PDA

View Full Version : Work at 24bit 96KHz?


VRW
12-29-2013, 02:44 AM
Hi there, just a short question.
How many of you do record/mix etc. at 24bit/96KHz these days just for audio (no film etc.).
Does it make sense generally?
Asking also because the "Mastered for iTunes" thing is kinda topic out there though as far as I know most of the
digital distributors still demand the files for uploading as 16bit/44.1KHz.

Im used to record at 24bit/44.1KHz-at first because Im only doing audio and secondly it feels kinda safer and less complicated
to me when converting it finally just from 24bit/44.1KHz to the standard 16bit/44.1KHz for the final product.

A further question is, that I read and heard it to be much safer and more reliable to use 88.2 etc. if the final product is at 44.1
instead of converting from 96KHz etc. to 44.1 (because of the linear math thing or something).
Is this still right for the average producer (the really high end guys of course
do have the equipment as well as the skills to do almost everything ;), I certainly know)?

So Im just curious whats the actual situation these days? :cool:
What do you think about it?

Thanks in advance, best regards,

VRW



MacMini i7Quad, 16GB Ram, MacOS 10.9.1, Apogee Quartet, Genelec Active, Yamaha NS10, Pro Tools 11.0.3, Logic 10.0.5, Waves, MCDSP, Duende Native, Softube, HOFA, IK Multimedia etc.

JFreak
12-29-2013, 03:18 AM
I most often work with 24bit and 48kHz sessions. In case client pays me extra, I'm using 96kHz but no way I'm halving my system resources for free. There is one exception, and that is monitoring latency. If the artist is too picky about the latency, I might just record at 96k to make latency smaller, but then I convert to 48k for mixing.

I never do 44.1/88.2 sessions, instead I always convert to 48/96. Makes my life easier in the long run.

Sample Rate Conversion from 88.2 to 44.1 is not easier than 96 to 44.1 or 48 to 44.1 -- it is a myth that somehow 44.1*2=88.2 would make SRC easier. It doesn't work that way. In a properly done SRC, the actual waveform is reconstructed and then resampled to whatever target sampling rate you choose. Plugins also do SRC all the time. Many best-sounding plugs upconvert and process in higher sampling rate and then downconvert back.

WombatStudio.Org
12-29-2013, 05:49 AM
If the artist is too picky about the latency, I might just record at 96k to make latency smaller

Everything I've read suggests this would double the latency.

JFreak
12-29-2013, 06:02 AM
Everything I've read suggests this would double the latency.

Nope. Latency of ADDA conversion is cut to half if you double the sampling rate. That is because converters work in samples and not milliseconds. Same amount of samples is dealt with faster if you up the SR.

Raoul23
12-29-2013, 06:40 AM
I saw an interview with Mike Shipley (god bless him) where he said that he works in 96/24 because some plugins sounded better at 96.

Nice to no about the higher sample rate the lower the latency :)

albee1952
12-29-2013, 08:07 AM
99% of my work is 24bit/48K.

ArKay99
12-29-2013, 09:05 AM
I work at 24/96 all the time. I then mix 24/96k, then dither that to 16/44.1k for my CD mastering. It sounds better to me than if I work at 48k.

I did some basic tracks at 48k and at 96k. For 'similar' output meter ballistics the 96k recording sounds 'louder' and has more detail...reverb tails sound deeper and the stage around the sound seems more detailed. Maybe it's better transient response or the plugs I was using run at 96k so there was no src happening there, but it sounds better to me.

The downside for me is X-Form takes days to complete even small sections, and yes it uses much more cpu at comparable buffer sizes. And there is twice the audio data to deal with when working with it.

Maybe in the end, after it ends up 16/44.k and dynamically and sonically altered it makes no difference, I don't know. However, to me, if it sounds better before processing, it should sound better after the same processing...even if it's only a little bit. As stated above it's really a trade-off as to available resources vs. quality, vs. normal workflow.

lexaudio
12-29-2013, 08:59 PM
I most often work with 24bit and 48kHz sessions. In case client pays me extra, I'm using 96kHz but no way I'm halving my system resources for free. There is one exception, and that is monitoring latency. If the artist is too picky about the latency, I might just record at 96k to make latency smaller, but then I convert to 48k for mixing.

I never do 44.1/88.2 sessions, instead I always convert to 48/96. Makes my life easier in the long run.

Sample Rate Conversion from 88.2 to 44.1 is not easier than 96 to 44.1 or 48 to 44.1 -- it is a myth that somehow 44.1*2=88.2 would make SRC easier. It doesn't work that way. In a properly done SRC, the actual waveform is reconstructed and then resampled to whatever target sampling rate you choose. Plugins also do SRC all the time. Many best-sounding plugs upconvert and process in higher sampling rate and then downconvert back.

No disrespect J, but there is a difference.

Here is the bottom line. In a proper room you will hear the subtle differences of recording at 88 or 96, granted that you aren't recording with crap gear.
The big reason for recording at 88.2 is that it is perfect math. 88.2 half is 44.1. There are no rounding errors.

48k to 44.1, is so small, your never gonna notice.

However, all the mastering engineers I have worked with have all said, whatever you record at, let them SRC.
Not all SRC's are the same.
If PT was perfect at SRC, why aren't the mastering engineers using it?

Weiss is far better in SRC. There are others that are better. Granted, if you aren't in a room where you can hear it, it doesn't matter.
It ends up listened to on ear buds.

But, listen to Daft Punk RAM. Listen to how in translates from ear buds, to laptop, to fair speakers to your studio.

96 to 48 is perfect math. But ultimately you still have to goto 44.1. That is 2 times conversion with a smaller rounding error.

Granted, this is all probably so trivial. This and that, 24, 32. 48. 96.

From my own experience and recording, there is subtle difference between 88.2 and 96k.
Pretty much to the point you really can't tell.

However, the difference between and 44.1/48k and 88.2/96k you can hear, given the room and speakers as well as well recorded material.

I stopped recording at 48 and went 88.2. I really can't say if mathematically it makes that much a difference but it was what I chose to do, but its only one conversion versus two, so I deliver 88.2 32 bit float mixes, and the mastering engineer prefers it at least according to him.

So whatever path you chose, these are some facts. Besides. Lower latency recording, less upsampling in the plugin as as well.
I notice a difference.
Random listeners notice the difference when we recorded at 48k vs 88.

Again, the room has alot to do with it, converters, ect.

Bottom line, if you can't hear the difference, and it is subtle, then it isn't worth doing. 48k will be fine.
There is no need to sweat over what you can't hear.

propower
12-29-2013, 11:34 PM
Nope. Latency of ADDA conversion is cut to half if you double the sampling rate. That is because converters work in samples and not milliseconds. Same amount of samples is dealt with faster if you up the SR.

In the AVID I/O the A/D D/A conversion time is uniquely different at 44.1k and 96k giving more conversion time reduction at 96kHz than one might expect.. The conversion time is roughly = to the LLM delay.

44.1k conversion time = 1.9ms
96kHz = 0.47ms

Total latency in the system is the conversion time + (2X the chosen buffer)/(SR) + any delay for getting data on/off the relevant bus (PCI, FW USB etc). For PCI or TB systems the data bus delay can be assumed to be zero.

So for instance: Lowest possible latency with HD Native TB at 96kHz is 1.8ms (64 sample buffer) and at 44.1 it is 3.35ms (32 sample buffer). I work at 96k all the time solely for latency considerations. FWIW 44.1 on old HD systems was sonically fine for me before I went Native.

Meads
12-30-2013, 12:54 AM
There's a fantastic article, that everyone should read, who wants to know the deal about SRC. Seriously, this explains it spot on. It's longish, but worth the read.

http://www.trustmeimascientist.com/2013/02/04/the-science-of-sample-rates-when-higher-is-better-and-when-it-isnt/

dr_daw
12-30-2013, 12:23 PM
+1 Great article Meads, worth the read. There is another one that is Lavry's research paper that will clearly show the detailed math, with graphs to show what he is quoted on in this article.

To the OP, I began working with ADAT's and DA88's which is what brought me to the 48kHz train of thought. We then moved onto a Soundscape system which we clocked at 48kHz. I have thought about just giving 88.2 a try (because it's closer to that 60 - 70kHz), but I always go by the 'If it ain't broke, don't fix it' program.

Also, if I made the change I'd have to have sticky notes all over to remind me to change my SR back when I wanted to open an older project ;) 24bit 48kHz is burned into my brain! Gives me more room on my hard drives as well :P

YYR123
12-30-2013, 02:28 PM
Thumbs up! Now if I could just afford Lavry converters !!!

Matt Hepworth
12-31-2013, 09:10 AM
No disrespect J, but there is a difference.



Here is the bottom line. In a proper room you will hear the subtle differences of recording at 88 or 96, granted that you aren't recording with crap gear.

The big reason for recording at 88.2 is that it is perfect math. 88.2 half is 44.1. There are no rounding errors.



48k to 44.1, is so small, your never gonna notice.



However, all the mastering engineers I have worked with have all said, whatever you record at, let them SRC.

Not all SRC's are the same.

If PT was perfect at SRC, why aren't the mastering engineers using it?



Weiss is far better in SRC. There are others that are better. Granted, if you aren't in a room where you can hear it, it doesn't matter.

It ends up listened to on ear buds.



But, listen to Daft Punk RAM. Listen to how in translates from ear buds, to laptop, to fair speakers to your studio.



96 to 48 is perfect math. But ultimately you still have to goto 44.1. That is 2 times conversion with a smaller rounding error.



Granted, this is all probably so trivial. This and that, 24, 32. 48. 96.



From my own experience and recording, there is subtle difference between 88.2 and 96k.

Pretty much to the point you really can't tell.



However, the difference between and 44.1/48k and 88.2/96k you can hear, given the room and speakers as well as well recorded material.



I stopped recording at 48 and went 88.2. I really can't say if mathematically it makes that much a difference but it was what I chose to do, but its only one conversion versus two, so I deliver 88.2 32 bit float mixes, and the mastering engineer prefers it at least according to him.



So whatever path you chose, these are some facts. Besides. Lower latency recording, less upsampling in the plugin as as well.

I notice a difference.

Random listeners notice the difference when we recorded at 48k vs 88.



Again, the room has alot to do with it, converters, ect.



Bottom line, if you can't hear the difference, and it is subtle, then it isn't worth doing. 48k will be fine.

There is no need to sweat over what you can't hear.


Lots of good info here as well as some good advice.

But, some potentially INcorrect info about perfect math between certain sampling rates for the majority of users. What you stated is wrong 99% of the time for average users. Everyone should read up on synchronous versus asynchronous sample rate conversion.

For 99% of SRC (asynchronous), there's no mathematical advantage from 96 to 48 versus 88.2 to 48, because it's all UPSAMPLED first, to the LCM, then divided using the exact same algorithm to reach the target sample rate for the output.

JFreak is technically correct in most instances. You are technically correct as well, only in far fewer instances (synchronous SRC).

The Weiss is Asynchronous, I believe.