Setting contrast with Spears and Munsil disc - Page 9

Yea, it’s almost like they can make $ from them… Wait, I see you can buy SMPTE-274M for $1500.00 from the SMPTE store.


Here’s a little more including 10.5. Section 7 defines things like specific levels. Standard stuff like Y=16-235, CbCr =16-240, 0/255 are prohibited, etc.


Also, Poynton does similarly and says pretty much the same thing as the SMPTE spec in his “Mergering RGB and 4:2:2” document



Note in both places, overshoot (WTW) and undershoot (BTB) ARE linked or lumped together (much like two sides of the same coin) and the reason cited for the presence of BOTH is exactly the same. …due to processing and to accommodate the transients that result from filtering.

I understand the desire to keep BTB/WTW separate, but as noted here and in many other places, they DO seem to me to be closely linked in specs and by industry insiders. The original debates here were all about BTB, and now most everyone has forsaken them.

I’ll also note again that I don’t place filter ring, processing artifacts, or filter transients in the same classification as valid signals, and neither did SMPTE or Poynton as shown by the above references.

I think A/B side-by-side images can be very helpful in making the distinction.

Dave

How do you calibrate a display with an input that has a clamp at BTB/WTW? You could calibrate to levels for one.

They don’t come in if the input has clamped them.

Additional intensity over 100% won't cause discoloration if the input has clamped them at 100%, additional drive over 100% isn't possible then. Blue and green won’t reproduce any additional intensity over 100%- it won’t be there. Only the BTB/WTW pixels will be clamped. All the other 100% and lower pixels will stay at the exact same levels they were at. You'll get all of the 0-100% and only the 0-100%.

If additional signal intensity over 100% is causing discoloration, or unbalancing your whitepoint and effecting your overall perception of the image, then maybe the problem is with the signal intensity being over 100%?

Surefire way to eliminate BTB/WTW? Are you setting me up?

Dave

I guess I consider it the ‘moral equivalent’ of not toeing the line or foot-faulting. < I'm sorry, Smokey. You were over the line, that's a foul. This isn’t 'Nam. This is bowling. There are rules.>

We’ve seen reports and comments here (from other than me) that for much mainstream BD material, the studios can tow the line pretty well. If you were an engineer in a studio, and likely a member of the Society of Motion Picture and Television Engineers, the leading technical society for the motion imaging industry, what would you do, to ensure a proper exchange of picture information, given the following SMPTE directive? passively unbound excursions or try to toe the line?

There is not much on the HD I've checked so far, there are certainly pixels here and there there on all content I've looked at. As expected light sources and reflection highlights are where you see issues.


I don't think it's double standards just that material sourced from video cameras is perhaps more likely to have WTW. So far all the really visible examples have not been from a film type process.

Flash is just a wrapper over perfectly normally encoded video. It is the rather crappy flash player that insists on expanding the levels to 0-255 and not doing decent scaling. If you get the content down and play it in a decent player it does indeed look just fine, I find the playback in the flash player to be noisy as I refer to above with the compression pumping and mosquito artifacts extra noticeable due to the lack of a smooth expansion when compared with playback of the same content in anything else.

John

No gain/offest control, this is the output direct from the mpeg decoder in YCrCb space. I've passed test patterns through and the levels have matched up to what is expected.

John

This would make sense I guess.

Film to video transfers can have the appropriate time spent on them.

Live or near live broadcast footage can't really be tweaked to the same level on a shot by shot basis.

The easiest solution I guess is to let the white excurse to the maximum range on those occasions where it blows out and not risk a strict clamping to 235 further exacerbating the visibility of clipping that you won't be able to do anything about anyway.

Not saying this is the case but if that was the scenario then you would have to decide whether to calibrate to preserve level up to 254 for broadcast with an associate contrast knock down for "rigidly" mastered material.

Or vice versa and live with clipped broadcast material.

Or you have different calibration regimes according to each source.

Incidentally I did recalibrate last night with 235 mapping to peak white. I also reprofiled in the Upsilon mixer with a 235 custom white point.

I'll talk to you offline in more detail.

BD looked fine with 235. I did check if I was losing anything above 235 and on the odd occasion I when I did get suspicious and checked over the material there was literally nothing there above 235.

Broadcast TV on this setting looked a little harsh to me but nothing too terrible ( need to spend a bit longer watching it).

Curious.

http://www.bbc.co.uk/blogs/bbcintern...rom_the_h.html

What do you make of this then?

He seems to be suggesting that there is nothing above 235 but keep it if it gets your peak luminance to a level you are visually happy but correspondingly don't worry if you have to "clip" it to get it there. ( from the perspective of a display that doesn't have the ability to clip the signal independently of the screen itself)

This almost takes us back full circle.

Maybe its easier to think of it in this somewhat facile way:

235 is the "minimum" white point you should map to , that doesn't mean you can't or shouldn't go higher but you definitely shouldn't go lower...

You mean, here inside the castle we use proper test patterns and meters, but here's a compressed/messed up one for you plebs, and by the way just adjust it how you like anyway? Not sure I'll read too much into that.

But even so I agree that it is looking like it is going to be a trade off thing with most probably prefering extra contrast ratio on most content at the expense of a bit of "sparkle", as Poynton puts it, in the highlights.

John

I'm not sure that's doing what you think it's doing, to clip everything above 235 you need to set up an adjustment with contrast boost set to maximum. Otherwise it will still be outputiing everything above nominal just mapped into a lower range, not very intuitive I'm afraid.

John

Yep I've replied to you. I wouldn't say its not intuitive , in a way its more clever than I thought as it compensates for display hardware even if it clips.

let me summerize you're post....
"If you clamp..."

That is a surefire way to eliminate WTW.
BUT...
I've never seen a cable box that clips.

My Onkyo passes BTB and WTW and my toshiba has no functions for clipping the HDMI signal. The different color channels clip at different intensties. None of this is uncommon.


So what do you recomend when you can't clip?

I did some measurements on my RS20 projector last night. Setting Contrast to show all the 'white bars' up to ~ 250, compared to setting it to show only those above 235, gives up about 9% in brightness (i.e., lumens from a 100% IRE window). So one must decide whether or not ~10% extra brightness is worth clipping whites down to 235. Or, of course, one can choose to go halfway down, etc.

I think the question is whether there is anything up past 235 worth hanging on to.

I hate clipped looking whites and video ( even from film) always looks clipped to me regardless.

However sacrificing 10% extra brightness is a visually large impact , video is not linear after all. There is no way anyone can justify that if there is literally nothing worth hanging on to above 235.

What we need to do is categorically get to a point where we can say where white is.

Agree, absolutely. If there is anything worthwhile above 235, one is probably willing to sacrifice 10% in brightness; but if there is not, then one certainly doesn't want to throw the 10% away. So as you have been correctly emphasizing, the question to answer is whether or not that is anything significant above 235.

10% light output is actually NOT that signifigant imho.

If you walked into a room and I showed you a display outputing 120cd/m^2 and then again at 100cd/m^2 and it was a truly blind test, you'd probably be only slightly better than a coin flip at picking them correctly.

You eyes are not that sensitive to the absolute amount of luminance at the highend.

I am so glad this discussion evolved into a very serious discussion regarding where white is and what is given up when we decide to clip. I was amazed at the early tone of this conversation when it was so dogmatic.

The conversation has moved to an empirical one that has real meaning for a lot of us. In my case, my green channel runs out of luminance too early to take much advantage of the very beneficial and needed extra luminance, calibrating 235 as the highest white my display produces.

The earlier part of this thread seemed to theorectical and I struggled with really understanding the principle here until 16 and 235 were called reference black and white respectively. I am glad BTB has been part of the discussion as well as WTW.

Looked reasonably obvious to me. It wasn't earth shattering I grant you but I suspect it would be more impactful on contrast starved displays.

absolutely a big deal on contrast starved displays.

On a perfect display (no light blacks), the conversation would be completely irrelevant because the gain of clipping would be nothing.

If we all agree not to argue we can actually reach meaningful conclusions.

I'd really like to put this topic to bed as it constantly rises up again.
I am coming down on the side of 235 with the ever so slight conundrum as to what broadcast does as uncovered by John's clever software.

As ever try it and see is a good bet. I've recaled for 235 on my setup and am so far happy with the result for BD.

I now need to see what what this does for broadcast. It will either look like a clippy mess or be barely noticable if at all.

This seems to validate what I've subjectively noticed on BDs and DVDs for quite some time. I don't watch much broadcast material (and regrettably none of it is currently hd for my setup) so I'm interested primarily in the conclusions that will eventually be made in this thread for DVD/BD content in particular.

Will be very interested to hear your final conclusions. At present, I am setting Cst to see the white bars (dimly) up to ~ 242; this leaves the bars up to 235 fully resolved. For this I need Cst = +8, whereas I need to reduce it to +6 to see them all the way up to ~ 250. With this compromise I give up only ~ 5% brightness from what I would have if I clipped all the way down to 235 (Cst = +10).

There are obviously an endless variety of choices people can make depending on their situations.

Errm, I think you're conflating visual memory with luma descrimination unless you mean given two otherwise identical displays -- can I tell which one is at 100 and which one is at 110 (or 120).

Agreed, particularly if we are talking about once your eyes are adpated to the the display and the darker the background. White will look basically the same once adaptation is done, black may be slightly higher but we are talking tiny perceptual differences. The more background light there is the bigger the difference and the more likely you are to spot the drop.

If you skip the adaptation and try and do an a/b then obviously they will be different.

John

Did some hacking today and it's looking good for an ffmpeg based software analyser that takes hours rather than days to run later on in the week.

John

Excellent John that data will be very useful in cracking ths nut.

I'm back home now so I'll reprofile my display once my i1LT warms up based on what we talked about.

------

It would be good if some of the other participants in this thread could do some testing (those that can) and see if we can actually find anything on BD or dvd that has meaninful level above 235. The more data the better but try and verify your findings.

After we come to a consensus we can further investigate the practical implications with regard to display optimisation which is after all what everyone wants.

I'm hoping we wil just be able to take Occam's razor to this but we'll see. One step at a time.

Some years ago I remember watching a DVD that seemed to had meaningful information over 235. I will borrow it from my friend again to re-test it. Though, it might not be very useful, because I think it was a DVD from a movie captured with video cameras, and not from film, so according to one of your posts, it might be the reason... It's a french movie: "8 femmes".

Actually I'd love to play along: I have a PC and plenty of DVD's and Blu-rays. Is there free software I can install to join the search for signs of extra-terrestrial life above 235?

oh yeah if you had set a at 100 and set b at 120 side by side it would be obvious.

but if you walked in and then out and back in, no way you could tell.

The program I'll release later on in the week will be free, but to read the data off blu-ray in particular you need a solution to decrypt the disc, most of us use anydvd-hd for this, that is paid software.

John