[Ardour-Users] Slice marks makes noise

[Adriano Petrosillo]

On Sun, 28 Oct 2018 13:04:23 -0400
Paul Davis <paul at linuxaudiosystems.com> wrote:

> On Sat, Oct 27, 2018 at 9:55 PM Al Thompson <althompson58 at gmail.com> wrote:
> 
> > But they do in a sine wave, which is the example given.
> >
> 
> They don't exist in the overwhelming majority of sine waves.
> 
> A true "zero crossing edit" requires that there is a sample of value zero
> where the edit/cut can be applied.
> 
> Almost all sine waves will instead have a pair of adjacent samples slightly
> and slightly below zero. How "slightly" depends on the sine wave frequency
> and the sample rate.
> 
> If you cut at one of the adjacent samples, you are actually creating a
> discontinuity in the output, the audible impact of which will depend on a
> variety of factors. Obviously, the if chosen sample's value is very close
> to zero, then the effect will be fairly similar to the "true zero crossing"
> case.
> 
> Apply this to real world audio, with much more complex signals than a
> single frequency sine wave... the results of the "zero crossing edit" not
> actually using a zero value are much more likely to be an audible
> discontinuity (aka "pop" or "click").
> 
> This is why Ardour doesn't bother with attempting to "find a zero
> crossing", and instead applies a predictable and controllable fade on each
> side of the cut. Even this cannot be guaranteed to work - there is always a
> hypothetical pathological case where it fails - but it is much more
> reliable than the "zero crossing' woo that somehow made into digital
> editing lore some years back.

Technically, finding the closest value to zero and applying a short fade there could minimise clicks though. The fade will progressively decrease sample values until, while not zero, they will become close to zero, and if they're already close to zero, they will/might be practically inaudible. This *should* be able to decrease the length of the fade required to minimise large value discontinuities. It's academic though. Nobody slices sine waves, and real world fades are usually quite long (because you don't need them to be short, as in, a dozen or two of samples).

-- 
Adriano Petrosillo <ampetrosillo at gmail.com>