor will act more strongly than my Max sidechain.

After running the recordings through this Sound I have a pretty decent mixdown. I don't have access to the original tracks of the mix. So I had to devise something that would work on the stereo mixdown handed to me. This Sound does the trick. Let's hear it for Kyma!!

[If you need a broad bandpass you can set a low Q value. But that kind of filter will always have rolloff on either side of the center frequency. Another trick, borrowed from old fashioned radio circuits, is to stagger-tune the two bandpass filters. If you set them so that the -3dB levels meet at the same frequency then you have an almost flat passband.]

-- DavidMcClain - 07 Jan 2004

Well, I'm frequently wrong... I am correct that subtracting one BPF output from the input signal is the same as a band reject filter from the same family. What I'm wrong about was the possibility of subtracting a twice filtered signal from the input without causing bumps and wiggles in the difference spectrum. You can't do that.

My mistake was in being too quick to generalize without fully checking it out. I ran the filter network through a signal simulator where I can examine the detailed spectral results and found that using two BPF's causes the same kinds of problems as using the ganged Butterworth LPF/HPF bandpass filters.

So here is a corrected frequency selective compressor with only one BPF filter in it.

-- DavidMcClain - 07 Jan 2004

... and here it is again, this time along with an FIR version that absolutely avoids these phase cancellation problems. The FIR version just uses complementary pairs of GraphicEQ? Sounds in the two branches. Since I found myself tending toward about a 1-octave wide band anyway, this FIR version works better than the IIR one anyway. The IIR version would be most useful in high-Q situations where you need 1/3 octave or narrower correction bandwidths.

-- DavidMcClain - 07 Jan 2004

Sonically it's working but where can I find the missing "gif" i.e. png files? Tweaking this sound, I just realised that there is no bass in Pantera songs! Didn't try FIR yet.

-- KarlMousseau - 07 Jan 2004

 
 
© 2003-2014 by the contributing authors. / You are TWikiGuest