What, exactly, is a compander?

[mth5044]

And when/what are they used for? I was under the impression they were some kind of volume adjustment where one section changes the volume then it comes back in to set the volume back to normal. I read the datasheet for the compander used in the PT80, but I can't make anything out of it.

In an aplication, does it matter where a compander relative to a buffer? In my ever evolving delay project, I was going to run:

1/2 TL072 buffer -> 1/2 compander -> 4 parallel pt2399 circuits (w/o buffer, just a loose schematic in the PT2399 datasheet)(feedback control inside the compander section, unlike the PT80 which seems to take the signal after the compander and feed it back into it again) ->1/2 compander ->1/2 TL072.

Is this apilcation OK, or with the compander mess with the buffer before the 4 way signal split?

[JKowalski]

It is used to compress the signal, making it more friendly to digital conversion (such as in the PT2399)...

Wikipedia is your friend:

http://en.wikipedia.org/wiki/Companding

As far as your project, that should work fine. I think the PT80 was made to replicate certain styles of delay sounds...? And feeding the signal back through the compander might add some special distortion to the waveform that increases in intensity every repeat, which gives the sound the designers were attempting to get.

[mth5044]

Thanks JKowalski. That makes more sense now. I'll just pluck the compander out of the PT80 

Any progress on your delay? That image of your 27PDT switches still haunts my dreams 

[JKowalski]

Ahaha, no progress. It's not gonna happen for a long while, I think.

I drew up a mini board for the PT2399, a while back, it's a little PCB that includes the entire delay section, with pins for Input, Output, Feedback Out, Feedback In, Power, Ground, and Delay Time. I was planning on doing something like a long board with four plug in PT2399 PCB modules, for effective use of space (PCBs on top of a motherboard PCB) and ease of troubleshooting.

You know, it would be very nice if I could do SMD miniboards... I don't know though, because I have plenty of DIP PT2399s and this thing would be in a rack mount anyways, probably.

[Mark Hammer]

Some circuits have the properties that they generate noise (i.e., add some noise on top of the signal).  If the circuit has high headroom and can tolerate a monster input, then the user can maximize the signal to noise ratio by simply assuring that the input signal is much larger than any noise that might be acquired.

But what if the circuit has limited dynamic properties and can't handle large input levels without distorting?  Generally, some form of compression is used such that the signal level is kept as high as is feasible without exceeding the tolerable maximum.

Of course, that can play havoc with the dynamics, and make the output sound unrealistic.  Enter companding.  This provides a complementary action such that the signal is compressed prior to the "risky" processing (the point in the path where noise might be acquired and distortion might occur) and then dynamically expanded by the identical amount such the the original signal dynamics are restored.  This, of course, avoids the risk of clipping, but has a noise-management effect as well.  Since the compression makes the input signal louder on average by boosting the soft parts, when the signal is expanded, those boosted soft parts are made quieter.  That expansion is applied not only to the original audio signal, but to any of the noise that was accumulated as well.  This has the effect of minimizing the level of any noise acquired.

Companding is used most often with BBD devices, but is also used with some digital delays, and with some phasers using either OTAs or FETs that may have limited signal-handling capability.

[alanlan]

Companding as I understand it is a static process not a dynamic process as used in conventional audio compressors and expanders.  I used a companding ADC/DAC pair once for a delay project.  It was 8 bits but gives a pseudo 12 bit performance by bending the scale i.e. it's non-linear.  I suppose we could get into an argument about what the term "companding" actuall means so suffice to say that there are at least 2 completely unrelated processes I know of which may be given the same term by different people.

I think the DAC was a DAC-76 made by PMI (Precision Monolithics Inc - anyone remember?)

[mth5044]

Some circuits have the properties that they generate noise (i.e., add some noise on top of the signal).  If the circuit has high headroom and can tolerate a monster input, then the user can maximize the signal to noise ratio by simply assuring that the input signal is much larger than any noise that might be acquired.

But what if the circuit has limited dynamic properties and can't handle large input levels without distorting?  Generally, some form of compression is used such that the signal level is kept as high as is feasible without exceeding the tolerable maximum.

Of course, that can play havoc with the dynamics, and make the output sound unrealistic.  Enter companding.  This provides a complementary action such that the signal is compressed prior to the "risky" processing (the point in the path where noise might be acquired and distortion might occur) and then dynamically expanded by the identical amount such the the original signal dynamics are restored.  This, of course, avoids the risk of clipping, but has a noise-management effect as well.  Since the compression makes the input signal louder on average by boosting the soft parts, when the signal is expanded, those boosted soft parts are made quieter.  That expansion is applied not only to the original audio signal, but to any of the noise that was accumulated as well.  This has the effect of minimizing the level of any noise acquired.

Companding is used most often with BBD devices, but is also used with some digital delays, and with some phasers using either OTAs or FETs that may have limited signal-handling capability.

Very clear, thanks!

A slightly related question - what sets the amount that the compander brings the signal down to? Is it a chip type basis or does the supporting circuitry play into it?

[mth5044]

Companding as I understand it is a static process not a dynamic process as used in conventional audio compressors and expanders.  I used a companding ADC/DAC pair once for a delay project.  It was 8 bits but gives a pseudo 12 bit performance by bending the scale i.e. it's non-linear.  I suppose we could get into an argument about what the term "companding" actuall means so suffice to say that there are at least 2 completely unrelated processes I know of which may be given the same term by different people.

I think the DAC was a DAC-76 made by PMI (Precision Monolithics Inc - anyone remember?)

This is slightly over my head, but I'm sure someone understands it 

[head_spaz]

Compressor + Expander = Compander

Music is dynamic... meaning that there are loud and quiet sections in music. Crecendos and such.
In the old days, magnetic tape recorders were tasked with capturing live performances.
But tape technology has its drawbacks. One of its weaknesses is that magnetic mylar film has a high noise floor which introduces a high frequency hiss into the recording, and can be easily heard when the volume of the music goes soft to a whisper level.

Another drawback is that magnetic tape has limited headroom, because the magnetic properties of the mylar film can be easily saturated magnetically, which acts like a diode clipping the signal peaks, or a kid banging his head against the ceiling when he's jumping in bed.

To address these issues, the recording industry invented a workaround to address both deficiencies. They came up with compression... and expansion devices.
How it works is the music is sent to the compressor to amplify only the low level volumes, while at the same time attenuating the high level signals. This means that the soft music is amplified over the noise floor, which drowns out the low level hiss. Since it is recorded "hotter" onto the tape... during playback the deck controls are turned down to match the original sound levels... so the noise floor is far below the music signals... the hiss drops to undetectable levels.

The compressor also attenuates the loudest content to a lower signal level so that tape saturation is prevented... which eliminates the clipping.

In your mind - picture the raw dynamics of music as a raging sea with huge roaring waves. After compression, the huge waves are made small, and the low level tide is raised to cover the gritty "hissy" sand on the beach. This preconditioned signal that the tape "hears" and records ends up as calm ocean of music that is neither too loud, nor too soft. It's just right... as far as the tape is concerned. On playback though... loud parts, soft parts, they all sound about the same volume wise. It sounds pretty lifeless to the human ear.
The compressor is used for preconditioning the music content that is to be recorded.

And this is where the expander comes in. It is used during the playback of that recorded content.
The expander's job is to take the calm ocean and turn it back into the raging sea that it was before, but without the clipping of tape saturation, and without the hiss from the noise floor because the compressor took care of that. The expander makes sure that the low levels are attenuated down to their original levels, while it amplifies the higher levels even louder to regain its original level... so that during playback, the musical dynamics are restored with all of the headroom of the original content.

So... in the conventional sense... a compander has two seperate functions. The compressor is used on the record side of the tape.... and the expander is used on the playback side.
The only problem with this method is that a tape recorded using compression sounds pretty lifleless, unless its played back thru an expander. I think that's why companders didn't capture the market, and are not so well known.
Dolby however, made a huge impact because decks offered it built-in as a standard equipment.

I guess a simpler anology might have been to compare it with a zip file.
Zip it (Compress data)
Archive it (Media storage - tape)
Unzip it (Expand it back to the original data)

Hope all this makes sense.

Compressors, expanders, and companders have become quite common today, in all sorts of devices, but they're still somewhat of a mystery to average joe.

[oldschoolanalog]

Courtesy of StephenGiles:
http://www.ka-electronics.com/images/pdf/NE570_Applications.pdf
Might be of some help.
PS: UTSF...

Search UTSF

[Ben N]

Great article. There was also an old magazine article project for a fuzz around here that had the halves of a 570/571 wrapped around a silicon transistor. The compressor saturated the transistor to produce the fuzz, while the operation of the expander was switch-selectable to reintroduce the original dynamics via the sidechained input signal, or to produce a gated effect, IIRC. I never built it, but just looking at the schematic and reading the article gave a pretty good idea of the basic operation of companders. I think Craig Anderton used them in a few projects as well.

[mth5044]

Thanks for all the info, I really appreciate it. Sounds like something that could be introduced to a variety of circuits.

PS: UTSF...

Search UTSF

UTSF = use the search function? 

[Ben N]

UTSF? 

[puretube]

The real fun begins,
when you add a switch to decide
whether the "output"-voltage will control the expander-part,
or if you use the input-derived control-voltage... 

[StephenGiles]

The real fun begins,
when you add a switch to decide
whether the "output"-voltage will control the expander-part,
or if you use the input-derived control-voltage... 

or a mixture of the two processed by a bounce circuit!!

[mth5044]

Hey you smart people quiet down over there!   

But seriously, no idea what you guys are talking about. To the search button!

[puretube]

Scuse my ignorance: of course the fun only really starts, when there is a time-delay between the occurence of the input-side
and the output-side
of the control-voltage deriving circuit-parts 

[mth5044]

Scuse my ignorance: of course the fun only really starts, when there is a time-delay between the occurence of the input-side
and the output-side
of the control-voltage deriving circuit-parts 

Lol ok, if you say so 




Ok, so if I put a compander around a complete delay circuit which for feedback, is feeding back a single repeat unto itself through a resistance to make the level decrease each time it goes around. Will the expander see the decreased signal and try to expand that signal back to what it was before the compressed stage? I mean, will it bring the levels of the repeats back up so that each delay is the same level as the signal into the compressor? Or will each delay remain at a decreasing level, but expanded by the factor at which the compander works?

Did that make sense?