CRI of HMI and Tungsten

[John DeLuca]

Will the higher CRI (color rendering index) of HMI pop color better than classic tungsten lights with a lower CRI? There is a huge difference to my eye, but will it make any difference with video color, or will it only be obvious with real film? Can anyone point me to a large/printable color chart, so I can record side by side tests and check the values in photoshop? I have a strong feeling HMI does indeed render video color better, esp in the blue channel.

[Jaron Berman]

I think there's a little confusion here. CRI is an opinion of how well a discharge source conforms to the spectral response of a blackbody emitter. If you put a tungsten lamp on that chart, it's automatically 100, because it behaves exactly as a blackbody emitter - it has a smooth, continuous spectrum. BTW, a blackbody is a "ideal" emitter, basically a superheated lump of carbon, and the Kelvin temperature (3200k, 5500k) refers to the temperature at which that particular spectrum is emitted. Discharge sources (xenon, fluorescent, hmi) don't have complete spectrums. Xenon is the closest to complete, but it still has spikes and voids. HMI and fluorescent can both approximate fairly well, but have far less continuous spectral curves than xenon, not to mention tungsten or daylight. Picture the tungsten response chart as a very gentle curve, where the HMI chart looks like a seismograph.

So while your eyes may think that HMI is the "better" source, as for actual spectral response, Tungsten is more complete and closer to the output of true sunlight sources, albeit at lower color temperature. Our eyes are more sensitive to blues/greens than yellows and reds, so the HMI has a more "poppy" look (also due to pronounced "spikes" at certain colors").

Obviously, this is a pretty quick explanation....definitely look into it more, but in terms of completeness of spectrum, tungsten is FAR better than HMI, Fluorescent, and even Xenon.

[Jaron Berman]

Oops, almost forgot:

SO WHAT?

Well, you can't invent frequencies of light that don't exist. That's why you can't ever "white balance" correctly under sodium vapor lamps - they have extremely incomplete spectrums. White balancing boosts and lowers the individual gains of each channel, R, G, B, in order to make "white" light. If there is no specific frequency of light cast on your subject, then no matter how much the camera tries to boost that particular frequency, you can't multiply 0. With tungsten, the light energy is concentrated less in the blues and more in the yellow-reds. So, the camera lowers the R channel and boosts the B channel. BUT, you start with light in ALL frequencies, so at least you have a fighting chance. HMI may appear nearly correct on a color chart, but you may find that when trying to match certain colors, it's just not possible. Just remember, HMI approximates daylight, but has a number of spikes and dips in its response. A video camera can smooth it out a bit, but cannot create frequencies that don't exist.

[John DeLuca]

Ah, that makes perfect sense. I guess I just like the "look" of electronic flash, but didn’t know if it would make a difference with video color or not. I will defiantly look into it more. BTW- what is your opinion on the Alzo HID lights?

[David W. Jones]

Originally Posted by John DeLuca

BTW- what is your opinion on the Alzo HID lights?

While I have no first hand experience using their products, a few things come to mind.
#1.
Ever heard the old saying "You get what you pay for"?
A full light kit costs about the same as a replacement lamp for one of my HMIs.
#2.
When a so called Professional light company calls a LAMP a BULB,
it sends me the wrong signal.

[Jaron Berman]

Interesting products. For certain special uses, they may be pretty cool, definitely a cheap way to add a daylight source. But, one thing to note is the CRI - 85. That's actually a bit lower than most standard industrial fluorescent tubes. Also, depending on what application this particular globe was originally intended, there may be output spikes. Again, our eyes are more sensitive to blue/green than red, so with industrial Flo's, HID's, etc..., they often boost the green/blue output - it artificially "brightens" the light - more lumens per watt (eventhough you have to filter that spike out in order to get smooth enough reponse to use on film/video... negating much of that "efficiency" advantage).

One reason why Kino flo's or Arri or K5600 HMI lamps are so much more expensive than their consumer market equivalents is for that reason. To make a lamp that's acceptable to film/video requirements takes a LOT of research. They basically try to smooth the spectral response of their light sources to eliminate as many of the spikes and dips as possible. Kino flo's run about 95+ in terms of CRI, and I believe Arri HMI's are all about the same 95+.

[Glenn Chan]

All light sources emit light at different wavelengths. The most useful description of this is a spectral power distribution graph (SPD); essentially, it just shows how much light is emitted at each wavelength.

The color of an object is determined by (A) its reflective properties and (B) the SPD of the light hitting it. If the SPD of your light sources differ, then the color of the object will be different. This is a form of metamerism. This happens with daylight versus tungsten, daylight versus 6500K (Daylight is not similar to a black body radiator in terms of SPD), and many other combinations of lighting.

In other industries, metamerism index provides a method of objectively measuring metamerism. Note that few people in the film/video industry really pay attention to metamerism (although this is something that a SMPTE committee is sort of investigating).

Lights with very weird SPD graphs (i.e. with lots of spikes in them) will likely exhibit more metamerism than other lights. They will also tend to have a low CRI rating (CRI is an objective measurement of how similar the SPD is to that of a black body radiator). But IMO, CRI is a little silly since it doesn't tell the whole story. Unfortunately, a lot of manufacturers tend only to provide the CRI figure.

Anyways, metamerism may not be worth worrying about. In film/video, you don't need extreme color accuracy. Film stocks, video cameras, and consumer monitors are all compromised in terms of color accuracy. And actually, color accuracy doesn't exist in the real world due to metamerism. The best you can do is to reduce metamerism. We tend not to notice color inaccuracies.

In some shows, you may also notice that closeups are lit differently than the wide. i.e. in CSI, the closeups may have a hair light that doesn't appear in the wide. It's essentially an intentional inaccuracy.

2- What you will likely be more concerned about is the effective color of the light source. In 3CCD cameras, this is determined by the SPD of the light and the spectral response of the three CCDs. Essentially, the light will have too much or too little red, blue, and green in relation to a reference color.

A practical example:
You may be shooting with both fluorescent (5000K, mystery CRI) and your own tungsten lights (3200K, 100CRI). You may find that after gelling your tungsten lights with CTB (converting from 3200K to 5000K), the fluorescents are a little green compared to your tungsten lights.

With HMIs, you may find them a little too magenta or too green (it depends).

3- In practical shooting, it will be difficult to get perfectly consistent lighting since the SPD and intensity of daylight changes throughout the day. As well, cloud cover can make very drastic changes to the light! In other situations, you are shooting in a location lit with overhead fluorescents (with medium CRI), and you simply have to live with it (and add plusgreen + CTB/CTO gels to your lights if need be).

The practical thing to do is to have a sense of the light's natural color (i.e. some fluroscents tend to be like 5000K, except greener) and to have some CTB, CTO, and plugreen/minusgreen gel handy (depends on the situation).

[John DeLuca]

Wow, you guys are amazing :-) I noticed a section in my electronic flash book about SPD graphs so I will have to read more into that.

[Richard Andrewski]

Originally Posted by Jaron Berman

Interesting products. For certain special uses, they may be pretty cool, definitely a cheap way to add a daylight source. But, one thing to note is the CRI - 85. That's actually a bit lower than most standard industrial fluorescent tubes. Also, depending on what application this particular globe was originally intended, there may be output spikes. Again, our eyes are more sensitive to blue/green than red, so with industrial Flo's, HID's, etc..., they often boost the green/blue output - it artificially "brightens" the light - more lumens per watt (eventhough you have to filter that spike out in order to get smooth enough reponse to use on film/video... negating much of that "efficiency" advantage).

One reason why Kino flo's or Arri or K5600 HMI lamps are so much more expensive than their consumer market equivalents is for that reason. To make a lamp that's acceptable to film/video requirements takes a LOT of research. They basically try to smooth the spectral response of their light sources to eliminate as many of the spikes and dips as possible. Kino flo's run about 95+ in terms of CRI, and I believe Arri HMI's are all about the same 95+.

Most of the tubes produced today fall in the range of CRI 75 to 80--90% of all the applications for fluorescent lighting don't care about CRI. Those that do care, for now anyway, are a small minority--that's why the higher CRI tubes aren't so readily available now and why they are more expensive. The special phosphor powders used for high CRI bulbs are rare and the process on how to use them is not well known. Simple supply and demand.

For anyone interested in CRI, I have an article I wrote on my website about the subject of Color Temperature and CRI and how you measure it as well as all the issues involved in producing fluorescent tubes as I have just gone through this process in sourcing tubes in China for my company.

http://www.coollights.biz/wordpress/archives/13

I also have another article I wrote which talks about why fluorescent tubes have mercury in them,

http://www.coollights.biz/wordpress/archives/14

Some people think if its not CRI 90 or above you shouldn't even consider the bulb. That's a blanket statement and not entirely accurate. I've done entire productions in CRI 80 5500K or 3500K or 3000K bulbs and not seen any ill effects whatsoever. It all depends upon what you're doing and the medium you're doing it in. Digital video and photography are not so susceptible to this "green spike" as everyone seems to believe. Film on the other hand may be--but film is being used less and less today and what medium are most of the people on this forum using now? How many are using actual film? I think its probably a small minority.

[Jaron Berman]

Nice webpage Richard, has lots of good information. I agree that CRI isn't the end-all be-all of rating lamps, which is why I called it an "opinion." While it's loosely based on measurements, it's pretty loose. CRI is a good rough guideline, especially when deciding on lamps which you may intend to use on faces. It's true, the hues in the face can be reproduced fairly closely using low cri lamps, however the difference is immediately noticeable when compared side-by-side. Soft white flo's, 6500K flo's, and even 5500 or 3200's with low CRI's all have their places... but when I'm keying skintones, I like to start with the largest color pallette available to me. Again, low CRI doesn't mean it won't work, BUT it generally means you're sacrificing tonality when compared to "more perfect" emitters. If you were to key your leading lady's face with an arri hmi next to this budget hmi, the difference would be immediately noticeable, even if matched for brightness and color temperature. When tungsten can be used, I prefer it. When daylight can be used, I prefer it. For those times when neither is an option, I like to use the tools which best approximate those sources, and sacrifice as little quality, flexibility and tonality as possible. On a budget, there are certainly solutions which work. This discussion board is full of comparisons, tips, and suggestions of that nature. But, it should be clearly noted that not all units are born equal.

As a footnote to the websites mentioned: In terms of mercury content, there is no realistic maximum to how much mercury can be in a lamp. The lifespan ratings, and EPA compliance mandate the usable amount. The construction of the electrodes determines the lifespan of the lamp, not the mercury content. Phosphors deplete over time, but it is generally due to the corrosion and "bastard" burning of the electrodes. It should be noted also that while fluorescent lifespan is impressive relative to tungsten, it is nothing compared to the theoretical maximums. Fluorescent lamps can and have been manufactured in a manner that allows them to light relatively consistently for 60+ years CONTINUOUSLY. As we become "greener," it's clear that fluorescent (and other) technology will slowly get better and longer lasting, and eventually replace tungsten.

[John DeLuca]

I think what it really comes down to is how well it mixes with daylight. The HMI jokers look exactly like electronic flash, while the majority of florescent tubes look a bit off.

[Richard Andrewski]

Originally Posted by Jaron Berman

It should be noted also that while fluorescent lifespan is impressive relative to tungsten, it is nothing compared to the theoretical maximums. Fluorescent lamps can and have been manufactured in a manner that allows them to light relatively consistently for 60+ years CONTINUOUSLY. As we become "greener," it's clear that fluorescent (and other) technology will slowly get better and longer lasting, and eventually replace tungsten.

Yes but wait until the "electrodeless" fluorescent lamp becomes more popular. LVD lamps are the potential fluorescent choice of the future. Lifespan increases dramatically because of the lack of electrodes. The only issue to watch for there is radio frequency interference from the induction method of energizing the lamp.

Cold cathode neon may have some promise too if the CRI can be raised to a level interesting for use in media production. The life of those would be higher also.

[Richard Andrewski]

Originally Posted by John DeLuca

I think what it really comes down to is how well it mixes with daylight. The HMI jokers look exactly like electronic flash, while the majority of florescent tubes look a bit off.

Well not all of them. Some of the simple choices can really be the best sometimes. I like these 5500K 30 watt bulbs I got at Home Depot. I'm using them all the time now and have really found it to be like you're not even aware there is any artificial light in the room--both for general use and for media production.

[Greg Boston]

Originally Posted by Jaron Berman

A video camera can smooth it out a bit, but cannot create frequencies that don't exist.

That's true Jason, but today's video camera is much better at this because of the ability to store custom color matrixes that account for those missing segments of the spectrum. I have a Fluorescent matrix preset on my camera and used it at a meeting once under standard office flo's. I was very impressed with the color balance it achieved.

In audio terminology, it would be like having a 200 band eq vs. a 10 band eq of years past. More specialized, and less generalized.

-gb-

[John DeLuca]

I just did a photo shoot with both HMI and Tungsten light sources available using the Nikon D2x. The other company shooting videography were using two 650w Qt Arris (dragged the shutter a bit while the subject held still), and we had the HMI lights (with UV gel) for photography. We also had Paul Buff strobes, but I wanted to see if I could tell any difference in quality using continuous light. Top link HMI, bottom link Qt Arris. Both have custom white balances with the Expo disk. The posted files are as good as we could get them in camera with no post correction, and we had no control over the videographers light positioning (so it may be a bad example, but I wanted to post it anyway, because I have noticed a difference in controlled situations too). I realize the HMI has a bit better in camera exposer, but it still looks better even when both files are corrected.

http://i50.photobucket.com/albums/f3...prema01542.jpg

http://i50.photobucket.com/albums/f3...prema01522.jpg