Upsampling..........

Hi Dave,
my wife would get it because we're both engineers ! In fact we often have nerd type conversations regarding audio, even though she's not interested in hifi, and she calls me a geek whenever I tinker with my system.

By the way I've never known the difference between upsampling and oversampling .... I suspect that they may be the same thing, but that the marketroids decided that upsampling is a more marketable term.

As an engineer I must say .... "Damn those marketroids !".

It's been a while, but I think it goes like this ....

Upsampling does not increase the detail, nor does it increase the bandwidth of the signal, but it alters the sampling frequency, and so it pushes the aliased signals (see Nyquist's theory of sampled signal) to higher frequencies.

This allows the anti-aliasing low-pass filter to have a slower rolloff. It is easier to design a filter with linear phase and lower passband ripple if the rolloff is less steep.

So the whole point of upsampling was to simplify the anti-aliasing low pass filter, so that it causes less degradation to audible frequencies.

Performing any kind of interpolation would be akin to doing some of the filtering in the digital domain. I believe the interpolation bit may be the difference between oversampling (old technology) and upsampling, but I'm not exactly sure.

Funny you should ask about the algorithm. I was thinking about this last night (because I really have no life !)

A perfect CD player would have an anti-aliasing filter that was a perfect brick wall. Below 22kHz signals would be completely untouched, and above 22kHz would be infinite attenuation.

In the time domain this equates to a sin(x)/x impulse response.

So my guess is that a good place to start for the interpolation algorithm would be the sin(x)/x waveform.