Anti-aliasing module

[Tim Kleinert]

The G2 oscillators have good anti-alias algorithms, which allows them to go into quite high pitch ranges before the annoying intermodulations kick in. However, this is not necessarily the case whenever you start to do DIY oscillators.

A simple example where aliasing is a problem is the derivation of a variable width pulsewave from a sawtooth oscillator via a comparator module -just the way it is done in the analogue realm. Using the hereby generated logic gate pulse as an audio signal readily produces intrusive artefacts caused by the intermodulation of the gate frequency with the G2 sampling frequency.

My knowledge on the maths involved in anti-aliasing technology is limited, so actually I don't know if it is at all possible to anti-alias a signal post facto. Maybe this would only be possible strictly for oscillators, and the module would require a pitch control signal and syncing pulse, so that it "knows" what the phase and frequency of the oscillator is. I'm just hypothesizing here...

However, if it is possible, I think it would be great to have as a separate module, in order to take care of such issues.

[blue hell]

To anti alias a signal one must make certain that no frequencies above half the sample rate are present.

Once those frequencies got in its not possible to remove them any more as such frequencies are not really present but they are reflected back rather into the representable spectrum (and so they have become 'indistinguishable' from the real signal).

So it is not possible to make a module that removes aliassing, all audio mudules will have to prevent it internally.

Maybe it would be possible for Clavia to make anti-aliasing an option on certain modules as DSP resources are needed to implement it and it's not always needed.

Jan.

[Tim Kleinert]

Well, I suspected that, but I wasn't sure.

In that case, this wish doesn't make much sense, does it...

And adding an "anti alias" option for logic signals is asking for a bit much, I guess...


oh well...

[jksuperstar]

Throwing in a simple low-pass filter fully opened up should prevent any aliasing (this would put the cut-off ~20kHz, less than the 1/4 rate point). Where to put it depends on your system. 6db would be plenty, though for some strong feedback systems or fast edges (squares, sawtooth) a 12db might be needed, or use a lower cutoff. For controlling aliasing of Blue or Yellow signals, set the cutoff between 6-12kHz. (the control signals run at 24kHz). In a digtial system is that those sharp edges can easily be described without aliasing...all you need to do is filter out the high frequencies/harmonics before *using* them where they would "fold" back into an audible range...so in your example, put the pulse train threw the lo-pass I mentioned before using it elsewhere. The lo-pass should *not* follow the keyboard, and should be static at 20kHz!

In essence, the lo-pass *is* an anti-aliasing filter. At least, there is nothing else (DSP wise) that clavia could do to a module's output. The rest would be handled by designing the oscillator (or the whole system) properly.

There's a good paper on designing oscillators here:
http://www.harmony-central.com/Computer/Programming/blit.pdf

[mosc]

[ian-s]

[Tim Kleinert]

g2ian wrote:

[jksuperstar]