Compression in Audacity

I’m new to compression for ACX compliance and other purposes and am trying to understand the basics so please help relieve my ignorance.

Compression based on RMS vs. on peaks:
The online manual says compression based on RMS is downward while compression based on peaks is upward. I started with a mono track with ACX check yielding overall peak -5.3 and overall RMS -31.0, duplicated it twice (within same project), and compressed both copies with threshold -12.0, compressing copy A based on RMS and copy B based on peaks. ACX checks on the copies after compression yielded: for A, peak -6.0, RMS -31.0; for B, peak -2.1, RMS -23.9.
– Questions:

  1. Why didn’t compressing copy A based on RMS change the RMS level? Is this because the starting RMS of -31.0 was below the -12.0 threshold so didn’t trigger compression? If so, why did the overall peak level decline from -5.3 to -6.0?
  2. Was the pre-to-post level increase in copy B (compressed based on peaks) attributable to the fact that the initial peak of -5.3 exceeded the -12.0 threshold?

Make-up gain:
The make-up gain option seems to allow gain only to 0db.
– Questions:

  1. Is any other target db level available, and if so how set it? (ACX wants peak and RMS levels below 0db so it’s hard to believe Audacity wouldn’t offer ACX-compliant make-up gain.)
  2. If using make-up gain, is the gain based on RMS or peak level? Or does this depend on which of these was used as the basis for compression? (Make-up gain based on RMS would seem to yield peaks exceeding 0db and thus clipping, which wouldn’t be good.)

Compression vs. Limiting:
The compression and limiting effects seem very similar. Is the key difference that the latter works only on maximum rather than on both max and min levels? If not so, please clarify.

Thanks in advance.

The Audacity compressor is rather quirky (not a ‘typical’ compressor, if there is such a thing).

It did, but by less than 0.1 dB.
‘A’ started with a peak level at -5.3 dB (so the highest peak(s) were -5.3 dB), and the compressor pushed down those peaks a tiny bit (0.7 dB) to -6 dB. The amount of compression is so small that the ‘average’ over the entire track is insignificant - possibly a change of -31.03 dB to -31.04 dB.

RMS (root mean square) is a kind of average. When you look at the waveform in Audacity, you may have noticed that the inner part of the waveform is pale blue, while the peaks are darker blue. The pale blue region represents the RMS level, where the level is averaged over several cycles of the waveform. You can see that the average changes during the course of track, generally being higher during the louder parts. It’s like a rolling average, where the averaging window is fairly small. In a few places, that average slightly exceeds -12 dB.

On the other hand, the Contrast tool measures the RMS average across the entire selection (the whole track, assuming that you have selected the entire track). The overall average can be expressed as a single number - in this case, -31 dB.

Yes.


The “make-up gain” control is badly named. It’s really just “normalizing” to 0 dB (amplifying so that the peak is 0 dB).

No, it just amplifies to 0 dB, though you can of course apply the Amplify or Normalize effects after.

Audacity isn’t made specifically for Amazon, and Amazon do not contribute financially, technically or in any other way, to the development or support of Audacity.
Audacity is a multi-track audio editor, used by millions of people for all sorts of things, including copying vinyl records to digital format, recording live music, producing musical compositions, monitoring seismic data, analyzing wildlife sounds, and many other things. Audiobook production is one of the more common niche uses, but still just a niche.

Yes, it’s “peak normalization”.

Yes they are. A “limiter” is really a particular kind of “dynamic compressor”.

The main differences are:
1)

  • Limiters are generally used for limiting excessively high peaks.
  • Compressors are generally used for reducing dynamic range.
    Note that if you limit peaks, then you are reducing the dynamic range, and if you reduce the dynamic range, then you are reducing the difference between high peaks and the average level, so really they are very similar in what they do, but the emphasis in why you do it is slightly different.

Limiters generally act much faster than compressors. Both limiters and compressors are kinds of “automatic level adjustment”, where the “gain” is adjusted according to the input level. Compressors generally adjust the level quiet slowly so as to achieve a smooth adjustment of the level so as to reduce the range between loud parts and quiet parts. Limiters on the other hand tend to react very quickly to peaks, so as to reduce high peaks without affecting much else.

Thanks, that’s very helpful. If I understand you right, to adjust the track’s overall peak to, say, -6.0 rather than 0.0 after compression, should I use the separate Normalize effect rather than the Compression effect’s make-up gain option?

Relatedly, for editing a mono voice-over (spoken word) track, is the preferred sequence Compress then Normalize? If the opposite holds, or if the preferred sequence depends on other things, please clarify.

The first thing you should know is that compression takes decades to truly master its effects. Many will disagree with me but I really do not care. When you compress for the ACX ,you will expand your noise floor. So if your NF is not around -130dbs before compression, then you are cutting your own throat. This is mathematics. You will also want to make all your cuts using EQ before you do any compression.

If you truly do understand how compression works then you will never exceed a 3 to 1 ration nor exceed a -6db reduction, as this will add distortion. If you can can not meet the specs with out using compression, then your set up needs to be evaluated, as far as meeting the ACX specs. If you want to make the track sound fuller or fatter, then you can do this with side chains by adjusting each track.

Audacity wouldn’t offer ACX-compliant make-up gain

Audacity does publish an ACX correction suite of tools.

Also included is ACX-Check, a tool to measure success.

If you start with a reasonable recording in a quiet room, it should be possible to create AudioBook chapters without a big fuss. The quiet room is usually the problem.

We also publish a technique for you to post a sample voice clip for evaluation here on the forum. Record your voice clean with no corrections. We’ll see how much work we have to do to make it ACX compliant.

http://www.kozco.com/tech/audacity/TestClip/Record_A_Clip.html

Quick note, you do have to be able to read. There is no acting filter or correction.

Let us know.

Koz

We can tell you what the tools are doing and why they work. That’s not posted in the document.

Koz