-
Ensure Premiere CS3 broadcasts color safe media?
How do I ensure that the media I am exporting will all be safe for broadcast color, without applying filters redundantly (conforming to DV/601 color range 16-235)?
Most sources say to simply apply the “broadcast colors” effect, but I have some issues with that solution.
1.) On the waveform monitor, corrected video is limited from 0-100 IRE. However, if i sample the images color, some colors still read up to 255 (broadcast safe colors should only reach 235).
2.) when importing digital media created only in after effects, the colors should not be broadcasts safe, yet they too range from 0-100 on the waveform monitor (any letterbox generated in premiere does show below 0 IRE). The broadcast colors effect provides no changes to the image at all.
So how do I know what to look for to ensure I will export to broadcast safe (note, I will not have access to magic bullet, or any 3rd party software/hardware)? Also, what is the Broadcast colors range I need to limit to on the waveform and vector-scope?
Lastly: the setup(7.5IRE) check box on the waveform monitor. Does it do anything? (I understand that 7.5 is the broadcast black, but why is the check box there?)
Thanks -
6 Replies
-
Ah, software “scopes”. You are partly right, but– only luminance should be limited to 100 IRE. Chrominance can modulate to 110 IRE and still be NTSC legal, so some RBG components can go above 235.
Here’s a good primer (more docs on parent page):
LeaderInstruments_vol01_no01.pdf
7.5 setup is somewhat tricky: it only applies to analog output. I think the switch is in the Premiere ‘scope for those of us used to looking at a signal with setup applied (analog NTSC). You should be using the full 0-255 range for most video/graphics; only certain types of hardware require the 16-235 limit.
In addition to Leader Instruments, Tektronix has some good documents on video standards, too. As would SMPTE.Hope it helps. HD is different, of course.
-jeff
-
I’d echo Jeff’s comments.
What happens when these various color ranges and definitions are converting is typically not well disclosed. In a computer, the material can be handled as color difference (video component…Y’PrPb or Y’, R-Y’, B’-Y for analog or Y’CbCr for digital) within the program itself, but a computer displays RGB.
The issue with broadcast safe filters is that they sometimes aren’t really ‘scaling’ anything. They’re often the chroma or luma version of hedge trimmers. Particularly in the case of reds and yellows, the quickest way to make visual mud or reds or visual…uh, greenish brownish stuff of yellows is to apply a broadcast safe filter.
I use levels quite a lot for simple, quick color correction. I also use Colorista a fair amount. I also have SpeedGrade for a stand alone color correction application. In all cases, I depend on an external scope to really tell me where the signal is. The software scopes can tell you what levels are in the file, but if you’re headed to tape, that conversion by your video card could be having some effect, intended or not, on your footage and the software scopes won’t know this because your video output is downstream from them.
If you’re going out to standard definition tape (lots of spots still shipped out on BetaSP in lots of markets yet…), used NTSC (standard def) scopes aren’t terribly expensive, or hard to come by, though I’d try to buy a used scope from somewhere that buys stuff from auctions, etc. and goes through them and verifies proper operation.
If you are not going out to tape and you have to output a file and you’re dependent on software scopes…you have to know a couple things. First, those full range RGB values can usually be scaled by whomever is taking them to tape. Many video IO cards (AJA, BlackMagic, Matrox) take care of the scaling, including -scaling- the luma range so that 0 black is at 7.5 where appropriate.
Second, when you’re making your final adjustments, look for luma overshoot on the waveform monitor and chroma oversaturation on the vestorscope. Many people know that video levels in excess of 100 on a waveform monitor are bad and you need to stay below that. You may even notice that the software hard-cuts luma levels there so you might be able to save some detail by backing that off somewhat. On color, you can put color bars on the timeline and call up a vestorscope and note that each color should be hitting a target zone around the vectorscope. The two scales you read on a vectorscope are
1.) The distance from the center of the scope indicates saturation…the further the signature is from the center, the more saturated it is…the outer ring is the absolute danger zone, but not every color has the same saturation tolerance as you’ll notice that each color target is not equidistant from the center point. I typically try to never push a given color beyond it’s relevant target box, even if it is still inside the ring.
2.) The ‘clock’ position around the display indicates the specific hue (sometimes called ‘phase’ by us old analog-trained guys) the color information is showing. You can see on the targets that there is a target for R=Red, B=Blue, and G=green…then you’ll notice that the secondary colors in between them also have targets, Y=Yellow, Cy=Cyan, Mg=Magenta.
Most vectorscopes also have a line right around 10:30-11 O’clock which is rarely explained. Sometimes it’s a leftover “I” axis indicator (There was a way of handling color that Quantel developed one time where it actually rotated the phase axes on NTSC material to increase color clarity and the two axes were called “I” and “Q”…but I digress…), but the intended purpose on many scopes is to be a skin tone reference. Believe it or not, all the races of the world have what is a very small range of actual skin hue…the saturation is different of course, but while there is a little deviation, most human skin tone when represented accurately should exist in a very narrow pocket on the vectorscope.
You might say that you wouldn’t need a vectorscope to see skin tone that was out of whack, but keep in mind that over the years professionals have had to always question whether their eyes were telling them the truth…mostly because analog monitors could drift over time, or somebody could decide to arbitrarily adjust a monitor and sabotage it’s accuracy. A vectorscope (and waveform monitor) that was properly adjusted couldn’t lie.
Even though we aren’t in quite the situation with our digital displays these days, two different displays could still easily show us two different interpretations when fed the same image. Our display card could have some settings that affect how that image is presented to the display…or our video in/out card could affect the image as it passes through.
With HD, we’re now even more constricted as many people don’t realize that even though we gained all these pixels, Rec 709 colorspace (HD) is actually more constricted than Rec 609 colorspace, giving us even less dynamic range in saturation and luminance.
Long post…sorry. Hopefully something in there is useful for you.
TimK,
Director, Consultant
Kolb Productions, -
As usual- good explanantion, Tim.
[Tim Kolb] “many people don’t realize that even though we gained all these pixels, Rec 709 colorspace (HD) is actually more constricted than Rec 609 colorspace,”
To perhaps lead us even further afield — I for one didn’t know that! I thought HD was able to use full-gamut RGB. Time for a refresher look at some SMPTE docs, I guess!
-jeff
-
[Jeff Brown] “I thought HD was able to use full-gamut RGB. Time for a refresher look at some SMPTE docs, I guess!”
Well…that’s another issue. SMPTE C is actually a subset of NTSC. Rec 709 is actually a bit wider (a very small amount) than SMPTE C, but does fall inside of full NTSC in a couple areas…
Lots of people with lots of time on their hands I think…
Also, even many RGB colorspaces can’t contain the complete video colorspace. There are certain colors that can only exist in one or the other, which explains why converting is a problem and converting back is a nightmare.
TimK,
Director, Consultant
Kolb Productions, -
Thank you Tim and Jeff for your robust technical knowledge. It will save me lots of tinkering time.
If I may, I ran a few more tests and would like to reiterate what I have read and make sure I know what to do. The client says that Luminance and Chrominance need to be within NTSC SD broadcast safe upon delivery via digital file type.1. If exporting to a file, DVD, or other digital media that will be sent to broadcast aim to keep the IRE on the Premiere waveform between 0 and 100. This is the best way to prepare for SD NTSC broadcast without affording 3rd party software or downstream hardware monitors. You can turn on setup7.5 IRE checkbox if your used to looking at analog signals, but manually setting levels to 7.5 is redundant (as the toggled on setup7.5 checkbox will show you your baseline IRE would really be closer to 15 analog IRE)
2. “Broadcast color” effect is not the end all solution to ensure a proper color space and “levels” seems to be a better choice.
QUESTION: levels effect seems to do the same thing to luminance when applied. It limits all IRE values over 100 to 100 immediately. Do you use levels because it applies this “limit” to luminance without effecting chromatic levels?3. Try to keep all colors on the vectorscope within the target range whenever possible for broadcast (IE the closer box).
CASE STUDY: “levels” and “luma curves”.
The following test was done with footage that ranged from 7.5 IRE to 110 IRE (note the setup7.5IRE toggle is not on and black should be 0 IRE). In this footage, an interior close-up of a subject is near a visible window with exterior light blowing out the background.“Levels” limits the waveform to 100 IRE. We do not see this change because the computers RGB color space cannot display color over 100IRE and the trimmed data is not usually missed. But what if we apply a Luma curve adjustment below the levels to play with contrast or brightness after the fact? The data above 100 IRE is lost and if we adjust the Luma curve we will be forever absent of the data. White areas will stay white, and if the brightness is lowered those whites are quickly replaced with matte grays.
I then switched the order of the effects. With the luma curve I maintained the mid-tones while adjusting the shadows and highlights till they were contained between 0-100 IRE. The levels adjustment layer was applied below to ensure there were no peaks over 100 IRE. This produced a very rich image that retained varying values outside the window previously crushed by the levels effect.
Placing the two instances side by side, the image with “levels” applied on top had more hot spots and less exterior detail.There is a good chance this is common knowledge to you both, and levels is only applied at the final stage of the process for safety anyways, but for others viewing this thread, and trying to learn how to push image quality, this may be useful (especially if the average video is pushing 110 IRE).
I hope I am right about all this. If not, please feel free to correct my deductions.
LASTY: sorry this is so long, but in case I would like to learn more about HDTV standards or rec 706 colorspace, where would be a good place to look?
Thanks
Keith -
Spewing it was such a good question and I wanted to know the answer 🙁
Mac Pro 2 x 2.4Ghz, 16GB RAM, GTX 660
Reply to this Discussion! Login or Sign Up