White Point Issues regarding SMPTE 196, for Film & Digital Projection Presentation to the ASC Technology Committee, 10/15/03 David Richards
Background The white point or color temperature for film exhibition has changed substantially since the inception of color motion pictures. Following is a thumbnail history:
1930s The lighting technology used for film projection at the time color prints were popularized in theatres was carbon arc, with a color temperature between about 5000 and 5500°K. This is what would have been seen in theatres in the “heyday” of Technicolor (Figure 1).
1960s Screens became larger, requiring more light. Dichroic-coated reflectors and/or heat filters were added to the light path to protect the film from damage (still using carbon arc as the source). These filter devices remove some of the visible red along with the IR, thus the color temperature increased somewhat, to between 5400- 6000°K. The filters also shift the white point toward cyan due to the missing red energy. Manufacturing tolerances associated with these coatings introduced a variation, or lack of precision along the red-cyan axis in the resulting white point obtained on the screen (Figure 2).
1970s Xenon lamps were introduced (also with heat filters), causing the color temperature to rise further. The native color temp of raw xenon lamp output is about 6100°K, or about 500°K higher than carbon arc. After IR filtering, this would tend to shift to above 6500°K (and also toward cyan, which doesn’t affect the color temperature measurement). The same problem with variation is found, just as with carbon arc, resulting from the dichroic filter coatings. However for xenon sources additional filter coatings were added to remove some of the blue light as well as the red, to shift the color temperature back down toward a more neutral color temperature of about 5600-6100°K, to better match carbon arc systems. This introduced additional variation along the Blue-Yellow axis, resulting in a fairly poorly-controlled white point. However this is still the practice today (Figure 3).
.370 DAYLIGHT LOCUS .360 D50 B.B. CARBON ARC LOCUS 4900 .350 D55
5400 .340 D60 y 61 .330 D65 6000
XENON-NATIVE .320 6500
.310 .300 .310 .320 .330 .340 .350 .360 .370
x Figure 1: Various white points in CIE 1931 x,y chromaticity space 1 Moving image Technologies, LLC .380
.370 DAYLIGHT CARBON ARC-IR LOCUS .360 D50 B.B. LOCUS
4900 .350 D55 RED-CYAN AXIS
5400 .340 D60 y 61 XENON-IR .330 D65 6000
.320 6500 RED-CYAN AXIS
.310 .300 .310 .320 .330 .340 .350 .360 .370 .380
x Figure 2: Carbon Arc and Xenon sources filtered for IR
.380 BLUE-YELLOW AXIS .370 DAYLIGHT CARBON ARC-IR LOCUS .360 D50 B.B. LOCUS
4900 .350 XENON D55 FINAL 5400 .340 D60 y 61 D65 .330 6000 XENON-IR
.320 6500 RED-CYAN AXIS
.310 .300 .310 .320 .330 .340 .350 .360 .370
x Figure 3: “Blue cut” added to Xenon reflector to lower C.T. Note: This represents the light source and reflector only, and doesn’t include the contribution of the lens, port glass, and screen.
The arrival of digital projection systems has opened our eyes to the current situation with film systems, but it should be understood that there’s no fundamental difference between the ideal white point for a xenon-based film system and that for a xenon-based digital system. Whatever recommendations are presented here, and whatever decisions are made by the industry, should be the same, or fairly close to it, for both technologies.
2 Moving image Technologies, LLC Problems with current Color Temperature specification
1. Specifying a color temperature is outmoded. When the light source is a glowing material similar to the ideal blackbody (like an incandescent filament), the color temperature follows the blackbody locus, and a measured color temperature will correlate to the appearance of white on the screen. However discontinuous light sources such as discharge lamps (including xenon) or laser sources, permit white points distant from the blackbody locus, and color temperature is no longer a useful metric – a two- dimensional coordinate system is required to accurately characterize the white point. This is especially true when filtering is present in the illumination system.
2. Dichroic coating is an extremely complex process with many individual steps, hence a lot of variables. To really control the process tightly would make the equipment cost much more than exhibitors are willing to pay. Therefore they are not too tightly controlled in practice. There are substantial differences in reflectors from one production batch to another. This is clear from the tolerance that appears in the current 196 spec: 5400 °K -200°, +600°. This is a pretty large range!
3. The effect of the “blue cut” xenon reflector lowers the color temperature, but also shifts the white point further off the blackbody locus toward green. This green shift is difficult to correct, and in fact is not corrected in practice. All modern cinema systems appear slightly green compared to a more ideal blackbody light source (or the sun). I believe that adopting a slightly higher color temperature as standard would reduce the impact of process variations, and allow tighter control of the white point in practical film systems.
4. Now that carbon arc is effectively obsolete as a projection light source, there’s no longer a need to match its old white point. Realizing this, some lamphouse manufacturers have relaxed their filter coatings for better light output, while others still attempt to maintain the SMPTE-specified white point. As a result, the range of color temperature encountered in the field today is fairly wide, between about 5500 to 6500°K. Most of these must be called “correlated” color temperatures because they are almost always substantially off the daylight locus toward green (see Figure 4).
Note: Up until now I’ve been showing both the blackbody and daylight curves in the figures, from now on I’ll just show the daylight curve, for clarity.
3 Moving image Technologies, LLC 0.38
Sample Mean 0.37
0.36
y 0.35 D-55
0.34
0.33
D-65 0.32 0.3 0.31 0.32 0.33 0.34 0.35 0.36 x
Figure 4: White points measured in film cinemas Represents measurements of over 100 actual theatres and screening rooms. (data from THX)1
Recommendations: Following are the recommendations of the author in two areas: white point, and tolerance.
A. Potential, practical white point choices:
1. D65. This is a point established by the CIE, very commonly used as a white reference in many industries dealing with color. Con: although we might not necessarily have a problem with projectors going up to D65, we must realize that whatever point we select will have a tolerance applied to it. With a realistic tolerance applied to D65, we’ll have white points in theatres up to 7000°K (and probably beyond). This is quite blue.
2. D60/61. This is a compromise, centrally located on the daylight locus between D65 and present D55. Although points such as D60 and D61 are not recognized as “official” reference points by the CIE, they are nevertheless “real” points on the daylight locus, documented in the colorimetric literature, and available for people to look up in references if necessary.
3. Texas Instruments: x=.314, y=.351. Some would have us accept that actual white points in theatres will always be green-shifted, as they are now. We’ve been living a lie for decades now, pretending that theatres were being calibrated to D55, when they weren’t. Do we pick a new point on the daylight curve and continue the pretence of a standard no one will follow, or should we recognize reality by choosing a target shifted toward green? The latter choice is the tack that TI has taken with their DLP Cinema™. They did extensive work that showed this better matches existing film systems, and gives the best results as far as light efficiency and dynamic range with xenon lamps.
4. Some compromise between all of the above.
4 Moving image Technologies, LLC
5. Another way to approach the problem of white point is not to pick a target point at all, but instead to select the tolerance boundary, or range of white points that we’ll accept, which may be a more important issue anyway. Selecting a boundary then establishes a de facto “virtual” white point at the center of that area. We’ll have to consider what acceptable tolerances are anyway, so in lieu of any clear industry consensus on the white point value, this is the method I’ll proceed with here.
B. Tolerance range The actual target white point is meaningless without a tolerance defined. First of all, how many classes, or levels of precision are needed for the various theatrical applications? We know the digital systems are capable of being calibrated much more accurately than film systems, so I’ll propose 3 performance classes, as follows: Class A: Best attainable accuracy, for digital color timing. Class B: Minimum acceptable limit for commercial theatrical digital presentation. Class C: Minimum acceptable limit for commercial theatrical film presentation.
I’ll assume the tolerance for Class A is the tightest that’s possible to measure with modern instrumentation, and won’t address it any further here. Instead, I’m going to propose a boundary for Class C, then simply narrow down that area for Class B. My intention is to end up with a generous tolerance for film in commercial theatres, extending from roughly 5500 to 6500°K. This will disenfranchise a minimum of the equipment and systems currently in use. But I also want to recognize the extensive work TI has done with regard to xenon light systems.
In Figure 5 I’ve drawn a triangle between D55, D65, and the white point selected by Texas Instruments, just as a starting point to establish a potential boundary of acceptability. The 1931 CIE x,y chart isn’t perceptually uniform, for this purpose we should use the 1976 CIE u’ v’ space. In Figure 5b I’ve shown the same triangle defined in Fig 5a, along with a circle that just encloses all three points.
.360 .500 D50 D45 T.I. DLP CINEMATM .350 .490 D50 DAYLIGHT D55 LOCUS D55 TI .340 D60 v' .480 y 61 D60 .330 D65 .470 D65 .320 .460 D75 D
.310 .450 .170 .180 .190 .200 .210 .220 .230 .240 .300 .310 .320 .330 .340 .350 .360 .370
x u' Figure 5a: D55, D65, & TI triangle, x, y Fig. 5b: Same triangle in u’, v’ space
Since meters are commonly available that measure in x, y rather than u’ v’, chromaticity coordinates are more usually specified in the former system. We can translate the circle from Figure 5b back into x,y space, where it becomes an ellipse.
5 Moving image Technologies, LLC .370 .370
.360 .360 D50 D50 TI TI .350 .350 DAYLIGHT D55 D55 LOCUS
.340 D60 .340 D60 y 61 y 61 .330 .330 D65 D65 D D .320 .320
.310 .310 .300 .310 .320 .330 .340 .350 .360 .300 .310 .320 .330 .340 .350 .360
x x Figure 6a: Perceptually equal tolerance applied Figure 6b: Finding center point
It can be seen that the circle in Figure 5b becomes an ellipse in Figure 6a, about twice as long on one axis as the other. This shows that our eye is less sensitive to hue changes along the daylight locus than it is at right angles to it. Its axis is rotated almost exactly 45 degrees to the x, y grid. Unfortunately, due to this rotation it becomes difficult to specify a tolerance that’s perceptually uniform, because the tolerance would have to written differently for each of the four quadrants. We’d rather not have to do trigonometry after making a measurement in order to find out whether a system is within acceptable limits or not.
The center point of the boundary created with this method is x=.320, y=.342. The resulting tolerance shown is ±.012 on the x axis, and ±.013 on the y axis. Film needs a large tolerance space because there’s no mechanism to adjust it in the theatre. But we don’t want to allow it to go too green either. A way to address this problem is to apply an asymmetrical tolerance to the white point for film. This offers an interesting possibility… notice that the TI white point is almost exactly in the center of the upper left quadrant of the tolerance box I’ve derived here. If we simply tighten and reposition the boundary slightly on the y axis, a simple fractional relationship is created between the white point I’ve proposed for film, and the white point TI has selected for their digital technology (Figure 7).
.360 D50 TI .350 B D55
.340 D60 61 y C .330 D65 D .320
.310 .300 .310 .320 .330 .340 .350 .360
x Figure 7: Potential tolerances for Class B (digital) and Class C (film)
6 Moving image Technologies, LLC This change makes the tolerance box for film exactly four times the area of digital. This spec could be written for digital and film projection as follows:
Film Digital x 0.320 ±.012 0.314 ±.006 y 0.345 ±.012 0.351 ±.006
Or alternatively: Film Digital x 0.314 +.018, -.006 0.314 ±.006 y 0.351 +.006, -.018 0.351 ±.006
The two specifications have exactly the same result on the screen. Reducing the target y value by .001, to .350, would make the respective boundaries match even better those I derived in Figure 6. But of course this is all little more than superstitious numerology, and many different numbers with other interesting relationships could also be derived for this.
Conclusion The current white point specified for film, D55, is impractical and outdated. It is also being largely ignored in the industry. Even in studio and post-production screening rooms where the customers are told the theatre is calibrated to 5400ºK or D55, in reality they typically aren’t very close. The technicians who set up those theatres had to go through dozens of reflectors to try to find one that’s even close to the spec. What happened to all those reflectors that weren’t close, that the technician discarded? They aren’t thrown away, they eventually make their way into commercial theatres.
I have presented here realistic, achievable proposals for a new white point and tolerance ranges. Clearly, D55 is not the optimum point and should be abandoned. Likewise, D65 is also not very desirable as a target point, being very near the boundary of what is perceived as a “neutral white” at the opposite end of the range. However I have no strong feelings about the exact white point selected, as long as it’s within the approximate range outlined here. I should mention that the difference between any points from D55 to D65 is fairly trivial, and only the true “golden-eyed” would be able to see much if any difference in actual theatres (assuming no other white reference is present).
The more important issue is, should we pick a white point on the daylight locus, which we know we can never exactly meet with a real product? Or do we submit to that reality and pick a point shifted off the locus? I expect this to be a bone of contention, with firmly divided opinions. Back in Figure 6 I showed that our perception of hue on this axis (North and South of the daylight locus) is even more critical than the color temperature number, which is all we ever used to consider in the past. So I think this warrants some serious consideration.
Speaking on behalf of a manufacturer of xenon illumination systems, I selected the point presented above as my preference from the standpoint of an attainable white point with good repeatability and efficiency. However I also recognize the “altruist” view and even the political aspects of the discussion that may push us toward some sort of “ideal” that may or may not be truly attainable in practice. I look forward to further, rational discussion on this topic.
7 Moving image Technologies, LLC References: 1 “The White Paper: Considerations for Choosing White Point Chromaticity for Digital Cinema”, Matt Cowan and Loren Nielsen, 143rd SMPTE Technical Conference, Nov. 4-7, 2001. [The reader is strongly encouraged to consult this paper, which provides an excellent overview of the prevailing situation in theatres today].
Author bio: David Richards is a co-founder and Principal in Moving image Technologies (MiT). Prior to co- founding MiT, Mr. Richards was employed by Christie Digital Systems, in Engineering and Management roles from 1997-2003. He serves on several of the SMPTE DC28 committees as well as the F2 Film Technology committee and P3 Projection Technology committee. He is past chair of the SMPTE Hollywood section, and was Program Chair for the first and second SMPTE Film Conferences, held in 1997 and 1998. He is the author of several papers and articles for various trade publications.
8 Moving image Technologies, LLC