Help: DAW - Mixer - DAW routing and bit depth

[a1236040]

Hello, the question is simple

The scenario:
I work with Nuendo3 with about 20 tracks. I process every tracks with good plugins as UAD, Waves and URS. Now I sum OTB via Speck Xtramix setting the volume and pan controls into the mixer (I live at zero those on the DAW), finally I record the mixdown on new track into nuendo.

The question:
Will I do better to record and work at 32bit or 24bit ?


I know that converters work at 24bit, but Nuendo internally works at 32bit and many plugins work at 32 or 64bit but I fear the redundance in the transformation of the bit depth in the various steps with quantization noise results.

This is my routing:

file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |-> Mixer mixbuss -> Analog Comp. -> In ADC -> new track Nuendo
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |




Thank you

[Hammer Mark]

I work with 32-bit float files ITB based on the recommendation from the Tischmeyer Internal Mixing DVDs. It had something to do with headroom -- that, if necessary, you could go above 0 dB on the channel without digital clipping. Also, I think that some of the UAD plugs up-convert to 32-bit, so you save that step if all of your files are already there.

[a1236040]

Thank you for your reply but I go out (OTB) for summing ....

[Thescort]

If the numbers in your daw do not top out at 0, they do not likely reflect dBFS. In digital, no matter the bit depth, you cannot exceed 0dBFS.
The way I have interpreted "digital headroom" is more like dynamic range:
The idea of headroom in digital is different than analog. Since no matter what, the user is limited to the same scale regardless of bit depth. Headroom could be thought, instead of "how far can you push it" to, "at how low a level can you effectively record".
In digital you need to consider:
Noise floor,
Amplitude quantization,
errors/error correction
and other things directly or indirectly related to Bit Depth.

In digital, you want to record at the highest level possible (so as to occupy as many bits as possible) while leaving yourself room so you NEVER touch 0dBFS. THe longer the word, the lower level you can effectively record essentially.

The bigger word lengths mean more extraneous bits which makes losing bits less detrimental. It also helps various methods of error correction to have a larger number of bits. Your noise floor is lower so you get a greater s:n allowing for lower levels. You can occupy more bits at lower levels so you get greater accuracy and fewer quantization errors at lower levels.

Larger bit depth allows you to more accurately record at lower levels ultimately giving you are greater usable dynamic range.

There is more, but I might not be totally accurate and I think the point is clear. The advantages of floating point are also immense (mathematically speaking at least). Especially with plugins and ESPECIALLY with summing (either groups or your mix buss) since it helps reduce errors caused by rounding! BUt since you sum outside the box that may not be relevant.

There are very few converters that actually convert to 32 bit float. A DAW will operate at 32 bit float to accommodate all the math involved in processing sounds in plugs, and in summing. Its probably always a good idea to record at 24 bit and use a good dither when dropping to 16 bits. However, you always must consider that you will be dithering down to 16 bit for cd use anyway, and you should judge the quality of the 16 bit end product, not the 32bit-float. A lot can happen when dropping your bit depth.

I think I just ran out of steam. Hopefully this helps, and hopefully its accurate, although I know there are a lot of great minds who can help fill in where I strayed from the path.

[a1236040]

Than you for your post

I record at -18/-12dBfs living headroom for plugin processing, after I sum out making the levels with mixer-faders and recording the sum into new track.

The only dubt is : 24 or 32bit ?
And if I use 32bit, is it better to put a dither (for 24bit) on every track that go to the mixer ?

When mastering I use a good M+bit dither.

[FeatheredSerpent]

All UAD plugs operate at 32bit, your cpu and OS (normally) are at 32bit, and it will actually take less cpu load if you run a project at 32bit, so if you have the space for the larger files, go for it.
All your processing will benefit (in theory).

As for dithering, you only need to do that with the stereo master file, not for each individual channel, so you can come out into your desk without dither, record the stereo mix back in at 32bit then dither that final file.

I'm going on what I have learned from reading about this stuff, and opinions tend to change on it depending on who you are talking to, so this is not set in stone at all, but what I have written makes complete sense to me (at this point in time )

[a1236040]

FeatheredSerpent:
Yes, I have a lot of space on my HD


And I totally agree with you


Thanks

[Thescort]

COnsider that most converters only convert to 24 bit. You might want to look into what it really means to run a project at 32 bit. Unfortunately I don't know the answer.

[a1236040]

Originally Posted by Thescort

COnsider that most converters only convert to 24 bit. You might want to look into what it really means to run a project at 32 bit. Unfortunately I don't know the answer.

It is the only my real dubt

[rashman]

Originally Posted by Hammer Mark

I work with 32-bit float files ITB based on the recommendation from the Tischmeyer Internal Mixing DVDs...

Originally Posted by FeatheredSerpent

All UAD plugs operate at 32bit, your cpu and OS (normally) are at 32bit, and it will actually take less cpu load if you run a project at 32bit, so if you have the space for the larger files, go for it.
All your processing will benefit (in theory).

As for dithering, you only need to do that with the stereo master file, not for each individual channel, so you can come out into your desk without dither, record the stereo mix back in at 32bit then dither that final file.

I'm going on what I have learned from reading about this stuff, and opinions tend to change on it depending on who you are talking to, so this is not set in stone at all, but what I have written makes complete sense to me (at this point in time )

I saw that DVD too. He also says something about 32 bit files giving the CPU less load. But is'nt Nuendo internally do all processes at 32bit Foat and therefore the data send to CPU and back is already 32 bit regardless of the File format? Am I making scense? Anybody on that...?

Regarding Dithering : if you're sending out each track individually, then there can't be any benefit in dithering each buss. But if you're sending out Groups of tracks to each buss and then out to Analogue, then you've already doing summ Somming ITB and dithering each Buss could make scense.
Arash

[a1236040]

This is my routing:

file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |-> Mixer mixbuss -> Analog Comp. -> In ADC -> new track Nuendo
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |
file-track Nuendo -> Plugin process -> Out DAC -> InLine Mixer CH -> |



24bit or 32bit project ?

If 32bit project: is it necessary to put the dither on every nuendo track (maybe Apogee UV22HR for 24bit) ?

Does someone work in this way?

Thank you

[narcoman]

NO DAW works at 24bit through the mixer. Nuendo is floating point. The only difference you're talking here is the difference between the stored audio format. That is down to personal preference because at ANY given stage the file is ALWAYS a 24bit audio file with 8 bit exponent. Sure - less mucking about with formats - but ultimately they are the SAME. PT uses a 48bit fixed point architecture in the mixer (much better than 32bit float for many reasons) - in fact I think it might be 56 bit these days. However - it's passing of audio to plugins is 24bit , so care must be exercised in NOT clipping.

Your audio is going in to your DAW at 24bit. You're working at a floating point bit depth (32 or 64 - can't remember what you said). Once the digital audio is being processed within your system - you're passing stuff around at 32bit (or 64) anyway. So at this stage there is no difference. The time it DOES make a difference is storing audio at intermediate stages for further work.

[a1236040]

Please, dont' make more confusion.... I use Nuendo and not ProTools and I have read that 32bit float is better than 48bit fix... Nuendo works internally at 32bit float and the same for the mixbuss.
So, I will appreciate if some great mixer engineer can respond. Thank you

[Faderjockey]

I'm not %100 on this..

But my feelings... IF you were mixing only ITB then I'd say 32bit.

but since your going to the Speck I'd say 24bit just to make less math with the converters..

Plus I would think (not sure you do this) if you get files from clients or other studios most are going to be 24bit. So you'll save time there as well not needing to convert it all.

I only see it helpful when your doing all the math in the box..
Again not %100 sure. It's just my feeling.

You could always do both when you have time to see for yourself if it's worth it..
I'm going to say doubt it.

[kt66]

There is NO advantage to recording at 32 bit/float unless you are recording with plug-ins on the Nuendo input tracks (which I'm fairly certain you are NOT doing and would generally be a bad thing to do) or you are performing offline plug-in processing on your audio files - even then you are highly unlikely to see any difference unless you do crazy things (such as reduce the level by 100 db in an offline process and then boost it back up after the fact). You will only be creating larger audio files and wasting disk space. There is no added headroom etc. by using 32/float files with the way you are working.

Nuendo mixes and processes plug-ins using floating point math no matter what the bit-depth of the audio files - your 24 bit files are processed at the full floating point resolution (assuming your plug-ins are floating point - and most VST plugs are).

Stick to 24 bit. Dither if you like, but the effects of the dither should be well below the limit of your hearing and even further below the noise present on the tracks from various sources. I only bother dithering when going to 16 bit or lower.

[Mike Brown]

Originally Posted by a1236040

I have read that 32bit float is better than 48bit fix...

So it must be true!

I would store that audio at 24.

More than enough usable dynamic range.

[a1236040]

Thank you at all for the reply.

I don't receive audio file from client, I work for me and tracking into Nuendo. I haven't problem of disk space and power on my PC, I only want better sonic results and understand if tracking at 32bit is better than 24.
When I worked ITB, tracking to 32bit did a great difference than 24bit.
The only issue now is if I go out to mixer with a 32bit audio and the DAC works at 24bit, will this truncation give me distortion?
I have see many studios that work in this way, as described above in my post, and I hope that someone of them can reply.

[kt66]

Originally Posted by a1236040

When I worked ITB, tracking to 32bit did a great difference than 24bit.

Tracking at 32 bit should not make ANY difference under the vast majority of circumstances - how did you actually determine it was better???? I'm sorry, but I just have to disagree with this statement as you should have seen no difference. Sadly, there is an awful lot of confusion about the use of 32/float files in Nuendo and very few people seem to actually understand how it works and what it means to their operation. I've argued this point with others before and they ended doing a null test between two mixes using Nuendo - one using 32/float files and the other using 24 bit fixed and guess what? The mixes nulled completely! No difference!

24 bit audio files have more resolution than even the best converters can actually provide and unless you are doing some kind of crazy offline processing of your audio files, THERE IS NO DIFFERENCE with using 32/float files in Nuendo - only larger audio files. Even if you have plenty of drive space, why make larger files for no good reason?

[FeatheredSerpent]

But it's with the processing where the benefit is to be had I thought.

The dither question - If you're working with a 32 bit project, even if it is all completely itb, do you need to have a dithering plug-in on your master outs?
No.
So, you don't need to dither those individual channels either if you are sending them otb.

With 32 bit plugs (UAD are 32 bit, Powercore plugs are 24 bit, don't know about the others), and you are inserting them on the tracks before they come otb (I do the exact same thing), then the depth of processing detail will be enhanced if the files are at 32 bit, because as someone pointed out in another thread, this gives a far greater resolution for processing, the recorded files themselves won't benefit, but you will get a more detailed and accurate level of processing on those files.
Then again, people are saying that even with a 24 bit file, it is being upsampled for processing anyway, in which case why would we even need a 32 bit option for projects in the first place? This is something I don't know.

As for the bit depth of the mix engines, and 32 point float being better than 48 fixed or whatever, I don't know about that, BUT..
My Spirit 328 mixer has a 56 bit mix engine, and I can hear a difference from when a mix is summed internally in cubase, and when all the individual channels are routed out to the desk and mixed there, the mixes sound bigger and more full of life.

The desk though has made me feel much happier about mixing in general and so I don't know how much of that bigger more pleasing sound is a placebo effect from having a nice big piece of mixing hardware in front of me, and how much is a true real life benefit, but the first time I used the desk I heard a difference, and I can still hear it.

So for me the sonic benefit for just a file itself being recorded at 32 bit is very much in doubt, but the increase in processing accuracy seems to be a solid fact, and pair that with the supposed cpu economy and it makes sense to use 32 bit, especially if you have plenty of storage.

When transferring projects to someone that is using 24 bit, then it isn't any hassle at all to batch convert them with
Voxengo

once you've bounced the files.

[kt66]

Originally Posted by FeatheredSerpent

So for me the sonic benefit for just a file itself being recorded at 32 bit is very much in doubt, but the increase in processing accuracy seems to be a solid fact, and pair that with the supposed cpu economy and it makes sense to use 32 bit, especially if you have plenty of storage.

The processing accuracy increase is only for OFFLINE processing of the audio files. I tried to make this as clear as possible. Plug-ins on track inserts are processed using floating point math just the same whether you are using 24 bit or 32/float tracks - this is a fact and not my opinion. Plus, the *increase* in processing accuracy that you would see from using 32/float files when you may possibly be doing some offline processing is in the vast majority of cases, going to be negligible.

Finally, the alleged better CPU economy that some are talking about is pretty much non-existent on modern computers. The only people who saw any benefit had very slow older systems/drives etc. and the difference was quite small at best. I saw absolutely no difference in CPU usage between 24 and 32/float files with any of my machines.

All of what I am saying can be proven with null tests (performed correctly) - please try it yourself if you do not believe me as many seem to be just guessing at things based on lots of sketchy information. This is not about which microphone or preamp one prefers, but simply about math.

[FeatheredSerpent]

Well I'm open to everything, I know I don't have many concrete answers at all.
But when you say with offline processing only, what is the difference with software using floating point maths offline or directly as with an insert?

It's still the same maths on the same type of file isn't it?
I am not aware that an offline process will give you a different/better result than that same plug running live on an insert, I've never read nor heard anything to that effect, not saying it's not true though. I don't expect you to give me a full run down on exactly how floating point works (well, I want you to but I don't expect you to!)

And I think also that when you talk about marginal improvements, then everything discussed on the forum is pretty much about marginal improvements, but if it gets us closer to what we would hear in the analogue domain then it's all good isn't it?

It's like with 44.1 compared to 96 sample rates, on the surface it's a marginal improvement for most people, but we're talking about esoteric values a lot of the time, and how we sunconciously perceive stuff such as frequencies over and above the standard 20khz range for example.

[Peter Z]

It is true that Tischmeyer recommends 32bit in all of his books and articles - especially for ITB mastering.

But one question from a1236040 has not been answered yet. If he mixes/sums OTB does he need dither on every track? I think so, but i don't know.

I want to know too. I have some older 32bit cubase 32bit projects and switched to a Toft ATB now. I want to remix the old songs. Dither all tracks??

Thank you for help!
Peter

[a1236040]

Peter Z: do you work at 24bit with Toft now ?

[Thescort]

You will only need dither if you are actually digitally converting a file from a higher bit depth to a lower one. If you are sending the file out of your converters and recording them to another medium (that is to say: D-A conversion, and then A-D conversion at the same or different depth) you shouldn't need dither.

[Peter Z]

a1236040:

Future projects i will record at 24 bits. I have much less ITB processing now so i think there will be no/not much difference.

I was just worrying about my existig 32bit projects but i think i will not care about dithering. Thesort might be right. If i convert to analog and resample the 2-bus mix to digital there should be no need to dither. It's only a digital/digital issue.

So it doesn't matter in what format we record. That's fine!

Peter

[rashman]

Originally Posted by kt66

The processing accuracy increase is only for OFFLINE processing of the audio files. I tried to make this as clear as possible. Plug-ins on track inserts are processed using floating point math just the same whether you are using 24 bit or 32/float tracks - this is a fact and not my opinion. Plus, the *increase* in processing accuracy that you would see from using 32/float files when you may possibly be doing some offline processing is in the vast majority of cases, going to be negligible.

Very True and well put. I was going to say the same thing but couldn't have explaind that well.thumbsup

Originally Posted by kt66

Finally, the alleged better CPU economy that some are talking about is pretty much non-existent on modern computers. The only people who saw any benefit had very slow older systems/drives etc. and the difference was quite small at best. I saw absolutely no difference in CPU usage between 24 and 32/float files with any of my machines.

Thanks for that info kt66. I've had this question for some time now. I feel I can trust you on that.

Originally Posted by FeatheredSerpent

But when you say with offline processing only, what is the difference with software using floating point maths offline or directly as with an insert?

It's still the same maths on the same type of file isn't it?
I am not aware that an offline process will give you a different/better result than that same plug running live on an insert....

kt66 could you answer that too? I can't explain very well but I think the difference is only for cases (in offline processing) when you're gonna overload the dynamic range provided by 24 bit (wich would clip if rendered on a 24bit file) or in cases when you're gonna have extreamely lo final levels (printed on the file) that artifacts could be created even in 24bit resolution. correct me if I'm wrong.

Originally Posted by Thescort

You will only need dither if you are actually digitally converting a file from a higher bit depth to a lower one. If you are sending the file out of your converters and recording them to another medium (that is to say: D-A conversion, and then A-D conversion at the same or different depth) you shouldn't need dither.

I suppose theorically (aside from any difference that could practically be heared or not) it would make sence to dither to 24bit at the last stage of any signal sent to D/As. The Daw internally processes data in 32bit float/48bit/64bit/whatever, and the converter is a 24bit D/A so recieves only 24bit data stream. Therefore the very last stage at the last bus/output of any daw must be sending a 24bit signal to the converters. Hence a down sampling should pe performed at the very last satge and so a 24bit dithering could theoretically make sceence. Again correct me please if I'm wrong.
Arash

[trapper]

"But when you say with offline processing only, what is the difference with software using floating point maths offline or directly as with an insert?

It's still the same maths on the same type of file isn't it?
I am not aware that an offline process will give you a different/better result than that same plug running live on an insert...."

Processing a 24bit offline with VST plugins means that ANY processing will be unceremoniously truncated to 24 bit. VSTs use either 32bit or 64bit for processing so if u apply an offline process to a 24bit file, the host program (e.g. wavelab, nuendo, any vst host) will convert the 24bit fixed int file into a 32bit float in ur pc/mac's memory, pass it to the vst effect and, once processed, the host will truncate the 32bit result back down to 24bit fixed int. This is the reason why I personally up any audio files (regardless of their bit depth) to 32bit when I apply effects offline as this will maintain the integrity of the processing done to the file in question. Other than this and gain operations, there is no real advantage to up the bit depth of a 8/16/24bit file as once the file is used in ur sequencer, it will be converted to 32bit automatically, so u will essentially be wasting disk space.

Hope that helps

[trapper]

"PT uses a 48bit fixed point architecture in the mixer (much better than 32bit float for many reasons)"

"I have read that 32bit float is better than 48bit fix..."

hah if it was that simple, EVERYONE would be fixed/float

I'd imagine the reason PT is fixed int is fixed point int DSP chips were faster, cheaper and more readily available than their slower, more expensive floating point counterparts back in the day

[Hammer Mark]

I'm curious why the OP would bother with OTB summing in the first place. I mean I get it if you have nice OB gear or a cool analog console, but it seems to me that DAWs do a perfectly acceptable job of adjusting volume and pan.

[a1236040]

Originally Posted by Hammer Mark

I'm curious why the OP would bother with OTB summing in the first place. I mean I get it if you have nice OB gear or a cool analog console, but it seems to me that DAWs do a perfectly acceptable job of adjusting volume and pan.

Simply for to give analog impact without losing in resolution.
If you low the volume digitally you lose the bit resolution, if you low the volume with analog fader you don't lose any digital bit into original audio signal.
I put my tracks into DAW at -3dBfs and to set the volume level with analog fader for maintain total bit resolution and better analog glue.