Color Meter Weirdness

[Michael Panfeld]

Hi: I recently acquired a Sekonic C-500 color meter. I sent it off to have the sensor replaced and recalibrated (there was a recall of earlier serial #'s). I was excited to test it out against my trusty Milonta Color Meter II (which was also recently calibrated).

The Sekonic has two modes: film and digital. Its supposed to be some fancy new 4-sensor arrangement for use with digital camera sensors.

For the sun, tungsten and HMI light sources, the two meters tracked fairly well together. That is in measuring both color temp and green/magenta shifts. Didn't matter much if I had the Sekonic in film or digital mode for these three light sources.

However, for LED and fluoro sources, there were huge gaps when using the Sekonic's digital mode. I have an LED panel that measured 5600K and no green shift on the Sekonic in digital mode, but mesured huge green shifts and around 8,400K in the Sekonic film mode. This large discrepancy was pretty much confirmed with the Minolta, although the latter had slightly different readings.

1) So my first question is can someone explain a bit more about the "digital" mode on the Sekonic meter? I am looking for something beyond the company's literature and marketing materials (which are usually what is copied and pasted into blog reviews, etc).

2) Is this meter's digital mode (or some similarly-configured meter) how light manufacturers are getting away with claiming 5600K color temps and minimal green that turns out to not be true when you get the products? I'll note that I have returned LED fixtures and CFL & 55 watt fluoro lamps becuase they registered 8,000K and very green on the Minolta meter (which was before I got the Sekonic).

3) Finally, what do these readings, and the meters two modes mean for a cinematographer, a colorist, and someone just trying to buy and use lighting gear in today's fast-changing marketplace?

Thanks

 

[Michael Panfeld]

bump

 

[Michael Panfeld]

Anybody want to chime in here?

I hope to go to NAB. If I do, I will be bringing my meters and a HV30. I'd love to test the lights on the spot and get the reps reactions if the meter has divergent readings.

If anybody is going to NAB, please do the same.

 

[Graeme Nattress]

I don't get it..... Correlated Colour Temperature is not altered by what format, Film or Digital, you're shooting on! The whole goal of a colour temp meter should be to report the temperature in Kelvin, and give an indication if the light source is above or below the Plankian locus and thus has some tint.

Graeme

 

[Michael Panfeld]

I don't get it..... Correlated Colour Temperature is not altered by what format, Film or Digital, you're shooting on! The whole goal of a colour temp meter should be to report the temperature in Kelvin, and give an indication if the light source is above or below the Plankian locus and thus has some tint.

Graeme

I agree, color temp is color temp, or at least it should be. But its actually not, according to the Sekonic's readings. I am hoping somebody can explain the results I am getting.

The two settings yield the same color temp for the following sources: the sun, HMIs, and tungsten

The two settings yield widely divergent color temps for the following sources: LEDs and fluoros.

Moreover, the Minolta Color Meter II's readings matchup with the "Film" setting's readings on the Sekonic, which is always way way bluer than the Sekonic's "digital" setting reading.

 

[Graeme Nattress]

I think that person would be Sekonic....

Graeme

 

[Richard Andrewski]

You have to know the limitations of color meters, how they react to "CCT sources" and also what that button on your meter means before you start assuming lighting manufacturers are at fault for whatever you are seeing.

We've talked about this subject before in the past elsewhere, which is basically that color meters aren't good for sources measured with CCT:


http://www.dvinfo.net/forum/photon-management/128213-led-lights-6.html

http://www.dvinfo.net/forum/photon-management/236502-my-view-led-lights.html

http://www.dvinfo.net/forum/photon-management/467263-flo-vs-led.html

http://www.dvinfo.net/forum/photon-management/470188-cri-really-dead.html



And elsewhere as well. What these posts are all saying is that there’s a reason that lighting manufacturers use the combo of integrating spheres and spectroradiographic equipment and not color meters. This is what I mean:

Thus, there is no pocket solution that you can just aim, press and expect consistent and accurate readings every time (for sources with correlated color temperature like LEDs and Flos). I know that’s what people may want or expect but that doesn’t mean it’s possible or practical with the measurement technology available and the popular sources being read by that technology. If nothing else, too many possibilities for contamination with the pocket solution. Paint color on walls, daylight coming in windows, practical fixtures all around, etc. An integrating sphere has a very good and expensive digital sensor in it that’s often calibrated with the tungsten bulb supplied by the sphere manufacturer. It’s a sealed environment with a neutral white and uniform coating on the inside. No light leakage in or out. The “light” equivalent of an anechoic chamber. It’s the only acceptable and repeatable solution for measurement and therefore it’s the standard with lighting manufacturers of all kinds. Then you have the issue of these color meters can’t consistently and accurately read two of the most important sources today: LED and Fluorescent. LED and Fluorescent have a lot in common in how they work (phosphor radiated by UV) so this is to be expected that they both have the same behavior in trying to measure them. It is in fact the spikey, simulated full spectrum behavior of these type sources which makes them harder to read accurately and consistently by a pocket solution out of a predictable environment. That doesn't mean these sources are unusable and they are very usable when the CRI is high enough or you can correct with a bit of minus green if necessary.

It’s not a big deal for even older color meters (as long as they are calibrated) to work correctly with full spectrum sources like real daylight, real tungsten or carbon arc. Even HMI as observed is close enough to full spectrum to show up more consistently in readings, although you can get different readings based on whether an instrument is in flood or spot mode. You still have potential contamination issues between competing sources in an environment, the walls and their color, etc. So again, there’s no reliable 100% “pocket solution” even for HMI.

Film or Digital mode? 99% of the members on this forum should be using the “Digital” mode. That's because we are discussing digital still cameras, digital video cameras and even a “digital film camera” and all the peripherals used around them including lighting and how they respond to lighting. Some people here have no doubt used or are still using both real film and digital (the other 1% of forum members), but they would hopefully understand already what these buttons are for on a color meter and what the differences are in response between real film and digital to how they react to various kinds of lighting. The lion’s share of the other 99% didn’t even know there was a difference between film and digital however and some will never have used film of any kind.

Digital mode closely mimics the response of digital media and human perception of lighting. Integrating spheres use a digital sensor and model for human perception as well. Why wouldn’t they? Most of the lighting in the world is made for humans and not specifically for cameras. Lighting manufacturers of all kinds use these spheres and again, that is the only serious standard for repeatable light measurement until someone invents something better and gets it adopted by everyone.

Film mode on a color meter mimics the response of real film. Film is more sensitive to green spikes than digital is. This is something people using real film have always had to deal with for as long as fluorescent and more recently LEDs have been around (HMIs too for that matter). People will often report green spikes with even some of the highest CRI flo tubes when using real film. The same tube may show no issues at all in digital. Many report that Kino bulbs seem a trifle too “magenta”. There’s a good reason for that. Kino has been a choice in the film world precisely because they do overcorrect. You may feel it’s too much for digital use however and will seem a bit too pink and a lot of people have said this on forums such as this one. This is not a conspiracy; it’s simply the reality of the differences between the two major competing types of media. I would go so far as to say there is actually a film CRI and a digital CRI because of this sensitivity.

Digital has white balance which can take out a lot of that kind of thing and can decide at the push of a button what white is based on the type of lighting you tell your camera it is experiencing. Not so with film: you have daylight or tungsten balanced film and have to stick pretty close to that and pay attention to the lighting in a far more deliberate sense than most using digital mediums may have to.

Used vs. new color meters. If all that isn't enough, then consider a used color meter and the potential for being even more off than a new one when reading spikey CCT sources. So you go shopping for color meters on Ebay and you see the new ones running in the range of $1000 to $1500 and all those used ones for one third to one half the price--what a deal! Why would you get the new one? These “bargains” are mostly there because at some point the owner discovers these older models aren’t that accurate for LEDs or fluorescent, which is what many people are moving to today. He may not even know why but still, there it is and has to be dealt with. The unknown answer may be that it’s an older “RB” (Red/Blue) type that was invented before the digital era and not a newer “RGB” type which has at least some hope of reading newer sources (flo and LED) somewhat. There may also be the issue that they aren’t calibrated and the owner finds out there is no easy way to calibrate his defunct meter. Or both being old and non-calibrated which should hopefully be the best bargain to be found. So these sellers have decided color meters aren’t necessary because they seem to be so inaccurate, or they are going to upgrade to newer RGB models like the Kenko KCM 3100 (which is my meter of choice) and want to pass the meter off to the next “beginner” for him to discover all this later. So until this new owner starts to suspect that it’s the meter and not their lighting that is the issue, it can provide for hours of “perplexity”.

 

[Ron Smith]

Epic. Excellent post Richard. That being said, I'm very happy with my C500 Sekonic and it is quite accurate (in a controlled situation of course)

Also, Michael Panfeld, why don't you do a test with the RED camera and see if the readings on your color meter match what you see when you white balance your red to the subject being lit by the LED or flo?

Let us know what you find....

 

[Michael Panfeld]

uhh, correct me if I am wrong, but you and other manufacturers state that their fixtures have a certain color temperature. In your case, I believe its 3200K and 5600K. That has to be measured in some way, and one part of my question is what are you using to arrive at the advertised color temp.

Integrating spheres are used for measuring total light output (lumens) of a light source. Spectra radiographic gear measures the breakdown of light amongst the wavelengths (http://www.spectralevolution.com/files/Spectral_Evolution_Si-based_spectrometer_series.pdf). Neither of these devices measures color temp. Only a color meter does that. So all of that defensive narrative, its mostly gobbledy-gook. IT ALSO SEEMS GEARED TOWARDS SUBTLY DISCREDITING MY OBSERVATIONS AND QUESTIONS, which I do not appreciate.

I am not stupid or inexperienced. Nor do I have some defective cheapo light meter that is uncalibrated, as you imply. Just the opposite. I have one of those $1300 meters, as well as an older model. Both are freshly calibrated. Both are getting consistent readings. So your last paragraph is quite offensive.

Also offensive is your suggestion that my observations are haphazard. Again, just the opposite. I performed a controlled experiment by placing both the light source and meter on fixed stands so that their position did not change. I also did this is an enclosed room, covered with duvetyne on all 6 sides. I would actually argue that an integrating sphere with all black is a more accurate way to measure color temp than the white interior of the sphere in your pic, because the white paint will impart a tint to the light it reflects.

You also go on to bash my understanding of the digital mode on my meter. Sekonic has published almost zero information on the digital mode beyond the obvious statements. And yet you suggest that pros should know the difference. How is that possible if Sekonic keeps all the cards to itself?

The fact is that light only has one color temp. So, when the digital mode gives a radically different reading, it means that it is simulating how a camera sensor would perceive the light. The problem with that is twofold. First, not all sensors are the same. Second, there are all kinds of assumptions that Sekonic must have made when it programmed its processor to interpret the data. What are those assumptions? How do they relate to the MX sensor (which I saw tonight at Ketch Rossi's demo in DC)? How do they relate to the Canon 5D sensor, Af100 etc... If I have more information, I will be better able to use this tool. That's all I am asking for, Richard.

On the other hand, in spite of all the issues with assumptions, you advertise a certain Kelvin rating for your fixtures. There is no asterisk. There is no caveat. There is no statement of how that was measured. So the consumer must simply take your word for it. And yet, the color meters and simple logic tell me that those advertised Kelvin ratings do not match at least one version of reality.

So yes, some of the light sources that I measured were Coollight LEDs. I really did measure these at about 8500K in film mode. But now that I have read your explanation, I can safely remove all that CTO and minus green gel from the LED lights. That's because the painfully obvious strong blue light that I can plainly see (that would be human perception) when it is placed next to an HMI that is supposedly a slightly cooler Kelvin than your LED's advertised color temp, is just an illusion that will not be captured by the digital sensor.

Except that editors tell me, and I have read posts about all the green casts and color correction that footage from digital, not film, cameras needs when shooting with LEDs or fluoros.

If this sounds a bit like a sarcastic rant, it is. I asked for some help understanding a complex topic, to which a) you did not directly answer any of my questions and b) you gave misdirection. If there is a conspiracy, I was certainly not implying one. Lighting manufacturers, acting individually and responding to market forces do not constitute a conspiracy. And the marketing forces are this: if you tell people that a color meter will show that a light source is 8500K, and if you shine a "5600" LED next to an HMI or some sunlight, and its obviously bluer than those sources, you are going to have a hard time selling lights. So its actually against your economic interests to bring all of this complexity into your marketing materials. That's why they invented the phrase caveat emptor. My post was imply trying to shed some light on this.

If anybody actually has some answers to my specific questions, I would love to hear it.

 

[Richard Andrewski]

Integrating spheres and their peripheral gear measure lumens, correlated color temperature, CRI, give a decent print out of spectrum of the light among other things. Color meters only show color temperature reliably for full spectrum sources, don't give CRI, nor do they give lumens and all the other items that can be found on the printout test report. You can see a scanned test report in our blog from a sphere and its electronics.

http://www.coollights.biz/wordpress/archives/category/cool-lights-bulb-testing-results

 

[Graeme Nattress]

Michael, Richard's response is entirely reasonable, but I don't think he's answering your question quite directly. Let me see if I can explain.

So Colour Temp Meters only work well with continuous light sources. These sources produce a colour temperature that lines along the black body radiator curve and are the only light sources that you can say has a real colour temp. All other light sources have what is called a "correlated colour temperature" which is when you figure out what the colour temp of the light would be if it was continuous and also describe how far above or below the black body radiator curve they are.

An integrating sphere is used for many things. If you want to accurately measure the CCT of a lightsource, they're a useful tool. However, what you don't use is a colour temp meter as in my experience they're less than accurate. What you use is a spectroradiometer, probably costing more than your digital cinema camera, and highly calibrated. They measure the entire spectrum then accurately calculate the colour coordinates of the light source and will provide an accurate and repeatable measure of CCT.

The spectroradiometer knows nothing about the camera you're going to be using. All it cares about is the spectrum that it's pointed at. If you want to know what a light source is like - you need to see it's spectrum, not what it's reading on a CTM. I like that link above showing the spectrum of the lightsource and the spectroradiometer test results. That is exactly the kind of info it's good to see.

Graeme

 

[Conrad Hunziker]

Heres some great information by an independent third body, the Science and Technology Council of the Academy of Motion Picture Arts and Sciences (the science arm of the guys who put on the Oscars). They have been doing some recent work into the problems and possible solutions to SSL (Solid State Lighting: LEDs). I highly recommend this video as a primer to the issues with SSL, but their whole site goes into much detail.

 

[Stephen Strangways]

I have an LED panel that measured 5600K and no green shift on the Sekonic in digital mode, but mesured huge green shifts and around 8,400K in the Sekonic film mode.

The majority of unfiltered LEDs do have a green shift. It seems to be a common perception that film cameras and the RED ONE are more sensitive to the green spikes in fluos than other digital cameras. So is it possible the Sekonic digital/film modes are aimed at that perception? If so, I'd expect the Sekonic in film mode might better match the RED ONE, though I would of course test it by shooting some test footage, and seeing where tint is best set on the RED ONE footage to match a colour chart. Does it require as large a minus green shift as the Sekonic meter is suggesting in digital mode?

The fact is that light only has one color temp.

It's not really that straight-forward. Here's the spectral energy distribution of the sun:

Here's a tungsten bulb and daylight:

And here's the CoolLights 5600k, quite typical overall for a fluo:

Whenever you get into discontinuous spectrum lighting like LEDs or fluos, things get really interesting because we are so far away from the theoretically perfect black-body radiator that kelvin colour temperatures are based on.