[kaseki]

Note the non-Plankian shape of the visible part of the solar spectrum at sea level. So the existential question is, should the image viewed on the monitor look as we would see it in the sunlight, or in the light of a reasonable quality flash, or do these sources fall sufficiently close together near the perfect Plankian 6500K blackbody point on the CIE Chromaticity diagram that few will notice? And if the latter, for either illumination of objects indoors or for illumination of the color checker, how good does the CRI have to be for an LED lamp to be indistinguishable from the flash result or the sun illumination result?

Attached Images

 

[swanlefitte]

Note the non-Plankian shape of the visible part of the solar spectrum at sea level. So the existential question is, should the image viewed on the monitor look as we would see it in the sunlight, or in the light of a reasonable quality flash, or do these sources fall sufficiently close together near the perfect Plankian 6500K blackbody point on the CIE Chromaticity diagram that few will notice? And if the latter, for either illumination of objects indoors or for illumination of the color checker, how good does the CRI have to be for an LED lamp to be indistinguishable from the flash result or the sun illumination result?

Haha! That's why I suggested taking multiple readings from noon on. Fit the difference to your existential light crisis. Will vary per person.

[Class A]

I can't compare a physical color chart illuminated by a flash with an on-monitor rendition of a chart.

Yeah, sure, but I think this particular thread is focused on making images look right on the monitor, not on real vs monitor comparison techniques.

Monitor renditions of flash photographed charts vs. sunlight photographed charts might well look very similar,

That's what I meant.

So the existential question is, should the image viewed on the monitor look as we would see it in the sunlight, or in the light of a reasonable quality flash, or do these sources fall sufficiently close together near the perfect Plankian 6500K blackbody point on the CIE Chromaticity diagram that few will notice?

I'd say they'd fall reasonably close together.

A dual-illuminant profile (cf. the OP's post) won't do all lighting situations justice anyhow, so I think the sun on a cloudless day is as good as it needs to be with respect to a perfect Plankian 6500K blackbody, even given all the atmospheric absorption. I think the OP was interested in using the camera for photography, not as a scientific instrument.

Interesting information, though, kaseki, very interesting indeed.

[kaseki]

,,,

Interesting information, though, kaseki, very interesting indeed.

I may have mentioned this in some other thread, but the page I scanned above (because I forgot where my electronic copy was stored) came from the RCA Electro-Optics Handbook, (c) 1974. Upon the death and dismemberment of RCA, Burle* republished this paperback as the Burle Electro-Optics Handbook. Eventually, Burle put the entire book on-line as a PDF. For basic photometry and radiometry, this compendium of the subject matter (as it was at publication) is extremely useful. The PDF should still exist on-line somewhere.** Paperback versions are available from Amazon, maybe even Burle.

P.S. A CDROM copy is available on Ebay.
* Now part of PHOTONIS
** http://www.clearlyvisiblepresentations.com/Burle_Electro-Optics_Handbook.pdf

---

Last edited by kaseki; 02-14-2019 at 08:21 AM.

[photoptimist]

Both LEDs and fluorescent lights suck for photographic lighting due to the lumps and spikes in the spectrum. Worse, both technologies tend to show oscillations in color and brightness in synch with the mains frequency.

If you want a light with a true "black body" color temperature, you'll need to use a light source that glows from heat such as the sun or a halogen/incandescent lamp.

To get a second illuminant, try putting a 80A tungsten filter on the lens.

[kaseki]

Both LEDs and fluorescent lights suck for photographic lighting due to the lumps and spikes in the spectrum. Worse, both technologies tend to show oscillations in color and brightness in synch with the mains frequency.

If you want a light with a true "black body" color temperature, you'll need to use a light source that glows from heat such as the sun or a halogen/incandescent lamp.

To get a second illuminant, try putting a 80A tungsten filter on the lens.

In this case, the tungsten source needs sufficient mass to not present any 50/60 Hz ripple above the error one is going to get anyway from all the steps between illumination and presentation.

[BigMackCam]

Given the additional issues mentioned...

Does anyone have any idea how Adobe and other major software vendors produce their dual-illuminant profiles? I can't imagine they wait for just the right lighting conditions outdoors. Surely they must use indoor lighting of some sort to photograph the relevant colour charts for profile creation? If that's the case, I really wonder what lighting they use...

[kaseki]

Given the additional issues mentioned...

Does anyone have any idea how Adobe and other major software vendors produce their dual-illuminant profiles? I can't imagine they wait for just the right lighting conditions outdoors. Surely they must use indoor lighting of some sort to photograph the relevant colour charts for profile creation? If that's the case, I really wonder what lighting they use...

Me too.

However, I can imagine an outdoor scene on a tropical island beach -- white sand, no nearby green palms, with umbrella, chairs, table, mint julep at hand, spectrophotometer, computer controlled alarm, cameras and photographer waiting for alarm to ring to take photos.

[photoptimist]

Given the additional issues mentioned...

Does anyone have any idea how Adobe and other major software vendors produce their dual-illuminant profiles? I can't imagine they wait for just the right lighting conditions outdoors. Surely they must use indoor lighting of some sort to photograph the relevant colour charts for profile creation? If that's the case, I really wonder what lighting they use...

Just a guess: flash and a tungsten source would be the logic pairing. Dual illuminant profiles seem to use a daylight and low-temperature lighting to build profiles for two ends of a temperature spectrum. In processing the picture, the software tries to estimate the color temperature of the light source and then interpolate the profiles.

If they wanted to do it "right," then they would use a monochromator to create a nearly monochromatic source of light, build a profile for each illuminant wavelength, and then use some really nasty inverse-problem math to estimate the maximum-likelihood illuminant spectrum from the RAW data. It's not for the faint of heart.

[Dartmoor Dave]

This thread is making me wonder about the possibility of shooting the same test card under multiple light sources that all have a nominal 6500K but not flat spectrums, then averaging the results. Since Adobe Profile Editor shows you the values it's using for each colour patch and allows you to edit them, creating an averaged profile wouldn't be too hard.

I might give it a try over the weekend.

[kaseki]

...

If they wanted to do it "right," then they would use a monochromator to create a nearly monochromatic source of light, build a profile for each illuminant wavelength, and then use some really nasty inverse-problem math to estimate the maximum-likelihood illuminant spectrum from the RAW data. It's not for the faint of heart.

And it won't be for those needing high frame rates unless their camera is tied by umbilical to a multi nVidia GPU PC.

[kaseki]

Given the additional issues mentioned...

Does anyone have any idea how Adobe and other major software vendors produce their dual-illuminant profiles? I can't imagine they wait for just the right lighting conditions outdoors. Surely they must use indoor lighting of some sort to photograph the relevant colour charts for profile creation? If that's the case, I really wonder what lighting they use...

Here is an example of indoor lighting: Harmony Room Product Support

[Dartmoor Dave]

It's been a cold wet morning, so I've played around with some profiles. This first set is three profiles created using Adobe Profile Editor from the same test card shot under different lighting. The top one is mid-day summer daylight from last year, the middle is flash, and the bottom is a daylight balanced LED light (I accidentally put LCD on the comparison, but it's actually LED).



As you can see, there are some differences presumably caused by the different spectra of the different light sources (and perhaps slight exposure differences, although I tried to match them as close as possible). I usually shoot in daylight using the same daylight profile shown here. I never personally use flash, but if you are a regular flash shooter I think it would be best to create a separate profile purely for your flash shots, as the differences between the daylight and flash examples above seem to justify it.

Now here's the test card that was originally shot in daylight, with the daylight profile applied at the top and an averaged profile at the bottom. The averaged profile was created by averaging the values for each colour patch in the daylight, flash and LED profiles. (When you're creating custom .dcp profiles using APE, you only need to worry about the 24 patches on a standard Macbeth/ColorCheck card. The software then twists the colour space around those values automatically, and also extrapolates the tungsten profile from your 6500K original.)



It seems to me that just these three profiles combined averages out to something very close to the straight daylight profile on its own. I'd suggest that the more different artificial light sources you average out, the closer you'll get to something that's indistinguishable from daylight. In which case, if you want one general purpose profile, you might as well just shoot the test card in daylight.

And how does any of this apply in the real world? Here's a random shot converted from raw to jpeg using my normal daylight profile on top, and the averaged profile at the bottom. Even just averaging out three different light sources seems to me to result in a difference that's negligible in real-world photography. I'd expect it to be impossible to detect any difference at all if you averaged out enough different illuminants.



Ah well, that kept me happy for a couple of hours on a wet Saturday morning. The methodology was definitely more rough-and-ready than lab-grade, but I hope it gets the basic idea across.
(Note: all test cards were shot using the Pentax K-S1 and SMC Takumar 24mm/3.5, so the equipment used is the same in all examples.)

[kaseki]

Thank you. I see on the lightest gray square in every card a pinkish cast that I will assume for now is due to my desk monitor being an uncalibrated and several years old HP ZR2440w. I have a just-arrived and factory-calibrated BenQ to set up and I will look again with that. I agree that it is very difficult to distinguish these profile compensations when the image of some place one has never been is viewed and the photographer's goal is realism rather than impressionism.

On the card labeled "Daylight LCD": Is that a camera-taken image of an LCD monitor screen, or did you mean "Daylight LED"?

[BigMackCam]

It's been a cold wet morning, so I've played around with some profiles. This first set is three profiles created using Adobe Profile Editor from the same test card shot under different lighting. The top one is mid-day summer daylight from last year, the middle is flash, and the bottom is a daylight balanced LED light (I accidentally put LCD on the comparison, but it's actually LED).
...
Ah well, that kept me happy for a couple of hours on a wet Saturday morning. The methodology was definitely more rough-and-ready than lab-grade, but I hope it gets the basic idea across.
(Note: all test cards were shot using the Pentax K-S1 and SMC Takumar 24mm/3.5, so the equipment used is the same in all examples.)

Dave - thanks for your efforts here, and for posting the results. Whilst there are certainly differences, they are smaller than I would have expected.

Can I ask, what was the daylight balanced LED light that you used? And how did you go about creating the averaged profile?

I see on the lightest gray square in every card a pinkish cast that I will assume for now is due to my desk monitor being an uncalibrated and several years old HP ZR2440w.

I thought I could also see a tiny slight pink / magenta cast in the greys of the flash image when viewed on my older calibrated laptop, but when I look again I don't see it. I haven't yet viewed the images on my newer calibrated laptop and desktop monitor. That said, the sample outdoor image processed with different profiles looks great, and the clouds look completely natural to my eye...