Rendering distinctions

Hello my friends .
I have a question to Renoise developers and the team : Lately , i made a project on Renoise 2.8 [i like to work on it more because it uses less CPU and act faster in some conditions] , The render interpolation algorithm : ARGURU’s Sinc ; is really fantastic . It works so slow in comparing with the Precise algorithm , but the results surprised me! I rendered the same song from 2.8 , then 3.x …
The Arguru’s sinc showed much detailed results than that one in 3.x

My question is : Is it possible to use this algorithm in 3.x ???


[It works only in older versions. Slow but so effective, especially on the high pass zone and reverb sounds]

Thank you for answering

I’ve no experience with the 2.8 resampling algorithm.

But I just want to point out to you that with 3.1 it is possible to render your project with a sampling rate of 192k, regardless of what your soundcard supports if you use the Offline mode. That is 4 times 48k! I use it all the time, it is a bit slow, but will 4x oversample everything, even VSTs, leading to really superior highs in the result if you use the HQ interpolation, which by itself isn’t so very strong. The aliasing will be mostly happening in the inaudible domain and can be discarded by a simple 15-20 khz lowpass filter on master or by post processing. If I need a 44.1 or 48k 16 bit output, I render 192k 32bit anyways and resample the results with conversion software after rendering.

I’ve no experience with the 2.8 resampling algorithm.

But I just want to point out to you that with 3.1 it is possible to render your project with a sampling rate of 192k, regardless of what your soundcard supports if you use the Offline mode. That is 4 times 48k! I use it all the time, it is a bit slow, but will 4x oversample everything, even VSTs, leading to really superior highs in the result if you use the HQ interpolation, which by itself isn’t so very strong. The aliasing will be mostly happening in the inaudible domain and can be discarded by a simple 15-20 khz lowpass filter on master or by post processing. If I need a 44.1 or 48k 16 bit output, I render 192k 32bit anyways and resample the results with conversion software after rendering.

Why do i need a 48k or 192 ? The standard is always 44100 16 - You will always need to Dither when choosing over 44100 , unless you will lose a lot of details in there .
I was asking about the algorithm itself T_T it really sounds different while choosing same parameters under other algorithms
Unfortunately , i do not know if it is possible to use the Arguru sinc in 3x . Even in 1.9 it was possible ))

I just wanted to show you another option for generating higher quality sounds.

And yes, 44.1/16 is the standard for distributing mastered music. Higher qualities is mostly niche audiophile business. But nothing will stop you using higher rates in intermediate steps in the production. As I said, after rendering 192k I normally convert back to 44.1/16 at some step, or 48 if I want to work with the samples in realtime.

The rendering in such high resolution will run all the plugins and instruments and samples and their algorithms in a much higher rate, allowing better sound quality output in many cases. Yes, that quality is preserved even when converting your finished stuff to 44.1/16. It is all about the highs having lots of room to breathe and spill the inevitable aliasing mostly into the inaudible domain instead of reflecting the frequencies into the audible range messing with your sound in a way that cannot be corrected any more. The conversion/resampling to 44.1 afterwards will just chop away the (inaudible) aliasing, leaving only the crisp sound that is wanted. If you render 44.1, the aliasing will be much sooner reflected into the audible range, degrading your sounds without cure. The interpolation filters will work in a different way, they filter the sounds before they get converted to their target rate, limiting the aliasing that would be generated at a low sample rates, but this has different implications and drawbacks and will only affect renoise sample instruments, not filters, distortions etc.

I just wanted to show you another option for generating higher quality sounds.

And yes, 44.1/16 is the standard for distributing mastered music. Higher qualities is mostly niche audiophile business. But nothing will stop you using higher rates in intermediate steps in the production. As I said, after rendering 192k I normally convert back to 44.1/16 at some step, or 48 if I want to work with the samples in realtime.

The rendering in such high resolution will run all the plugins and instruments and samples and their algorithms in a much higher rate, allowing better sound quality output in many cases. Yes, that quality is preserved even when converting your finished stuff to 44.1/16. It is all about the highs having lots of room to breathe and spill the inevitable aliasing mostly into the inaudible domain instead of reflecting the frequencies into the audible range messing with your sound in a way that cannot be corrected any more. The conversion/resampling to 44.1 afterwards will just chop away the (inaudible) aliasing, leaving only the crisp sound that is wanted. If you render 44.1, the aliasing will be much sooner reflected into the audible range, degrading your sounds without cure. The interpolation filters will work in a different way, they filter the sounds before they get converted to their target rate, limiting the aliasing that would be generated at a low sample rates, but this has different implications and drawbacks and will only affect renoise sample instruments, not filters, distortions etc.

I think that Arguru’s algorithm can solve these problems in 44100 16 . But im just trying to find out a way to use it in Renoise 3,x
It works only in older versions .

Hello my friends .
I have a question to Renoise developers and the team : Lately , i made a project on Renoise 2.8 [i like to work on it more because it uses less CPU and act faster in some conditions] , The render interpolation algorithm : ARGURU’s Sinc ; is really fantastic . It works so slow in comparing with the Precise algorithm , but the results surprised me! I rendered the same song from 2.8 , then 3.x …
The Arguru’s sinc showed much detailed results than that one in 3.x

HQ mode actually is an improved version of arguru sinc. You can also setup it for each sample individually, the render setting will overwrite it, if HQ selected, AFAIK.

And yes, 44.1/16 is the standard for distributing mastered music. Higher qualities is mostly niche audiophile business. But nothing will stop you using higher rates in intermediate steps in the production. As I said, after rendering 192k I normally convert back to 44.1/16 at some step, or 48 if I want to work with the samples in realtime.

U sure? Because all playing devices today have built-in 48kHz. So if you buy a m4a in itunes (I never did), will it be 44,1kHz actually? At least I have seen 48kHz in m4a quite often, but almost never in mp3.

I dont think that it is an improved version of Arguru’s sinc / the Arguru sinc shows better results

Ps : all the distribution companies and aggregates request 44100 16 ONLY ( itunes - beatport - junodownload…etc )

HQ mode actually is an improved version of arguru sinc. You can also setup it for each sample individually, the render setting will overwrite it, if HQ selected, AFAIK.

U sure? Because all playing devices today have built-in 48kHz. So if you buy a m4a in itunes (I never did), will it be 44,1kHz actually? At least I have seen 48kHz in m4a quite often, but almost never in mp3.

Since 3.0 the interpolation must now be set on a per-sample basis. See: Sample Properties

The default interpolation mode is Cubic (as it always has been), but you can optionally change this to Linear, Sinc, or even disable it completely.

We now also have Anti-Aliasing on a per-sample basis, which will enable band-limited oversampling to further improve the interpolation results.

The “Precise/HQ” render option does not override all samples to use Sinc interpolation, rather it simply enables a higher quality version of whatever interpolation mode the sample is set to.

i.e. If a sample is set to Cubic then HQ will enable a higher precision Cubic interpolation, if it’s Sinc then HQ enables an even higher precision Sinc interpolation, and so on.

So if you really feel that Sinc interpolation can improve a certain sample, you will need to manually set that mode yourself on a per-sample basis.

Do keep in mind that Sinc is not simply “better” for everything. There are actually a few weird edge cases where it can make certain samples sound worse. So always use your ears first to test things :slight_smile:

Since 3.0 the interpolation must now be set on a per-sample basis. See: Sample Properties

The default interpolation mode is Cubic (as it always has been), but you can optionally change this to Linear, Sinc, or even disable it completely.

We now also have Anti-Aliasing on a per-sample basis, which will enable band-limited oversampling to further improve the interpolation results.

The “Precise/HQ” render option does not override all samples to use Sinc interpolation, rather it simply enables a higher quality version of whatever interpolation mode the sample is set to.

i.e. If a sample is set to Cubic then HQ will enable a higher precision Cubic interpolation, if it’s Sinc then HQ enables an even higher precision Sinc interpolation, and so on.

So if you really feel that Sinc interpolation can improve a certain sample, you will need to manually set that mode yourself on a per-sample basis.

Do keep in mind that Sinc is not simply “better” for everything. There are actually a few weird edge cases where it can make certain samples sound worse. So always use your ears first to test things :slight_smile:

At last ! K.F is back :slight_smile: long time no speak pal !
Well i feel that Arguru’s Sinc is fundamentally affecting the 10k 19k region ( There is something special in it )
It is like using a summer instead of an exciter and feel that crunchy breeze that you cant explain.