For years I've noticed crossfades between two wavs don't behave predictably in Reaper (and presumably other DAWs).

Reaper defaults to a curved crossfade which I always assumed was designed to account for the logarithmic scale of loudness so the crossfade was inaudible. Generally this works fine, and in these cases switching to a linear X-fade then creates an audible 'dip' in the edit.

However, I've noticed some crossfades have an audible 'swell' and actually need switching from the default curve to a linear fade to work seamlessly.

I always thought the logarithmic shape was foolproof, but is it programme-dependent?

Certainly I've found particular instruments (e.g. pianos and organs) to be more fussy about the length of the fade, but this is a different issue!

In my experience, it's always program dependent.

I deal with this quite a bit looping samples, for instance. With complex material (mixes, loops, harmonically dense material), usually an equal power crossfade is the right choice, with a linear crossfade resulting in a level dip as you describe.

With simpler material, other crossfade types might work better - in short, I usually start with the crossfade type I expect will give the best results on the material I'm working on (based on past experience), and if it's not good, then I will try other types in order to get the results I need.

I think this is pretty normal with audio - there isn't a 100% successful recipe you can always apply, just a good set of ground rules, and ability to explore further should your ears tell you it's not sounding good.

An equal power cross-fade (given equal level signals) will have a gain of 0.707 (square root of 2) at the mid-point, whilst a linear one will have a gain of 0.5 at the mid-point, which will result in a dip.

Of course if one part of the fade is tailing off and the other is fading up in the fade area, it's always going to be hard to chose a fade type that's not noticeable.

I should clarify that one recurring example is a set of crossfades in continuously recorded material where I want to isolate a section for repair. In these examples, the equal power fade nearly always sounds like a bump, and yet surely this is the one time where it should sound seamless, because the two 'takes' are actually the same continuous take?!

It should. Unless the equal power fade isn't being implemented correctly in the software.

You could try out on some simple waveforms - e.g. a sine or triangle wave and see whether they fade correctly. Though if the two bits of audio aren't 100% in phase, then I'd imagine you could still get dips in the middle.

??? I have never heard a crossfade in any recording I have ever edited.

I use Samplitude / Sequoia - if I can hear an edit I adjust the edit points until I can't hear it at all.

It's not a symptom I recognise from my cross-fades in reaper I have to say. What duration of fade are we talking here?

I've not noticed a problem either - although the default setting on older versions of Reaper was to do a quick fade at the start and end of each clip rather than crossfade so sometimes I'll get a glitch due to that if I use an old template. I'll normally place edits just before a transient but that's a hangover from the days of tape.

If I want to repair something in another program (like RX or Audition) I'll work on the whole file rather than just a piece of it.

mjfe2 wrote:I should clarify that one recurring example is a set of crossfades in continuously recorded material where I want to isolate a section for repair. In these examples, the equal power fade nearly always sounds like a bump, and yet surely this is the one time where it should sound seamless, because the two 'takes' are actually the same continuous take?!

This condition is the one guaranteed situation where you should use the linear cross-fade option. It is exactly akin to the angled cut when editing analogue tape! You have identical and time-aligned (coherent) material on both sides of the cut, so at the middle of the cross-fade the level of each half should be -6dB to ensure the summation of two identical signal halves is the same level as the original, increasing/decreasing linearly pro-rata either side of the centre of the crossfade.

Where you are editing two non-identical pieces together (eg different takes of the same musical piece) then although the linear cross-fade might work in some cases, it probably won't... And that's when the other fade curve options (and tweaking the level of incoming vs outgoing material) becomes more necessary and appropriate.

The constant power curve option, which gives -3dB attenuation to each part at the midpoint of the cross-fade, is often a good place to start as it will maintain consistent signal amplitude through the cross-fade with most similar but non-coherent material.