The problem with parallel compression that you didn't know you had. What it sounds like and how to fix it.

If you're parallel compressing, and you should be because it's such a useful effect, then you have a problem that you need to consider. If you don't then your mix won't turn out the way you expect it to.

I've written two articles on parallel compression recently - How to automate tracks that have parallel compression and What does parallel compression on vocals sound like?

The first describes the problem I'm talking about, and the second demonstrates parallel compression with audio clips.

But I haven't actually demonstrated the problem yet, so I'll do that in a moment. But first a recap of the problem.

Let's suppose that you parallel compress like this (click the image for a larger view)...

This is one of the standard ways of parallel compressing. Another is to duplicate the track rather than using an aux but that method is not problem-free either and I'll cover it on another day.

What we can see is that there's a vocal on one track. That track is bussed to an aux track which is compressed and blended in with the vocal. The result is parallel compression that brings up the lower levels while leaving higher levels mostly unchanged. It sounds nice.

The problem briefly is that if you want to change the level of the vocal, then the parallel compressed version will sound different as well as being different in level. This is because you've changed the proportions of uncompressed and compressed signal. This applies whether you use a post-fade or pre-fade send.

If you don't realize this, then you're not in full control of your mix, and that's important.

If on the other hand you are aware of what's happening, you can make whatever adjustments that are necessary, or make an informed judgment to leave things as they are.

So what's at stake here?

Well, I have two examples. In one I have an audio clip made with the settings shown in the screenshot above. In the other, I have lowered the level of the track named Vocal by 10 dB, with a post-fade aux send. I have then corrected the levels so that both examples are at the same level, as measured by the Waves WLM Plus loudness meter. So here goes...

Vocal at 0 dB (Example 1)...

Vocal at -10 dB, level corrected (Example 2)...

Can you hear the difference?

If you can then either you have very sharp ears, or you are experienced in audio and you're attuned to such small changes, many small changes adding up to big differences over the course of a full mix.

But I can demonstrate more clearly...

In the next example I have put Example 1 in the left channel and Example 2 in the right. You should listen on headphones.

What you will hear is that the stereo image shifts according to the relative levels of the signals. The image shift correlates to differences between the two signals, demonstrating that the two examples are indeed different. So lowering the level of the vocal track has changed the sound of the parallel compression.

If you are just changing the fader level, then you can either ignore the difference, or compensate for it. But if you are automating the fader, then the sound of the parallel compression will change as the song plays through. Clearly this is a problem that needs to be resolved.

OK, so now you know about this potential issue, take a look at How to automate tracks that have parallel compression and you'll see one way of handling this.

Audio - Such fun!