Rendering 24 bit tracks w/ 32 or 64 bit effects.

[Frippertronix]

If your unprocessed track is 24 bit, is there a reason to render or capture audio above 24 bit if you are using 32 or 64 bit internal processing effects (i.e. capture it at 32 or 32 FP).

Thanks!

Dave

[greendoor]

The internal audio streams between a VST host and effects are 32 bits float. Whether the effect uses 64 bits internally won't give you any more resolution than 32 bits, but those 32 bits might sound better for it.
I believe you would be recording with a 24 bit soundcard? In that case, the 32 bit files don't have any more resolution than a 24 bit file, but they just have some headroom which means you can't clip that file during processing. So there is probably no advantage in rendering any higher than 24 bits. But I use 32 bits myself if I know i'm going to re-import into SX.

[Frippertronix]

The internal audio streams between a VST host and effects are 32 bits float. Whether the effect uses 64 bits internally won't give you any more resolution than 32 bits, but those 32 bits might sound better for it.
I believe you would be recording with a 24 bit soundcard? In that case, the 32 bit files don't have any more resolution than a 24 bit file, but they just have some headroom which means you can't clip that file during processing. So there is probably no advantage in rendering any higher than 24 bits. But I use 32 bits myself if I know i'm going to re-import into SX.

If the rendering is being done after the processing (which, as you say, is done at 32 bit FP), then shouldn't you do the final render at 32 bit FP to avoid having any clipping of the final track?

[greendoor]

Thats why I use 32 bits for my project, and export my mix at 32 bits. But at the moment I ensure that my mix file doesn't clip - I leave headroom for the mastering process. Since I do that, I would probably not lose any quality by using 24 bits. But I figure the diskspace difference is so small, and Cubase has to convert everything to 32 bits anyway, so why not stay in 32 bits as long as you can? But - recently i've been trialling a new idea (based on the logic behind analog summing). With 32 bits, there is actually no need to restrict the mix file to 0dB. So instead of lowering the master fader (or track faders) and throwing resolution away - why not let the mix exceed 0db? (It doesn't clip - only clips if you feed that to your 24 bit fixed D/A soundcard!) The mastering process invariably requires compression and limiting - so why not keep as much resolution as is available, so the final compressor and limiter is working with higher resolution material. In other words - why throw away 6 db of resolution at mixdown, only to apply 6db of zero-content make-up gain at mastering? It's a radical theory that I haven't heard anyone else suggest before - but try it, it works! Whether you hear any difference in the final result is about as debatable as whether analog summing boxes are worthwhile. But the theory sounds good - somebody correct me please if i'm wrong!

[C00kie]

If your unprocessed track is 24 bit, is there a reason to render or capture audio above 24 bit if you are using 32 or 64 bit internal processing effects

Nope, I actually do not see any reason at all. You host should do all nescessary bit depth conversions for you, so I see nothing to worry about.

[kritikon]

If your unprocessed track is 24 bit, is there a reason to render or capture audio above 24 bit if you are using 32 or 64 bit internal processing effects

Well, your dry sound is not going to gain anything, but your FX will be recorded at higher resolution, so actually there is a potential reason to render at 32bit. With something like reverb tails, 32bit should technically be better than 24bit - same with long delays. I can't see much of a reason to bother with FX like choruses, flangers, valve sims, Eq etc though. Depends if you can actually hear the difference or not. I have some 32bit reverb impulses, and they sound better to me...BUT ...I can't definitively say whether it is the increased bit depth that improves the smoothness of the tails, or simply that they are recorded better than most of the other 24bit ones. Also - they are those Noisevault impulses of an EMT plate and a Quantec ... so it also could just be that they are very fine reverbs that make them sound better

So ... yes ... some FX will benefit from higher bit depth -> although I'm sure 24bit is plenty good enough for most applications.

[greendoor]

If you are taking this seriously, like mastering engineers do, then you actually want to avoid as many bit depth calculations as possible. Something is lost each time - which is why I keep everything 32 bit float until the final resample to 16 bit CD. The difference will be in the very low resolution stuff. Most peoples soundcards and listening environments will be too noisy to notice any difference. But there are still people who have good gear and can hear every little detail. MP3 users need not apply.

[kritikon]

Yep...I'll concur with Greendoor here - I use 32 bit nowadays for anything generated within my PC - obviously you can't record externally at 32bit yet, but anything else done inside your DAW will be calculated at 32bit float so it just means less risk of loss if you use 32bit for everything. It doesn't use any more CPU, and it's no extra hassle but at least it guarantees better precision.

If ever you do any bouncing down, even to 24bit, you are still going from a higher bit depth to a lower one, which technically means dithering is needed, although you should only use dither once at the end. So any bounce can introduce flaws. Personally I can't hear it, but I'd rather just not take the chance when it's so easy to avoid. It's not like the argument over high sampling rates - with high sample rates you have real drawbacks such as increased CPU load to take into consideration. High bit depth doesn't really have any drawback (larger files, but that's almost a moot point as HDs are so big nowadays and CDRs are so cheap). Just set your render options to 32bit, leave it and don't worry about it any more.

If you really have to keep your file size down for whatever reason, use 24bit but try not to bounce things down would be my opinion. That's exactly why I use 32bit nowadays - I do alot of rerendering with FX - especially things like reverbs (to save CPU). It's not unknown for me to rerender something several times and the more you render, the more likely lower bitdepths will have an effect.

[Left Headphone]

Upsampling is good... Ok!.... Ok!

[martian]

Thats why I use 32 bits for my project, and export my mix at 32 bits. But at the moment I ensure that my mix file doesn't clip - I leave headroom for the mastering process. Since I do that, I would probably not lose any quality by using 24 bits. But I figure the diskspace difference is so small, and Cubase has to convert everything to 32 bits anyway, so why not stay in 32 bits as long as you can? But - recently i've been trialling a new idea (based on the logic behind analog summing). With 32 bits, there is actually no need to restrict the mix file to 0dB. So instead of lowering the master fader (or track faders) and throwing resolution away - why not let the mix exceed 0db? (It doesn't clip - only clips if you feed that to your 24 bit fixed D/A soundcard!) The mastering process invariably requires compression and limiting - so why not keep as much resolution as is available, so the final compressor and limiter is working with higher resolution material. In other words - why throw away 6 db of resolution at mixdown, only to apply 6db of zero-content make-up gain at mastering? It's a radical theory that I haven't heard anyone else suggest before - but try it, it works! Whether you hear any difference in the final result is about as debatable as whether analog summing boxes are worthwhile. But the theory sounds good - somebody correct me please if i'm wrong!

sorry to be negative but seeing as the human hear cant apreciate resolution much beyond 20bit, does seem ultimately pointless.

[griels]

sorry to be negative but seeing as the human hear cant apreciate resolution much beyond 20bit, does seem ultimately pointless.

Probably, although in theory there could be initially imperceptible elements of an audio signal that were made perceptible by digital effect processing (heavy multiband compression for example). And each processing stage introduces quantisation errors which can be worked around by using extra bitdepth. In practice I suspect 32 bit is overkill though...

[Frippertronix]

sorry to be negative but seeing as the human hear cant apreciate resolution much beyond 20bit, does seem ultimately pointless.

Probably, although in theory there could be initially imperceptible elements of an audio signal that were made perceptible by digital effect processing (heavy multiband compression for example). And each processing stage introduces quantisation errors which can be worked around by using extra bitdepth. In practice I suspect 32 bit is overkill though...

Does 32 bit use significantly more hard memory than 24?

[griels]

Does 32 bit use significantly more hard memory than 24?

A third more hard disc space... It's up to you whether you think that's significant...

Bear in mind that it'll use a third more hard disc bandwidth too, which might be vaguely significant if you're bouncing a gazillion tracks...

[greendoor]

Upsampling is good... Ok!.... Ok!

Um no - upsampling just adds insignificant bits. Our raw 24 bit input is converted to 32 bit float in our DAW, which is unavoidable. My point is, from that point on, don't down-sample until you absolutely have to.

Bear in mind that our raw tracks are frequently compressed and limited several times before the final product is made. Compression multiplies - so if you have a 2:1 followed by 5:1 followed by a 10:1 compression ratio - that's actually 100:1 compression. Small defects around the noise floor could be brought up to everyone's attention.

Sometimes the theory doesn't work. For example, in theory I shouldn't hear any difference between 44.1 or 96 kHz sampling because my ears can't hear about 20kHz. BUT - I find that my VSTi synths sound so much better at 96kHz, and it's nothing to do with the higher frequency range, it's something to do with aliasing well inside the 20kHz range that goes away when I use 96kHz. So I don't trust theory - it's what my ears tell me that counts.

[useruseruser]

If your unprocessed track is 24 bit, is there a reason to render or capture audio above 24 bit if you are using 32 or 64 bit internal processing effects (i.e. capture it at 32 or 32 FP).

Thanks!

Dave

Not really. You wouldn't be able to hear the difference.

But you probably would hear the difference of a 24bit calculated effect and a 64bit calculated effect (internal processing).