192 Khz?

Did a search and haven’t seen this asked before… so here goes…

Would it be possible (without too much difficulty) to implement higher frequency rates in Renoise than 96kHz?

This has absolutely nothing to do with perceived audio quality of the final mix. (As far as I’m concerned, once you hit 48kHz, the human ear can’t tell the difference. Well, maybe there are a few x-men out there who can.)

However, when using a tracker it is very common to use techniques such as pitch-shifting. When you do a lot of this, the benefits of higher frequency rates become pretty clear, as you can lower sounds by two octaves and still retain information above the threshold of human hearing. I also run bidule which supports higher refresh rates and it would be very useful to run bidule as a vst/vsti at higher frequency rates.

I am a bit unsure about this topic: does your soundcard support 192khz? I doubt it does, so the final output will be at 96khz at most. So you are asking for Renoise to oversample every audio source (samples and plugins) and then downsample again? If so, apart from the CPU overhead, how would the benefit of oversampling be noticeable for your purpose after the downsampling?

I’m definitely missing the point here

Some Soundcards support 192khz (and many claim they do, but really don’t).
The only reason renoise does not support higher frequencies then 96khz is afaik some incompatible internal fx (like mpReverb)?
There is a huge quality gain going from 44.1khz to 96khz, especially if you do a lot of sample processing (which most people do in renoise).
You can pitch far more with far less aliasing noise for instance.
In total it really makes a huge difference. Even if you use samples with lower samplerate to begin with.
Fx like filters just sound far better and with less artifacts using higher samplerate.
Try for instance to play some high pitched saw waveform sample in renoise. Render using sinc at 96khz and then at 44.1khz. You should hear a massive difference in aliasing and other artifacts. This really adds up when you play a lot of stuff together.

I wonder how the same test would sound if renoise could render 192khz…

still, what I am trying to say is that the final render, I mean the render you will publish to people, will probably be at 44Khz, so in the end using 192Khz could turn into a disadvantage

Yes, the end result and final samplerate is a whole different issue.
No need for insane samplerate there. But as Shane Turner said, it is inside renoise this is interesting, before you do a final mixdown render.
And as I pointed out, it could also be useful when generating new instruments inside renoise. Oversampling is a well known trick that vsti’s and other effects have used for a long time.
You should anyway always do a final render at highest possible samplerate. After that you reduce samplerate to this rendered file.

Some people also like to render separate tracks as well to mix the tracks externally. Then you can also probably benefit for something higher then 96k.

If I had the hardware and CPU grunt to do 192khz I damn would do it. I’ve now got some really good gear, especially the ADC and DAC, but it’s running at 96khz (still pretty amazing for digital). Final mixes are rendered at 96khz 32bit float. On the way out of the DAC I dither down to 24bit using Ozone, and connect my the DAC output to my tube compressor for mastering. I have a separate computer and card that’s connected to the really good ADC running that 44.lkhz 16bit sampling capturing the compressor output. The converters are Myteks, and they do another even better model at 192khz. Suffice to say that one day if I can afford that I’ll be upgrading to that (and hoping I have a ballsy CPU rig to hack it). Would be great to have that clock option in Renoise.

as far as I know, downsampling to 44khz from a 48khz multiplier can result into a quality problem

You are way off with this one I’m afraid… Humans can’t hear anything at all above just 20 kHz, and even then you can’t really hear them, they just cause an unpleasant feeling.

It is actually around 12-15 kHz that sounds become indistinguishable from each other. Hence, a CD’s limit of 22.05 kHz will perfectly suffice, as will 48kHz (half the sample rate of renoise).

I can understand wanting to keep more quality when pitch shifting, but just shifting it any amount at all means that you lose enough fidelity that it makes little difference.

Personally, I don’t even use 96kHz, mine is set to 88.2khz, which is double the CD sample rate, meaning a very simple mix down process (and of course save processor power anyway). There are various web pages about doing this.

Interesting stuff.
What about downsampling 88.2 khz for dvd’s 48 khz ?
Should I keep dvd’s rate in mind when I’m doing sounds for it?
And what about going from 44.1 to 48 khz?

AFAIK this is just a myth. And is more about how hard the math will be. The quality should not suffer. Should not be an issue these days :)

If you are just listening to a single file, then a 44.1 will do just fine. But the moment you wanna mix together stuff or do any processing to these files you wanna have the highest possible samplerate and bitrate.
How much of an difference will there be? It really depends on what material you do and how you process stuff. The extreme is to generate the sounds yourself using waveforms in renoise. You will clearly hear and see a difference between a 44.1 render and 96khz render (especially if you use sinc interpolation). You get rid of a lot of noise, ringing effects and other nasty stuff.
The same is true when recording audio at high samplerate. It’s better to sample with high samplerate, then reduce sample/bitrate if you have to.

Of course… and I do.

I can qualify with out a doubt that ANY form of downsampling is destructive to the audio and is a major cause of loss of realism in dynamics as well as causing all sorts of unpleasant ‘knots’ in the spectrum going up through the harmonics. It’s a case of mathematics trying to emulate an organic process, and falling short. Any professional mastering operation will simply not mathematically downsample for a final master. They will do it by the method I outlined above: going out a DAC at a high sample rate, running through outboard analogue gear, and then into a separate ADC running at 44.1khz (or 48k for DVD audio etc). This simply isn’t myth, it’s fact that it sounds better. A-B the two methods and see for yourself.

Sure, we can’t physically hear above 15k-16khz, but the implications of having a higher sample rate in digital audio is absolutely imperative to a quality sound. If you can try this test: Pick any of your songs in progress and do an A-B. Run the mix at 44.khz and listen to the upper mids and highs, and in particular notice a greyish haze to the sound. Now switch your card to 96khz and reinitialize. Play the mix back at 96k and again pay attention to the upper mids and highs. You should notice that the greyish haze is less present, and the texture of the sound is much smoother and open. The difference will also depend on the quality of your converters, some nastier converters have a more knotted and bent representation of the sound. Use the higher rate. Digital is a flawed medium, you got to give it every chance possible to overcome these shortcomings.

Anyone who thinks that their favorite commercial music finished by a professional mastering studio is all done at 44.1khz or uses mathematical downsampling is just kidding themselves. We’re forking out pretty serious money to get the best converters we can afford so that end result is as smooth as possible. There’s always at least some form of high end analogue involved, which always is a way more organic process than leaving it to chance with number crunching.

I fully agree. Of course all downsampling reduces quality.
The ‘myth’ I was talking about is the factor of digital downsampling.
Like, many people say you should use 88.2 instead of 96 if you know you gonna publish something in 44.1 (no analog gear intended to be used). In that case the only thing you save is some cpu usage when converting… Sure there must be some interpolation involved when downsampling from 48 to 44.1. But that one-time interpolation is nothing compared to the benefits you get from higher samplerate used in the entire project where you probably use interpolations all over the place.
So my point, like your point is to use the highest possible samplerate until the final stage (and then using whatever gear you have to downsample). And don’t worry about downsampling factors and stuff.

Let’s star making music analogically to avoid all this samplerate hassle \o/

I’m not really that knowledgeable about the issue but why to work with higher frequency rates than the final media your music will be on, when downsampling is destructive? Media are mostly 44.1kHz or 48kHz. Does the advantage of better pitch-shifting beats the loss of quality in downsampling?

If the samplerate is limited, you won’t know what the song actually sounds like. At 44,1khz the highest freq possible to replay is 22,05khz, and as a sinewave it will get rendered basically as a pulsewave, which wouldn’t be what is suppoed to be. Even at quarter of the samplerate for a sinewave you’d still get only couple of samples per wavelenght which isn’t sufficent as it still has some noticable aliasing especially without dithering/interpolation.
Downsampling the work will damage the audio, but it won’t be any worse than if you would’ve been making the song on the lower samplerate to begin with, but instead whilst making the song you get a better representation of the actua audio.
Gaah I’m way too tired, shouldn’t be typing this late. Hope the post isn’t completely unreadable…

Now this is debatable. But you have a good point in the last thing about hearing the actual quality the song as you make it. Because there can be a quite large difference for very heavy processed sounds.

But, back on topic, many people would like to resample stuff inside renoise. To create instruments. Render selection. Render new instruments etc. It would be nice to have the highest possible sample rate available. So to let the user decide what is needed for the project…

Here are 3 files illustrating in a simple way the huge difference.
Three renderings:
44.1khz 16bit cubic interpolation.
44.1khz 32bit sinc interpolation.
96khz 32 bit sinc interpolation.

Listen in renoise and have a look at the Master Spectrum as well.
It is a simple instrument made in renoise. Hand drawn 44.1khz saw in a hurry (hence the DC offset of the samples). With a slow instrument filter envelope and some vibrato.

You hear a very distinct difference because it is ‘extreme’ manipulation of a sample (that is very common to do in renoise btw). When you only pitch samples a few semitones the difference is far less distinctive. But it’s there! You really do notice this when you add up many channels.
I urge everyone to try to render a song they have made into separate 96khz tracks, then try to sum these tracks on a good analog mixer (best option), or in a digital DAW for further processing/mixing (renoise is not a good choice here because of the lack of disk streaming). You should get better overall quality.

Sometimes you actually want the aliasing to be a part of the sound design. So for these specific tracks in a song you should render at working samplerate and cubic interpolation and then upsample them to match the mixing project samplerate (if the software requires you to do so).

Yeps. :)

If you want to cross over into the pure analogue domain you can have nice endless argument with tape gurus about which tape speed is the best! Oh no not more endless arguments! :P :w00t:

Moral of the story: do your research and pick the best method to suit your project, budget and patience. No one method is perfect, but service ‘the idea’ in the best way possible.

I kind of thought I explained the reason why in my first post, but maybe I’ll try rephrasing it… This has nothing to do with the final mix and everything to do with generating sounds and samples with an extremely wide bandwidth… sound design. Working in high frequency rates from the start till the end of the project. You can do some pretty sweet stuff in renoise with sound design and musique concrete practices which all involve lots of pitch shifting, for example, never mind when incorporating your own synth patches…

I can’t hear above 17khz so 44.1khz technically goes higher up than I can hear (nyquist puts it at 22.05khz) but as I said in my opening paragraph this has nothing to do with the perceived audio quality of the final mix and everything to do with the processing that goes on up until the final mix.

Also having experimented with converters (my mastering engineer acquaintence tells me wavelab is probably the best for the affordable market) when sampling down from 96-44 or 48-44, I personally can’t really tell the difference. Honestly I never downsample to 44.1 anyway because I have no interest in making CDs. 96 is nice when I’m DJing/mixing my samples/loops in ableton because I do everything in 96 or higher and can pitch shift them significantly without loss of clarity. By this I mean samples/loops I’ve generated in renoise, bidule or other software (usually a combination thereof)… although I have taken close-mic’d recordings of percussion with decent mics that have frequency content at 30-40khz easily.

Anyway, this isn’t a HUGE deal to me (especially if it is particularly difficult to implement), it would just be really nice to have, especially because renoise is a major part of my workflow :)

I prefer to render my tracks as i worked with them, so usually i just render them as 48 khz, cubic interpolation, 24 bit… lots of problemes are dissapeared since i’ve start useing this method.