Film-Tech Forum ARCHIVE: Color correcting

This is topic Color correcting in forum Digital Cinema Forum at Film-Tech Forum ARCHIVE.

To visit this topic, use this URL:
https://ft-forum.com/ft/cgi-bin/ubb/ultimatebb.cgi?ubb=get_topic;f=16;t=003137

Posted by Richard May (Member # 2627) on 10-10-2017, 04:10 PM:

When checking red, green, blue, black and white, what values are you looking for? Do you check x,y or X,Y,Z? I was always curious about this?

Posted by Harold Hallikainen (Member # 5405) on 10-10-2017, 05:33 PM:

SMPTE st0431-1-2006.pdf calls for xyY values of 0.314, 0.351, 48.0 with Y in cd/m^2 for peak white. 48 cd/m^2 is equivalent to 14.0 fL. Luminance tolerance is 21% for theaters and 7.3% for review rooms. Chromaticity tolerance is .006 for theaters and .002 for review rooms.

The LSS-100P ( https://www.qsc.com/cinema/products/test-and-measurement/usl-lss-100p/ ) measures XYZ in cd/m^2 and then displays what the user wants (typically xyY with Y in either fL or cd/m^2).

Harold

Posted by Richard May (Member # 2627) on 10-10-2017, 08:38 PM:

What values are you looking for in the primary colors?

Posted by Harold Hallikainen (Member # 5405) on 10-10-2017, 09:44 PM:

SMPTE does not have a specification on this that I am aware of. Others may have more information.

Harold

Posted by Leo Enticknap (Member # 534) on 10-10-2017, 10:56 PM:

The color calibration procedure (at least on NEC and Barco projectors - I haven't done it on a Christie or Sony) consists of projecting red, green, blue, then white test patterns, and taking a reading from the screen for each one using a photospectrometer. You enter the X and the Y for each color that that the meter gives you into the software. At the end of the process, the software figures out what changes it has to make to each color to get you to 0.314 and 0.351, does it, and then you measure the white again, to confirm that the adjustments have been successful.

This should ideally be done whenever any significant change to the optical system takes place (xenon bulb is replaced, lens is taken out and the back of it cleaned, etc.), but the meters are so expensive that in reality, it's usually only done at planned maintenance service calls.

Posted by Harold Hallikainen (Member # 5405) on 10-10-2017, 11:06 PM:

Of course, with the previously mentioned LSS, you can do daily automated tests of peak white chromaticity and luminance....

Harold
(designer of the LSS)

Posted by Tom Bert (Member # 5719) on 10-11-2017, 06:17 AM:

Do note that:
x=X/(X+Y+Z)
y=Y/(X+Y+Z)

Posted by Marcel Birgelen (Member # 6801) on 10-11-2017, 08:34 AM:

Harold, I know it's not your job, but can you name a figure for a current list price for the LSS-100?

Posted by Leo Enticknap (Member # 534) on 10-11-2017, 10:27 AM:

I've heard rumors (Maybe Tom could confirm?) that Barco is looking in to building a photospectrometer into their projectors, so that color calibration becomes like the CLO system for ensuring that overall luminance stays the same throughout a bulb's service life: a one-time during installation, set-and-forget procedure, and the projector automatically makes adjustments to keep the color calibration on target thereafter.

If this is true, this strikes me as a great idea. Building the cost of a $10k measuring device into a $100k capital investment means that a theater is more likely to buy it than as a separate item, the presentation quality boost from which 99.99% of moviegoers in a typical suburban 'plex will not notice or appreciate.

Posted by Steve Guttag (Member # 268) on 10-11-2017, 11:01 AM:

LSS-100P is $804.00USD list. Naturally, street prices are substantially lower.

As for primary colors the "native" colors of a xenon projector are typically in the ballpark of: (x,y)

Red (.640, .320)
Green (.265, .690)
Blue (.150, .060)

However MANY things will affect that including the mirrors/reflectors and everything else in the light path including the port, screen and the notch filter (in fact, the notch filter is normally adjusted to minimize color correction of Red or Green or both (it is a yellow notch).

DCI target colors for xenon projectors are (x, y)

Red: (.640, .320)
Green: (.280, .640)
Blue: (.160, .100)
White: (.314, .351)

So with this, you have the three points of the triangle as well as the center (white).

Posted by Tom Bert (Member # 5719) on 10-11-2017, 02:00 PM:

quote: Leo Enticknap
I've heard rumors (Maybe Tom could confirm?) that Barco is looking in to building a photospectrometer into their projectors, so that color calibration becomes like the CLO system for ensuring that overall luminance stays the same throughout a bulb's service life: a one-time during installation, set-and-forget procedure, and the projector automatically makes adjustments to keep the color calibration on target thereafter.

I can confirm that we currently do NOT have such a feature on our roadmap :-) Wondering where you heard the rumour from.

A really built-IN sensor would yield some practical problems, e.g. the fact that you'd be measuring from behind the porthole.

In the open&agnostic philosophy that we develop products, we are open to interface with 3rd party measurement devices.

Posted by Harold Hallikainen (Member # 5405) on 10-11-2017, 02:27 PM:

I like the idea of test equipment being independent of the equipment being measured. That is, I don't think the LSS should be used as a sensor to automatically adjust the projector. With it being an independent instrument, drift in either the LSS or the projector will be caught since it's unlikely they would both drift the same amount in the same direction. If the projector automatically adjusted based on LSS readings, and the LSS readings were incorrect (someone bumped it so it is no longer pointing at the screen), the projector would misadjust itself in attempting to correct the situation. I'd rather just get an alarm that something is wrong. By the way, on DLP projectors, I've found that color is very stable. The amount of each color component emitted for each pixel is determined by pulse width modulating the mirror for that pixel. PWM is very stable and predicatable, so the colors are stable. Any instability would be due to drift in the optics, filters, light source color, etc. You can see some projector stability measurements (and LSS stability measurements) at http://ftp.uslinc.com/ftp/Products/LSS-100P/Documents/Technical/ColorMeasurement/ColorMeasurment%20LSS_160525.pdf .

Harold

Posted by Richard May (Member # 2627) on 10-12-2017, 07:43 AM:

Thanks everyone. Thanks for the x,y coordinates Steve. That's what I was wondering about.

Posted by Lonny Jennings (Member # 365) on 10-12-2017, 11:01 AM:

Make note, however, that Steve's numbers for the three primary colors are not a "standard". It's just what you nominally would have if the projector is reproducing the whitepoint at the Standard .314, 351. The standard doesn't care about the parts to get there, just the final result.

Posted by Richard May (Member # 2627) on 10-12-2017, 01:45 PM:

I understand. At least I have a general starting point.

Posted by Dave Macaulay (Member # 813) on 10-14-2017, 04:52 PM:

Yes, the primaries measured in color calibration are what the projector natively produces. You can't change them, they are what the light engine filters pass. DCI approved projectors will have primaries within a few percent of the "official" spec primaries. If a primary is very different from spec towards the middle of the colorspace triangle, the projector will not be able to project the full color gamut. And it can not project that corrected primary to spec... you can't go any redder than the primary red.
The only projector I've tested that attain the DCI/SMPTE spec primaries is the Barco P model. All others have some - red and green usually - that measure slightly within the color triangle.
With Barco, checking corrected colors uses a special TCGD called "color verification". Its target primaries are not the DCI/SMPTE primaries (probably because except for a P model the projector can't possibly project those?). I assume corrected color verication target x-y values on other projectors are the same.
There's no point in checking corrected primaries in my opinion. If you get 0.314/0.351 for corrected white, all is as it should be.

Posted by Mark Gulbrandsen (Member # 72) on 10-15-2017, 12:18 PM:

quote: Leo Enticknap
I've heard rumors (Maybe Tom could confirm?) that Barco is looking in to building a photospectrometer into their projectors, so that color calibration becomes like the CLO system for ensuring that overall luminance stays the same throughout a bulb's service life: a one-time during installation, set-and-forget procedure, and the projector automatically makes adjustments to keep the color calibration on target thereafter.

It's about time someone did this!!

Posted by Mike Babb (Member # 1379) on 10-15-2017, 12:48 PM:

Personally I think USL and Qalif are on the right track for people who want to verify light and color on a more regular basis. I don't find colors drifting so much that the public would notice, but light levels are a constant problem, so tracking footlamberts and x/y would be valuable to those who care. Not sure the projector guys need to get into integrating it into their unit.

Posted by Steve Guttag (Member # 268) on 10-15-2017, 02:27 PM:

quote: Dave Macaulay
There's no point in checking corrected primaries in my opinion. If you get 0.314/0.351 for corrected white, all is as it should be.

That is wrong. If you input the wrong value or your meter hadn't sufficiently stabilized when reading it (easy to do on the USL PCA100) reading the corrected primary colors will spot the error (had it happen).

A common misconception is the belief that if you attain .314/.351 you are on the money everywhere. Color correcting is like using just the bass and treble controls to tune a room. With the white point, you also have a mid-range control too but it is nothing like tuning with several parametric or 1/3 octave EQ.

Just because you check one point doesn't mean the rest are good.

Even worse are the fools that pull the notch filter on the likes of the NC900 and "think" they color corrected it back to DCI because, after all it reads .314/.351 when they are done. That is as stupid as using bass/mid/treble controls to tune out a spike in the response at 10KHz. Yeah, you can move it but it is all relative to everything else you are moving and the spike will still be there when you are done.

For those interested in the whole color correction thing and yet more things to measure, check out the DCI spec sheet, section 8.3.4. It defines the minimum area of the triangle that the projector can reproduce. That once corrected that it should hold its white point over a wide range of stimuli too.

I do have to end this with. Remember, in the film days, you got what you got, unless you were a studio with a pile of reflectors to get what you wanted and nobody really cared in the cinema. It is great we have this new tool to be better and more uniform but there are many factors in putting on a show and getting the perfect color reproduction is just one of them. Most DCinema projectors (xenon, at least) out of the box will be more uniform from unit to unit than film projectors ever were.

Posted by Leo Enticknap (Member # 534) on 10-16-2017, 08:08 AM:

In the film days, another big variable was the print itself. This was especially true in the final decade or so of the film days, when quality control in the mass production labs really deteriorated, and the sensitometry and densitometry checking was simply not done to the standard it had been a generation earlier.

I take Harold's point that having test and measuring equipment made and sold by totally separate entities to the makers and sellers of the equipment being tested is an important check and balance, but I still believe that built-in, automated color measuring would be a step forward for the industry. This shouldn't replace regular checks by a tech using an independently made and calibrated photospectrometer (in the same way that checking the integrity of the CLO sensor is part of the formal maintenance regime as laid down in the service manual), but in multiplex type houses where color calibration by a tech using a $10K Photo Research or Jeti meter isn't done very often and realistically never will be, it would be a boost to presentation quality.

There would be problems implementing it (e.g. port glass - you'd either have to do setup calibrations with the glass both removed and in place, or have the meter in the auditorium and connected by LAN to the projector), but none would be insurmountable.

Though, given that the two main factors that cause the color temperature of a projector's output to change over relatively short-term time periods - the xenon arc bulb and the reflector - will go away if the industry does convert to laser, this may end up being a solution looking for a problem. The problem certainly exists at present, though.

Posted by Mark Gulbrandsen (Member # 72) on 10-16-2017, 10:54 AM:

quote: Mike Babb
Not sure the projector guys need to get into integrating it into their unit.

Should be a DCI requirement in projectors that use mercury lamps!!! Those drift in level and color the most, although the average Joe wouldn't notice it.

Mark

Posted by Gordon McLeod (Member # 33) on 10-16-2017, 06:33 PM:

but mark you praise the 900c [Smile]

Posted by Leo Enticknap (Member # 534) on 10-17-2017, 08:04 AM:

A specific selling point of mercury lamp projectors is that you don't need a trained tech to swap out the lamps in them. If Mark is right and the color drifts more with these bulbs over a typical operating life than with xenon arc bulbs, then it's a double whammy for presentation quality: the color will drift more, but will be less likely to be corrected regularly, as the projector will go longer between maintenance visits by a tech.

Posted by Steve Guttag (Member # 268) on 10-17-2017, 09:06 AM:

I think another problem with putting auto color correction into projectors is how well they would do it. Barco is the only one of the DLPs that does CLO well. Somehow they are the only ones that figured out you needed a different calibration based on lens zoom and allowed one to have it read properly.

Christie is the only one that has figured out auto xenon adjustment. Barco has it but it is notoriously slow and inaccurate.

An NEC, well, even their light meter can, and often does, mislead one into a lower and less even light output!

So putting the overhead of making the color stable (at least as far as getting out of the projector) is going to be less than ideal and will vary by projector company.

And think about this. So you calibrate the projector using your $8,000 - $20,000 meter and now you are going to have something with FAR less accuracy and stability maintain that calibration? And if it isn't as accurate, then what was the point?

The Qualif people would love to sell you their magic $15,000 or so box that will measure your system each day and apply the appropriate color calibration to ensure you are always spot on. But who is going to dump that sort of cash per-screen? And if you do, don't forget about the meter recalibration costs (plus the labor to mount/unmount the system annually). And for what? Has ANYONE (outside of a studio screening) EVER had a paying customer complain about their color correction (and I don't mean a 2D/3D PCF file issue but actual you were at .318/.349 type color complaint)?

We'd be better served by getting the light levels right. People know when the image is too dark though they won't know the difference between 14fL and 13fL...by the time you get down to 8-9fL, they know it is too dark (and, hopefully, before).

I also hope that this insanity of using flat gain screens goes away. If your screen gain is above 1.3 and flat, you don't have even light. You can make something from nothing. If you don't have enough light with a matte-white screen, the solution is a projector that has more light, not to put in a gain screen. There is no magic there. A gain screen takes the light from the sides/corners and puts it in the middle and ensure that everyone gets a different hotspot. The overall light level goes DOWN with a gain screen, it is just the middle that goes up.

That said, I've come to the conclusion that with low-contrast projectors (all S2K), using a mild gain will have fewer complaints about the low contrast. It isn't that the picture suddenly got contrast, it is that you lowered the gain such that people don't see the "grey" of the S2K chip. Again, the solution is to NOT use the S2K projectors and if you are going to use them, think about the laser versions though that will remove their cost advantage of them.