The square wave itself is constructed using plain line segments, and for 
that reason isn't bandlimited (in such cases, not slightly but 
massively) - it is already aliasing, and filters can't do anything about 
that. Try vco or vco2 to create a bandlimited signal. In digital audio, 
we can't really have any waveform with corners, abrupt instantaneous 
changes, or "literally" straight lines. We sort of get away with it with 
piece-wise amplitude envelopes, most of the time, but not for audio rate 
oscillators.

Richard Dobson


On 20/10/2015 12:50, Peter Burgess wrote:
> I seem to have an issue with aliasing that I can't get rid of, at least,
> it sounds like aliasing. I'm currently working at 96 kHz sample rate,
> and I am passing a square wave oscillator through 4 x 80 pole
> butterworth filters at 20 kHz. The filters make no notable difference to
> the aliasing.
>
> I have tried changing the sample rate, the ksmps and the cut-off
> frequency and the of the filters and the buffer sizes to see if any of
> them make a difference. The interesting thing is that if I reduce the
> cut-off into the audio range, it filters both the un-aliased and aliased
> frequencies as if the aliased frequencies are already added to the
> signal before it reaches the filter.
>
> Incase it makes a difference, I am running csound 6.05 on Ubuntu, and I
> have tried running my .csd both in csoundQt and the terminal. I have
> attached an example .csd to show what I mean.
>

Two comments

1: oscili is more prone to errors than poscili or oscil3 etc

2: your squarewave has al the aliases in it already.  Needs to be band 
limited

Now over to the experts....

On Tue, 20 Oct 2015, Peter Burgess wrote:

> I seem to have an issue with aliasing that I can't get rid of, at least, it
> sounds like aliasing. I'm currently working at 96 kHz sample rate, and I am
> passing a square wave oscillator through 4 x 80 pole butterworth filters at 20
> kHz. The filters make no notable difference to the aliasing.
> 
> I have tried changing the sample rate, the ksmps and the cut-off frequency and
> the of the filters and the buffer sizes to see if any of them make a
> difference. The interesting thing is that if I reduce the cut-off into the
> audio range, it filters both the un-aliased and aliased frequencies as if the
> aliased frequencies are already added to the signal before it reaches the
> filter.
> 
> Incase it makes a difference, I am running csound 6.05 on Ubuntu, and I have
> tried running my .csd both in csoundQt and the terminal. I have attached an
> example .csd to show what I mean.
> 
> Pete
> 

After you generate an aliased signal, it is unlikely that you can fix it with a filter. You need to
generate a bandlimited or alias-suppressed signal to start with.

========================
Dr Victor Lazzarini
Dean of Arts, Celtic Studies and Philosophy,
Maynooth University,
Maynooth, Co Kildare, Ireland
Tel: 00 353 7086936
Fax: 00 353 1 7086952 

> On 20 Oct 2015, at 12:50, Peter Burgess  wrote:
> 
> I seem to have an issue with aliasing that I can't get rid of, at least, it sounds like aliasing. I'm currently working at 96 kHz sample rate, and I am passing a square wave oscillator through 4 x 80 pole butterworth filters at 20 kHz. The filters make no notable difference to the aliasing.
> 
> I have tried changing the sample rate, the ksmps and the cut-off frequency and the of the filters and the buffer sizes to see if any of them make a difference. The interesting thing is that if I reduce the cut-off into the audio range, it filters both the un-aliased and aliased frequencies as if the aliased frequencies are already added to the signal before it reaches the filter.
> 
> Incase it makes a difference, I am running csound 6.05 on Ubuntu, and I have tried running my .csd both in csoundQt and the terminal. I have attached an example .csd to show what I mean.
> 
> Pete

Oh, one minor adjustment, the gen30 function needs a much bigger table number to properly prove that it works