Re: [Csnd-dev] [EXTERNAL] [Csnd-dev] UDO local sr

Make sure you have libsamplerate for the best quality conversion. There’s a simple fallback converter, but Secret Rabbit Code is better.

The limitations are:

- can’t setksmps alongside it
- can’t use a-rate or k-rate arrays (that may be possibly changed in the future).
- if you use global arrays or bus channels it won’t work either. All access has to be
through xin and xout, because that’s where the conversion happens.

Testing is helpful, thanks for that.
========================
Prof. Victor Lazzarini
Maynooth University
Ireland

> On 20 Mar 2024, at 11:27, Anders Genell <[email protected]> wrote:
> 
> Ah, brilliant!
> I just saw the flurry of emails connected to the code submissions and it said something about not allowing a-rate, but that was likely a misinterpretation of mine.
> 
> I will try to build and test to see if my scenario seems to work as I hope.
> 
> Thank you!
> 
> Regards,
> Anders
> 
> 
> On Wed, Mar 20, 2024 at 10:51 AM Victor Lazzarini <[email protected]> wrote:
> Hi Anders,
> 
> There is now an undersample opcode to go along with the oversample. I am not sure what you mean about 2., it does allow
> a-rate opcodes.
> 
> best
> ========================
> Prof. Victor Lazzarini
> Maynooth University
> Ireland
> 
> > On 20 Mar 2024, at 08:55, Anders Genell <[email protected]> wrote:
> > 
> > *Warning*
> > This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.
> > This is mainly for Victor Lazzarini, 
> > 
> > I saw you added code in a PR to implement local sample rate in UDOs which I think is absolutely fantastic.
> > It seems there are two limitations that I really hope could be addressed in the future:
> > 1. The code allows for upsampling but not for downsampling
> > 2. The code does not allow for a-rate opcodes
> > 
> > The use case I am hoping will be possible in the future is when having a reasonably high global sr (say 48kHz or even 96kHz) where I would like to apply relatively narrow band filters with center frequencies at say 20Hz, 25Hz 32Hz or the like. Such filters tend to be somewhat unstable at such high SR if they also need to be reasonably steep, and by downsampling before filtering a better result is often achieved. This is commonly done in signal processing for analysis in e.g. Matlab or GNU octave.
> > The idea is e.g. to be able to manipulate individual bands, like in the pvs group of opcodes, and combining the individual bands back to one signal afterwords, so just docking global reduction of SR would not be a viable route, unless of course starting several instances of csound and using channel opcodes to combine everything in the end.
> > 
> > Would this be a possibility in some future iteration of the newly implemented code?
> > 
> > Regards,
> > Anders