[music-dsp] Waveform Interpolation

[andy butler]

>> I'm thinking you'd have to "lo-pass" the sinc to make the results alias 
>> free.

Peter advises:-
> 
> That means changing the sinc kernel (tuning the filter), the algorithm 
> is the same.
> 
>> Or so it looks from here anyway, don't know if that's a practical idea,
>> but that's what I was expecting to see in the example.
> 
> Yep that's an expensive but practical idea 

:-) I'm on the right track then.
(even if it won't take me anywhere)

>> Nigel says:-
>>> Linear interpolation is simply a poor low pass filter 
>> I'm aware I'm on a list here with folks who have far greater expertise 
>> than I do, but surely that's nonsense. Or at least it's only true in the 
>> case where the linear interpolation is used for a constant fractional 
>> delay.
> 
> I too see all interpolation as filtering.

I don't see how polynomial interpolation a filter, is it?

> If you just simply resample a 
> signal with zero order hold, you'll get alias images all over the 
> spectrum. Now if you do *any* kind of interpolation (including linear), 
> it will smooth/filter that aliasing somewhat. 

That makes sense.
Now I'm thinking that the zero order hold re-sampling
could be modelled as phase modulation by a saw tooth wave
in terms of the artefacts generated.
...and that low passing those artefacts gets you a model
of linear interpolation.


thank you

andy butler