[Ardour-Users] click free editing and zero crossings?

[Ross Johnson]

Jörn Nettingsmeier wrote:
> where does this myth come from that edits at zero crossings are
> something particularly desirable? editing at a zero crossing *does*
> *not* mean you get a click-free junction.
>   
 From power electrics I guess and terminating transmission lines, which 
admittedly is not the same as audio. But notwithstanding your valid 
example and argument, there are many cases where cutting at a zero 
crossing is the best option even if it is not perfect.

> we're in a discrete world, so yes, the wave form is discontinous at
> every sample. that's not an issue. the important question is: are there
> so many new overtones generated by an edit (and not masked by content)
> that it begins to sound bad? or, to put it another way: does the
> waveform change direction in an extreme way?
>   
I suspect we really want to cut where the signal contains the least 
energy across the band of frequencies where the effects are the most 
undesireable. I do this quite often by ear by scrubbing back and forth 
near where I want to cut.
> to see why zero-crossing edits don't always work, imagine some region
> containing a simple low-frequency sine wave. say it approaches its end
> point (which is at a zero crossing) from the positive half-cycle. now
> splice another similar sine wave to the end of the first region. say it
> begins at a zero crossing, but it moves upward, into the positive
> half-cycle, immediately. presto: a nice zero-crossing edit with a
> horrible click.
>
>        *       |       *
>    *       *   |   *       *
>  *           * | *           *
> *-------------*+*-------------*
>                |
>                |
>                |
>
> the only way around this is a crossfade (which basically means a
> low-pass filter at the edit point).
>   
Clearly, if you were to splice these two signals you (or rather some 
clever software) might choose the next zero crossing of the trailing 
section to cut at (180 degrees further along), then splice. Generally I 
cut at a zero-crossing and then apply a fade, or crossfade, even if the 
fade is only a few samples long. My, often misleading, intuition tells 
me this introduces lower harmonic distortion than an arbitrary cut with 
fade.
> /me leaves soapbox now.
>
>
>