What do extreme compression ratios do to the amplitudes?

[soundmodel]

What do extreme compression ratios (such as 50:1 or 100:1) do to the amplitudes?

Because I perceive that limiting is just a bit more aggressive form of compressor. However if e.g.

100:1

reads that

whatever goes over threshold is reduced to 1/100th in output.

However,

I find it difficult to picture, what the output of an entire sound file would look like.

[Mayae]

What do extreme compression ratios (such as 50:1 or 100:1) do to the amplitudes?

Because I perceive that limiting is just a bit more aggressive form of compressor. However if e.g.

100:1

reads that

whatever goes over threshold is reduced to 1/100th in output.

However,

I find it difficult to picture, what the output of an entire sound file would look like.

Do you have an oscilloscope? If not, get one. Otherwise, bounce a track and look at the waveform..

[aciddose]

Because I perceive that limiting is just a bit more aggressive form of compressor.

That is true.

A limiter is designed to respond quickly (or even "look ahead" in time by delaying the input signal) and provide an absolute limit to amplitude by using an infinite ratio.

The ratio is simple. If you have a threshold of 0 dB and the input signal is +20 dB, an infinite ratio (1 - 1/inf) will apply -20 dB to the signal and the output will be 0 dB.

With a ratio of 1/2 the signal will be reduced by -10 dB.

With 1/8th you'll get -17.5 dB.

So all you need to do is:
output = (1 - ratio) * (threshold - input)

For example:
threshold = 0 dB
input = 6 dB
ratio = 1/2
output = (1 - 1/2) * (0 - 6) = -3.0 dB

[Vertion]

What do extreme compression ratios (such as 50:1 or 100:1) do to the amplitudes?

Because I perceive that limiting is just a bit more aggressive form of compressor. However if e.g.

100:1

reads that

whatever goes over threshold is reduced to 1/100th in output.

However,

I find it difficult to picture, what the output of an entire sound file would look like.

It sounds like this:

https://www.youtube.com/watch?v=zLa5sOkyzqs

[soundmodel]

I think I did some thinking:

Extreme ratio such as 100:1 with very low attack and (not too low) release:

So if everything above threshold goes to 1/100th above threshold, then the gain reduction is first of all drastic. But secondly, if the threshold is very low, then e.g. a signal going 100 dB over threshold
would need to be reduced to 1 dB over threshold.

Therefore, suppose the threshold is very low, attack and release times are set as mentioned before and the ratio is e.g. 100:1.
Then since 16-bit audio has only 96 dB of dynamic range and what we specified is basically a very fast (and extreme) limiter,
then ~everything above the threshold would be reduced to 1 dB over threshold.

So what one would expect to see in output is 16-bit audio with approximately unity gain (perhaps with small variations that occur between attack and release envelopes entering and exiting). I.e. that there’s barely no dynamic range at all, but everything has been ”normalized” to the same gain.

[Gamma-UT]

If you simply tried it out with a 'scope plugin as suggested above, you're going to do a lot better. Because, for one, you'll probably quickly find out attack and release often don't give you what you think you're going to get - not least because there's no hard and fast definition of attack and release used in the industry (though there are common rules of thumb).

In any case, with both low attack and release you might be surprised at the (clicky) results. With low attack and a longer release, there will be less of a surprise – welcome to flat-top audio waveforms.

[soundmodel]

Oh yeah, but I do remember from practice that low attack and low release tends to produce distortion, rather than "clean" limiting.

[Richard_Synapse]

So what one would expect to see in output is 16-bit audio with approximately unity gain (perhaps with small variations that occur between attack and release envelopes entering and exiting). I.e. that there’s barely no dynamic range at all, but everything has been ”normalized” to the same gain.

What you describe is typically the effect of a digital brickwall limiter (which works entirely different). An analog-style compressor/limiter does not crush the dynamic range completely, in fact it can have the opposite effect, e.g. Daft Punk style pumping.

Richard

[soundmodel]

Can you elaborate on that?

How do brickwall limiters differ from analog-style compressor/limiters?

[Richard_Synapse]

Classic compressors/limiters look at the current level (possibly averaged over some time, digital compressors may use a peak latch which is where sliding min/max algorithms come into play), then adjust the level to the compression curve, smoothed with the Attack and Release parameters to avoid distortion.

A brickwall limiter digitally limits to exactly 0dB via a lookahead window, typically around .5-1ms. It is not based on or inspired by analog electronics in any way.

Richard

[Miles1981]

brickwall = attack = 0ms, usually peak (because you want the peak to be limited, not the RMS)
analog = attack > 0ms

[soundmodel]

So does a brickwall limiter cut off the top of the waveform? If there's no envelope time to "shape" the peaks?

So theoretically doing low threshold extreme brickwall limiting would produce some sort of pulse wave as an output?

[nix808]

yeppy-
but it does this: \_/

otherwise hard clipping causes noise, so it cuts it square, but a transition/soft curve period is requisite or some audio will act like like it's clipping, and sound wrecked

[Miles1981]

I said attack, not release.

And no, it wouldn't produce pulse like PWM (as in a binary signal). It would produce something really low in sound and heavily distorted. Even if a long release, you would end up with something like the original signal increasing in amplitude and then squashed again.

[Richard_Synapse]

So does a brickwall limiter cut off the top of the waveform? If there's no envelope time to "shape" the peaks?

There is time, the lookahead window of say 1ms. This is why Brickwall Limiters work so well, and beat classic compressors by an order of magnitude if the goal is to maximize loudness.

Richard