Up Rezing Audio in Post?

[Peter Moretti]

When doing a lot of color correcting, it's a good idea to use a 10-bit project even if the source footage's color depth is only 8-bit.

Is there a similar philosophy with audio? Would it be beneficial to convert 16-bit sound to 24-bits before doing all the audio sweetening and post work?

Thanks much.

[Petri Kaipiainen]

All better audio editors use internally 32 bit processing anyway, so there is no point in you uprezing the files first. You could say the editor converts everything to 32 bit without asking, then outputs the original resolution again.

[Gary Nattrass]

I record all my audio at 16 bit 48k but when using pro tools to post the session I do everything at 24 bit.

The reason being that is gives more headroom on the mix buses so there is more bandwidth and less chance of distortion when compressing mixes etc. Mixing in the box has lots of sources going down to one stereo bus so if they are all at 24 bit bandwidth it helps.

I still export as 16 bit 48k but it is good to have the internal workings of the dub at 24 bits.

Hope this helps.

[Petri Kaipiainen]

Originally Posted by Gary Nattrass

The reason being that is gives more headroom on the mix buses so there is more bandwidth and less chance of distortion when compressing mixes etc.

???

How does converting 16 to 24 bits give more headroom? Converison just adds 8 zeros to each sample, data itself stays at the 16 most significat bits as before, that is, just as close to 0 dBFS as it was recorded.

Internal processing at 32 bits (48 bits in PT) takes care of all possible rounding errors, be it 16 bit or 24 bit file, so I still see no reason to upconvert a 16 bit file.

[Stan Harkleroad]

If you record in 24 bit it gives a lower noise floor and allows you to record at a lower level and amplify without adding considerable noise. Converting 16 to 24 doesn't give you that same advantage. You're not going to add useful information to your file by converting to 24 bit. I would see it as similar to converting MP3 to WAV. It might add digital information to the file and bloat it but it's not actually going to make it better than it was before.

[Gary Nattrass]

Originally Posted by Petri Kaipiainen

???

How does converting 16 to 24 bits give more headroom? Converison just adds 8 zeros to each sample, data itself stays at the 16 most significat bits as before, that is, just as close to 0 dBFS as it was recorded.

Internal processing at 32 bits (48 bits in PT) takes care of all possible rounding errors, be it 16 bit or 24 bit file, so I still see no reason to upconvert a 16 bit file.

No I am not upconverting at all just setting the session to be 24 rather than 16 bits the system is then running at 24 bit audio resolution but the files are still 16 bit, it the same as taking a 16mm print and upscaling it to 35mm it is still 16mm resoloution but within a 35mm frame.
I appreciate that internally the computers processor is running faster but setting the session to 24 rather than 16 regardless of the source material bit rate allows a better bandwidth in the mix buses once you start combining multiple tracks down to a stereo mix bus.

[Petri Kaipiainen]

I am not talking about processor data bus widths, but about the fact that audio editors "uprez" ALL files to 32 bit word length to avoid rounding errors. Also 24 bit files. That is 8 bits more than uprezing 16 bit file to 24 bits.

My way: 16 bit file -> editor converts to 32 bits for processing.
Your way: 16 bit file -> you uprez to 24 -> editor converts to 32 bits for processing.

What is the point?

Everything happens at 32bit wordlength anyway inside the editor.

[Gary Nattrass]

Sorry but I am not up-resing at all just setting the pro tools system to be 24 bit when using 16 bit files.

My master audio tracks are 16 bit 48k and are exported via OMF as 16 bit 48k.

They are imported into the system at 16 bit 48k but the pro tools system is running at 24 bit resolution.

The mixing is then in 24 bit domain and whatever the internal computer is running at but the mixdown master mix is still 16 bit at 48k but with all the session buses working at 24 bit resolution.

When I worked at AMS Neve we used to get this confusion all the time and people think that extra bits are added on or truncated in systems, they are not! 16 bits of information will still be 16 in a 24 bit or even 32 bit res system. Also in the case of AES EBU people think that it is 24 bit audio, it is not it is only 20 bits with 4 bits for parity timecode and user bits.

[Marco Leavitt]

I have heard that post houses routinely uprez to 24 bit, but I don't understand the benefit of it. I did ask a professional sounddesigner (who did the post audio on Mad Men), and he had never heard of anyone uprezing like that.

[Petri Kaipiainen]

My undestanding is that Pro Tools handles everything in 48 bit fixed point presicion, so setting 16bit files to be processed in 24 bit project would not change anything at all.

Ah, well...

[Gary Nattrass]

Originally Posted by Petri Kaipiainen

My undestanding is that Pro Tools handles everything in 48 bit fixed point presicion, so setting 16bit files to be processed in 24 bit project would not change anything at all.

Ah, well...

Think of it as putting a half pint in a pint pot there is less chance of spilling it! or 500ml into a 1000ml euro glass

[Petri Kaipiainen]

Pro Tools has only a two pint glass, it does not help pouring your half pint to a pint glass first as PT then pours it into a two pint glass anyway.

By the way, there is a major fault in you analogy of different size glasses and 16 - 24 bit audio: your 16 bit audio will slosh at the top of the bigger (24bit) glass, not at the bottom. Water fills the glass from the bottom, but 16 bit audio file converted or processed at 24 or 32 bits stays at the top, no sloshing/headroom added at all. All extra space added is at the bottom.

[Gary Nattrass]

Originally Posted by Petri Kaipiainen

Pro Tools has only a two pint glass, it does not help pouring your half pint to a pint glass first as PT then pours it into a two pint glass anyway.

By the way, there is a major fault in you analogy of different size glasses and 16 - 24 bit audio: your 16 bit audio will slosh at the top of the bigger (24bit) glass, not at the bottom. Water fills the glass from the bottom, but 16 bit audio file converted or processed at 24 or 32 bits stays at the top, no sloshing/headroom added at all. All extra space added is at the bottom.

Thanks for that it gives me an excuse to have two pints tonight!

[Chris Rackauckas]

All that it does is add to the dynamic range. The main thing you want is something that edits in 32 bit floating point. That this does is it extends the dynamic range so that the top of the dynamic range is far above 0dbfs.

Why does this matter? It has nothing to do with the sound really, if you aren't stupid it will sound the same. But what it will do is get rid of all internal clipping and the only point you will be able to clip is at the master fader where it dithers it back down to 24bit (or in your case 16). This means you can run a compressor overboard, but it won't clip in the system and therefore your audio quality is preserved... until it needs to go to 24 bit again. That's alright because all you do is pull down the master fader and you are fine. Without 32bit float, you have to go through and find out WHICH STEP is clipping to solve it instead of just bringing down the final before dithering. So there is an advantage for floating point.

The advantage for 24bit vs 16bit after it's already been recorded in 16bit can only come from increasing the dynamic range, and this can only be done by an expander. If you aren't frequently using expanders (usually, because of the use of compression, you are killing the dynamic range anyways, and not increasing it), there should be no point in just going to 24bit. But making sure you edit in 32bit floating point within your system is a must.