8 bit vs 32bit editing modes: When to use and why

Cliff Etzel wrote on 2/2/2009, 2:20 PM
This question is in parallel to the Vegas for Broadcast - Very frustrated thread

I've not had a need yet to use 32bit float mode as far as I can tell up to this point in time - would anyone care to elaborate on when to use 32bit mode versus 8 bit and why? Is it something to look at if one may be looking at broadcast as a distribution outlet when editing with Vegas Pro?

Cliff Etzel - Solo Video Journalist
bluprojekt | solo video journalism blog

Comments

farss wrote on 2/2/2009, 3:00 PM
The only time I'd use it is working with 10bit material such as Digital Betacam and even then only if grading camera original tapes. Why?
Because it means I can use all the data the camera recorded.
To use it you will have to wrangle certain technical issues that Vegas introduces, refer to articles by Glenn on this.

In the end for submitting content for broadcast you're more likely to get into trouble with 32bit mode than 8 bit mode.
It has no bearing on the issue of chroma sampling raised in that other thread that I'm aware of. If you are shooting HDV and delivering SD you can wrangle better image quality by rendering to the 8 bit Sony YUV codec and have a post house dub the file to a tape format such as Digital Betacam. Again you don't need 32bit mode to achieve this.

Hope this helps rather than confounds.

Bob.
Cliff Etzel wrote on 2/2/2009, 3:14 PM
Bob that helps immensely - thanks for breaking that down into layman's terms.

Another question - you recommend just using the straight SONY YUV codec - not the SONY YUV 10 bit one - just want to make sure I'm understanding correctly. Is there anything gained from using the 10 bit codec or is it a moot point?

Cliff Etzel - Solo Video Journalist
bluprojekt | solo video journalism blog
farss wrote on 2/2/2009, 3:32 PM
The 10bit one will result in bigger files.
As far as I know no broadcaster transmits 10bit.
10bit means more detail at the bottom end / darker part of an image.
I'd sweat about every pixel and every bit of detail if I was trying to sell something like Blue Planet to the BBC, pretty unlikely scenario for most of us :)
In general when you try to sell content first you have to have content that their marketing people think is worthy. THEN it can have to pass muster technically. If your content is truly outstanding and they want it they may even let their technical bods work with you to resolve technical issues.

Here's a realy simple suggestion, no maths involved I promise.

Why does it matter to broadcasters?

They switch programs and spots using automation. Generally no humans involved. They want an easy time of it, how can you help them and avoid getting content rejected?

Easy.
Watch and LISTEN to their stuff. Same TV, same system. Playout some of your content. Does it look the same, does it SOUND the same. If you have to reach for the volume control or adjust your TV in anyway you could have a problem.

I emphasise LISTEN because I have been making a mistake of having my audio too hot. What I was doing was perfect for corporate work where you need the sound to punch through. Broadcast is not like this unless you're making one of those horrid BOGOF TVCs.

Bob.
Coursedesign wrote on 2/2/2009, 3:37 PM
Using a 10-bit codec in Vegas in 8-bit mode is actually OK for straight cuts only.

For anything else it's a waste, as every user-initiated operation is truncated to 8-bit afterwards.

The 32-bit support was not really added for supporting 10-bit DigiBeta, but for linear gamma compositing and effects to take Vegas a step closer to After Effects functionality without having to leave the NLE.

People who shoot 10-bit every day (Digibeta, D5, etc.) should use an NLE that supports 10-bit, instead of processing every frame with slow 32-bit arithmetic and dramatically increased disk throughput and storage needs.

GlennChan wrote on 2/2/2009, 6:49 PM
Pictures of 32-bit advantages here:
http://www.glennchan.info/articles/vegas/linlight/linlight.htm

Actually for some of the effects you don't even need 32-bit mode (as plugins can work internally in 32bit and yield the same result as a 32-bit project).

- I don't get into banding artifacts, since they usually aren't a problem as most images contain noise that hides them.

- No NLE works in 10-bit internally. Processing will be 8, 16, or 32 bit unless they are using custom hardware. It's best to do processing at a high bit depth, i.e. 32bit.

Storage-wise, the sony 10-bit YUV codec should be the same space as other systems which can capture 10-bit.
Coursedesign wrote on 2/2/2009, 10:09 PM
- No NLE works in 10-bit internally. Processing will be 8, 16, or 32 bit unless they are using custom hardware.

True, but after each frame has been crunched in fast 16-bit arithmetic, it goes back to compact 10-bit storage. And when 32-bit arithmetic is needed for more complex work, the storage and disk bandwidth need is about two thirds less, based on 10-bit vs. 32-bit storage.

It's best to do processing at a high bit depth, i.e. 32bit.

As you noted, internal processing is done at high-bit 4:4:4 anyway. The question is what happens when each frame is stored.

Storage-wise, the sony 10-bit YUV codec should be the same space as other systems which can capture 10-bit.

Of course.

The problem in Vegas as that if you put 10-bit footage on an 8-bit time line, you can only do straight cuts, or your footage will be truncated to 8-bit immy.

If you put your 10-bit footage on a 32-bit timeline, there is no bit loss (other than potential YUV-RGB-YUV conversion loss), but everything slows to a crawl and your disks fill up (if you have enough, otherwise everything just stops).