Allow Cubase to export projects in 44.1 kHz at 96 kHz FOR REAL

Surprised no one so far has posted a link to the “Digital Show and Tell” video by Monty. It’s about time :wink:

1 Like

More frequent samples means the curves can be fit more accurately. \It is really a rather basic point. If you don’t get that, there really isn’t going to be much useful conversation. If you want to argue that this better curve fitting isn’t necessary, you may be right. I don’t think it is necessary for my work. But denying how digital sampling works does not help the rest of your arguments.

You are being obtuse to the point of frustration at this point.

Please provide proof or a link to any scientific paper that would support this claim.
All you have written so far are wishy-washy analogies that are false.

But how does digital audio sampling work? Please explain with references to actual science.

1 Like

Oversampling is a very nice feature from Reaper. Today most “analog” plugins have saturation and harmonic distortion, but they don’t take care of the aftereffects very well, they have problems with aliasing. This SHOULD be handled by the plugin, usually it is not a problem with 96k projects. But is not so easy. With a plugin causing harmonics a other plugin down the chain might get aliasing, so you need a high-cut on each plugin. The best way do handle this is to have a oversampling option per plugin. You can still run the things does not need the cpu “waste” in oversampling mode. It should also have the dynamics so it can be in mixing/ASIO guard but not in realtime paths. My solution to this is to convert the hole project when it getting finalized. But it is not work-flow efficient. It should be a interactive work-flow at all time.

2 Likes

I’m not a mathematician, neither do I understand the sampling theorem, but what I’ve understood is that the theorem proves that it doesn’t matter how many points you use beyond double the frequency you want to sample.

It’s as if you were to take points (0,0) and (1,1) and draw a line. It’s y=x. Filling in 2,2 3,3 4,4 5,5 won’t make the line any more accurate.

1 Like

No, you keep making this point and you simply ignore the corrections people are giving you. I don’t understand your attitude at all.

Increasing the sample rate from 44.1kHz to 96kHz makes *no difference * to how well a 20kHz bandwidth limited signal is represented.

1 Like

Except it is demonstrably NOT a straight line. We deal in complex waveforms that look more like
Not Sine

AD conversion consists of converting the natural sound wave (which is usually a composite of dozens or hundreds of individual waves, each having its own harmonic series) into a set of of digital points that APPROXIMATE the shape you see e.g. in this graphic. The whole notion of “frequency” is misleading because the composite wave embodies lots and lots of different frequencies which may or may not line up, or which may be heard as nuance or color rather than distinct pitches.

Nonetheless, the sample rate is certainly fast enough to express the prominent pitches (frequencies) from the total wave – if they are well encoded within the wave. But is the coding good enough to adequately reproduce the many overtones and other effects that don’t strike the ear as “pitches” per se?

Maybe yes, maybe no. I don’t know. Some people believe there are nuances that are lost when good ears are listening through good speakers.

Perhaps it becomes more clear by zooming in further to where the individual data points are visible:
Screenshot 2022-05-17 161836
You can clearly see in the highlighted section that this sampling loses some information. Enough to make a difference? I don’t know. But denying that there is information loss makes no sense to me.

Digital does not work like a tape recorder.
44.1K sampling rate has the same resolution as 88.2K sampling rate.
88.2k has more information above the frequency that 44.1k is able to reproduce.
Think of it as both share the same information if you would overlay them, the 88.2k one simply has an extended frequency range. Thus, having more information to reproduce 20k to 40k frequency.
To display it crudely:
0 Hz------------------- 20kHz ----------------- 40kHz
44.1k dadadududadadududada
88.2k dadadududadadududadadadadududadadududada

There are absolutely reasons to work at higher samplerates. Some converters sound different at different samplerates (Some of the first RME comes to mind)
Plugins can sound radical different, I personally have had guitar pedal overdrive/distortion plugins change the sound so radically, that I thought I had the wrong patch loaded.

My suggestion would be work at 96k, use Freeze and/or Render In Place as much as necessary and complete a project that way. Then take a critical listen, AB double blind and make an informed discission.

4 Likes

There is no maybe, it is a scientific fact.
The “many overtones”, as you say, will be perfectly represented if they are within half the sampling frequency. AD converters doesn’t care if what it is sampling is an over tone or a fundamental.
You should ask those “some people” what exactly ARE the nuances they seem to be missing.

@ggmanestraki’s example with two points and a line is not only a brilliant and accurate, but also happens to be Euclid’s first postulate. I dare you to disproof it!

You seem to show genuine interest in the subject, so I’d like to encourage you to pick up a book and and gain some knowledge. (And I don’t mean that in a condescending way. :slight_smile: )

2 Likes

D/A and A/D | Digital Show and Tell (Monty Montgomery @ xiph.org) - YouTube

2 Likes