D60 issue - DCI white gamut violated by projectors with "Use White Clip" feature


(Chris Davies) #1

There’s a function labelled “Use White Clip” in certain DLP Cinema projectors that can cause clipping and tinting of D60 whites when viewing DCI X’Y’Z’ images such as DCPs produced in ACES workflows.

I first noticed this behavior several years ago during a private DCP screening booked in a nearby commercial cinema. The content had been mastered on a projector calibrated to P3D60, which of course falls within the minimum gamut of a correctly-calibrated D-cinema projector. Needless to say I was surprised and disturbed to see that D60 approaching 48 nits in DCI X’Y’Z’ space appeared to be out-of-gamut in projection. Since that first discovery I have taken every possible opportunity to inspect the running projector configuration at other commercial cinemas and festival venues. In the majority of auditoriums running projectors from the same manufacturer I found that the “Use White Clip” setting was enabled.

At my post facility we use the same series of projectors for our grading theaters. As a rule we disable any unclear or unnecessary processing features as part of our deployment and calibration of any displays, so we have always operated with the “Use White Clip” option disabled. The manufacturer’s documentation is vague regarding the feature. Based on my minimal experimentation, my best guess is that “Use White Clip” increases the maximum light output achievable for DCI calibration white (.314,.351) by removing the headroom that would normally be available if the projector were set up so that X’,Y’,Z’ = 4095,4095,4095 = 52.37Y,0.333x,0.333y. Instead, it might be that .314,.351 (or possibly the “Virtual White” defined in RP 431-2) is mapped to the brightest possible point of the projector’s measured native gamut (aka MCGD) to increase contrast and maximize light on screen. If I’m correct in my assumption, this feature violates the minimum white gamut and the results would vary with each unit’s measured native primaries and white point. Again, I really don’t know why this feature exists and what it really does to the running calibration, but I can see that it causes more or less clipping depending on the chromaticity of X’Y’Z’ signals approaching 48 nits. DCI calibration white never appears to clip with this feature enabled.

I contacted the manufacturer hoping to learn more, and was told that this setting is enabled by default for factory presets configured for X’Y’Z’ (e.g. IMB inputs), and that it’s meant to avoid unnecessary loss of brightness and contrast that can occur with X’Y’Z’ sources (paraphrasing here). They couldn’t confirm whether this feature is specific to the manufacturer, or if is part of the TI light engine processing common to all DLP Cinema projectors.

Has anybody else encountered this in the wild? And can anyone shed more light on why this feature exists and why it causes clipping of correctly-prepared DCPs?


(Charles Boileau) #2

Hi @ChrisD. Just had a look at our IMB setup and it seems that we have it on. Never noticed that this was a problem. I’m guessing that once you “print” to a mastering format (and then to DCP) you don’t get the “protection” that the RRT/ODT has against clipping highlights (and reverting to DCI or D65 WPs).

I tried turning it on and off and the WP goes becomes super pink. So I’m guessing I would need to recalibrate to get the proper WP. Am I correct?

If you assumption is true it would mean that you would most likely have the same problem if you had calibrated to D65… Correct?

Thanks!


(Tobias Schaarschmidt) #3

Hi Chris. When you had this issue, did you by any chance use Resolve with its DCDM ODT activated in the Colormanagement Tab in Project Settings?

There are 2 DCDM ODTs in ACES 1.0.3 - “ODT.Academy.DCDM.ctl” and “ODT.Academy.DCDM_P3D60.ctl”. The first one produces out of gamut values in the Highlights which can cause these Artefacts that you described when this setting in the DCI Projector that you mentioned is activated. The second ODT does not produce any out of gamut values since it has a P3D60 gamut clamping build in (so this is the one that should be used!).

I just found out after a little test: The first one is implemented in Resolve - so Resolve really should not be used to create DCDM exports with the Build in DCDM ODT until BMD fixes that (I will write an Email to them now).


(Charles Boileau) #4

Hi again! It would be great to hear back from you @ChrisD. Thanks!


(Tobias Schaarschmidt) #5

The Resolve 14.2.1 Release fixes the issue. Now there is access to the X’Y’Z’ DCDM ODT with P3 Gamut clamp as per SMPTE Specs.


(Chris Davies) #6

Hi @CharlesBoileau and @Tobias, thanks for the feedback!

In our case we aren’t using Resolve to master DCDMs but it’s great to have that extra info if we ever fire it up.

What I can report is that I connected to our Christie projector with the TI DLP Cinema software and found that the “White Clip” feature is integral to DLP Cinema color processing (it shows up in the TI control software, not just in the Christie menus). I haven’t had the time to take measurements with it enabled vs. disabled, but my best guess now is that the mapping from DCI XYZ space to projector-native space is modified with the “Use White Clip” option to reduce the white gamut headroom, maximizing the light output (i.e. absolute contrast) for some given x,y value. The “Use White Clip” option exposes additional entry boxes for x,y values and the default value appears to be 0.314, 0.351.

What does this mean in practical terms? If “Use White Clip” is enabled on some DCI projector, and the White Clip values in that projector are set to 0.314, 0.351, we will probably see uneven clipping on the bright values that are not close to 0.314, 0.351. When it clips we will see tinting (uneven R,G,B) as D60 Y’ values excurse the available gamut (where White Clip is set to DCI calibration white).


(Charles Boileau) #7

Thanks! That’s what I thought… But in “real world” scenario is this a real problem? Would the common observer see this? PLus, considering that all theatres probably have this on.

I get that a mastering room has to at it’s best. But I always try to strive to get a “real world” example for our clients.

We’re most likely purchasing a laser phosphor projector from Barco in the coming weeks. I’ll see if this is present in the menus. It might not be an issue with laser as the light output is “better”.

What’s your thought on this?

Thanks!


(Tobias Schaarschmidt) #8

Whether the common observer is able to see that or not depends on the way the projector handles those out of gamut values. In the best case there will only be some highlight clipping that might go unnoticed. But there might also be projectors out there that could display those out of gamut values as completely weird colors - which will be very noticeable. As for Laser phosphor projectors i can not say anything yet - i would have to have one or two days alone with such a device and my calibration equipment :slight_smile:


(Nick Shaw) #9

I’m guessing that it’s related to the fact that DCI reference white is X’Y’Z’ = [3794, 3960, 3890], and the max output of the DCDM ODT is X’Y’Z’ = [3887, 3960, 3973].

IF the result of switching the white clip on in the projector is to clamp it in X’Y’Z’ to [3794, 3960, 3890] then the maximum output of the DCDM ODT will be clipped to DCI reference white, meaning it can’t hit D60 white at 48nits. It also means that as it approaches maximum output, X will clip first, then Z, and finally Y. To an observer adapted to D60 white, this would mean peak whites were skewed to the greenish DCI white, with possible colour artefacts in near peak pixels. Is this what you observe?


(Tobias Schaarschmidt) #10

I actually get much higher max Values out of the DCDM ODT (3983, 4058, 4071 in my test). The max values of the DCDM ODT with P3 gamut clamp are 3887, 3960, 3974 (D60, because it was designed with an assumed observer adapted White of D60 in mind)) The problem is the Y’ component. it must not exceed a code value of 3960. This Limit is due to a SMPTE spec (i do not recall exactly which one). It was specified this way so that you can have different creative White Points in Color grading (D55, D60, D61, D65, DCI) all with a Peak Luminance of 48nits - for all these White Points the Peak Y’ code value would be 3960 whereas the other 2 components can have higher values than that depending on the chosen creative White Point - but would still be SMPTE/DCI compliant. I have to look for the source of that info - i found that maybe 2 years ago when i ran into that issue with clipping highlights due to a non SMPTE/DCI compliant P3D60 to X’Y’Z’ colour transform.


(Nick Shaw) #11

You’re right. When I test the DCDM ODT with ctlrender I get values up to 4057 for Y’. I was just looking at the output of OCIOdisplay in Nuke, and there both DCDM ODTs clip Y’ at 3960. While that might be a perfectly reasonable thing to do in terms of producing a compliant DCDM, it’s not what the CTL ODT does.

So the best approach would seem to be to stick to the P3 gamut clipped ODT, as you already suggested.


(Jim Houston) #12

Yes, you are right. It is Annex G of SMPTE 432-1:2010 which is the section describing the inclusion of the 52.37 constant in the encoding equations. I fought in the committee to get this in, so kind of remember where it should be. Technically, you can have a higher value for the code for Y, but in most cases I think that is going to get clipped as it is not required that projectors produce a higher luminance than 48nits. (mirror-white or internal device 1.0 is usually calculated from the 48nit assumption for the signal. Whether or not you are driving the projector harder with a lamp setting is unrelated to the internal processing)

If you look at 431-2, in Table A4, you can find the max values of the white gamut all at 48 nits. (in white-1 -2 -3 )