Confused about Compression

[KingTuck]

I'm sure this is a dumb question. Hopefully someone can help me clear it up

Okay, so let's say I'm trying to compress a sound that has a pronounced transient like most percussive sounds. Let's say, for the sake of example, a snare drum. The idea is to have the compressor let the transient pass through, and compress the parts of the signal after the transient.

So the attack on the compressor is set up to let the transient pass through, but not so long that it leaves the entire signal uncompressed. Well, this is where my fundamental confusion comes from. With the transient left untouched in the signal, and the compressor jumping in to attenuate the volume of the sustain afterwards so it's more in line with the release/tail of the snare drum, there's now a larger gap between the peak of the transient and the sustain of the snare drum, correct?

Just to play with numbers. Let's say that the transient is something we'll call LOUD 1. The tippy top of the signal in terms of volume. The sustain of the snare drum is LOUD 2. The tail is LOUD 3, the quietest part. The compressor is set up so the attack time lets it leave our LOUD 1 transient unprocessed, and kicks in during our LOUD 2 sustain. It applies gain reduction enough that LOUD 2 is now comparable/closer to our LOUD 3 tail in terms of volume.

Now however, the difference between the volume of the transient and the sustain and tail of the snare drum is greater than it was before. When applying make up gain to get the body of the snare to where it was before, I'm forced to drag up the more pronounced transient too. Forcing it louder than it ever was before.

I suspect I'm simply thinking of this wrong somewhere. If someone could help out I would be grateful.

[Keith99]

In this scenario there is no need for make up gain if the transient is not being compressed

[KingTuck]

In this scenario there is no need for make up gain if the transient is not being compressed

I'm confused. Wouldn't I just be making the whole thing perceivably quieter? If I'm pushing the sustain to be more in line with the tail, the whole thing will end up quieter if I don't apply make up gain.

In any case of compression where by the transient is left untouched, how will the compressed signal ever be louder than it was before? What am I missing here? Thanks for the help btw.

There's something about this I find it hard to understand thus far. If you leave the transient as it was, absolutely any gain reduction will only make the signal quieter, right? Indeed the sustain of an instrument might be more in line with the tail of the instrument, but when compared to the transient it will always be on the whole quieter. Now obviously this can't be correct otherwise compression wouldn't be useful for sounds with pronounced transients. What bit of info am I missing?

[Gamma-UT]

Psychoacoustics. The brain reacts to sudden loud noises. You don't have to sustain the loudness to gain the psychoacoustic effect - this is partially due to the way that the muscles around the ear tighten in response to loud sounds to protect the ear canal and partly conditioning. So, you can let a snare smack and then push the rest of the sound out of the way for something else to come in so you aren't filling up the spectrum with noise. The brain will perceive the snare to be ringing even though its tail will suddenly be much quieter.

Normally, you want to clip or limit the transient as well to gain the most 'smack' out of the sound and gain some loudness, so you use two compressors or a clipper and a compressor to do the job. The transient clipping stage is where you can push the loudness up (if you want).

[geroyannis]

KingTuck you are right, that's how compression works.

So like Gamma-UT said you can clip a part of the transient and then boost the whole signal.

Or, you could use parallel compression. Take a send of your track and compress it really hard with a fast attack and release to catch the transient. Basically make it sound squashed really bad. Then add some of this to your original track. So, you keep the transient from the original track but now you have a fuller body from the compressed track.
A compressor with a wet mix control can basically do the same.

[KingTuck]

What does it mean to clip a part of the transient? Do you mean push it up in a limiter so the transient becomes more in line with the body of the instrument?

I was messing around trying to compress a percussion sound I had earlier. No matter where I left the attack, the transient click'd. Not in a pleasant way, it produces the same audible effect as when a sample clearly doesn't end on a zero crossing. So I loaded up Transient Master and pushed down the attack of the instrument before it was being fed into the compressor. This got rid of the click. All in all it sounded pretty strong. Is this idea related to what you two are saying?

As I take it when I see images of before/after compression, the notable part is that the sustain of the percussion is much louder than it had been before. Presumably this necessitates the same concept you two are talking about?

Something I'm wondering about too here. If I pound down on a sample with a quick attack and quick release (mostly for experiment purposes), I can much more audibly hear the tail of the sample. It seems to be genuinely louder. Now there isn't any make up gain being applied, so I believe it is technically the same volume it would have been before. Am I right in thinking that? Is that effect truly a psychoacoustic phenomenon? If so, that's very cool. Thanks for the help you two!

[geroyannis]

What does it mean to clip a part of the transient? Do you mean push it up in a limiter so the transient becomes more in line with the body of the instrument?

Not with a limiter but with a clipper. I'm using the free GClip for that purpose.

Something I'm wondering about too here. If I pound down on a sample with a quick attack and quick release (mostly for experiment purposes), I can much more audibly hear the tail of the sample. It seems to be genuinely louder. Now there isn't any make up gain being applied, so I believe it is technically the same volume it would have been before. Am I right in thinking that? Is that effect truly a psychoacoustic phenomenon?

Pretty much.

[KingTuck]

Interesting. I don't suppose anybody knows of a place I can read up about this phenomenon?

Final question, hopefully , I checked out GClip and I'm a bit unsure how what it does is different from any limiter with an ultra fast attack time. What's it doing that a dedicated transient designer plugin or ultra fast limiter isn't?

[geroyannis]

Well, a hard clipper is not really a dynamics processor but rather a waveshaper, essentially distortion. Maybe this will help:
http://www.soundonsound.com/sos/aug13/a ... 0813-2.htm

[Gamma-UT]

A limiter or compressor turns the gain down until some time after the level dips back below the trigger threshold. A limiter is just a compressor with a ratio of infinity:1. It will not just reduce the peak volume but the volume of the tail for a certain amount of time afterwards, determined by the release value. With fast release times, you get pumping. A hard clipper simply lops the top off the transient without affecting the rest of the signal. How much gets lopped off depends on the threshold.

For drums and picked guitars, the clipping is often practically inaudible as the initial peaks are so short-lived the ear doesn't really catch them. This is why clipping often gets used during mastering as a loudenating process. But as you turn the threshold down, you hear the clipper's actions as it eats more into the signal.

Soft clippers behave more like tape in that they allow some signal above the threshold to pass through, but highly attenuated – similar to the action of tape saturation.

[KingTuck]

Hmmm, let's say for example I'm using the Fruity Limiter in FL Studio. I can turn the attack down to 0 (as in actually zero, not just ultra fast). The release time to zero. And put the lookahead on zero. When I do this aren't I technically doing the exact same thing? I don't use many limiters normally so I'm unsure if 0.00ms is something that's very uncommon and why people would say clipper =/= limiter.

I can certainly achieve that exact same kind of "harmonic distortion" when doing this, and keep the transient perceivably as it should be. I suspect this would fall under the hard clipping style. But I also feel I could set up the FL limiter to achieve the soft clipping output too.

[nineofkings]

A hard limiter with 0 attack and 0 release becomes a clipper.

[geroyannis]

A hard limiter with 0 attack and 0 release becomes a clipper.

I'm not quite sure. A limiter just drops the volume of the signal, while a clipper chops off the peaks, actually altering the shape of the affected signal's waveform.

[KingTuck]

A hard limiter with 0 attack and 0 release becomes a clipper.

Awesome. Thanks everyone for the help on this.

[KingTuck]

A hard limiter with 0 attack and 0 release becomes a clipper.

I'm not quite sure. A limiter just drops the volume of the signal, while a clipper chops off the peaks, actually altering the shape of the affected signal's waveform.

I think that a limiter with no attack, release or lookahead necessarily has to cut off the peaks (it doesn't have time to react any other way). In the example image you posted if the limiter is set up without the attack/release/lookahead, anything above the threshold will now have the exact same amplitude equal to the threshold. Necessarily distorting the waveshape in the way a dedicated clipper plugin would. There's no way for it to maintain it's exact original waveshape.

If I'm wrong, please correct me! Just how I've put together the logic of all this so far.