Compressor or Limiter?

[amz-fx]

Your compressor might be a limiter...

Check out my 3 part series of compressor pedal tests in the AMZ Lab Notebook.

Thanks, Jack

[Mark Hammer]

Nice coverage of the compression universe.  I guess the only gripe I would have with the 3-parter is that the graphs essentially address instantaneous response to transient peaks.  That is, an input signal of X amplitude will only exit as a signal of Y amplitude.

That's accurate, and not uninformative in any way.  But unless there's something I'm not getting, what differentiates a limiter from a compressor is what they do after the transient peak has passed.  The limiter says "Hey, not MY job.  I'm only here to make sure that nothing larger than THIS amplitude leaves the output jack.", while the compressor does that as well as helping out low-level signals.  As I understand it, limiters will not have a problem with "breathing" (this slowly-increasing hiss after a peak has passed and followed by silence), where compressors will, because compressors treat residual hiss at the input as equivalent to soft passages needing "help".

Admittedly, that what-happens-next aspect of compression is very difficult to illustrate with the simple transfer function graphs used, since it involves plotting gain recovery of a standardized input signal over time.  I can't fault you for not doing the impossible.   

Electronic Musician magazine had a great feature article on compressors and compression (including how to set them for different signal sources) back in February 2001, called "The Big Squeeze".  If you can find it in a music store or someone's selloff of back issues, scoop it up.

[amz-fx]

The limiter says "Hey, not MY job.  I'm only here to make sure that nothing larger than THIS amplitude leaves the output jack.", while the compressor does that as well as helping out low-level signals. 

That is not what is happening with the compressors. They do not change the ratio of the low level signals at all for most of the pedals tested. This is easy to see in a Dynacomp in that the low level signals do not turn on the transistor detectors at all, so they are not having an influence on the dynamics. There may be gain but is simple linear gain... if it is 10db gain when there is no control voltage from the detector then a -40db signal gives -30db out, -30db in gives -20db out, and so on. It is just a boosted signal but not a compressed one; there is a difference.

The only pedal in all of those graphs that is compressing the signals below the threshold is the last one in Part 3 (Q&D 2).

These comp/limit pedals are only changing the input-to-output ratios of the peaks above the threshold, and many are doing it at a high ratio like 10:1,and therefore functioning as a limiter. They are clamping hard on the output even as the input signal increases. The Keeley Pro, which can do lower ratios such as 4:1 or 2:1 is a peak compressor... it is only compressing the peaks and handling signals below the threshold in a linear 1:1 relationship. It is there in the graphs... below the threshold input=output... 1:1 is no compression or limiting at all. Only the peaks above the threshold have the slope of the line changed.

Admittedly, that what-happens-next aspect of compression is very difficult to illustrate with the simple transfer function graphs used, since it involves plotting gain recovery of a standardized input signal over time.  I can't fault you for not doing the impossible. 

Actually I have a test rig that can dynamically measure compressor characteristics. Maybe I will do some more tests that show further working of these pedals, once I am no longer tired of shuffling compressor pedals around.

Best regards, Jack

[EBK]

A really nice full range compressor for guitar is detailed here:
http://www.thatcorp.com/pedals/4316%20Battery-Powered%20One%20Knob%20Squeezer.pdf



I have this on my bass pedalboard, and it is almost always clicked on.

I have an LED bar graph on mine that shows when it is amplifying and attenuating, and by how much.

[amz-fx]

From a Thatcorp Design Note:



The ratio is showing the change in the slope of the line as I described above.

Best regards, Jack

[EBK]

Isn't that what my graph showed?  When the single knob is turn down all the way, it acts like a peak limiter, but as you turn it up, it alters the ratio both above and below the threshold.  I forgot to mention that the different curves represented different knob settings.

[amz-fx]

Isn't that what my graph showed?  When the single knob is turn down all the way, it acts like a peak limiter, but as you turn it up, it alters the ratio both above and below the threshold.  I forgot to mention that the different curves represented different knob settings.

When the knob is all the way down, it is a peak compressor. As you turn it up, the ratio of the signals below the threshold begin to take on a compression characteristic and the peaks above the threshold begins to be more and more compressed. (This is basically what you said.)

That is an interesting circuit. If I wasn't tired of compressors, I might build one.

Best regards, Jack

[PRR]

....limiters will not have a problem with "breathing"...  ... .....  the compressor does that as well as helping out low-level signals.

An abused limiter can breathe like a frisky Corgi.

Lifting small signals is an "Expander".

Blesser wrote it all clearly.


All of these functions can be put in one box, so what do you call it?

[EBK]

Lifting small signals is an "Expander".

That doesn't seem correct.
I thought an expander was the opposite of a compressor, i.e., it lowers quiet sounds and lifts louder sounds, increasing the dynamic range of the signal.
Maybe I'm simply misunderstanding where the threshold is in your statement.

[Mark Hammer]

The SSM2166, used in the excellent Q & D compressor, includes downward expansion.  That's a big part of what makes it one of the quietest compressors I own.

Companders using the NE570/571 chip, can be configured for compression OR expansion.  Since such chips are used for reducing the noise from a medium that is prone to hiss, compression is generally applied to signals above some designated level to make the best of a limited headroom, and the expansion is intended to that same content to restore original dynamics.

When expansion is included in a processor largely focussed on compression, it tends to be downward expansion, for the purposes of keeping hiss out.

[PRR]

I think Jack's point is that many of the boxes called "compressors" in this market have fairly hard knees and flat plateaus, much like the safety part of a broadcast/film limiter.

His 'AMZ Q&D 2 compressor, based on the SSM2166' can be run as a dBx type full-range compressor.

[Mark Hammer]

I think Jack's point is that many of the boxes called "compressors" in this market have fairly hard knees and flat plateaus, much like the safety part of a broadcast/film limiter.

True, but the reason why guitarists treat them as "sustainers" is because their functioning/design often brings the signal back UP to the knee.  I'm sure not all do, but enough that our focus as users of them is on the overall constancy of the level produced, rather than simply the absence of unwanted peaks.

Hopefully, Jack can find enough time and patience to measure that signal-amplitude-over-time aspect, and illustrate it in a manner digestible for members here.  Certainly, the role o gain-recovery time is something I find poorly understood by many players, sales-staff, and even writers/reviewers.  The dynamic "feel" of a compressor has a lot to do with how it rebounds from what it had to do to constrain a peak.

[amz-fx]

True, but the reason why guitarists treat them as "sustainers" is because their functioning/design often brings the signal back UP to the knee.

I may not be understanding you correctly, but the signal is not being brought up to the knee by the compression process. You can turn up the volume of the output, but the whole response curve moves up, however it retains its shape...  i.e. the part above the threshold is still limited at 10:1 and the part below the threshold is still 1:1 except it has some gain. The dynamics of the low level signals are not changed though they can be increased in volume exactly as with a booster pedal.



In the image above, you can see that below the threshold the response of the buffer and the Dynacomp are exactly the same. If you pick softly or let the notes decay down to a soft volume below the threshold, they will not sound any different than through a buffer (except with maybe a little more noise and distortion caused by the CA3080). The ratio of input to output remains 1:1

I did not start out to do tests on compression pedals as my initial intent was just to be able to measure my Q&D pedal. It was when I started tested a couple of other pedals to compare it to that my interest in the whole thing got stimulated.

The SSM2166 chip used in the Q&D 2 has interesting properties because it has a noise gate on very low level signals, and a hard limiter on the strongest. In between is the compression range. All three of these ranges have different ratios of input-to-output.

Best regards, Jack

[amz-fx]

Here is a buffer signal (gain = 1) vs. a boosted signal (gain = +10db)



The signal with gain is higher on the graph, so it is stronger (more volume).

However, the slope of the two lines is exactly the same and the ratio of both is 1:1 meaning there is no compression. A 10db increase in either input results in a 10db increase in its output. When you compress a signal, you change the slope of the line. For example, a 10db change in the input will produce only a 5db change in the output if the compression ratio is 2:1  (10/5). If a 10db increase in the input results in only a 1db increase in the output, then the ratio is 10:1 and it is limiting.

This is the design of peak compression (or peak limiting) and is the way that almost all vst compressor plugins work, as well as the majority of the hardware guitar compressor pedals.

Best regards, Jack

Edit: fixed typo

[Mark Hammer]

1) The 2166 is such a great chip.  I am dismayed that I don't see it incorporated into more products.  Mind you, I say that never having worked with any of the THAT chips.  Perhaps they might make me say "Good riddance!" to the 2166.  Who knows.

2) Perhaps my focus is too much on circuits that behave like AGCs.  Even the Whisper, a sort of further development of Anderton's compressor that you co-designed with Thomas Henry, applies gain, pulls back on that gain in response to peaks, and restores the gain as the LDR recovers.  That's what I'm talking about.  The transfer function IS most assuredly 1:1, but the impact on the resulting output signal is that the restoration of gain brings softer signals up higher.  I sort of misspoke when I said "up to the knee".  But the idea is that gain WILL be applied to the quiet parts, *unless* that quiet part somehow momentarily exceeds the threshold.

Is that making more sense now?

[PRR]

....the reason why guitarists treat them as "sustainers" is because their functioning/design often brings the signal back UP to the knee.  ....the overall constancy of the level ...

If the plateau is "flat" to a few dB, it "sustains".

If the plateau has enough droop to hear, we hear it decay. Maybe not as much as in real life. But our sense of level is not near as good as sense of pitch, so slightly-sloped compression may not sound like much.

It was instructive to listen to the coded signal in dbx tape noise reduction. (Sadly that's all gone except in tape master restoration studios.) Everything is in proportion, just not original proportion. Technically it extends decay (RT60 is now RT30); the ear doesn't really hear it that way.

[amz-fx]

1) The 2166 is such a great chip.  I am dismayed that I don't see it incorporated into more products.  Mind you, I say that never having worked with any of the THAT chips.  Perhaps they might make me say "Good riddance!" to the 2166.  Who knows.

2) Perhaps my focus is too much on circuits that behave like AGCs.  Even the Whisper, a sort of further development of Anderton's compressor that you co-designed with Thomas Henry, applies gain, pulls back on that gain in response to peaks, and restores the gain as the LDR recovers.  That's what I'm talking about.  The transfer function IS most assuredly 1:1, but the impact on the resulting output signal is that the restoration of gain brings softer signals up higher.  I sort of misspoke when I said "up to the knee".  But the idea is that gain WILL be applied to the quiet parts, *unless* that quiet part somehow momentarily exceeds the threshold.

(1) The whole exercise started out as a way for me to learn more about the Q&D/2166, and as I tested it, the differences between it and the vintage compressor designs stood out. In the end, I achieved a much better understanding of the details of its workings and I am designing an upgraded version. Disappointingly, there are no ssm2166 chips in stock at Mouser and the lead time is 11 months!

(2) Ah ha! As I was thinking about it, I began to suspect that this was what you were referencing. It is true that the gain is applied to all the low level signals and a volume pot is used to trim the audio output to a useful level. This is a flaw in the simple designs that are more than 40 years old and not necessarily a characteristic of compressors. The CA3080 in the Dynacomp and the op amp in the Orange Squeezer are not only providing the drive for the full-wave rectifier, but also are the audio output. Whatever gain is used to drive the rectifier has to be corrected or trimmed for the audio. This is cheap, but poor design.     If the rectifier gain stage were separated from the audio path, these designs could be improved and residual low-level noise reduced.

(3) The Keeley Pro is the best of all of the designs that I tested, and the Q&D was next. I think with some redesign, I can improve the performance of the Q&D, but the chips in the K-Pro are just superior, though more complicated to use. The Q&D has a performance/simplicity advantage since so much of the design is all on one chip, which makes it simple to work with. Disclaimer: Though I do not claim to have designed any specific part of the Keeley Pro, Robert sent me the schematic as it was being designed and asked for my input. I sent him a number of comments on compressor design and the THAT chips, which he may have used (or not).

I should also note that the Empress compressor is not the current version that they sell, but is the prior design in the larger box. It is a very nice sounding compressor, though I was able to dial in the sounds on the Keeley Pro quicker and more easily.

A couple of last notes...  I made a compressor with the NE570 but did not include the data (just an oversight). I also did some testing of distortion that each pedal gives at specific input levels. Maybe I can post that data at the end of the last article if I get some time.

Thanks for your input!

Best regards, Jack

[Vivek]

Dear Jack,

Thank you for your excellent articles and experiments !

[ElectricDruid]

That image really shows why there's so much interaction between controls on a compressor, which (I'd say) is what makes them difficult to use effectively. After you tweak a few things, most people have no idea what it's actually doing.

If we could design a compressor with a in/out function that went from a straight line (no compression) to a big curve (lots of compression) there wouldn't actually be any specific threshold point. This would eliminate one control and also give a more transparent result, as Jack noted about the Empress, which basically seems to do this.

[Mark Hammer]

And many thanks for your output, Jack!