[sdiy] simple "level separator"?

[Magnus Danielson]

From: "MED" <teenagewasteland at prodigy.net>
Subject: Re: [sdiy] simple "level separator"?
Date: Wed, 30 Apr 2003 16:33:21 -0500

MED,

> > Actually, this point to one of the type of curcuits which is not very
> often
> > discussed here, namely dynamic processors (expanders, noise gates,
> compressors,
> > duck processors, deessers, etc. etc.).
> >
> > Cheers,
> > Magnus
> 
> How *does* an expander actually work?
> -MED "now that you mention it..."

Oh, why would you ask *THAT*? ? ? ;O)

There are two great principles of dynamic processors, the feed forward and the
feed backwards principles. All designs I've seen fall into either of the
categories, even if I easilly could invent more categories.

These categories is easilly understood if you realize that there is two main
blocks in a dynamic processor, the signal/gain chain and the control chain.

The signal/gain chain is basically the VCA, but for more speciallized cases can
filters be involved too, but for the moment let's just consider a VCA.

The control chain listens to a signal, detects the signal level by some
mechanism and then process this detected level to generate a control signal to
the VCA.

The feed forward principle simply have the input signal of the control chain
before the VCA, so that the control chain process the signal in the forward
direction of the signal until it joins in the VCA.

The feed backward principle simply have the input signal of the control chain
after the VCA, so that the control chain process the signal in the backward
direction of the signal until it joins in the VCA.

Depending on which category a dynamic processor belongs to, the details of the
control chain differs to achieve the same effect.

Now, we can have different principles like expansion, compression and noise
gating. There are more, but these will be sufficient to understand the
generics.

An expander would expand the dynamic range of a signal, so that a 3 dB change
in signal level on the input would result in say a 6 dB change in the output
signal. To acheive this one can mentally consider it as a mapping of an input
signal level onto an output signal level. In order to acheive the difference
in signal level we naturally use a VCA, and we then must make sure that we
acheive a gain or damping as appropriate. There is always a point where no
change in gain is acheived, for an expander, above this point you have a net
gain in signal level and below a net loss in signal level. In practice, you
have say a full-wave rectifier, followed by lowpass filter (these together
will detect the signal level), then by controling the DC offset and gain of
the signal level you generate the VCA control signal.

Now, this simple expander (or compressor if you configure it so) will have a
number of deficiencies. For instance, you may consider the speed of reaction
too slow. You might change into a more elaborative filter, or change to a more
complex detector. You might also want to make the time constant (the reaction
time of the filter) changeable so that you can adjust it to needs. You can even
make the reaction times differ depending on the direction of change, so that a
rising signal level (attack) has a different reaction time than the falling
signal level (release). Such a diffrentiation is easy to explain the need for,
since we often tend to hit something (quite rise in amplitude - poles way over
to the right side of the s-diagram) and then let the resonance act (slow
decay in amplitude - poles on the left side of the jw-axis, but very near the
jw-axis).

So, compressors work just the same way really, just that you process the signal
so that the net gain reduces for higher signal levels. The noise gate simply
has a non-linear processing so that input signals below a certain signal level
receives a high damping where as signals above a certain level experiences a
unit gain. Noise gates are for obvious reasons always feed forward.
A limiter works similar to noise gates, but below the threshold you experience
unity gain and then above the threshold the compression increases with
increased input level such that the output signal level stops rising or rises
very slowly. Feed backward limiters are trivial to make.

There are *many* subtle points which can be discussed forever and where I
master only a few on a systems level. It supprises me that althought there is
a richness in functionallity, most of the time fairly simple devices to make
so that it should be fairly ontopic since many people use this stuff. Actually,
sometimes it seems underused and considered expensive, where I tend to think it
should be such a simple thing to do really. Equalizers seems to get the same
kind of mental treatment even if they are not as a wide palette of designs.

Cheers,
Magnus - who reengineered catastroph limiters to PA speakers capable of 139 db(SPL) on 1 meters distance of a single cabinette - we never used less than 4 a side!