16-235 Broadcast safe?

[Giroud Francois]

I am afraid that by jumping right in the luminance level (0 vs 7.5 ire or 16-235 range) you miss the real subject of the question.
What makes a signal broadcast compliant is not only luminance, but color space.
You could get a valid luminance value (i.e 230) totally non-broadcast (like RGB as 230-0-0) since a pure red could be out of the ntsc color space cube.
This does not happen in PAL where color space is different.
And yes it is in a way related to the 7.5 ire problem but not always.
That is why you got the option in some program creating video from pure calculated pictures (most are titling or 3d programs) to remove forbidden colors.

[Graeme Nattress]

The 16-235 range is for the luminance of the Y'CrCb and it not RGB. Yes, you still have to worry about pushing the colour saturation too far, and indeed, you still have to worry about it in PAL, but PAL is a lot more tollerant of really red reds or blue blues, where as NTSC smears them all over the place.

Graeme

[Tony Hall]

Doesn't the 16-235 range apply for creating RGB images for video programs? Not too long ago, I read in videography or DV magazine that you can use the levels command in photoshop to set your output levels to 16 for the darks and 235 for white before taking them into other programs.

I'm pretty sure that that range would apply to all color spaces with 256 levels of brightness.

[Sean McHenry]

I was thinking about this other issue on the way in to work this morning. The idea that you get more dynamic range with the digital level set at 16 is true but again only if you stay digital, which you can't do forever.

If you shoot with 16 as the lowest black level and 236 as the peak white, and shoot some scenes with peak white values that might clip at 236, that's pretty normal. Take that footage and let a deck bump the outgoing black level by using the "add setup" and you will have peak whites out of NTSC spec.

If 16 digital = 0 IRE and 236 digital = 100 IRE after the D/A conversion, using only "add setup", you have simply shifted the levels up by 7.5 and you now have black at 7.5 and your peak whites at 107.5. Interestingly, this is right about where broadcast sets the clipping before the transmitter.

This would then mean that if I am shooting digital, I need to realize my peak whites need to be 7.5 below where I might want to shoot them, peak white in digital at 236, as the analog conversion using the simple "add setup" is going to drive them higher by 7.5 IRE and therefore out of bounds.

All in all, following the specs or not, this is a mess.

Sean

[Graeme Nattress]

Adding setup doesn't work quite that way, Sean. The whites stay where they are and only the blacks move - so no you don't loose anything by shooting correct 601. Remember DigiBeta works exactly like DV in this regard, as does DVCPro50. D1, D2, D3 and every other digital format. The only difference being that they have proc amps, effectively, on their analogue outputs that map 16 to 7.5IRE and 235 to 100IRE. It's not so much an addition as a combination of addition and multiplication - a two fixed point mapping.

It's not really a mess, just that all the problems you've thought of have been dealt with, and it works just fine. You're just gowing through the similar thought processes that the engineers that invented the digital video standards went through, finding the same problems, but they got there ahead of you and made sure it all works.

Remember, most workflows these days are mostly digital. With perhaps some analogue footage coming in at the beginning where you must map the 7.5IRE to 16 and 100IRE to 235. Most decent digitizing cards have features that work just like a proc amp, and you can use the bars on the start of the tape to make sure that all the levels end up at the right place.

Then it's digital through your NLE and tape, until it gets broadcast, where it gets sucked onto a media server via SDI, and finally get turned into analogue for broadcast, at which point, the DtoA converter maps black 16 to 7.5IRE and 235white to 100IRE and as you can see, everything just works out fine!

Graeme

[Sean McHenry]

I know we all abandon this as we didn't seem to be getting anywhere but I wanted to post this note from Avid on the issue of setup.

Quoted from the 1-05-2005 Avid Reseller Report

QuickTime Reference RGB/601
The QuickTime Reference movie export within the Media Composer Adrenaline HD and Avid Xpress Pro HD applications now allows the user to select the RGB range of their sequence export.

Graphics uses the full 8-bit RGB scale from 0–255 while digital video uses the SMPTE standard of 16-235, leaving “headroom” and “foot room” at both ends of the scale. Even though it would appear to be a small difference, 16 points of RGB is roughly the equivalent of ½ stop in exposure terms. As you can see from the swatch below, the black and white values of 0 vs. 16 and 235 vs. 255 are quite different when seen in the context of each other:
[Graphics Range 0-225] [Digital Video Range 16-235]

When creating streaming video for computer use, it is better to encode for the full range of 0-255 rather than to encode for the video output range (16-235) so blacks don’t appear “washed out.” There are also some software-based MPEG2 encoders that assume the range is 0-255 rather than the SMPTE standard of 16-235.

Users can choose either RGB (0-255) or 601/709 (16-235) in the Media Composer Adrenaline HD or Avid Xpress Pro HD export settings window (see below). This feature gives users full control over their RGB levels for all program types, regardless of destination – web, DVD, or television broadcast.


It would seem Avids recommendation to use 0-255 would only be for graphics media. But they also say 16-235 is 601. If I thought there was a chance I would need a DVD or multimedia from my project, I would bring in the footage 0-255 to have the option. I can always compress the range on output for NTSC.

Sean

[Ignacio Rodriguez]

> This is potentially due to using less expensive
> D/A processing. I haven't looked into that any further.

I have confirmed this, comparing the output of my PDX10 (the same cam Graeme has) and an old but very high-end DV VTR (so old it doesn't have Firewire). The video coming from the camera had some banding, suggesting that the DA conversion being used does not use the full DV bit depth.

Of course setup was not used, in either the camera or the deck, as I live in an NTSC country where, just like in Japan, setup is a non-issue. Or at least should be.

I must say though that some informatin on this thread has me somewhat puzzled. Do Canada and Mexico also use setup? I was under the impression that setup was only used in the US. Since many TV sets used here are imported from Mexico, they might be configured tu used setup, or not. I don't know.

I have seen bars on local TV stations with apparent differences in black levels. So in the end it seems to be a mess here too :-(

[Graeme Nattress]

Canada uses setup. Don't know about Mexico.

I've not seen banding on the PDX10 - looks nice here. Any banding due to using 16-235 will be apparant on any 8bit camera though.

Graeme

[Bill Ravens]

My process agrees with Graeme. When I shoot DV, everything stays at 16-235 RGB(by default), with black at 0 IRE. It is wrong to add 7.5 IRE to the setup, if you work in DV, because when this is played back on a deck that automatically adds 7.5, blacks will be washed out. You have no way of controlling what deck is used to playback the data.

Regardless of which method one uses, I think it's clear that tapes/DVD's be labelled with what setup level was used.

I ALWAYS encode an NTSC color bar in a hidden menu on all my DVD's. It's simply astonishing the variability I've seen in playback equipment. Many professionals don't really understand what's going on with their equipment. People are argueing about pluge bars, which vary by +/- 3.5 IRE, when viewers TV's are set anywhere from 0 to 15 IRE.

To compound the problem, calibrating chroma and phase(hue) is practically impossible. Many, many professionals are running their sets way too hot, becasue it looks good to over saturate. The problem is, it's totally subjective.

This whole system is a total mess.

[Ignacio Rodriguez]

> I've not seen banding on the PDX10 - looks nice here.
> Any banding due to using 16-235 will be apparant on
> any 8bit camera though.

Oh it's very slight. I really can't see it on my TV. But when viewing on a pro monitor using S-video and comparing it with the output of a high-end DA, there IS a difference. Trust me.

Or try it. Does your NLE have hardware to output S-video by itself? If it does, it probably has a better DA than than the camera. Perhaps it can also be seen comparing the cam's output to a DVD player's output, but then there is the added difference of recompression. And, as Bill has mentioned, DVD players quality can vary a lot too.

[Graeme Nattress]

I can stick DV out of the Decklink Pro HD - I don't watch watch anything out of the camera itself, so I've never compared the analogue outputs on the camera to the DSR-25 deck I use or the Decklink's outputs.

Graeme

[Ignacio Rodriguez]

So Glenn, Graeme, setup is also the norm in Canada? Any word on Mexico?

[Sean McHenry]

I think what I was attempting to point out is Avids use of the nomenclature...

"Graphics uses the full 8-bit RGB scale from 0–255 while digital video uses the SMPTE standard of 16-235"

I shoot digital video, not graphics. I suppose Avid could be wrong.

Sean

[Graeme Nattress]

0-255 is not suitable for digital video use, and what Apple does, like it or not, with FCP, is to convert all 0-255 range graphics into video range on import, and if you export to a RGB codec (ie for web, multimedia) it converts the video back to 0-255.

No I would really like FCP to give you some choice so that, if you wanted to, you could make illegal super-blacks or use 16-235 images without having first to convert them to 0-255.

Graeme

[Sean McHenry]

I admit, I could be hearing this backwards but isn't that the opposite of what we were looking at last time we started down this road? Wasn't 0-255 the system that some folks were pushing as the choice (601 standard) to use for digital video?

Isn't 16-235 NTSC video with setup?

I tried to put it all behind me but I got sucked back in when Avid released the info above indicating, to me at least, that they too thought 16-235 was the proper way to utilize video inside the Avid system. And that Avid advocated sending out digital video at 16-235. So, if you are supposed to edit at 16-235, and send digital video out at 16-235, doesn't it make sense to shoot at 16-235, ie., with setup?

Avid, in the paper noted above specifically stated to use 16-235 for digital video.

Sean