Converting bit depth/sample rate when rendering

[sirvoice]

Hi guys,

I am rendering a project for mixing (all stems individually) and have a couple questions. Apologies in advance for the noobness coming.

Unfortunately some sounds were recorded at 44k and some at 48k. If i render the whole project at 44k should that sound fine? Live says it will render project at 48k then downsample to 44, which it then does. Does that mean it is upsampling then downsampling...Will this effect sound quality?


Also, when rendering @ 24bit after rendering Live say "converting bit depth" of each track and does so. But in the session they say they were recorded at 24...So there should be no conversion....Any thoughts here?

[Stromkraft]

Hi guys,

I am rendering a project for mixing (all stems individually) and have a couple questions. Apologies in advance for the noobness coming.

Unfortunately some sounds were recorded at 44k and some at 48k. If i render the whole project at 44k should that sound fine? Live says it will render project at 48k then downsample to 44, which it then does. Does that mean it is upsampling then downsampling...Will this effect sound quality?


Also, when rendering @ 24bit after rendering Live say "converting bit depth" of each track and does so. But in the session they say they were recorded at 24...So there should be no conversion....Any thoughts here?

It's all 32 bit when it hits the master. And the summing process is 64bit.

[Tarekith]

If you ultimately need the files at 44.1, then just set that as the render option and you're good to go

[rev_engineer]

Sorry to revive an old thread, but...

Also, when rendering @ 24bit after rendering Live say "converting bit depth" of each track and does so. But in the session they say they were recorded at 24...So there should be no conversion....Any thoughts here?

I've been trying to find an answer to this as well. I'm rendering 24bit recordings with export set to 24bit, but the dialog always says "converting bit depth". Is it something I need to worry about? Surely there should be no conversion required?

[Angstrom]

Sorry to revive an old thread, but...

Also, when rendering @ 24bit after rendering Live say "converting bit depth" of each track and does so. But in the session they say they were recorded at 24...So there should be no conversion....Any thoughts here?

I've been trying to find an answer to this as well. I'm rendering 24bit recordings with export set to 24bit, but the dialog always says "converting bit depth". Is it something I need to worry about? Surely there should be no conversion required?

I'm not 100% sure about this case but internally Live mixes at 64bit and 32bit word length. Anywhere "Where signals are mixed" is 64 bit, if I recall correctly.

So in that case even when using only 24 bit depth waves, if the are mixed together at different volumes then the resultant word-length is longer (eg dynamic range now spans perhaps 32bit wordlength) . So when you render that theres some low dbs to deal with. At that point it's a choice to either truncate (cut off) the bits below the boundary or more usually- dither to smooth the transition to 0.

This is my understanding of the process you are seeing.

[rev_engineer]

That makes sense, thanks for your thoughts. My concern is that I've been asked to provide rendered tracks without dithering ready for mastering, but if the audio is undergoing bit depth conversion on export will it not need to be dithered?

[Stromkraft]

That makes sense, thanks for your thoughts. My concern is that I've been asked to provide rendered tracks without dithering ready for mastering, but if the audio is undergoing bit depth conversion on export will it not need to be dithered?

There's bit conversion going on all the time, in the mixer, in your plug-ins and all bits aren't used all the time anyway. You typically only need to dither once (there are a few that will disagree, but there you go).

Check out Bitter to make some of this more visible.

[fishmonkey]

That makes sense, thanks for your thoughts. My concern is that I've been asked to provide rendered tracks without dithering ready for mastering, but if the audio is undergoing bit depth conversion on export will it not need to be dithered?

if you are working at 24 bit, whether you dither or not when rendering the tracks out at 24 bit will make very little, if any, audible difference. try it and see.

if you were rendering out at 16 bit that would be another matter, but your mastering people will presumably be dealing with that end of things...

[Tarekith]

If your rendered 24bit wav is going to be the final format (ie, MFiT+), then I go ahead and dither the final master, even though it's 24bit. If it's an intermediary stage format (ie, to be mastered later), then I'd be more inclined to leave the dither off.

It's one of those ongoing internet debates, some people swear you only dither once to 16bit, other people say you should dither anytime you change bit depths. Personally, at anything 24bit or over I can't hear it either way, so I don't stress about it too much

[Angstrom]

I only dither when reducing wordlength. I consider the initial render to be one of those cases - as we are coming from a situation where there is (supposedly) potentially 64 bits of information. More often there is a lot less than that.

I think someone in this thread suggested The Bitter meter and If you slap something like that on your master you'll see what range is being produced in the set. If the displayed range is greater than your (output file) render depth then I consider a dither is a good idea. So if you are rendering from Live to a 24 bit wave and that meter shows 40 bits being used, I'd dither.

That's how I do it anyway.

[Stromkraft]

As a side-note: anyone think they know something the philosophy of different plug-in makers regarding if, when and how they apply dithering when reducing word length after oversampling in the plug-ins? I'm sure I've read a few accounts on this, but can't remember or find these right now.

[Angstrom]

oversampling is different. Oversampling relates to samplerate , not dynamic range.
Sample rate is : how many samples go by in a second
wordlength/bit-depth is: how big are those samples

oversampling is used to accurately calculate high frequency content, much higher than you are listening to. For an effect such as a distortion or cross modulation where the sidebands can shoot up to 100kHz (way above your hearing threshold of 22kHz) it's a good idea to calculate those properly by doing it at a rate which can handle such high frequency content!. So if you want to calculate audio in real time at 100kHz you need to oversample the process up to near that rate.

After it's been oversampled and then calculated it is reduced back down to the listening sample rate ... it's not through dithering, it's through interpolation. Which is pretty much like a lowpass filter.

Don't go quoting me on this as I'm explaining/phrasing it all very loosely!
Also, I'm an idiot.


tl;dr Oversampling is for sample rates, Dithering is for bit-depths

[Stromkraft]

Don't go quoting me on this as I'm explaining/phrasing it all very loosely!
Also, I'm an idiot[/b]

No, you're not! I'm the idiot this time, cause I knew this and forgot. I never learn to post before coffee. Mixing up oversampling and wordlength…

However, I did have a point. what I meant was whether anyone knew something about different philosophies on if, when and plug-in makers go from X bits to 48bit or even 64bit, say, for internal processing and then out again to X bit wordlengths.

Maybe it's not too late for coffee, after all?

[jestermgee]

Just for some extra resources if interested (I'm certainly not an expert):

http://thestereobus.com/2008/01/12/samp ... tal-audio/

A basic video that explains the principals and maths behind samplerate/bitdepth/interpolation

https://www.youtube.com/watch?v=W2-FP7twy8s

[fishmonkey]

A basic video that explains the principals and maths behind samplerate/bitdepth/interpolation

https://www.youtube.com/watch?v=W2-FP7twy8s

it's tempting to make analogies between digital audio and digital images, but there isn't really a completely straightforward one. in the video, the analogy made between sample rate as the pixel resolutions of the X & Y axes in a colour 2D image is not a good one.

probably the most direct analogy would be to consider a 2D grayscale image. in this case the sample rate corresponds to the number of pixels used for a given image size. the bit depth would then correspond to the number of bits used to represent the number of grayscale levels.