Random thoughts on compression

spantini

spantini

COO of me, inc.
This is not a problem in search of a solution.

I had this random thought pop into my head this morning: If I compress an audio wave at 4:1, render that to WAV, then re-compress that at 4:1 and render that to WAV, would that be the same as using 8:1 compression to begin with? I don't know any of the math behind this stuff, so I have no idea.

I did notice the second 4:1 render's audio wave was different than the first one. Not much, but noticeable. It didn't look as if it was doubled or halved - nothing that large. I also see that the file sizes as shown in the file's Properties show the re-compressed file as slightly larger.

I didn't bother stacking two compressor FX on the original track - would that produce the same results as above?
 
I think you'd need to stack a 4:1 then a 2:1 to get similar compression to a single 8:1.

If you have a peak 16 over the threshold, 8:1 is going to tame that to 2,
where as 16-over through a 4:1 is 4, through another 4:1 is 1.
 
Here's a snapshot of the original waveform (#1) and three separate renders (#2 #3 #4).

#1. No Compression
#2. 4:1 Compression
#3. 4:1 + 4:1 stacked Compression
#4. 8:1 Compression

I find it interesting that #2 and # 4 are nearly identical. All compressor settings other than ratio are identical. Threshold -32 with -4dB reduction.

compression test.jpg
 
Last edited:
I would guess that the threshold will play a part in things as well. For example, you could have the threshold set to start compressing everything above -10dB. If you do any volume adjustments since your highest peak should now be well below 0, you could be compressing things that weren't within the window the second time around.

With your setting at -32dB, you are probably compressing most of the signals. Maybe if you zoom in on portions of the wave file, you might see more of a difference that the zoomed out view.
 
I tried doing what you said at 4:1, 8:1, 16:1 and 4+4:1. Here are the zoomed in peaks. I think it's a case of peaks that were just above the threshold compressed on the first 3, but when the level drops below -32, the second 4:1 didn't do anything. Look at the peak with the pointer. The bottom one is 4:1+4:1. So I'm guessing it won't be the same as 16 or 8 for some peaks. Whether it sounds more or less compressed would probably depend on the source material.

COMPRESSING.jpg
 
Make-up gain is also part of the "calculus" because it will also determine how much of the first compressor output is affected by the 2nd, similarly, attack and release, whether controllable or built-in, will depend on the input, and so it's probably safe to say that serial compression is almost never exactly the same, though "it depends" is probably what your ears might say.

It's very rare that I use 2 compressors on a single track, unless one is for ducking. Of course, some degree of subsequent compression is always there somewhere in a [summing/folder] bus, but that is not specific to a single track.
 
I was tempted to add makeup gain but left that alone, along with most of the default settings. I adjusted the Thresholds to -32 for a -4dB cut, then the Wet Outputs to average -16, Dry faders all the way down, then the different Ratios for each render.
 
spantini said:
This is not a problem in search of a solution.

I had this random thought pop into my head this morning: If I compress an audio wave at 4:1, render that to WAV, then re-compress that at 4:1 and render that to WAV, would that be the same as using 8:1 compression to begin with? I don't know any of the math behind this stuff, so I have no idea.

I did notice the second 4:1 render's audio wave was different than the first one. Not much, but noticeable. It didn't look as if it was doubled or halved - nothing that large. I also see that the file sizes as shown in the file's Properties show the re-compressed file as slightly larger.

I didn't bother stacking two compressor FX on the original track - would that produce the same results as above?
Click to expand...
What do you have the Threshold set at - and the attack and release -? - what it's showing is that Compression is different than how most people think about it - the Wav files are going to be somewhat identical I assume - because you are compressing just the peaking part of the wave - and 4:1 or 8:1 is not that different in how it looks - if it was a snare shot that really hammered down - you would see a definitely leveling difference between 4:1 and 8:1 - the higher compression would show more of a limiting line - a little less so for the 4:1.
 
The waveform should not change except for the peaks that go above the threshold. This is assuming no gain after compression and 'digital' compressor(not modeled on a 'color' compressor). If you add gain, the waveform shape should not change, only the amplitude. If you add a second compressor afterward (it doesn't matter if it's after a non dithered bounce or on the same session) it won't touch anything if the threshold is the same and there has been no added gain. The ratio only determines how much gain reduction will be applied when the threshold has been passed. You would have to have the threshold set fairly low to see a large difference between 4:1 and 8:1 realistically speaking.

However, once you add gain to the first compressed signal you have changed the peak levels and now the second compressor, even with the same settings, will compress any new peaks that go over the same threshold. This is often done with vocals: the first compressor is just to grab short peaks, the second (with different settings!) is used to even out a performance and/or to pull/push the vocal 'forward'.
 
Quick test with a hand clap and some compressors.
Image just shows the peak value for each track, and the track names are the compressor ratio used.
No makeup gain, fast attack...everything equal except ratio.

Looks like my thought was right, in so much as you'd multiply stacked ratios to find their single equivalent, rather than add.
There's very little difference, in peak value, between 4:1+4:1 and 16:1,
and between 4:1+2:1 and 8:1.

Presumably stacking compressors is going to magnify any unique properties of the compressor used.
How that affects the sound, as perceived by us? No idea. :ROFLMAO:


comps.jpg
 
Now that intrigues me. Stacking multiplies - does not add. This makes me think of volume increases on a home stereo: turning the knob halfway up doesn't mean 50% volume. I forget the general formula for that stuff, been years since I had my hand in.
 
spantini said:
This is not a problem in search of a solution.

I had this random thought pop into my head this morning: If I compress an audio wave at 4:1, render that to WAV, then re-compress that at 4:1 and render that to WAV, would that be the same as using 8:1 compression to begin with? I don't know any of the math behind this stuff, so I have no idea.

I didn't bother stacking two compressor FX on the original track - would that produce the same results as above?
Click to expand...
Stacking ? Compression ? Re-compression ? 8:1 ? 4:1 ? WAV ? Random thoughts ?
I barely speak English ! o_O
 
spantini said:
Now that intrigues me. Stacking multiplies - does not add. This makes me think of volume increases on a home stereo: turning the knob halfway up doesn't mean 50% volume. I forget the general formula for that stuff, been years since I had my hand in.
Click to expand...
Volume controls can be made with different tapers, so halfway up on one isn't necessarily the same number of dB (from WFO) as halfway up on another. Some home audio makers would build in a steep ramp at the beginning of the rotation to give the impression that a receiver was more powerful than others in a store demo display.

Generally speaking, a 2:1 power ratio is a 3 dB difference in SPL, as long as speakers aren't being driven near their limits.
 
It won't add like that.

Compressing something at 4:1 at a specific threshold at instant attack and release (for discussion purposes), will leave you with a wave that has 1/4 the dynamic range of the original above the threshold.

Now, if you take that processed wave and run it through the same compressor with the same settings, the resulting wave would have 1/4 the dynamic range of the processed signal above the threshold.

So, if you had a wave with 8db peaks above the threshold, you would send it through the first compressor and end up with 2db peaks above the threshold, then when that goes through the second compressor and end up with .5db peaks above the threshold.

However, if the threshold is different, or the attack and release are different, that complicates things a lot.
 
You must log in or register to reply here.
Back
Top