KVR Audio

greendoor · Post by **greendoor** » Sat Oct 29, 2005 11:44 am

I expect this to be contraversial, but i've found a way to test a compressor plugin to see what it's actually doing:

Take an audio track, and clone it. Phase invert the copied track, and they should cancel out perfectly to zero.

Now - insert a compressor on either track. With a 'good' compressor, you should be able to set the threshold at 0, and the output gain at 0. This means that the compressor is doing nothing - what goes in is what comes out - undamaged. In other words - you still hear perfect silence, because the two tracks are identical, just phase inverted.

With many compressor plugins I have, this is exactly what happens. As you lower the threshold, the compressor starts to work (with a ratio >1:1). So the the peaks of audio that are being compressed are now different in level from the original track, and therefore instead of hearing silence, you hear the parts of the audio that are being compressed.

This is a great way to hear the subtle differences in compressor parameters. Instead of being confused by hearing the whole processed sound, you are just hearing the differences between the original and the processed sound.

BUT - i've been surprised by some compressors that are not able to do this phase cancellation trick. That means that these compressors are 'damaging' the sound below the threshold.

I am very pleased with Kjaerhus compressors - these don't damage the signal when not compressing.

I was surprised that the Ultrafunk Sonitus compressor damages the signal - unless the Limiter is switched on.

All Voxengo compressors damage the uncompressed signal! I found it impossible to set any of my Voxengo compressors so that the signal going in was exactly the same signal coming out. Even Marqius set on 'Clean' - with the threshold at 0, ratio at 0, makeup gain at 0 - according to the meters the compressor is doing nothing, but it is impossible to phase cancel the two signals.

This means that Voxengo compressors are not capable of "first doing no harm". That doesn't mean they are bad, or that they don't sound very good. But it does mean that they are very coloured and NOT transparent, ever.

Kjaerhus would seem to have the best transparent compressors (by this test).

I've often wondered if some Voxengo plugins have some hidden filtering built in to sweeten the sound. I would prefer to apply my own eq or filters and have control over what I apply. I prefer a compressor to simply compress - and to do zero harm to the signal when not compressing.

This is easy to test for yourself - I would be interested in comments from people who have actually done this.

soulata · Post by **soulata** » Sat Oct 29, 2005 11:59 am

great post. I'll do that with the compressors I have.

k

SSL · Post by **SSL** » Sat Oct 29, 2005 12:06 pm

Or you can simply get a proper monitoring system instead.

diverdee · Post by **diverdee** » Sat Oct 29, 2005 12:10 pm

SSL wrote:Or you can simply get a proper monitoring system instead.

errrrmmmmmmmmm......

ow much ee it?

[edit] good call greendoor, on the test [/edit]

fritzman · Post by **fritzman** » Sat Oct 29, 2005 12:14 pm

greendoor wrote:I expect this to be contraversial, but i've found a way to test a compressor plugin to see what it's actually doing:

Take an audio track, and clone it. Phase invert the copied track, and they should cancel out perfectly to zero.

...

This means that Voxengo compressors are not capable of "first doing no harm". That doesn't mean they are bad, or that they don't sound very good. But it does mean that they are very coloured and NOT transparent, ever.

Hi greendoor,

interesting test. If all that you're after is that a compressor is mostly transparent when the parameters are set to do nothing then fine.
I'm not interested in a compressor that does "nothing" to my sound. I want every tool in my chain to improve the sound to my taste.
Do you hear these differences when actually using the compressors you where talking about? Do the ones that are not transparent according to your test really sound bad when you actually use them? Blind test is the right key word here ...

Best wishes, FRitz

Killbox · Post by **Killbox** » Sat Oct 29, 2005 12:14 pm

Really great topic!!! But i have one note.
With this trick you're not hearing the compressed signal, but the uncompressed, where the compression works

you can't hear the compression, cuz it has lower levels

Thank you again for this topic!!!

bduffy · Post by **bduffy** » Sat Oct 29, 2005 12:17 pm

Well, I think there are two things to consider here: yes, there are some analog/phase algorithms going on, but I think what you're hearing is latency introduced by the plugin. Probably necessary to process the analog algos. I noticed that when I used Voxengo Marquis on a drum bus that it flanged with the dry signal...so I'm not sure why I hear it with that and no other plugins. I assume it's the buffer latency it introduces. I don't if that's a bad thing, depending on usage.

Killbox · Post by **Killbox** » Sat Oct 29, 2005 12:31 pm

In my ears and eyes it looks like it's not only from the latency...
What will Alexey say?

It is interesting to me.

Aleksey Vaneev · Post by **Aleksey Vaneev** » Sat Oct 29, 2005 2:09 pm

greendoor, 'doing harm' is not used correctly. Most compressors usually do what you are describing - working as a 'wire' when no compression takes place. But this means they do not work as they should in my quality standards. Every compressor should add 1st order DC filter since compression adds DC offest. This DC filter automatically removes 'wire' behaviour.

You should try these tracks with analog and especially valve equipment - none work as 'wire'.

Please do not make wrong conclusions based on the fact discovered by greendoor. I regret he used word 'harm'. He should have used word 'adjustment' instead.

martian · Post by **martian** » Sat Oct 29, 2005 2:16 pm

BUT - i've been surprised by some compressors that are not able to do this phase cancellation trick. That means that these compressors are 'damaging' the sound below the threshold.

im confused, why would this mean they are 'damaging' the sound?

besides try your phasing test with something like the sonalksis eq, cancells out, but actually their is audiable change.

so your test is flawed from all angles.

listen instead.

martian · Post by **martian** » Sat Oct 29, 2005 2:17 pm

double.

sonicvent · Post by **sonicvent** » Sat Oct 29, 2005 2:23 pm

That means that these compressors are 'damaging' the sound below the threshold.

There's also knee to consider. Thresh @ 0 and a -6db knee will still compress. Great thread! thx greendoor

csl · Post by **csl** » Sat Oct 29, 2005 2:36 pm

harming the sound? many compressors are coveted exactly because of the character they impart, software or hardware. affecting the signal path when not compressing is hardly destructive, plus it's not as though you'd ever use one in this state! you can't win can you? in the early days compressor vsts would get stick for being lifeless and without character, and now we have component-modelled compressors that get stick for not being sample-by-sample transparent.

if it sounds good when its doing it job, then it is good.

DaveB · Post by **DaveB** » Sat Oct 29, 2005 2:40 pm

Simple "black box testing" cannot measure the quality of a complex compression algorithm. What one person calls damage another will call character.

There's only one way to measure compressor quality - use your ears. Does it sounds good? That's really all that matters.

Good thread though!

Dave
www.db-audioware.com

autloc · Post by **autloc** » Sat Oct 29, 2005 3:31 pm

I really like a good objective measure of quality. I've appreciated when people have analyzed various samplers' resampling algorithms because I know, at least for me, that the aliasing illustrated by an STFT is well inversely correlated with my perception of a resampler's quality.

I don't see that transparency at null settings is terribly relevant in a compressor, though, and given that many may introduce latency (at least for real-time computation), mixing the wet signal in with an inverted dry signal seems pretty irrelevant. Well, not all irrelevant - it may reveal some (limited) details about the algorithm being used - but it is not a useful indicator of quality.

How to test a compressor plugin ...