Switching to 96khz for mastering?

[Ichad.c]

I work at 44.1khz for tracking and mixing, for a lot of reasons.

I hope you have some seriously good reasons for tracking @ 44.1kHz. I'd recommend 48kHz - and the reasons are simple:
1)Keeps file sizes manageble
2)Leaves space at the top of the spectrum for the anti-aliasing Lowpass filters in your harware and any software that oversamples.

Or another aproach - OneRepublic - 'stop and stare'
1)Recorded 96kHz
2)Mixed 48kHz
3)Mastered to 44.1kHz

Also, Phase47 gives some really sound advice aswell.

P.S. Oversampling in DACs is common - my old portable sony disc-player has 64x oversampling - and it sounds awful.
My Sony D7 runs @ ~1.7MHz - no oversampling - but wait - it's a 1-bit converter.

Disregard the marketing hype, use your ears, make great music.

Regards
Andrew

[whyterabbyt]

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.

do you? changed your mind already?

http://www.kvraudio.com/forum/viewtopic ... 10#5096310

[jupiter8]

There are many good points so far in this thread @midnight.

Funny that apart from the "Try both and see for yourself" it's mostly crap advice as far as i can see.

[chokehold]

44.1 vs 48 kHz only determines how far "left and right" (low Hz, high kHz) of the frequency spectrum is stored.

44.1Hz is CD quality, but recording in and rendering to 48kHz rules out CD as a medium for distribution.
Rendering down (classic resampling) from 48kHz to 44.1 will always introduce artefacts, distortion, noise and errors. Maybe not obviously audible, but it's there and it degrades the signal.
Plus the additional HDD space the 48kHz waste on lots of unused and not audible frequency content.


Regarding this 96kHz -> 48kHz -> 44.1kHz approach, read some articles about resampling introducing artefacts, distortion, noise and errors.

You'll be alarmed how bad even the quite expensive DAWs and audio editors turn out to be at resampling, IIRC there's even a website dedicated to comparing the horrible effects resampling has in various DAWs and editors.

EDIT: found the website, here it is: http://src.infinitewave.ca/

Because of all this degradation, it's NOT ideal to resample over and over again! AT ALL! If you feel audiophile enough to record in a higher sample rate and then want to reduce it, ALWAYS (!) take 1/2s ... so if you want to create a CD or MP3 with 44.1 but choose to record on a higher sample rate, then use 88.2kHz!

Always compare resampling to image resizing.


So in order to get the best result for standard CD/mp3 distribution, the "native" format of 44.1kHz is the ideal choice.


Oversampling uses additional Bits to make calculations more precise.
To make it easily understandable, even if not 100% correct:

* Say you recorded a track in 24bit, and say 20 of those Bits go towards "storing the signal from 0dBfs to -96dBfs", with 4 Bits left for other information.

* Say 32bit would use 28 Bits for "storing the signal" from 0dBfs down to -112dBfs (or so), again with 4 Bits remaining for other info.

* As you recorded in 24bit, all of the signal below -96dBfs is noise floor.

* So to effectively use the 8 Bits more, the 24bit signal is shifted in the 32bit range, so that the 32bit signal will only calculate down to -96dBfs as well, like in the 24bit source signal.

* This leaves 8 Bits which are now unused, and not used to "go lower in volume".

* These unused 8 Bits of "volumes below -96dBfs" are instead used to calculate and store positive overshoot, to create additional headroom over 0dBfs.

* This means that while calculating a signal, not everything above 0dBfs will be chopped off, but rather calculated as well and then clipped at the output stage.

Why is this of any use?

Say you go into an EQ plugin with a 24bit signal that, for whatever reason, peaks at -0.2dBfs. If you now boost an EQ band by 1dB, you'd have clipping of +0.8dBfs, introducing ugly digital distortion.

If the EQ plugin operated at 32bit, it would use the first 24 Bits to process the signal normally down to -96dBfs, and then use the additional 8 Bits to calculate stuff above 0dBfs.

Like this, you could easily run a signal into the EQ that peaks at -0.2dBfs, and then boost a band for several dB - without having clipping.
...as long as you use a volume adjustment inside the EQ to lower the overall volume down again to peak at 0dBfs or below, before returning the signal back to the Host.


So as you see, oversampling has nothing to do with sample rates or resampling.

[jupiter8]

So as you see, oversampling has nothing to do with sample rates.

It doesn't ? That's news to me. I'd say it has everything to do with sample rate and nothing what so ever to do with bit depth.

[whyterabbyt]

Oversampling uses additional Bits to make calculations more precise.

No, it absolutely, unequivocally, does NOT.

[dreamkeeper]

chokehold - you should remove that POBS, before someone steps into it...

oops... too late! Now, how are going to pay the cleaning of all those shoes, huh?

[Aloysius]

Oh yes it DOES.

[whyterabbyt]

Oh yes it DOES.

who are you replying to? me, jupiter8, chokehold, or the kids at the pantomime?

[Aloysius]

Oh yes it DOES.

who are you replying to? me, jupiter8, chokehold, or the kids at the pantomime?

Kids at the panto!

[whyterabbyt]

Oh yes it DOES.

who are you replying to? me, jupiter8, chokehold, or the kids at the pantomime?

Kids at the panto!

oh no you're not...

[Aloysius]

Oh yes I am ... or am I???

[Doug1978]

^^ it's behind you...
(Whoops, sorry wrong thread)

@Jupiter8 - there is some good stuff in this thread, but also some confused thinking. However, I particularly liked Mutant's idea.

As for the OP, @midnight - as others have said, no there's not much point in specially mastering at 96khz in that situation.
Then again, if you think it makes your music better, just go for it.

[Amberience]

44.1 vs 48 kHz only determines how far "left and right" (low Hz, high kHz) of the frequency spectrum is stored...Oversampling uses additional Bits to make calculations more precise...So as you see, oversampling has nothing to do with sample rates or resampling.

I've never read so much crap in my life!

Sample rate affects frequency content of audio; oversampling directly relates to this.

Bit-depth affects dynamic content of audio; lower bit-depths cannot represent as wide a dynamic range as higher bit-depths.

[Compyfox]

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%.

This does not work due to one main reason:
Internal Oversampling.

So if you use the same plugins for the 88.2 file and for the 44.1 file, and with both OS is activated, you add "more" content to the 88.2 file than the 44.1 file.

If you want a total null test, no(!) OS for the 88.2 files, and 2x OS for the 44.1 ones.


Then again, in the original thread there was a link to a YouTube video which took a closer look at what's happening. The difference was in the higher frequencies - and they were very, very, very marginal and in non-hearable ranges. Especially with mid-end to highend converters.


So if you want to do something right: 48kHz/24bit at least, then go from there. But the better the DAC, the less you hear a difference.

As for the OP, @midnight - as others have said, no there's not much point in specially mastering at 96khz in that situation.
Then again, if you think it makes your music better, just go for it.

This is not directed at you, but at the general discussion in here.


Again, your main advantages in higher sampling rates would be smoother EQ courves and less aliasing with modules that introduce any form of distortion.

If you (the user) work with a 96kHz file right from the start while mastering, with 4x OS you get 384 kHz internally(!) to work with. If you work in 48kHz and use 8x OS (which is standard in most plugins by now), you get the same value. And according to an article on Wikipedia (see: http://en.wikipedia.org/wiki/Sampling_rate ), 384kHz is higher than the editing sampling rate of SACD (other known as Digital eXtreme Definition, or DXD in short) , before it got converted into 1-Bit (which would have been 2,88MHz).

Was there something added in terms of content? I don't think so.
Can you really hear the difference? I don't think so either.

Do you think that mastering studios constantly work in 192kHz in order to create 96kHz DTS-HD Master Audio files? Think about it for a second.


If you work in higher sampling rates right from the start, you tax both your CPU and HDD. So you need to do workarounds - and Oversampling is(!) a simple yet effective solution.


But the most important question still remains... does it matter if we still talk about masterings with RMS values of -6dB(FS)?



Two sides of the medal.