Compressor confusion

  • Thread starter Thread starter singlespeak
  • Start date Start date
Hehe...Well, it was funny.

Some people need to loosen up. All wound up and shit. :D
 
singlespeak said:
That's why people don't throw birthday parties for me anymore.
Click to expand...
Yeah, I know the feeling. It's the same thing here, except I started using my own "hose" instead of a garden hose, if you get my drift. That somehow seemed to offend more than a few people....

G.
 
But it was an interesting discussion about compression and how it works.
 
SouthSIDE Glen said:
They automatically push up the volume by an amount proportional to the amount of gain reduction applied in the compression stage. Theoretically this would mean that - if it were a linear relation - when you set the gain reduction ratio to infinity-to-1, it would supply an infinite amount of makeup gain and send the signal level up through the roof,
Click to expand...

Except that gain reduction isn't infinite with a ratio of ∞:1, it's equal to the amount the signal exceeds the threshold. If make up gain is tied to gain reduction then 10dB over threshold = 10dB reduction = 10dB make up.
 
bouldersoundguy said:
Except that gain reduction isn't infinite with a ratio of ∞:1, it's equal to the amount the signal exceeds the threshold. If make up gain is tied to gain reduction then 10dB over threshold = 10dB reduction = 10dB make up.
Click to expand...
On paper, brick wall limiting is by definition a ratio of inf:1. If x>0 dB over threshold in = 0 dB over threshold out, that's infinite compression.

I know that in the physical world that what we call "brick wall limiting" really doesn't meet that ideal, some signal usually manages to eek through, but I was talking theoretical definition, not physical circuit design shortcomings.

And I believe the automatic makeup gain is not tied to the number of dBs lost in compression, it's tied to the compression ratio setting. Correct me if I'm mistaken about that, but none of the ways I can think up that it would have to work if it were tied to the actual signal and the actual decibels lost make any sense to me.

G.
 
SouthSIDE Glen said:
On paper, brick wall limiting is by definition a ratio of inf:1. If x>0 dB over threshold in = 0 dB over threshold out, that's infinite compression.
Click to expand...

But "infinite compression" still equals finite gain reduction, and if make up is tied to reduction it will be finite as well.

SouthSIDE Glen said:
I know that in the physical world that what we call "brick wall limiting" really doesn't meet that ideal, some signal usually manages to eek through, but I was talking theoretical definition, not physical circuit design shortcomings.
Click to expand...

I'm with you there.

SouthSIDE Glen said:
And I believe the automatic makeup gain is not tied to the number of dBs lost in compression, it's tied to the compression ratio setting. Correct me if I'm mistaken about that, but none of the ways I can think up that it would have to work if it were tied to the actual signal and the actual decibels lost make any sense to me.
Click to expand...

It's funny, tying it to ratio (while disregarding threshold) makes no sense to me. To tie it to reduction would be a matter of looking ahead, finding the peak amount of reduction and plugging that number into the make up gain. But that's not to say that I know of one that works like that.

I just checked several dynamics processors with automatic gain compensation. One is a brick-wall "mastering limiter" with look-ahead that simply raises gain the same amount as the threshold is lowered. Two others meant for individual tracks must have some sort of formulas more complex than just matching reduction. Make up is less than the change in threshold but more than the reduction (∞:1, fastest attack/release, track with -10dB peaks), and seems to have limits to how much is applied. Pulling the threshold down a little, not enough for the signal to exceed it, resulted in a little boost (say -6dB threshold yielded 3dB or so of gain). Pulling it down to light or moderate reduction resulted in more boost (threshold between -12 and -20dB, less than 10dB gain). Pulling it far enough down resulted in a leveling off of the make up gain (still 10-12dB of gain). So for these plugins make up is not tied to directly reduction, threshold or ratio, and there is definitely something in the code to prevent stupid amounts of make up gain even with stupid amounts of reduction. I don't think they are trying to maintain a constant RMS because many settings make it sound noticeably louder. I think they are using a fairly sophisticated algorithm that looks at ratio and threshold.

None of which changes the fact that automatic gain compensation almost always sucks on individual tracks.
 
Makeup gain is likely tied to a hypothetical "loudest possible signal" that could get through. Let's assume instant attack and release to simplify.

So... Say you have a 1:1 ratio and a threshold of...well threshold doesn't matter at 1:1. The loudest hypothetical signal that can get through is 0 (full scale), so there is no makeup gain.

Now say the ratio is 2:1 and the threshold is -10. The loudest hypothetical signal that could get through would have to try to cross 10 db of the "compression zone" while being hit with a 2:1 ratio. So it could only ever reach -5 at best. So the makeup gain is 5 to "make up" the 5 db of the "compression zone" that could not be crossed.

2:1 ratio with a threshold of -20...makeup gain is 10.
4:1 ratio with a threshold of -20...makeup gain is 15.
3:1 ratio with a threshold of -9...makeup gain is 6.
9:1 ratio with a threshold of -9...makeup gain is 8.
brickwall ratio with a threshold of -9...makeup gain is 9.

So makeup gain isn't tied to your actual signal. It's based on a hypothetical loudest signal possible. Not that I know for sure, but that's the only thing that makes logical sense to me.
 
I'll second Chibi's assumptions. It would be interesting to add attack speed into the equation since a fast initial transient would violate the results derived from just threshold and ratio.
 
The equation definitely changes when you add attack and release, but I'm willing to bet that the principal remains the same. As a test, drastically change attack speed while leaving everything else the same and see if the automatic makeup gain changes. I'm willing to bet makeup gain goes down as attack time goes up.
 
bouldersoundguy said:
But "infinite compression" still equals finite gain reduction, and if make up is tied to reduction it will be finite as well.
Click to expand...
Finite when using number of dBs as the scale. Infinite when using ratio as the scale.

Here's why the dBs idea didn't make sense to me:
bouldersoundguy said:
To tie it to reduction would be a matter of looking ahead, finding the peak amount of reduction and plugging that number into the make up gain.
Click to expand...
If you're talking a dynamic, peak-by-peak adjustment, the results would be a zero-sum cancellation of the compression; the output would equal the input. So that's out.

A static adjustment based upon a look-ahead to determine the maximum peak and adjusting based upon that is just simple peak normalization. Which would actually kind of make some (little) sense, but unfortunately doesn't match the symptoms described:
So I added a compressor and I was fiddling about with the controls and at one point I noticed that when I increase the ratio from 3:1 to 15:1, the volume actually peaks higher!
Click to expand...
This indicates that the pre-compression signal itself is not the factor on which the amount of makeup gain is calculated, because the signal did not change. All that changed was the reduction ratio. And if the look-ahead were post-compression, the peak output signal would stay the same, not increase.

Chibi's theory that it's a function of the combination of ratio and threshold might be true. The way to test that is to keep the same signal and reduction ratio, but change the threshold and see what happens. If that also changes the amount of makeup gain, then Chibi is on the right track.The next step then would be to actually make some measurements and calculations to see if the actual formula is as Chibi surmises.

I would have sworn that earlier in this thread or in another earlier thread asking about this, the OP had said that when they cranked up the gain reduction too far, that the makeup gain sent the signal into clipping. Maybe I read it wrong - or maybe I forgot to take the needles out from under my toenails before I came online that day - in which case it's my mistake and I've misled myself. I don't know, because I can't find the entry now offhand. But if that were actually true, that could easily be explained by it being connected to ratio only.

G.
 
Last edited:
Betting it has auto-makeup…ridiculous novice feature… drives me nuts.
 
You must log in or register to reply here.
Back
Top