What is "companding" and/or a "compander"?

[skiraly017]

The Stompbox Cookbook has a project called the Compander-Lite but I'm lost as to what the device actually does. Is it a delay? A doubler? A self-oscillator? Thanks for the help.

[Transmogrifox]

compressor/expander

It is often used in delays, reverbs, choruses and flangers where a signal is sampled by a lo-fi device such as a BBD or 10-bit A/D converter.  The compressor part levels the signal so you can get maximum bit resolution going into the sampler.  It increases signal/noise ratio and reduces distortion.

The expander part returns the signal to its original dynamics on the output side. 

[Mark Hammer]

There is more information provided here:  http://www.semiconductors.philips.com/acrobat_download/datasheets/NE570_3.pdf

Dig around for compandor application notes: http://www-eu4.semiconductors.com/acrobat_download/applicationnotes/AN174.pdf

"Compandor" (compressor/expandor) is a composite term, like "modem" (modulator/demodulator).  Companding was developed primarily for magnetic tape.  Tape, as a magnetic storage medium, has limits to how much signal it can "store" (you've probably heard of tape saturation).  In addition, tape introduces a certain amount of noise on playback.  Consequently, the problem with tape was that you couldn't improve the signal-to-noise ratio by simply recording a hotter signal.

By artificially restricting the dynamics of the input signal, what got laid down on tape could be "forcibly" set to the maximum level the tape could handle without distorting.  Keep in mind that restriction of dynamics also involved making sure the softest parts also had their levels boosted, and the loudest parts had their levels kept under control.

On playback, the dynamics were restored, such that the loud parts were made loud again, and the soft parts were made soft again.  because the tape noise was not compressed and the re-expanded, but simply expanded, tape noise would be sharply attenuated, producing vastly improved S/N ratios onplayback without having to run the tape at 15 inches per second or faster.  Because the dynamic peaks were squashed during compression, they would not push the tape past what it could cleanly handle.  Re-expanding them would restore true dynamics and deliver the full dynamic range of the input without having to listen to either tape hiss or saturation-based distortion.

Understand that companding was really an industry response to the rather severe limits imposed by cassette tape, which was thin and run slowly (1-7/8 ips).  S/N ratio was much less of a problem when working with 2" tape at 15ips.  The huge reduction in tape size, thickness, and speed brought huge challenges in terms of achieving a listenable sound.  No real surprise that Philips, as one of the first companies to come out with tape cassettes in the early 70's was also the driving force behind the NE570/571 compandor chips.

Of course, what was useful for tape did not have to resticted to tape alone.  Keep in mind the basic underlying premise of companding is that you have:
a) a storage medium or processing path that is limited with respect to signal-handling and tends to generate noise in itself
b) a high likelihood of high input signal levels and inconsistent dynamics
c) a need to reduce both noise floor and peak clipping on output

As you might expect, there are non-tape contexts that have similar properties.  For example, bucket-brigade devices, reverb springs, and some FET or OTA-based chips that can't handle high signal levels.  You will find NE570/571 compander chips on a variety of analog delay, flanger, and chorus pedals, as well as the Boss PH-2 and RPH-10 phaser, and some spring reverb units.  What they all share in common is that the BBD/spring/phase-shift stage generates noise and can't handle huge input signals without misbehaving in some manner.

Some people frown on companders and companding.  To some extent, I see their point.  It can be the case that the tracking of dynamics is not perfect and what gets called "restored dynamics" is close but not a perfect replica of the input dynamics.  Of course, given the alternative, it may be the smallest of problems to contend with.

The PAiA/Anderton Rocktave octave divider uses a compander chip to provide the sort of consistency in signal level that improves tracking by the 4013 flip flop. Since the octave dividing tracks better with less signal dynamics and worse when dynamics vary, Craig Anderton used a little dynamics mistracking trick with the 570 to have the chip apply slightly different dynamics on expansion tan were applied during compression.  This essentially "turns off" the octave signal as the input note decays to a point where sputtering and inconsistent division is likely.

Longer than what Trans wrote while I was working on this, but between the two of us, you are now officially smarter.

[skiraly017]

And I do feel smarter!

How it all breaks down...for me anyway.

[cd]

And I do feel smarter!

How it all breaks down...for me anyway.

How's the life of celibacy treating you?

[skiraly017]

How's the life of celibacy treating you?

Tequila often leads to non-celibacy moments. 

[davebungo]

There is another use for the term "companding".  The type of processing described in previous posts refers to the process of restricting or expanding the dynamic range of a signal by monitoring the signal level and modulating it accordingly, using a gain cell.  There used to be (not any more as far as I know) DACs (so called companding DACs) which had a built in non-linear transfer function which could achieve psuedo 12 bit resolution using only 8 bits.  I remember a device made by a company called PMI (Precision Monolithics Inc).  This did not rely on manipulating the signal level but more directly by deliberately having more bit changes at lower signal levels and less at higher signal levels.  The transfer function had an 'S' shape to it as a result.  This was reversed at the output of the process.  I do remember a delay line project which used these along with an 8 bit SRAM based memory.

[davebungo]

http://www.analog.com/en/prod/0,2877,DAC88,00.html

[wampcat1]

And I do feel smarter!

How it all breaks down...for me anyway.

How's the life of celibacy treating you?

[gez]

Are there still uses for companders in this day and age?  Not being sarcastic, this is a genuine question; are they still being manufactured?

I ask as they seem to have been dropped from the pages of most (all?) the usual suppliers in the UK (unless you want to pay an arm and a leg at those places that specialise in 'esoteric' stuff).

[Mark Hammer]

Still available, and still manufactured and supported from what I know (See SA570/571 in both DOP and surface mount form), though I think your suspicions about them being supplanted by other technologies is likely accurate.  I know that, for a period, cordless phones would pitch companding as a product feature, but that was 10 years ago when cellphones were the rare exception rather than the norm.  That's why I dragged the invention of cassette tape into the discussion (though I should have included telecommunications via wire into the picture too).  Companding was developed as a technological solution to a medium-limitations problem.  Once those types of medium-based products are supplanted by other technologies (non-degrading fibre-optic digital telephone transmission as an example), the raison d'être for companding starts to fade.  There ARE "legacy" products where I imagine companding is still important, but they are becoming fewer in number.

It's the same thing with BBDs.  As cheap, decent digital delay chips become available, BBDs fade into the background and the companies which once supported them (Reticon, Matsushita) drop support.