Compression and Phase Distortion

[ChuckU]

So I'm hearing "Compression can cause phase distortion. If you have a DAW, use fader automation to tame your levels, etc."

Well my attitude towards compression, (and this is mostly with plugins in Sonar as I don't have a decent hardware comp), has been if you don't know how to use a tool, then don't. Compression is something I try to be careful with as I still don't know what to listen for and how it helps. It's more obvious than it used to be but it's still a skill I need to develop.

Anyway, I just produced probably my best sounding drum track yet and to get the toms to have sufficient definition, I thought about compressing the overheads, so that the toms ringing in the overhead tracks would not muddy up the tom tracks. Damned if that didn't work the nuts. Listening back to the song, the toms were right up front and distinctive. But what I also noticed was the cymbals ever so slightly 'whooshing' in certain places. I identified this as phase distortion as different frequencies were getting squashed.

To minimize this, I played with some of the compressor parameters. Ratio I think did the most which makes sense, but some of the characteristics of the track that I liked were not as good.

So I'm wondering: is this a situation where multi-band compression would help? If so, can anyone recommend a plug-in and settings?

[Ethan Winer]

Chuck,

> So I'm hearing "Compression can cause phase distortion <

Not true. Maybe if you process only one side of a stereo mix you'll have timing issues, but any unwanted side effects of compression are related to changes in volume, not phase shift.

> I also noticed was the cymbals ever so slightly 'whooshing' in certain places <

This too has nothing to do with phase. You're merely hearing the effect of the compressor changing the volume, probably too quickly because the release time is set too short.

> is this a situation where multi-band compression would help? <

You'll do well to truly learn how to use regular compression first. See my short article that explains the basics, 8th in the list on my Articles page:

www.ethanwiner.com/articles.html

--Ethan

[ChuckU]

Thanks Ethan. I think I already downloaded this some time ago but it bears another look. Understanding the content intellectually and actaully applying the theory are worlds apart unless you can actually hear it and I think I'm starting to do that.

[masteringhouse]

Originally Posted by Ethan Winer

Chuck,

> So I'm hearing "Compression can cause phase distortion <

Not true. Maybe if you process only one side of a stereo mix you'll have timing issues, but any unwanted side effects of compression are related to changes in volume, not phase shift.

--Ethan

In the case of digital compression you could be introducing latency issues that when mixed with other tracks will cause a phase shift. The compression per se is not causing the phasing issue it's the latency, but regardless the result is the same. You can also get this problem with just about any other plugin as well.

Pro Tools has a time adjuster plugin to compensate for this on their TDM systems, without this you're kinda stuck.

So in theory it would be better to use automation, in practical usage this is next to impossible for the application you're talking about since you would be spending hours making the adjustments. On a vocal or guitar solo this makes more sense.

Pick you're evil and don't look back ...

[Ethan Winer]

Tom,

> you could be introducing latency issues that when mixed with other tracks will cause a phase shift. <

Yes. Of course this is time delay, not phase shift, and as you said on some DAW platforms it happens with any plug-in. My main point was to address the incorrect assumption that a "phasey" sound is phase shift because it's not.

Also, I use Sonar which has always included sample-accurate delay compensation, so I never even think about that. Indeed, the fact that some DAWs do not correct for that amazes me.

--Ethan

[masteringhouse]

Originally Posted by Ethan Winer

Tom,

> Yes. Of course this is time delay, not phase shift, and as you said on some DAW platforms it happens with any plug-in. My main point was to address the incorrect assumption that a "phasey" sound is phase shift because it's not.

--Ethan

I guess it's a question of semantics, but for me small incremental time variations such as that introduced by latency is a phase shift (AKA Linear phase distortion). If it's really bad (greater than the period of the sound wave) then it can start to appear as a delay.

Nevertheless I feel that this is an important point to bring up as may people new to digital audio do not recognize this and it can really degrade the sound of a mix, in particular with any instrument that uses multiple miking such as drums.

I would suggest that people try the following experiment in their DAWs.

1. Create a duplicate copy of a track.
2. Add a plugin to one of the tracks, but do not process the track. The idea is to check for time delay/phasing problems, do not bypass the plugin, just insert it.
3. Listen to the sum of the 2 tracks. If it just sounds 3 db louder than the original you're good, if not, then your DAW does not automatically compensate for latency.

Originally Posted by Ethan Winer

Also, I use Sonar which has always included sample-accurate delay compensation, so I never even think about that. Indeed, the fact that some DAWs do not correct for that amazes me.

Good to know, is there more info on this? Since plugins introduce varying amounts of latency how does Sonar handle this?

[Ethan Winer]

Tom,

> for me small incremental time variations such as that introduced by latency is a phase shift <

The difference between time delay and phase shift is that phase shift yields different delay times for different frequencies. Plain time delay is the same (number of millseconds) for all frequencies.

> Since plugins introduce varying amounts of latency how does Sonar handle this? <

I believe/assume Sonar provides a buffer of known length, then fills it with data which the plug-in processes. Some plug-ins can change the buffer length if the DAW supports that - for example to allow room for the addition of a reverb tail that extends past the end of the song - but I'm sure the new length is relayed to the host program.

--Ethan

[masteringhouse]

Originally Posted by Ethan Winer

The difference between time delay and phase shift is that phase shift yields different delay times for different frequencies. Plain time delay is the same (number of millseconds) for all frequencies.

--Ethan

Ethan -

What you are describing is the difference between non-linear and linear phase distortion respectively. When a sound is delayed by an amount less than the period of the frequency it's phase distortion regardless if all frequencies are affected in an equal amount, or if it's frequency dependent. This is one of the differences between using FIR and IIR filters.

Originally Posted by Ethan Winer

I believe/assume Sonar provides a buffer of known length, then fills it with data which the plug-in processes. Some plug-ins can change the buffer length if the DAW supports that - for example to allow room for the addition of a reverb tail that extends past the end of the song - but I'm sure the new length is relayed to the host program.

--Ethan

Buffering allows for the tracks with the least amount of latency to be delayed by the amount required for the high latency tracks to "catch-up" but this still doesn't explain how the delay times are calculated. Since this varies by plugin, the DAW will need to know the exact amount of latency produced by each plugin beforehand and delay all of the tracks appropriately.

Is this technique explained anywhere?

[Ethan Winer]

Tom,

> this still doesn't explain how the delay times are calculated ... Is this technique explained anywhere? <

Beats me. I have seen it discussed now and again in the Sonar users forum at www.cakewalk.com. But usually just the fact that Sonar does this and some other DAWs don't, more than the way it's coded or how the data is organized. Some of the senior Cakewalk support guys frequent that forum and answer questions, so you could probably get the real answer there.

--Ethan

[masteringhouse]

Originally Posted by Ethan Winer

Tom,

> this still doesn't explain how the delay times are calculated ... Is this technique explained anywhere? <

Beats me. I have seen it discussed now and again in the Sonar users forum at www.cakewalk.com. But usually just the fact that Sonar does this and some other DAWs don't, more than the way it's coded or how the data is organized. Some of the senior Cakewalk support guys frequent that forum and answer questions, so you could probably get the real answer there.

--Ethan

Thanks Ethan!

Apparently Pro Tools 6.4 has this feature now as well. The following document gives a few clues on how this works (at least for PT).

http://www.digidesign.com/support/re...Win_ReadMe.pdf

Apparently the plugin reports the latency delay to the application and the application adjusts for it in the I/O buss.

Good stuff!

[Ethan Winer]

Tom,

> Apparently the plugin reports the latency delay to the application and the application adjusts for it in the I/O buss. <

Yeah, that makes sense.

I was pretty surprised when I first learned that some DAWs don't do this because with Sonar I've always just taken it for granted.

--Ethan