Originally Posted by PlasmaPZ80U
How can you tell? Can you tell with actual program material or do you need test patterns to see the difference?
Bit of both really. I use test patterns to confirm I'm not clipping "anything" above 235 (its quite difficult with bright displays to actually see the last couple of steps up to 254 with a mid level apl...your eyes are not so great at differentiating levels towards peak white). Don't worry if you are clipping off 254 for example.
On program content (BD , dvd and dtv) I can usually tell almost immediately if its clipping above 235.
Originally Posted by PlasmaPZ80U
If video levels are only 16-235 and DVDs and Blu-ray Discs all have video levels, what is the real world significance of showing above 235? What would be the point of sacrificing valuable light output just to be able to show levels that are not supposed to exist in video content anyway? If it can be done without making any compromises I don't see any reason to clip above 235 intentionally, but otherwise it seems to be rather pointless and detrimental advice.
16-235 are reference levels not limit points ( as I'm sure has been discussed to death round these parts).
The mystery of the 16-235 limit has been floating about for years and I regularly come up against it in the workplace. Its often trotted out by people who are hugely experienced and respected in other areas and you will be able to find lots of comment and opinion masquerading as fact that seems to support the 16-235 limit especially in the semi/pro broadcast forums. People have a tendency to latch on to numbers and then misquote them without realising how they are applied in the real world.
If you have a look back through some of the comments by stacey and others regarding this issue you will find references to video range from well respected sources that explicitly mention the need for excursion beyond the reference range.
Thats all well and good but what does that mean in reality?
Remember what I said about how images work. If you have a display with a notionally massive contrast range you want to map material to it in such a way as to give your eye as many steps up the point where your eye cannot differentiate changes in brightness. Otherwise it looks clipped ( or more clipped than is necessary in the case of video which will always look clipped to a certain extent with certain types of material).
Similarly if you think about it if the display was capable of depicting tiny brightness changes towards black you could also display range below 16 and still maintain a notionally healthy visual black level even to the point where the brightness differences below 16 were pretty much undetectable to your eye , the point is the level is still there.
If you think about how a high dynamic range displays and imagery works it might make it clearer ; imagine you are outside a cave in moderate sunlight. If you look towards the sun your visual system hits a point where you cannot differentiate any increase in brightness level ( hopefully you have also not burnt your eyeballs out) if you look towards the cave you can't really see much differentiation towards black inside the cave as your eyes are essentially washed out by the ambient light levels , if you walk into the cave suddenly you can see more variation towards black. The relative black and white peak levels have not changed just your visual systems ability to differentiate them. If you have a display and material that presents high dynamic range its closer to how your eye is presented with brightness variation in the real world : its a total range that your eye is unable to differentiate all off in all situations your eye slides up and down the range according to environment .
Thats all very well but what about video?
If you accept that there exists level outside 16-235 your next question is what does it contain...garbage/noise , nothing or useful image intensity steps. Well it has to contain something otherwise why have reference points that aren't at the maximal limits of your available dynamic range ( although even then anyone that creates imagery and knows what they are doing never bangs off the end stops of their available intensity range but thats a slightly different issue).
So if you accept its 16-235 but with headroom/rolloff/excursion ( last one not so helpful a term) then what is the additional stuff for?
Clamping off noise in a technically imperfect engineered system:
Yes sure but the noise doesn't just bang in right after the ends of the reference range. The whole point is not to be able to guillotine off the noise but have a nice smooth organic roll off into the end stops of the dynamic range whilst losing the noise. Hard clips in images ( video or otherwise) are never good, pleasing imagery is all about smooth transitions.
So 16-235 is the reference range but what does that mean.
Simply its the bit you really need to keep , however it doesn't conversely mean that you should take a knife and just chop off the rest . How you deal with it is a function of your display capabilities and viewing environment as has been stated again and again.
Lets expand out the pot a bit.
Most video we are concerned with watching has originated from film in most cases ( or originally captured with more intensity steps than are ultimately presented to the end viewer..even stuff shot on video to be honest). So its "bigger" than the video it ends up as.
A negative film scan is usually 10bit. 0-1023 code values. A film scan also has a defined reference range although in this case its linked to the density steps (film imagery is a density range rather than an intensity range , the difference is mostly semantic as its still essentially a recorded intensity range despite the mechanics) .
black ref coincides with the Dmin point at code value 0098
white ref coincides with the Dmax point code value 0685
These are the points that will appear notionally black and peak white if you strike a one light print straight off a negative , they relate to the intensities you will actually see ( film is a bit fuzzy as its a chemical system so you actually see some variation above 685 and below 98...hint think why films generally looks nicer than video).
So whats outside the reference range? The Dmin is the point where the film essentially can't get any darker because of its chemical nature ( fact it does actually go darker hence the need for a reference level above the end stops of the range)
The Dmax is the point that registers as white and the clearest the print can get. Fact it actually goes higher than this hence why its the 90% white point not the end of the range. (allowing for fuzziness is an important part of most engineered systems...absolutes are usually frowned on...its all about tolerance)
Film negative also has massive headroom in the whites (its why it looks nice people). Everything above 685 is useful image that can be made visible by bringing it down to the Dmax point (think about video ...think about that range above 235 , think about how images work).
Luckily for me I work with film day in day out so I'm very familiar with material in mainstream blockbusters before it gets transferred into video (often I even have to convert stuff to video for client temps).
Historically the default method to convert a film scan to video is to chop it off at 98-685 and map this to ...16-235 (you will find white papers on this) and give it a gamma twist.
This creates video imagery that looks like garbage with 99% of imagery and has never been how you really convert film to video and still make it look nice.
If I have a picky client they normally ask me to match to video rushes that have already been generated through a proper film to video pipeline by the colorist(s). I would say 99% of the time pretty much everything from 95-1023 on the original negative has been mapped into 1-254 on video. At the high and low limits of the video it will have been very crushed together by necessity but the point is it hasn't simply been chopped off.
If I have to come up with something myself I never chop it off according to the reference limits. I grade it subjectively on a calibrated video system using quite arbitrary curves to give nice range in the blacks and whites but with a pleasing level of overall contrast. According to the image content I may clip off values outside the dmin dmax range.. like I say it depends on image content... a guy in a cave probably won't have much level if any above 685...if he lights a candle its a different story , likewise someone silhouetted in front of a raging inferno may not have any level below 250 ( doesn't mean I won't drag it down a bit to give better contrast).
I'm hoping most of what I've posted will be somewhat relevant, the key thing is to stop seeing images and video in terms of strict reference ranges. The ranges have a purpose ( usually to differentiate from formats that have different ranges) but they are not absolutes when it comes to how the imagery should look.
Video is a pretty pathetic intensity range and your eye likes big ranges. Video is not like the real world and most displays do not get anywhere near the levels of intensity variation your eye is used to seeing on a daily basis.
Back to whether to clip at 16-235 or not...
Lose everything below 17 to black: your display is unlikely to show level below this point in a useful manner.
Keep everything up to 254 as even with the maximal range on display video will often look clipped to your visual system regardless and you might as well mitigate this as much as is possible.
If your display in conjunction with environment is so lacking in contrast performance that keeping everything above 235 results in a visually unpleasant experience then chop some of the range above 235 off to free up some of the contrast.
If you prefer to watch with a hard clip in place at 16-235 fine knock yourself out no-one is going to kill you for it but please don't confuse this as being of absolute correctness.