Experience with color calibration between lamps and lasers, and their matching.

[Steve Guttag]

I color calibrate at 14fL. As a result, regardless of where I'm at in the range, it is going to be either right or very close. For the record, on lasers, I like to start them below 50% and I have some near 20%.

It will be a while before we really know how long these lasers are gong to last in real-time. I suspect having to re-laser a projector will be akin to buying a new battery for your EV...expensive. Plus, when companies rate their lasers, they keep using ½-intensity as their life...so, if you are starting out near 100%...or even 80%...are you good with 7fL (24Nits)? I think not. But it will be too expensive for most exhibitors.

I'm happy to see that the decay is moving slowly (particularly the ones near the bottom of the range)...so being closer to 50% should be safe but I'd keep below 75% still.

[Ryan Gallagher]

I color calibrate at 14fL. As a result, regardless of where I'm at in the range, it is going to be either right or very close. For the record, on lasers, I like to start them below 50% and I have some near 20%.

It will be a while before we really know how long these lasers are gong to last in real-time. I suspect having to re-laser a projector will be akin to buying a new battery for your EV...expensive. Plus, when companies rate their lasers, they keep using ½-intensity as their life...so, if you are starting out near 100%...or even 80%...are you good with 7fL (24Nits)? I think not. But it will be too expensive for most exhibitors.

I'm happy to see that the decay is moving slowly (particularly the ones near the bottom of the range)...so being closer to 50% should be safe but I'd keep below 75% still.

Yeah I found the comment about the tracking only really working in the upper range to be sort of a non-feature too, considering I was aware that the current "advice" is to spec near 50% or below on the install.

Ed had done exactly what you describe for the scope and flat formats he built. But then when he was gone and we started from flat macros and zoomed to something close to scope for F220, that is where we ran into the issue. I suppose if I had started from the scope macro set and applied a different picture crop and zoom, maybe we would not have needed much adjustment (and avoided most of the unwanted color drift).

[Stefan Scholz]

I remember the book "Е.М. Голдовский Техника кинопроекции", Goldovskij was a famous researcher and scientist in the field of cinema projection, and Technics of movie projection seemed to be a standard book on the subject in the 40s and 50s.
They did extensive research on color viewing, and after all it was not important to have a correct color reproduced, it is more important to have it psycially correct, meaning accepted by the majority of viewers.
Otherwise hydrotyped color films would have never worked, due to their oversatured color image.
Other more important factors, than color matching are brightness, displayed contrast ratio, light source etc.
A cameraman saw colors in extremely bright light - film had very low sensitivity. Screening the print a theatre, at light level well within the night vision - 50 nits or 16 fl is borderline night vision - leads to a noticable shift of the perceived color on all humans watching it.
Film grading as I know it, has been done on monitors calibrated to a color profile, at 150 nits.
We had a screening upfront a crew, the projector, lamp based, calibrated to 14 fl, and industry standards in hall No 1. Massive complaints, they have to redo the grading, it is well off what they know etc.
Then asking them into hall No2, which has a new Ultra high contrast Laser 4k projector at 27 nits. Same DCP, ansd suddenly they said, looks like expected. WEll, we are in the day vision range. Which makes a difference. After all we asked, at which setting they calibrate their monitors - 150 nits immediate answer.
Projected contrast also has an influence on the image, 14 to 16 fl, or around 50 nits peak brightness! This is as wrong as the 85 dBC SPL on the RLP in the room. A medium brightness of 30 nits or 10 fl must be kept, therefore, the higher the contrast ratio, the higher your peak brightness must be.
And last but not least, sound reproduction and quality of sound reproduction also interferes with the way humans perceive colors.

There are so many aspects, that I have no sympathy for color graders claimig a projection is off. Just like hearing, one of the last factors is your daily condition, if you are tired, awake, distracted.
Yes, early laser projectors had too narrow wavelength, wave length extension helped, and most problems are no longer prominent in a well calibrated setup on a matt white screen.

[Christos Gartaganis]

Christie has just released a new series of Hybrid Laser projectors which combine an RGB laser source and a specially designed phosphor wheel. They claim they have solved all known issues associated with laser speckle and color metamerism. They are also suitable for post productions houses where color accuracy is top priority. The contrast ratio can go as high as 6.000:1 with special UHC lenses available exclusively from Christie.

[Marco Giustini]

interesting, I found them on their website. Does anybody know how they work?

[Stefan Orlandic]

I apologize for bumping an old thread, but I came across it during some random browsing of this excellent forum.

I’ll be addressing mostly the first and second pages of this thread, where a number of misconceptions were presented regarding Barco, the SP4K platform, metamerism, and several related issues.

Having been part of a closed group of post-production professionals invited to Barco’s facilities in 2013 to test and analyze the first-generation laser projector prototype, I can say the color discrepancies were already substantial even then. So, 13 years later—and after six or more iterations of different laser illumination systems—we are still having essentially the same conversation.

We were the first post-production facility in the world to install a Barco SP4K RGB projector, and it has now been in our studio for more than five and a half years. To my knowledge, we have probably done more feature film color grading on this projector platform than anyone else in the world.

Barco had to revise official brochures and published specifications because of issues we raised, and likely spent seven figures on R&D, travel, consulting, and related efforts—much of it, unfortunately, to no useful end—after I spent nearly two years being a daily pain in the a@@.

It took Barco executives several years to fully understand the extent of the mistakes made with the SP4K platform—issues we had been briefing them on from day one.

To quote one of their most senior executives: “You are the only installation in the world, out of thousands, complaining about dozens of color and image quality issues.”

I may still be under a gentleman’s agreement with Barco not to publicly disclose all details and findings regarding their RGB laser projector platform, but I can share some broader observations.

I have written dozens of pages of reports, findings, measurements, conclusions, and recommendations. Unfortunately, most of them were read, acknowledged, and then effectively ignored.

Barco’s earlier 6P laser systems had their own metamerism-related issues as well, but the image tended to lean more toward magenta rather than the green bias seen on the SP4K platform. As a result, the image subjectively appeared less objectionable—or simply less obviously wrong.

Barco’s new MOC was designed for RGB laser projectors (SP4K-P, Freya+, Nerthus) using a different green laser plate than the standard digital cinema units. As some of you already know, that MOC cannot simply be applied to models using a different green laser configuration.

But even for the projectors it was supposedly designed for, Barco’s MOC is fundamentally flawed—not just in execution, but in the underlying methodology, workflow design, testing approach, conclusions, and the public data they chose to publish. Frankly, as a practical tool for serious post-production work, it achieves far less than many seem to believe.

More importantly, the entire process appears to have been built without properly measuring a range of other critical color parameters—some arguably far more important than simply identifying an alternative white point.

And for the record: slight changes in RGB coordinates within the MCGD result in essentially zero visible perceptual difference when it comes to metameric correction.

In practice, you need to create your own MOC based on perceptual matching against a trusted reference LCD display. In many cases, that means creating a different MOC for each individual project, because the issue goes much deeper than simply selecting a different white point. There is no universal MOC that solves every case.

Typically, the required re-adjustments (project to project) end up somewhere in the region of ±0.002 on both x and y relative to measured MCGD values of white point.

However, once you start shifting the white point toward a perceptual match, you introduce other problems: green gamut has a serious collapse and significant deviations in saturation sweeps within the color volume. The result is that certain colors begin to look clearly wrong or unnatural.

That is precisely why you need a proper LCD reference display as a sanity check (FSI DM240, DM242, etc.).

Color grading in P3-DCI white point using any MOC is, in our experience, largely impractical. Achieving a reliable perceptual match to an LCD reference is extremely difficult, and the workflow simply does not hold up well in real-world grading conditions.

That is why we ended up doing all of our grading in P3-D65, calibrated via perceptual matching to a properly calibrated reference LCD monitor.

We were also the first studio facility to test the final SP4K-P projector, and once again, my eight-page report—full of concrete findings, practical recommendations, and clear proposals on what needed to be fixed and how—was ignored. The result: another flawed product release.

The idea of implementing MOC only on post-production projectors, while not rolling it out via software update to commercial cinema units, is frankly ludicrous. You would then be watching two fundamentally different images between grading and exhibition, which completely defeats the purpose of any reference workflow.

A number of solid, realistic suggestions—well within the actual release timeframe of the new projector—were ignored, only for objectively worse decisions to be made instead.

And that’s not even getting into the absurd idea of offering the post-production projector only as a rental, at an obscene price point. It took Barco roughly 18 months to realize just how badly they had misjudged that strategy.

Regarding the RGB laser architecture: these are not single-wavelength primaries, as some seem to assume. Each primary consists of multiple wavelengths.

The newer green laser plate uses a completely different green laser compared to the earlier generations, and that change is substantial. The previous green implementation was severely underperforming relative to DCI-P3 specification.

Now, regarding the often exaggerated claim that observer metamerism makes reliable professional grading impossible:

That has simply not been our experience.

Across 5 professional colorists, 50+ cinematographers, 50+ directors, and 50+ film editors who have worked in our studio over the years, not a single person has reported perceiving the image differently in any meaningful operational sense when viewing our calibrated SP4K setup.

I’ve read countless online posts, assumptions, and theoretical discussions claiming otherwise, but in practice, that has simply not been the case with the SP4K platform—at least not in our real-world experience.

To be precise: I am not arguing that observer metamerism does not exist as a scientific phenomenon. It obviously does. What I am saying is that, in our professional environment, across thousands of supervised color grading sessions, it has not presented itself as a meaningful operational problem when the projector was properly calibrated to P3-D65 and perceptually matched to a trusted reference monitor.

That includes both male and female professionals, across a wide age range—from roughly 30 to 75 years old.

So while metamerism is absolutely real, the narrative that it automatically makes accurate professional evaluation impossible, or that everyone will inevitably perceive wildly different color, is in my view massively overstated.

In the end, Barco MOC achieves far less than its proponents seem to believe. There are absolutely more effective ways to achieve better measurement accuracy, greater calibration precision, and a much smaller perceptual mismatch—but not within Barco’s current way of thinking, and certainly not with commercial cinema units using the older green laser plate as they exist today.

[Harold Hallikainen]

Extensive comments! How does a reference LCD monitor work? What is the backlight source? Is it a typicaly blue/phosphor LED or something else? Is the light fairly broadband with RGB filters?

[Steve Guttag]

Stefan, very interesting information, indeed. I can't claim the controlled post-production environment that you have. I can say, I do perceive the SP4K series as redish, not green. My first introduction to metamerism was on a Barco demo where I perceived the image a red shifted while the person next to me saw it as green shifted. It has been long enough that I'm sure that we were not seeing the current green laser plates being used (or the multi-wavelength reds either).

I've also noted that on the Christie CP4415 with the UHC-P lens, the colors looked "right" to me. Nothing felt "cartoonish." Christie also has their RGBH projectors coming out, which introduce a phosphor wheel in addition to RGB laser. Have you tested one of them (or the CP4415)?

[Marcel Birgelen]

Extensive comments! How does a reference LCD monitor work? What is the backlight source? Is it a typicaly blue/phosphor LED or something else? Is the light fairly broadband with RGB filters?

I think Sony has some OLED based reference monitors, but the one we have (I think, an X1800), is based on a very expensive, wide-angle LCD and a white LED array as the backlight. They're sparse on the details. There is a ton of electronics in there, that you would not get in a cheap-o LCD monitor of equal size. I guess much of the magic is in getting the electronics to match the deficiencies of both the backlight array and the LCD array in front of it.

Regarding color grading on the SP4K: We've given up on that quite a time ago. That's why we keep a DP4K-23B around, not because I like the space it takes up, but simply because current laser technology is a form of "personal NTSC": Never The Same Color(tm). Every professional color grader will hate the guts out of your laser projector. Part of it might be bias, but I can see the "color defects" myself too on the SP4K machine, especially when you know what to look for.

I think they tried to use some clever software hacks to mitigate parts of the issue, but to me it seems like that sometimes leads to banding artifacts, especially in large areas filled with "soft gradients". I guess people are so hooked up to their highly compressed Netflix streams with horrible contrast in many dark scenes, especially on some less sophisticated OLED screens, that people don't even notice them anymore.

[Marco Giustini]

Thanks for taking the time to share your experience, Stefan. I have save this conversation in case it disappears!

What you say is very interesting - my colour science knowledge is limited so I have questions, which might be silly, apologies.

One is: how does Barco MOC work compared to just dial different coordinates for Red, Green, Blue and White? If observer metamerism does exist and does apply to the SP4K as you say, how can that be practically eliminated if it's caused by a narrow spectrum?

Also: you say that once the projector is perceptually matched to a reference LCD screen, then you can grade on that projector. But how would that be rolled out to cinemas on the field?

[Marcel Birgelen]

MOC is a hack and everybody knows it. Since part of the calibration is based on perception, it's clear that the only thing the picture will be calibrated to is to the one who did the calibration. That's why we've given up on RGB laser for any color-related work.

The elephant in the room is CIE1931, are we going to give up CIE1931, a bundle of color science used since 1931 and replace it with something that fixes or minimizes the issue all together? The implication being that all the current hard- and software and all associated workflows need a thorough update or complete replacement. There are color models like CIE2016 that cope far better with the problem, but due to the enormity of the problem, those will not find broad adoption.

Meanwhile, the problem might just be a temporary one, as it only affects the current generation of RGB-based laser projectors and a bunch of OLED screens with narrow-band RGB light-sources.

How long is this technology going to last? How easy is it to "fix" the issue for those technologies? Can't Barco et all switch over to laser sources with broader RGB spectra? How long are we still putting projectors in rooms? When will those be replaced by LED walls? LED wall produces, for example, are already working on making their products "broad spectrum" trying to avoid those issues.

[Harold Hallikainen]

I know enough about color to be dangerous, so here goes... As I understand it, CIE1931 represents the average XYZ filter response in a large number of people. There is considerable overlap in these filters making it difficult to stimulate the X filter without also stimulating the Y filter, which limits the ratio we can develop between these two filter outputs. To try to get independent control over these filters, we use light sources that are at wavelengths where the overlap is low. For example we might use a long wavelength red source to stimulate the X filter while not stimulating the Y filter much. But this places the light source wavelength on the "skirt" of the filter where minor variations in the center frequency of the filter or the skirt slope produces large variations in the output amplitude and the corresponding ratio between filters resulting in metamerism. We could try to use the center frequency of each filter, but then we have substantial filter crosstalk reducing the filter output ratios we can achieve.

So, how does a broadband source solve this? What is the spectral plot of RGB light sources (or xenon through filters) compared with the XYZ spectral plot? Do we have limited color range due to the light source simultaneously "illuminating" adjacent filters?

LEDs, of course, have somewhat wider spectrum than laser, but are still pretty narrow compared to black body sources. What wavelengths are used in LED displays? Are they spread through use of multiple LEDs or phosphors?

[Marcel Birgelen]

So, how does a broadband source solve this? What is the spectral plot of RGB light sources (or xenon through filters) compared with the XYZ spectral plot?

Here's a plot of a Xenon arc lamp I stole from the Internet, compared to the sun. You see that even with filter, xenon is a true broadband source, mimicking closely the light coming from the sun.


Here a comparison of the actual output of a DLP projector with both xenon arc lamp (dotted line) and that of an RGB laser machine (solid line):

Spectral-power-distributions-of-the-RGB-primaries-of-a-laser-solid-lines-and-a.webp
Below, the well-known XYZ plot:

Screenshot 2026-05-26 at 22.42.33.png
Broadband light sources, such as filtered xenon lamps or white LEDs, address the challenges of the CIE1931 XYZ color matching functions by distributing energy across a wide spectral range rather than targeting narrow, specific wavelengths. The CIE1931 XYZ filters overlap significantly, making it difficult to stimulate one filter without affecting the others. This overlap causes metamerism, where colors appear different under varying light sources, and skirt sensitivity, where small shifts in wavelength or filter alignment distort color accuracy. Broadband sources mitigate these issues by averaging their energy across the entire visible spectrum, reducing the impact of minor spectral mismatches and improving color consistency. However, this approach also means that adjacent filters are stimulated simultaneously, which can limit the maximum achievable contrast between colors and slightly reduce the gamut for highly saturated hues. (Read: HDR becomes harder/more expensive to achieve ).

So, while broadband sources sacrifice some color range and purity compared to narrowband sources like RGB LEDs, they're a practical balance for real-world applications. Narrowband sources can achieve purer, more saturated colors by targeting the peaks of the XYZ filters, but they are highly sensitive to variations in wavelength or filter characteristics. Broadband sources, on the other hand, provide greater robustness and stability, making them ideal for applications where color consistency is critical, such as color grading. Their smoother spectral power distributions result in more reliable color reproduction, even if they don’t reach the theoretical extremes of the CIE1931 color space.

LEDs, of course, have somewhat wider spectrum than laser, but are still pretty narrow compared to black body sources. What wavelengths are used in LED displays? Are they spread through use of multiple LEDs or phosphors?

We've never really had those issues with CRT displays, did we? I remember that it took years before we got the first "color calibrated" TFT screens. So, color science and phosphor seems to be a well-understood issue. Also striking: The issue is far less pronounced with laser-phosphor machines. So, I guess that coating LEDs in all kinds of phosphor will be the easiest solution and far less of a hack than using LEDs with multiple colors.

Then again, using lasers with multiple wavelengths might be a partial solution for e.g. Barco. The SP4K laster plate contains like 30 individual laser diodes. If they would all be tuned to slightly different frequencies, couldn't that solve part of the puzzle? You'd expect it would even reduce speckle in the final output. I guess the problem here becomes an inventory and supply and therefore cost issue. What used to be one bulk item, now becomes 30 individual items...

[Harold Hallikainen]

Thanks! I notice that the peak of red, whether filtered broadband or laser, is about 625 to 650 nm, while the peak of the X filter is about 600 nm. So, in either case, we are on the longer wavelength skirt which is necessary to reduce crosstalk into Y, but makes stuff sensitive to the skirt slope and center frequency.

The green peak is about 525 nm, while the Y peak is about 550nm. Moving the green emission to a shorter wavelength reduces crosstalk into the X filter.

The blue emission peak is 450 to 475 nm, while the Z peak is about 450 nm.

Back in high school, I wondered how we mixed red and blue light to get violet, and I now know that it is because of the "second hump" in the X filter at about 440 nm. It's interesting (to me) that the second X peak is around the same wavelength as the Z peak. I'd expect the second X peak to be at a shorter wavelength than the Z peak so as we swept from longer wavelengths to shorter, we'd see Z max out, then see it start to decrease as X increases.

To get a wider range of colors, I wonder if we could go to four primaries, each individually modulated. For "deeper red," we'd decrease the emmission at 600 nm and increase the 650 nm. But, as I learned recently, with laser projectors, the individual laser outputs are not modulated by the DMD, but, instead, the lasers are combined to "white" light, then filtered to RGB for the DMD chips. Anyone up for more filters and a four chip projector?

[Marcel Birgelen]

To get a wider range of colors, I wonder if we could go to four primaries, each individually modulated. For "deeper red," we'd decrease the emmission at 600 nm and increase the 650 nm. But, as I learned recently, with laser projectors, the individual laser outputs are not modulated by the DMD, but, instead, the lasers are combined to "white" light, then filtered to RGB for the DMD chips. Anyone up for more filters and a four chip projector?

The interesting thing is, those are laser diodes, you don't need an extra mirror device to do pulse-width modulation, you can do that with the diodes themselves. You could easily add some new primaries and modulate them alongside the others to manipulate the "white light" output signal going into the prism.