Kelvin Accuracy with Solux Lights

[bkkstudios]

<p>I have a Solux 3500K flood light on my desk. I have a Kenko KCM3100 color meter. I turn off all lights in my room. I put the color meter about 10" below the flood light. It measures 3150K. Moving it around in terms of distance from the Solux, I vary from 3100 to 3250 or so.</p>

<p>So my question is, does this measurement make sense, in terms the way I'm doing it? And secondly, why would the Solux be so far off it's theoretical 3500K?</p>

 

[bruce_watson1]

<p>What voltage is going to the lamp? Solux bulbs will only give the rated K temperature with 12VDC. Anything less and you'll get a lower K temp. Nearly all 12V fixtures are undervoltage to some degree, except, presumably, the fixtures you can buy from Solux. But those fixtures, in my experience (from five years ago or so), have so many other problems (paint blistered on mine, and they couldn't hold direction) that they aren't worth messing with.</p>

[digitaldog]

<p>The meter is useless. Its providing correlated color temperature which is a range of colors. It doesn’t surprise me the values don’t exactly match. The measurement doesn’t make sense so you are spot on here. The point you made that explains it is the term theoretical 3500K.</p>

<p>http://www.ppmag.com/reviews/200512_rodneycm.pdf<br /> What you really need is a Spectrophotometer that can read the SPD of the light.<br>

Lets take CCT 6500K and D65 and look at the differences to explain why your experiment doesn’t sync up (the numbers are different):<br>

a) D65 is a spectral power distribution (a certain amount of energy at each wavelength across the visible spectrum).<br /> b) D65 is a tristimulus value; the D65 spectrum, when viewed by the CIE standard observer, produces an XYZ triplet (or xyY if you prefer).<br /> c) 6500K blackbody radiator is a spectral power distribution.<br /> d) 6500K is blackbody tristimulus value; the 6500K blackbody spectrum, when viewed by the CIE standard observer, produces an XYZ triplet—similar to, but slightly different from, the one found in (b).<br /> e) Correlated color temperature takes a color's chromaticity coordinate (x,y) and finds the particular blackbody temperature whose chromaticity coordinate (d) is closest to it. Note that there are many different colors that have the same correlated color temperature. So a spectrum is very precise and unique. Its xyY is less precise and unique. Its CCT is even less precise and unique.<br /> D65 is a unique SPD (there exists only one). A color whose CCT is 6500K is not unique (there are infinitely many different xyY and SPDs that share it).</p>

[patricklavoie]

<p>interesting answer.. but my nose is bleeding ; )</p>

[Tim_Lookingbill]

<p>What color hue does the 3500K Solux look like lighting a spectrally flat white surface, Ziggy? Mind you I wouldn't know where to find such a white surface unless the paper you buy specifies it and then you'll have to trust them against another competitor who claims the same but shows a different hue all together.</p>

<p>It's kind of like asking how weather scientists arrived at zero when first inventing and then calibrating a precision thermometer. Was it humid or dry during the first calibration? Because that does influence temperature readings or at least the way it feels and it's always wet or dry somewhere in the world.</p>

<p>Isn't that a more practical and useful way to define the look of your light, Ziggy? Who cares if it matches up to specs measured by a spectrophotometer. How does it make your inkjet color AND B&W prints look? I'm sure on the noticeably warm side compared to a 6500K calibrated display.</p>

<p> </p>

[keith_b1]

<p>The Solux bulbs are incandescent, so they can be run on AC as well as DC. This opens up the possibility of varying the voltage up or down (via a Variac) to obtain a modest variation on CCT.<br>

Personally I use high-CRI 5500ºK fluorescents at my printing station because they're cheap.</p>

[digitaldog]

<blockquote>

<p>Personally I use high-CRI 5500ºK fluorescents at my printing station because they're cheap.</p>

</blockquote>

<p>But have awful spectral spikes and can cause havoc with papers that have high OBAs. And CRI isn't a very good metric of light quality. That said, I too use such bulbs in my GTI light booth but also use Solux for the above reasons (the quality of light can't be beat).</p>

[Tim_Lookingbill]

<p>Hey Keith, if you're into 5500K fluorescent bulbs, try the Alzo Digital CFL's. </p>

<p>http://www.amazon.com/gp/customer-media/product-gallery/B00198SWJI/ref=cm_ciu_pdp_images_all</p>

<p>I posted the images to the right of the pics of the bulbs. That's exactly how they look. They're probably the best and neutral CFL I've ever come across for color rendering.</p>

[bkkstudios]

<p>The flood light is AC. The smaller spot lights are DC. From a previous post, this should not matter, i.e., voltage drift should not affect color temperature. And certainly Solux sells both AC and DC lights, so their results should match their marketing.</p>

<p>While I understand, somewhat, your post Andy, from a layman's perspective, I don't see the error in my ways.</p>

<p>The Kenko, in my observation, gives the correct Kelvin temperature wherever I am, i.e., if I set my camera to the result it provides, my photos seems properly white-balanced. So I take from that, it is reasonably accurately measuring temperature for the purpose of white balance.</p>

<p>So, while there may be infinitely many K values for my Solux AC light, I'm still missing why it is off in this case, and not in others. To the point, it seems my Solux lights are off, not the color meter. Is my logic missing something?</p>

[Tim_Lookingbill]

<blockquote>

<p>Is my logic missing something?</p>

</blockquote>

<p>The logic that you didn't seem to read or understand (if you did read) the facts about the meaning of Kelvin in relation to photography that others here have posted on in this thread.</p>

<p>Appearance of Kelvin according to the numbers is never consistent between device manufacturers with regard to devices (and software) like lights, camera WB, Raw converter interpretation of the appearance of the K numbers, etc.</p>

<p>Ever notice thermometers in the store are all slightly off among a certain few compared to others? Which one's correct? It depends on who's defining it and everyone thinks theirs is the correct one. But all are correct in determining if it's hot or cold and that's all that matters.</p>

<p>I mean when it's hot (90 degrees), you don't feel a 3-5 degree discrepancy the same way as you would when it's say a comfortable 75 and jumps to 80 degrees. How do you quantify that from a consistent scientific standpoint?</p>

<p>So from a practical human functionality stand point which is what all this Kelvin color cast definitions should be centered around, consistency in the numbers only works if you stick to one manufacturer's definition of it. Once you start comparing outside of that device, the Kenko, all bets are off for having anything natural or man made match to its definition.</p>

<p>Hope you can understand that logic.</p>

[digitaldog]

<blockquote>

<p>Ever notice thermometers in the store are all slightly off among a certain few compared to others? Which one's correct? </p>

</blockquote>

<p>Its even worse in this case. A picture is worth a 1000 words:<br>

<img src="http://digitaldog.net/files/LinesOfCCT.jpg" alt="" width="500" height="559" /></p>

<p>You see values for Kelvin above. Lets look at 5000K. The black body curve represents the actual theoretical value for 5000K yet moving the other axis, you see a line of correlated color temperature that moves from magenta to green. Any point on that line marked 5000K on that axis is a legitimate color we can call or a device can measure as 5000K. Again, using correlated color temperature such as CCT5000K defines a <strong>range</strong> of colors. If I gave you an exact x/y value, or a standard Illuminate, you could see that exact color above. </p>

[bkkstudios]

<p>Thank-you Tom and Andy, and yes Tom, I do read.</p>

<p>While I understand the arguments, I remain puzzled by the big picture.</p>

<p>I would presume, and maybe this is faulty, that any manufacturer whose existence is based on color accurate products (i.e., Solux and Kenko), adhere to some standard definition of color for a set Kelvin (i.e., some fixed xyz coordinate as u note Andy for a particular K). If not, how do they produce products that are credible or work with other products?</p>

<p>Now, separate from that, my practical observation is that the Kenko works for WB settings on a Nikon and Canon. And when I measure my Nikon speedlight and Bowens lights with my Kenko, they are quite near the 5500K they are promoted to be. And ditto when I shoot and let C1 or Capture NX show the K value, it matches my Kenko. So it would seem, several companies have agreed on what 5500K means, or which xyz point it refers to within a few hundred degrees.</p>

<p>So it seems to me this is not so stupid an experiment, but maybe you are right that drift of hundreds is completely understandable.</p>

 

[digitaldog]

<blockquote>

<p>I would presume, and maybe this is faulty, that any manufacturer whose existence is based on color accurate products (i.e., Solux and Kenko), adhere to some standard definition of color for a set Kelvin.</p>

</blockquote>

<p>They can’t because Kelvin values on that line in the chart above are based on a <strong>theoretical object</strong> (the black body radiator) and all the stuff in bulbs don’t behave as an object that actually doesn’t exist (OK, we could maybe say the Sun gets close <g>). That is why whenever you see a Kelvin value, it should point out <em>CCT</em> there unless its really damn close (as measured with the correct instrument) to that black body curve. When you see CCT Kelvin, that is the clue that YMMV if an exact description of a color is necessary. </p>

 

<blockquote>

<p>Now, separate from that, my practical observation is that the Kenko works for WB settings on a Nikon and Canon. And when I measure my Nikon speedlight and Bowens lights with my Kenko, they are quite near the 5500K they are promoted to be. </p>

</blockquote>

<p>Again, if they match, if they don’t match, if they are kind of close, you cannot use these values as anything concrete or absolute because you are defining a <strong>range of colors</strong>. <br>

If you use a $1 ruler to measure an inch, then I take some lab grade super duper laser thingy that cost $100K and it tells me its not an inch, then you take a $3 ruler and it too tells you its an inch, are both inexpensive rulers values useful? As I said above, with CCT, the issue is even worse. Its a far larger range. </p>

[photomark]

Ziggy, companies like Solux do publish spectral power distribution data for their lights so you can know with

great detail how the light is expected to perform. Andrew is right, if you really want to test your light, you need a

spectrophotometer.

 

Color temperature was meant as a guide for the relative warmness/coolness of a light, not a specification.

When you are talking about tungsten lights, it actually worked pretty well. My little desk lamp has a CCT of

about 2700K and it gives a chromaticity of about x: .456, y:.413, which is really very close to the planckian

radiator for 2700K. In fact if you plot a spectral power distribution of this light, it falls almost exactly on the

black body curve until you get to the far red end of the spectrum. This is expected because incandescent

sources are almost blackbody radiators—just less efficient.

 

 

Having said that, I disagree with Andrew in that I think you should expect a good color meter to give more

accurate CCT values. While a particular CCT may represent a range of colors, one particular light source

should have one, and only one, CCT.

[digitaldog]

<blockquote>

<p>Having said that, I disagree with Andrew in that I think you should expect a good color meter to give more accurate CCT values. While a particular CCT may represent a range of colors, one particular light source should have one, and only one, CCT.</p>

</blockquote>

<p>Depending on the filters and the rest of the system, we would expect more accurate values but it still ranges and as you say, you really want to use something like a Spectrophotometer and some software that will convert the values.</p>

[Tim_Lookingbill]

<p>Ziggy, we're not implying your experiment is stupid. We're trying to help you understand you're trying to pin down a moving target, well really three moving targets...</p>

<p>1. Human's constantly adapting to any given Kelvin appearance to a degree...</p>

<p>2. Color science definitions that can't be tied to any real object that can reproduce it consistently...</p>

<p>3. Device manufacturers using their own proprietary hardware/materials to define their own appearance of Kelvin.</p>

<p>The Solux lamp wasn't designed to be used for photographic lighting. It was designed for viewing prints by humans, not cameras. The lamp's special filtering attributes applied to a tungsten based hot spot light enhances definition with both luminance and color detail (cool/warm/complementary) on reflected objects like prints that closely mimic the reflective qualities of the sun. This is based on my own observation using the 4700K Solux 50 watt.</p>

<p>Usually camera's pick up on the halogen/tungsten underpinnings of the Solux light which will beef up the complimentary (red/green-blue/yellow) reflected layering effect (Dutch master painters used to mix paint on canvas) producing images with more color detail, depth and definition (no plastic look to skin tone) when the red bias (a specific hue of red) from the filtered tungsten is neutralized during white balancing.</p>

<p>I'm guessing this is why 60 Minute's Ed Bradley (an African American) specifically preferred to be lit by the Solux by his cameraman. It seems to be very difficult to get nice looking unbiased browns in skin tone at least from my experience even using daylight. This light characteristic/quality has nothing to do with Kelvin numbers.</p>

<p>So there's more important and useful things in defining a light source that makes concern over matching Kelvin numbers seem secondary.</p>

[bkkstudios]

<p>Sorry for continuing, and thank-you for your patience, but I really want to get this right.</p>

<p>If I understand it, if the there are no spikes at different wavelengths, the K value is reasonably good. According to the Solux SPD, they have a very flat curve, so their K value should be close to a black body (as would be expected of a company whose life and death depends on color accuracy). See http://www.solux.net/edu9.htm. If I understand right, CRI is one measure of this flat curve, and Solux quote a CRI of 98. So if my understanding is right, the Kenko measure of a Solux light with those characteristics IS valid (assuming the Kendo measurements are valid).</p>

<p>In this particular case, since the Kenko is not emitting light, but rather measuring light, I assume it's CRI (if that even makes sense), in irrelevant.</p>

<p>Now, aside from this theoretical discussion, as you may have guessed, I'm a bit obsessed by color accuracy. Don't ask why. We all have obsessions. So forgetting what I do now with Solux, Kenko, and about 10 other gadgets related to this, how can I get the best color accuracy possible, and what instruments do I need for that? I see spectrophotometers are like 5-10K USD but look physcially impractical and not suited to photography.</p>

<p>The use case is I have is some fashion fabric with SPD xyz. I want to shoot it and print it, and have the results show up as XYZ, and give the customer some proof of that accuracy (other than our eyes).</p>

<p>Finally, a recommended book, oriented towards photographers would be good.</p>

<p>Once again, thank-you.</p>

<p></p>

[photomark]

I think you're making life too difficult for yourself. It doesn't really matter if the solux is off a bit or if your color meter is off. You

want a monitor with a white point and a luminance that gives you a good visual match to to the light you will be using to evaluate

prints, fabrics, etc. Your eyes are constantly adapting to the color of the illuminant and your brain is doing god knows what, so

chasing a particular color temperature is not always helpful. Andrew R. has some white papers floating around that explain it quite

well; maybe he'll be kind enough to post a link.

 

Also, accuracy is chimerical with color. You can specify a color absolutely, with a SPD or an XYZ coordinate, but that doesn't

guarantee a visual match unless you can guarantee the same viewing conditions. You see this every day—the light in your house

looks white until you see from outside at night through the window. But the light's color still has the same XYZ coordinates. There

really is no proof of accuracy other than our eyes—it is subjective. Until we have a really good color appearance model, we're

stuck with it. Here's a great example of accurate colors not matching in a single image:

http://www.lottolab.org/illusiondemos/Demo%2014.html

 

If you need dead-on colors with a fabric, you may be able to get LAB values of the fabric's color, but again, you may be chasing

your tail when it comes to real-world applications. Fabric is not always opaque, it looks different when lit from different angles and

against different backgrounds, it may be printed in CMYK etc. A few degrees kelvin in your work light is the least of your worries.

The best bet is to do what most other professional photographers are doing: calibrate your monitor, have smooth consistent

working light, and photograph a color checker on set so you have a known reference.

 

If you love dealing with numbers and color and want to start reading, all roads eventually lead to Wyszecki & Stiles "Color

Science" but it's a monster reference. R.G.W Hunt's "The Reproduction of Colour" is a little pricey, but quite good and much more

approachable than Wyszecki & Stiles. Neither is really written for photographers, though.

 

Also, you don't have to spend $5k on a spectrophotometer. The i1Pro will run you about $1k and it is very practical.