Tim Canfer

1. what method MAX uses to anti-alias (aa) the saw~ object? I assume that it would be either a filter at the end, or some kind of interpolation. 

In the help file it says it sounds analogue-esque, & while better then phasor~ it still sounds overly harsh to my ear.

I ask because I'm looking into the core reaktor saw (multiosc) which sounds better and actually, on a graph does not aa anywhere near as well! This seems to use a two point parabolic interpolation on the falling edge.

3. I am also wondering how, if possible, to do that in MAX...

1. what method MAX uses to anti-alias (aa) the saw~ object? I assume that it would be either a filter at the end, or some kind of interpolation. 
2. Does it use a lookup table?

how-does-saw-anti-alias

Saw~ probably uses a minBLEP algorithm. The basic idea of this kind of algorithm is to replace steps in a periodic waveform by precalculated bandlimited steps. If the period of the waveform is less than the length of the bandlimiting function, multiple BL steps must be added together for each sample-- the fact that saw~ takes much more CPU at high frequencies indicates that something like this is going on. search terms: eli brandt hard sync bandlimited BLIT BLEP.

You could make an antialiased sawtooth in MSP by putting a phasor~ and an appropriate filter in an upsampled poly~. 

Saw~ probably uses a minBLEP algorithm.  The basic idea of this kind of algorithm is to replace steps in a periodic waveform by precalculated bandlimited steps.  If the period of the waveform is less than the length of the bandlimiting function, multiple BL steps must be added together for each sample-- the fact that saw~ takes much more CPU at high frequencies indicates that something like this is going on.  search terms: eli brandt hard sync bandlimited BLIT BLEP.

You could make an antialiased sawtooth in MSP by putting a phasor~ and an appropriate filter in an upsampled poly~.  


> In the help file it says it sounds analogue-esque, & while better then phasor~ it still sounds overly harsh to my ear.

look at the mcgill modular book for tips on getting analog sound from your digital machine: 

http://www.cim.mcgill.ca/~clark/nordmodularbook/nm_book_toc.html

http://www.cim.mcgill.ca/~clark/nordmodularbook/nm_filters.html#analog

look at the mcgill modular book for tips on getting analog sound from your digital machine: http://www.cim.mcgill.ca/~clark/nordmodularbook/nm_book_toc.htmlhttp://www.cim.mcgill.ca/~clark/nordmodularbook/nm_filters.html#analog


that's great - I'll get reading - thanks.

I am finally having some mediocre success in trying to filter in the time domain (as it where), by using reaktor's sawAA method. But I'm having trouble understanding how to get a similar parabola. Anyway - here's the idea so far with a rudimentary interpolation scheme. Not as good as saw~ yet but I'm ever hopeful!

that's great - I'll get reading - thanks. 

How does saw~ anti-alias?