There does seem to be a few errors, unless I'm reading it wrong.
1) There's no 7.5 setup on NTSC DV. Adding and removing setup is the responsibility of the A<->D converters. Add setup to DV and you're pushing the blacks up too high making them look washed out.

2) We've yet to see a DVD player clip superwhites. What DVD players will do is effectively legalise their output. If you legalise your video prior to encoding then on playback your video is going to look kind of washed out. It gets worse than this, we've had content off a number of sources that has been 'legalised' and the best way we've found to stop it looking washed out is convert from Studio RGB to Computer RGB using the CC FX preset.
Bob.

Hi farss,

1- In regards to #1: I don't believe I said that NTSC DV has '7.5 setup'. I will look over the article to see if it is a little confusing.

Perhaps part of the confusion stems from Vegas' video scopes having a "7.5 IRE setup" setting...? (Which doesn't affect your video levels at all, much like how the headphone / control room volume setting on a mixer only affects monitoring levels.)

2- I think what may be happening here is that your MPEG2 encoder cannot generate superwhites in the first place.

It may be expecting black level at 0, white level at 255. In that particular case you cannot feed input values that are lower than black, because you can't go lower than 0 with the input values. This would explain why the video looks washed out (you may be feeding 16-235 input for an encoder than expects 0-255).

Which MPEG2 encoder are you using by the way?

3- My DVD player (an Apex) does clip superwhites. I tested by doing a variation on SMPTE bars.
In Vegas, Superimpose a rectangle of 255 255 255 RGB white over a portion of the 100% white bar in the bottom left. Superimpose a rectangle of 0 0 0 RGB black over part of the PLUGE.
On my DVD player, there is no distinguishable line between the rectangles and areas at black/white level which indicates clipping.

Tests I've done were with the vanilla MC encoder that comes with Vegas.
Encoding SMPTE PAL bars to mpeg-2 out of Vegas and bringing the mpeg-2 bars back into Vegas and checking that with scopes showed levels remained the same.
Burning same file to a DVD and playing that out and capturing it back via a ADVC 300 showed all levels were compressed to fit within 16 to 235. Running the same test using PTT to VHS via ADVC 300 showed no level shift.
My tests were far from scientific however two of my associates who work in broadcast have gotten the same results using very different equipment to create 'test' DVDs from TPGs. Measured composite outputs from the DVD players on Tektronics scopes revealed the same level shifts.
All of the above makes some sense to me. Given that DV in general is only 8 bit there's good technical reasons not to waste resolution by not using the full range of values in the digital realm. I notice that our SD Connect converters seem to take this approach by having an option to 'legalise' output. In other words you can feed them 0 to 255 in the digital realm and get full range legal anaolgue video on the outputs. Taking the opposite approach of compressing the dynamic range in the digital realm would seem a good way to cause image degradation. If the vision has to pass through multiple D->A and A->D conversions things would soon get pretty messy.
Bob.

Hi,

Thanks for the Tutorial.

A Question about PAL, is there any difference using PAL - in DVD, Broadcast and Codecs?

Also, is there a difference for HDV ?

Thanks

Udi

Glenn:

Do you consider the waveform and the histogram alternative tools for the same job? I would have been interested to have seen how the histogram looked and changed along with your waveform, as I generally have used the histogram (rather than the waveform) as it shows the black and white end points more precisely and gives me a visual of where I have set the gamma.

The best point I thought you made is never extend the black and white points beyond where the shadows and highlights loose detail; it is real easy to watch the scopes but not the picture!

Mike Stevens

The histogram is a good tool too. However, in this particular case, it wouldn't have worked too well.

The picture has a black bar on the bottom which does not contain any useful information. It doesn't matter if it clips as it's supposed to be black anyways. However, it does show up on the histogram and makes it hard to figure out where the video's black level is at. This black bar is actually below the video's black level.

Farss:

I will look into what you're saying.

The DV format does specify a luma value of 16 as digital black level and a luma value of 235 for digital white level. This does lower bit depth for proper video, and also allows for over/undershoot of the video signal.

farss:
What exactly do you mean by "legalise"?
To me, legalise means clipping off any illegal color values, such as values below proper black level and above proper white level. The blacker-than-black PLUGE bar for example would be clipped if you were to legalise the bars.

I think we may not be on the same page here in terms of what legalise means.

Glenn,
what we'd seen happening was what I'd assumed to be the correct way to 'legalise' video. That isn't to clip the values that are under 16 to 16 and those above 235 to 235, the result of doing that'd be pretty nasty. The results I was seeing was to effectively alter the gamma such that the values were remapped within 16 to 235, i.e. 0 becomes 16, 255 becomes 235, 128 stays the same, 16 becomes around 30 etc, in other words the dynamic range is compressed to fit, not clipped to fit.
My research started with two things. Firstly I was using a pretty dodgy monitor, feed it anything below legal black and it'd roll like crazy, how did I hit this problem, well any fade in Vegas goes right down to "0", so everytime I put a fade on the T/L this thing would roll. Same thing would happen with any video I printed to VHS with this monitor until I worked out how to deal with the issue in Vegas.
But the odd thing was, make a DVD of the exact same 'illegal' video and no more rolling around. However the material I had run the BC FX on looked a bit washed out, others here were saying the same thing, their video when made into a DVD looked washed out.
That's when I and others here did some further testing and we all seem to have gotten the same answer. Now if the video was being clipped to legal it wouldn't look washed out, it'd look, well, crushed!
Now I've made a few DVDs from material supplied by a broadcaster who insists on keeping everything 'legal', yes their DigiBeta masters are precisely Studio RGB, except if I leave it alone and just make a DVD using Vegas, the MC encoder and DVDA it looks pretty washed out. The answer is to add a Studio RGB to Computer RGB conversion prior to encoding.

One other thing to note, the Histogram display in the Vegas scopes isn't affected by the Studio RGB switch.
Also, leaving the scopes running with real time update sure looks impressive but they're a major CPU hog, if your preview frame is going through the floor best to turn them off. I think that's one very good reason why there's no scopes on capture!

Bob.

Perhaps part of it is this: Ideally, if you have a digibeta master and make a DVD out of it, the levels should remain the same without you having to touch anything. i.e. it shouldn't look washed out when you make a DVD.

Not all manufacturers follow standards though, which would likely explain why the levels change going to DVD.

On my system I ran a test and the color bars do clip when played back on my Noma DVD player (sorry, though it was Apex earlier). When I play it back on a computer, I see different results between powerDVD and Windows Media Player. PowerDVD passes values "straight through"... so 0 to 0, 16 to 16, etc. WMP does something like studioRGB to computer RGB conversion, so 16 in the DVD goes to 0 on the computer screen. Both are correct depending on how your display is setup.
PowerDVD shows all three PLUGE bars.
WMP and my Noma DVD Player do not.

This seems to suggest that some DVD players do clip superwhites and superblacks. The Joe Kane and Avia Calibration DVD FAQs indicate that this is true (at least for superblacks, or blacker-than-black colors).

I'm interested in why there are differences between our two setups.

what we'd seen happening was what I'd assumed to be the correct way to 'legalise' video. That isn't to clip the values that are under 16 to 16 and those above 235 to 235, the result of doing that'd be pretty nasty. The results I was seeing was to effectively alter the gamma such that the values were remapped within 16 to 235, i.e. 0 becomes 16, 255 becomes 235, 128 stays the same, 16 becomes around 30 etc, in other words the dynamic range is compressed to fit, not clipped to fit.
To me, I would call that converting (between two color spaces) instead of legalising?. Anyways, that's just semantics.

I agree, the whole thing is confusing enough even if everyone stuck to the same standards!
I'm just in the process of editing some material from a DVD authored by a company who I hope know what they're doing. Looking at the levels on it, the whites go to around 245 but the blacks only down to 16. So to match their black levels I set my blacks to 16 as well. Yet the preview in DVDA reveals that my blacks are decidedly grey compared to the black border in the preview window.
I also checked some footage from a Sony 570 DVCAM camera which records bars. Bars line up perfectly in Vegas with the scopes set for Studio RGB yet the white levels in the vision go well over 255, switching to Computer RGB, the whites are at 255 but the blacks are at 16.
Anyway, when I get some time I'm going to check this out in more detail with some real Tektronics scopes and a range of DVD players. In the interim I just do whatever it takes to get it looking right compared to commercial DVDs.
Bob.

This afternoon I took the FX1 and in soft late afternoon light took a series of shots at various exposure levels then came home and put them through the Vegas scopes. I was surprised to see the waveform monitors range, with a full range exposure, was 110% to -10%, not 0-100. On the color-bars the waveform went from 100% for the pure white to -5% for the pure black, while the histogram for the color-bars went from 237 to 4. Maybe Glen can tie this data into the meaning of this thread. It looks like the FX1 has these super blacks and whites that have a greater range than the color-bars it generates.

ALSO I'd like to solicit opinions as to how the Zebra setting effects maximising the dynamic range of the original recording. For daylight outdoors filming what Zebra level do you use? most courses or tutorials I have attended recommend 90 while my FX1 defaults to 100+

Mike Stevens

The FX1: I don't have that camera so I didn't know it recorded superblack information (from what you're saying).


2- The zebra % setting determines the exposure level where the zebras start appearing. At what % you want them is a bit of a personal preference.
What exposure level you want to aim for is also a personal preference.

If you want to capture the maximum dynamic range and tweak levels into legal range in post, then:
It's probably best to use a higher zebra setting (i.e. 100%).
Expose so that there are small bits of highlights that have zebras on them. This is likely optimal.
Chances are, the light in the scene exceeds your camera's exposure latitude so you need to sacrifice some image detail somewhere. The zebras is one tool to help show you where you might be losing detail in the highlights.

You may also find that it is ok to expose for the subject and let other elements in the scene have significant clipping. See:
My article on aggressive exposure
Look at the pictures and scroll down to the summary table at the bottom.

It's a matter of personal preference and workflow in that case. It also relates to the color correction you wish to do. If you don't have the time to tweak every shot, it may just be better to always expose for the subject regardless of what else is happening.

Exposing for the subject:
It helps to have the zebras kick in at 70-90%. You probably want highlights on the subject just hitting 80% or 90% (somewhere in that range), so that the highlights are in the 70-90% range. If your camera only zebras at 100%, then iris up until you hit 100%, then back off a little.

To know things for sure, you should test your camera to see how it reacts to different exposure levels and how that correlates to what you see in the viewfinder.

Glenn,
I just ran a rather simple test to see what is happening with DVD players.
I created two sets of bars using generated media that step from 0 to 16 and 235 to 255, encoded them to mpeg-2 using the MC encoder that ships with Vegas and burnt to a DVD.

The encoded files yield the same levels as the source.
Using my DSR-11 as a D->A converter I can distinguish both sets of bars on my monitor.

Playing these out of two different DVD players they're gone. The players are legalising the video by clipping the levels to legal values. One of these players is connected via composite and the other by component, same result.

What the DVD players are doing is the same as the BC FX in Vegas on Lenient.

I don't know how I got the results I did before, it was a long time ago though.

One concern though. Given that at least one Sony camera that I've looked at seems to be recording in the range 16 to 250 legalising that is a bit problematic. Simply applying the BC FX would leave the blacks alone but clip the whites causing a minor loss of detail. Using a computer RGB to studio RGB transformation would do the right thing to the whites but push the blacks up too high. Probably a custom template using the CCs would be the answer.
Bob.

farss:
You could use the Levels filter and do "output end" = 0.922 (or slightly higher) to take care of superwhites.

Black level will shift by an insignificant amount. If you wanted to fix that, you could slightly raise "Output start".


Another method is to use the color corrector. Have gain around 0.922 (or slightly higher), offset at around 1.3 (or slightly lower), "saturation" at 0.923 (or slightly higher). You may want to tweak "saturation" slightly since the CC filter will mess with actual saturation. The "saturation" slider in the CC filter should really be though of as "chroma" or "color difference signals/levels", which is related to saturation but not the same thing.

2- In both methods, Vegas is kind of annoying in that it doesn't have some setting to maintain blacks at 16, and for the CC filter to maintain saturation. Some simple algebra on the front-end would fix this (and a button to toggle that algebra on/off).

Would someone explain for me the purpose or the reality of what is happening when my Sony FX1 records (as it appears to be doing) these "super" blacks and whites. To me, all this can mean is that contrast is maximised, but that implies that on cameras that do not have super black and white contrast is lower than it could be. Are we saying super-black is the real black (no information) and that regular black is a little grey (say 16)?

Mike S

duplicate post deleted

Not sure if this helps:
Many digital video formats (not for computer display) are defined with black level at 16 and white level at 235.
Anything below black level shouldn't be seen so it'd be the same color as black level.
Same thing for white level... although some CRT-based displays will be able to display superwhites with distortion.

To go back to black level: that regular black is a little grey (say 16)?
Regular black should appear perfectly black (and the same color as super blacks).


The reason they have superblack and superwhite is to allow for under/overshoot of video, which can be caused by edge sharpening/enhancement among other things.

Anyone who lives in southern California will know of the grocer "Trader Joe's" who this week has stocked a fine Vouvray made in the Loire Valley by La Chateau Eleveurs and a bottle of which I have just consumed. The reason for this extravagance is that next week I undergo repeat quadruple heart by-pass surgery on old and diseased graphs that even the surgeon says is "dodgy" and involves removing my heart so he can turn it over to work on the back (doctors call that the posterior), so for the present time I have given up consuming the cheap plonk I usually drink for finer wines.
The relevance of this to Vegas is that the effects of the Vouvray has fussed up my once clear mind so that I now feel I understand Glen's last post and I thank him for it.

If I get this right, in the digital world luminance goes from 0 out of of 256 (computer bits for pure black to pure white) and that anything outside that range is not distorted like analog but is clipped - caput - gone -zero - nada. Now the people who broadcast video know that at levels below about 16 it's so dark it looks just like black and at levels above 235 or what ever it looks like pure white so, to maximise the range of contrast or for other silly reason they say 16 is black and 235 is white and so they clip at these levels before the digital thing get its chance. My camera that shoots these super blacks and whites thinks this is bull and shoots correctly and that's all there is to it.

If this is correct I suggest you all run to your local off-licence or liquor store and buy some Chateau Eleveurs because it sure seemed to sort me out, but if this is wrong don't bother to tell me because I've got other things to do this week, but I would still recommend you give the Vouvray a try because it really is good, but don't drink quite as much as I just did before you post here else you might find yourself on some blacklist.

Anyway, be that as it may, the big point that Glenn made and with which I have always agreed (except that in practice it is not so easy in the California sun with a 2.5" monitor) is that it's best when it looks best , and I suppose the real truth is we all need laptops and DVrack and when I get out of the hospital that will be my first purchase. Why do even good video cameras not have decent monitors and histograms!

Now the people who broadcast video know that at levels below about 16 it's so dark it looks just like black and at levels above 235 or what ever it looks like pure white so, to maximise the range of contrast or for other silly reason they say 16 is black and 235 is white and so they clip at these levels before the digital thing get its chance. My camera that shoots these super blacks and whites thinks this is bull and shoots correctly and that's all there is to it.
Not quite. :)

16 is defined as black. A bit of video at 16 should appear as the darkest black possible. A bit of video at 235 should be the whitest white possible (although for DVD and DV you can technically kind of go higher, less so for DVD than DV).

I'm not exactly sure why DV equipment records levels above 235 because technically that isn't following the standards. But they do, and it's somewhat useful. (Or maybe they do it because it's somewhat useful.)

If you want to kind of shoot like film, then you want to capture as much exposure latitude as possible and use up as many bits possible (to reduce rounding / quantitization error). The extra information in the superwhite gives you "bonus" exposure latitude which you can bring down into the legal range to play with.
In a similar vein, the superwhites give you headroom/safety against digital clipping.

Perhaps the commercial reason why DV cameras record superwhites is so that the camera appears better when hooking up to a TV for demos. The superwhites would make the picture brighter and have slightly higher contrast, which helps to grab the consumer's eye.

2- The few cameras I've examined closely don't record "superblacks". Not sure what's going on with your FX1... it may be compression artifacts or black bars on the sides that is popping some values that low.

3- The superwhite values are prone to get clipped when it gets to the audience. Broadcasters will either clip the superwhites or reject your master.

MH_Stevens: many, many good points. Good luck next week. Look forward to further insights when you are back in action!

This may give some insight into why the levels are defined the way they are:
http://www.tek.com/Measurement/cgi-bin/framed.pl?Document=/Measurement/App_Notes/Component_Puzzle/standards.html&FrameSet=television.

Bob.