Crossfades not working correctly

[m_bryant]

For the past few days, crossfades in audition haven't been working correctly. Instead of a smooth transition between two clips, I get a volume boost in the middle of the crossfade.

To demonstrate the issue, I've inserted a clip into a new multitrack session, added a split in the middle, and overlapped the two clips, as shown below.

This is a clip crossfading into an identical clip, so in theory, when it's mixed down it should be identical to the original clip. However, when you compare the original with the mixdown, you can see a volume bump where the crossfade is

Original:

Mixdown:

Any idea what would be causing this, and how to fix it?

[ryclark]

Have you had your crossfades working OK previously as you seem to imply by saying 'for the past few days'? Did something change between the times when crossfades worked OK and now that they don't?

[m_bryant]

Yes, they've always worked previously. I can't think of anything that has changed between then and now (although I assume something must have).

[ryclark]

You will always tend to get a bit of a lift in level if you are crossfading absolutely sample identical audio unless you are very careful with the crossfade curves. Normally a crossfade is expected to be used with audio clips that aren't actually identical so the addition of the two won't give a boost in level. You can change the type of crossfade curve from the Clip menu Fade In and Fade Out selections.

[SteveG_AudioMasters_]

I checked the behaviour in the present release against the behaviour in Audition 3 - they behave identically. The size of the blip, and whether it is louder or quieter at the crossover point rather depends upon the length of the crossfade.

[m_bryant]

Like I said, this is happening with all crossfades, not just identical clips. Using identical clips was just to illustrate the problem. The exact same thing happens with every single automatic crossfade. This has never happened before in the years I've been using Audition. The normal behaviour for automatic crossfades is to automatically create a crossfade curve that keeps a constant volume (assume the two clips are of roughly the same volume of course).

Here's the same thing happening with a crossfade between two separate clips of generated white noise:

No matter what clips I use, there is consistantly a rise of about 2 to 3 dB in the centre of the crossfade. The only way I've found to get a crossfade with a consistant volume is to manually set the crossfade to symmetrical and set the fade in/out value to 0... i.e. a straight-line crossfade:

Under normal circumstances, a straight-line cross-fade should actually result in a slight dip in volume, so clearly something is going wrong.

[SteveG_AudioMasters_]

No, there's nothing wrong at all. This isn't a simple thing to explain, but it all has to do with how 'coherent' the sources in your crossfade are, and whether you are considering the intensity or pressure of them, and the way energy is distributed. If you want to know how complicated this really is, have a look at this: http://www.sengpielaudio.com/calculator-coherentsources.htm. If you don't get that (and it wouldn't surprise me if most people didn't!), then try an experiment in audition doing exactly the same thing, only using pink noise and a linear crossfade value of 19 (the default). You should find then that there is no blip, whatever you do; the values as they are represent the best crossfade settings for real-world energy-distributed signals. But if you do this with tone, you'll get some very strange results, depending upon whether the tone values add in phase or not, so results will vary according to where you position your crossfade. What you're trying to eliminate is the effects of a dB addition, and in fact just a simple thought experiment should show you why, at any point on the crossfade, you have to be able to match these exactly, whilst removing the correct amount of energy from the signal. If you do your last crossfade (the straight-line one) with pink noise, you'll get a dip, because you won't have kept the 1/n noise level constant.

And despite what you think, this has always been the same in Audition, going right back to Audition 2, when it was introduced.