the 16x oversampling filter http://msp.ucsd.edu/techniques/latest/book-html/node194.html from this part of the book is too expensive to use with [clone]. [clone] also just really does not like any reblocking (closing the patch while its running will often crash pd so I'm trying to avoid using it with a dynamic patching system like netpd).

Is there a cheaper way to achieve something similar or some maths to try 8x or 4x oversampling to smooth out high pitches? It's a polyphonic casio cz synth I'm testing with really high midi notes

@fishcrystals the main alternative in this case would be to use polyblep I think.
I'm planning on making some objects to add polyblep residuals but I thought some vanilla polyblep logic already exists in some libraries, can't think of which right now. (I thought maybe https://github.com/MikeMorenoDSP/pd-mkmr/tree/master but can't see anything at 1st glance)

you could look at these papers for more info:
https://www.eurasip.org/Proceedings/Eusipco/Eusipco2015/papers/1570104119.pdf
https://dafx16.vutbr.cz/dafxpapers/18-DAFx-16_paper_33-PN.pdf

edit: actually cz synths had aliasing afaik..

@seb-harmonik.ar said:

edit: actually cz synths had aliasing afaik..

The 101/1000 called it good with a second order lowpass tuned to 17khz. The common VCF mod for the 101/1000 was just making that filter variable instead of fixed so you could get all the aliasing you wanted.

@seb-harmonik.ar they polyblep ones are here https://github.com/MikeMorenoDSP/pd-mkmr/tree/master/oscillators

& actually its also in there - the os.cz~ oscillator that I'm using though I wastefully hooked it up so you can slide through waves and everything in realtime, something I always wanted my cz101 to do

but it was mostly from playing around with really high notes in anticipation of what might happen in netpd .. later I realized the casios of that era actually modulo notes above and below their range to the top and lowest octave

nice tip @oid - sounds like if anything I should introduce more aliasing

@fishcrystals said:

sounds like if anything I should introduce more aliasing

If you want the CZ sound you want the aliasing and the 17khz filter, use fx to smear and hide the aliasing like we did in the old days. With my aliasing free 192khz interface I find the CZ oscillator seriously lacking and sterile. People go out of their way to avoid aliasing and noise but aliasing and noise are a big part of the sound of much of our classic gear and their effect on the sound is far more complex than just noise and aliasing especially when you add fx into the mix.

in my opinion the usual 16x oversampling is almost unavoidable for FM or PD synthesis.

the attempt to take care of the bandlimiting with prior components has to fail everywhere, where even sinewaves as sources will already hit the nyquist barrier.

@fishcrystals said:

too expensive to use with [clone].

Might be obvious, but you only need one final anti-aliasing lowpass filter for all the upsampled clone-instances.

[clone] also just really does not like any reblocking (closing the patch while its running will often crash

I can not reproduce this. Pd 0.54.1 Win10

or some maths to try 8x or 4x oversampling to smooth out high pitches?

Modify J07.oversampling.pd to do any oversampling-factor.

mostly nostalgia out of me - the cz 101 felt unique, probably out of its limitations. really wanted to slide through all its settings in real time or at least send sysex to randomize a patch while it was playing.

what interesting now about casio's take on FM/PM with 'phase distortion synthesis' is after listening to it do its thing through a patch randomizer I'm surprised how many real instruments it will attempt to sound like in kind of a plastic toy way. does a lot with very little

@lacuna figured it out, the crash happens in windows with 0.54.1 when you try closing the patch after sending [clone] the [resize < message - but only if the amount of voices is different than when it started

& thanks - it wasn't obvious to have a second subpatch connected to [clone] for the oversampling filter

also looking closer I see that all the math for calculating the filter is in J07.oversampling.pd