Switching to 96khz for mastering?

[@midnight]

Is it common to switch to a higher sampler rate for using mastering plugins?

I work at 44.1khz for tracking and mixing, for a lot of reasons. What about switching to 96khz after I render the stereo mix. Then I could run at a higher sample rate when I do my final master limiting with something like Ozone.

What do you guys think about that?

[Spitfire31]

I'd assume it to be completely pointless, unless your original material was also recorded at 96 kHz, which it apparently isn't.

Of course, we're not talking about internal upsampling in mastering plug-ins. That's a different story altogether.

/Joachim

[vata44]

I have always heard of this technique... and use it (from 48khz 24 bit).

Basically, my understanding is that the higher you do any processing in, the more smooth it will be throughout... which ultimately will make things appear more "analog" in a sense. Much less digital stepping, and more detailed interpolation if you will.

The way I read about it:

'Picture a perfect sine wave and the 44100 pictures it takes to create it in lets say 1 second... Then picture a perfect sine wave with 96,000 pictures of it in 1 second. Then consider a perfect sine wave with 192,000 pictures of it taken within 1 second. That last one is a very damn SMOOTH sine wave. So much more information for the computer to compute and draw your final wave forms with.'

Noticeable to my ears? WTFK (who the f--- knows)

I could be wrong, I am only following my ears and the advice of people I consider better than myself.

I simply follow it based on the practical theory alone.



The one thing I do know, is it will not kill you to try.

[Phase47]

The rule in mastering - and this is the truth - is you're only as good as your mix. Mastering ten years ago is alot different from today, and forget about twenty years ago... who remembers 1630s and DAT tapes, right?

My philosophy is to respect the mix when it comes to mastering. Mostly, that's minimalism when it comes to processing... as in, if something comes in at 44.1 I wont capriciously upsample it "just because" if it is going to end up at 44.1 when I'm done.

My Golden Rule: don't do more than you have to... do less if you can. Do no harm, etc.

[@midnight]

I'm talking about so I can take advantage of the higher sample rate for example when using a plugin like Ozone, which doesn't use oversampling in the limiter. So if I switch to 96khz, even though my project is a 44.1khz wav file, Ozone will be able to use 96khz for its DSP operation to generate less aliasing/more analog like waveform on the final output.

what do you guys think about that one?

[quayquay17]

'Picture a perfect sine wave and the 44100 pictures it takes to create it in lets say 1 second... Then picture a perfect sine wave with 96,000 pictures of it in 1 second. Then consider a perfect sine wave with 192,000 pictures of it taken within 1 second. That last one is a very damn SMOOTH sine wave. So much more information for the computer to compute and draw your final wave forms with.'

The difference here is that you will be taking 192,000 pictures of a sine wave made of 44100 pictures, so it won't actually be any smoother.

Think of it this way, a pro digital camera takes clearer pictures than a cellphone, but if I take a picture on my cellphone and then take a picture of the picture with my pro camera it will be a perfect crisp image of a really poor image.

We do need one of the DSP/sample rate gurus to chirp in about mastering effects and higher sample rates vs. oversamling.

[Sampleconstruct]

There is tons of info about this here:
http://www.kvraudio.com/forum/viewtopic ... highlight=

[jupiter8]

I have always heard of this technique... and use it (from 48khz 24 bit).

Basically, my understanding is that the higher you do any processing in, the more smooth it will be throughout... which ultimately will make things appear more "analog" in a sense. Much less digital stepping, and more detailed interpolation if you will.

The way I read about it:

'Picture a perfect sine wave and the 44100 pictures it takes to create it in lets say 1 second... Then picture a perfect sine wave with 96,000 pictures of it in 1 second. Then consider a perfect sine wave with 192,000 pictures of it taken within 1 second. That last one is a very damn SMOOTH sine wave. So much more information for the computer to compute and draw your final wave forms with.'

Errrmmmm no.
http://en.wikipedia.org/wiki/Nyquist%E2 ... ng_theorem

In essence, the theorem shows that a bandlimited analog signal can be perfectly reconstructed from an infinite sequence of samples if the sampling rate exceeds 2B samples per second, where B is the highest frequency of the original signal.

[gavriloP]

It is a problem that people "look" audio on computers because what you should be watching (if that's what you want to do) is an oscilloscope AFTER DA conversion, that is the analog outputs of your interface. Then you'd see that this "stepped" waveform is in fact totally smooth once it is in a realm of reality. Extra resolution is like more than two points for single line, redundant.

[Compyfox]

Here... we go... again.


What I usually do (and I think this is the "secret of the pro's" as well):
Depending on the production (and the final format), I upsample the desired production, or render from 48kHz to 96kHz out of Cubase/whateverhost directly. Then I go into Wavelab and go from there.

What I'm basically doing is creating a "higher samplerate" version to begin with. Now, if I use the tools, which are mostly at least 2x fixed OS up to 8x or even 16x OS, I have an editing capability more closer to the so called (and to some - superior) analog realm.


But what I'm actually doing is using a standard recording, oversample it, run it through gear (which is oversampled or not), then downsample it again.

Think of it like (OTB example):
rig 1 plays back on 48kHz -> external OS matrixes (they exist in hardware form! or better DAC's have them built in) -> editing console (in this example, OTB modules) -> rig 2 recording at 96kHz.

Only that you do all that ITB.


Math examples:

44,1kHz as basis:
2x OS = 88,2 kHz
4x OS = 176,4 kHz
8x OS = 352,8 kHz (getting close to SACD now)
16x OS = 705,6 kHz

48kHz as basis:
2x OS = 96 kHz
4x OS = 192 kHz
8x OS = 384 kHz (SACD again)
16x OS = 768 kHz

96kHz as basis:
2x OS = 192 kHz
4x OS = 384 kHz (SACD)
8x OS = 768 kHz
16x OS = 1,536 MHz (!)


BUT...
Some ADC/DAC (at least high class ones) do OS on both input and output as well. RME is a prime example. Some have an 64x OS on the input, and 128x OS on the output, or 128x on both. So even with 44kHz or 48kHz - ask yourself - do you really need to record at higher sampling rates which only taxes your CPU?

Example with 64x and 128x OS with a 48kHz recording:
64x OS = 3,072 MHz
128x OS = 6,144 MHz

Can it get much analog"er"?
Can you even hear the difference? Especially on earbuds?

I think gavriloP summed it up perfectly.



Focus on ideal worklevels rather than squashing everything to sh*t instead of worrying about sampling rates. Especially if you are not(!!!) in the video postpro or HD audio mastering realm.

The rest was pretty much handled in the "48 vs 96kHz" thread here in the FX section.

[penguinfromdeep]

'Picture a perfect sine wave and the 44100 pictures it takes to create it in lets say 1 second... Then picture a perfect sine wave with 96,000 pictures of it in 1 second. Then consider a perfect sine wave with 192,000 pictures of it taken within 1 second. That last one is a very damn SMOOTH sine wave. So much more information for the computer to compute and draw your final wave forms with.'

Errrmmmm no.
http://en.wikipedia.org/wiki/Nyquist%E2 ... ng_theorem

Yeah the picture (or cinema) analogy regarding sample rates is not really good, there was a good article in Sound on Sound magazine that debunked a lot of digital audio myths, was kind of eye opener for me at least. Sampling theorem is perfect!

[Mushy Mushy]

What's wrong with the OPs original thread?
Maybe 2x 48khz threads = 96khz

[Mutant]

To actually hear the difference between what the mastering plugin sounds like at 44.1 and 88.2 you can always render both with the same settings and then upsample the 44.1 file and invert the phase on it and mix it with the other one 50% to 50%.

[ObsoleteAcc99]

[DELETED]

[Doug1978]

There are many good points so far in this thread @midnight.
I'm inclined to agree that there is not a great deal of point switching to a higher sample rate when you do your 'final master limiting with something like Ozone'.

HOWEVER, I often do the reverse of what your original post states - and for this, I can see a point in downsampling from 96khz to 44.1khz.
For some of my synth tracks, I will record at 96khz, export the audio at this rate and then downsample using Voxengo's R8brain Pro to 44.1khz.

The point is that some softsynths sound different at higher sample rates than at lower ones. Even though the final sample rate is the same, I can detect a difference in clarity and depth in the synth audio track via this approach.
(If someone REALLY wants me to prove it by sticking an audio file here I will).

There is a final reason for why I sometimes export at 96khz and downsample to 44.1khz via R8brain Pro and that is because I use Nebula for some of my mixing duties.
Some of the best Nebula programs (for instance AlexB's console programs) are different beasts at 44.1 and 96khz.

But I never master my own stuff (at either 44.1 or 96khz) if I want to publicly release it - that I give to a professional.

Ultimately, there is no right answer. It is a subjective decision as to what you hear (or think you hear) is best.