Why am I seeing a difference between 256bit and millions of colours+ with 8bit footage?

Daniel Smith · May 22, 2008

Hi.

I recently shot a bit of HD on a HDX900, recording onto DVCProHD tape, when I digitised it I saved a clip as un-compressed but with "Millions of Color +" (ie. 32bit). It's an 8-bit camera (on DVCProHD tape) so it only goes up to 128 on the binary scale so 256 should cover it, however, when I tried saving it as 256 it came out all grainy. I'll post examples if need be.

Any ideas why this is? Am I confusing bits somewhere along the line here?

cheers.

Dan.

Edited May 22, 2008 by Daniel Ashley-Smith

Daniel Smith · May 22, 2008

Sorry, just to save any confusion, when I say 256 bit I don't mean 256 bits along the binary scale I mean 256 colours (ie. 256bit combinations total, which goes up to 128 on the binary scale which is 8-bit)

Edited May 22, 2008 by Daniel Ashley-Smith

David Auner aac · May 22, 2008

I recently shot a bit of HD on a HDX900, recording onto DVCProHD tape, when I digitised it I saved a clip as un-compressed but with "Millions of Color +" (ie. 32bit). It's an 8-bit camera (on DVCProHD tape) so it only goes up to 128 on the binary scale so 256 should cover it, however, when I tried saving it as 256 it came out all grainy. I'll post examples if need be.

Hi Daniel,

32 bit in this case refers to all channels (32bit ~ 4.3 billion values) Strange thing though: 32bit isn't divisible by three, but 36bit would be. Is "Millions of Color +" indicating that the clip would be saved as RGB + Alpha (4x8=32 bit) and not to RGB?

And DVCPro being 8bit refers to each single channel. So the material you shot is 24bit ~ 16,7 million colors. But saving as 256 bit saves the clip as indexed color yielding an adaptable or fixed 256 color scheme, same as a GIF file on the web.

Cheers, Dave

Daniel Smith · May 22, 2008

Hi Daniel,

32 bit in this case refers to all channels (32bit ~ 4.3 billion values) Strange thing though: 32bit isn't divisible by three, but 36bit would be. Is "Millions of Color +" indicating that the clip would be saved as RGB + Alpha (4x8=32 bit) and not to RGB?

And DVCPro being 8bit refers to each single channel. So the material you shot is 24bit ~ 16,7 million colors. But saving as 256 bit saves the clip as indexed color yielding an adaptable or fixed 256 color scheme, same as a GIF file on the web.

Cheers, Dave

Ah, that clears up a lot :) cheers David. I've exported a 24bit version and it's fine.

Only other thing is, how is HD-SDI exported? Including the alpha channel (4x10) that's 40 bits, excluding the alpha channel it's 30 bits, will the editing software just up-rez the bit rate to 32 and automatically exclude the alpha channel?

Cheers.

Edited May 22, 2008 by Daniel Ashley-Smith

Phil Rhodes · May 22, 2008

8-bit generally means "256 levels per RGB channel", hence 16.7 million different colours.

10-bit, by comparison, means "1024 levels per RGB channel." Go away now and work on your binary arithmetic until you understand why this is so, then come back and tell me how many discrete colours a 10-bit device can resolve.

HD-SDI is capable of carrying a variety of data types, in both 10 and 8 bit. You will, somewhere, have a choice, if you are on an NLE that supports the SDI hardware you're using. It rarely includes alpha channel.

Some NLEs are good at 10-bit; others can be persuaded to do it, others still won't touch it with a bargepole. It can be tough to check that your workflow is not being invisibly truncated to 8-bit at some point.

P

Scott Fritzshall · May 22, 2008

There are a handful of file formats that are 10bit, but generally when you're dealing with them in post, you interpret them as 16bit. With your software set properly, it will map the 10 bits across 16 bits when processing it, and since very few file formats are 10bit (most are 8, 16, or 32), you will tend to write them out as 16bit files as well. DPX is a notable exception.

Phil, your point about the data getting truncated is a good one, and one I had meant to ask about- a friend of mine is investigating a workflow involving shooting on a Varicam and recording straight to a hard drive via HD-SDI in order to bypass the DVCPro processing and compression. Is this output 8bit or 10bit?