Pages: [1]   Go Down

Author Topic: Daylight simulators for camera profiling  (Read 2740 times)

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Daylight simulators for camera profiling
« on: February 26, 2015, 03:54:50 am »

I'm tired of having to wait for the perfect daylight outside and would like to have a small experimental setup to do some camera profiling and related experiments.

I want a full spectrum without spikes, and it seems like Solux 4700K lamps is the only thing out there, unless you get those new expensive multi-channel leds like gamma scientific RS-5B http://www.gamma-sci.com/products/rs5b/.

However, the Solux seems to vary a bit so you need a precise dimmer with overdrive capacity, preferably up to 16 volt. If you don't worry about burning out the lamp soon it seems that you can get 5000K with very high similarity to a real D50 spectrum (without UV) with this lamp, but this is when running it in overdrive.

I would have preferred D65, but the only thing I've found there in terms of Halogen is Ushio Whitestar and I don't think their filtered lamp is without spikes (ie spectrum match with D65 is poor), but I'm not sure. Making a camera profile from D50 is okay though, that's actually what I've been using most of the time anyway.

I think I can do with one lamp only, there will be issues with uniformity for sure but a flatfield shot should solve that issue.

I'm currently looking for a fixture and a power source to be able to run that Solux 4700K lamp in say 8-16 volts with precise voltage control. I have a photospectrometer to measure the actual output.

I find it a bit surprising though that there are no LED solutions out there with good spectrums, except that gamma scientific. I've found this research project too with even more precision: http://www.nist.gov/pml/div685/grp03/vision_lighting.cfm it feels like someone should have made a LED-based D65 lamp with 10 or so LEDs by now to make a full spectrum without spikes... I haven't found any though, Solux halogens seems still to be the king in (affordable and readily available) daylight simulation.

I am satisfied with fluorescent "daylight" lamps for print viewing, but when it comes to camera profiling I don't want a spiky spectrum.

Any tips, ideas, experience?
« Last Edit: February 26, 2015, 04:03:55 am by torger »
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Daylight simulators for camera profiling
« Reply #1 on: February 26, 2015, 04:39:25 am »

I find it a bit surprising though that there are no LED solutions out there with good spectrums, except that gamma scientific.

Hi,

Your best bet would be something based on an incandescent lightsource (with filtration). LEDs do not have a broad spectrum, they rely on fluorescence and other techniques to achieve something more broad band and of different color (they will therefore also drift over time). It will usually result in something not close to any Blackbody radiation smoothness which would IMHO be the best for profiling.

When in the field, for a quick visual check of the lighting conditions, I use this Pocket Spectroscope. I gives a quick impression of what to expect from a lightsource (it even shows the Fraunhoffer absorption lines from our atmosphere in daylight), although human vision is not that good at an absolute quantification. It also clearly shows the peaks in Daylight quality specialized fluorescent tubes for the graphic industry, and it shows the deficiency of LEDs in Red, and irregular emission in Blue. It's great to make on the spot judgements of spectral emission continuity (it does assume the observer has good color vision ;) ).

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Daylight simulators for camera profiling
« Reply #2 on: February 26, 2015, 04:46:53 am »

I found this, a recent 22 channel led commercial product, only €5850 :)

http://image-engineering-shop.de/shop/article_iQ-LED/IQ-LED.html?sessid=padw1UN8nZY6OK2HNkoLqTIFGBltZWo9F6y55d61bjlFOSCBwYuyQ78hWyMyaf7j&shop_param=cid%3D566%26aid%3DiQ-LED%26

The principle of these are that you have so many LEDs that that combined you get a nice smooth spectrum, like this:



Programmable too, that's cool. But with prices €6000 - €10000 it's not really a product for the home lab :-\
« Last Edit: February 26, 2015, 04:50:36 am by torger »
Logged

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Daylight simulators for camera profiling
« Reply #3 on: February 26, 2015, 07:15:57 am »

I'm currently looking at this: Solux 4700K 35w black back lamp + a cheap dc lab power supply 0-15 volts with enough wattage (~€40) + a rail fixture which I attach to a light stand with tape, and make some soldering hack to feed the rail from the dc lab power supply instead of using the regular one. Then tune the voltage until I reach 5000K (measuring the spectrum/temp using argyll+colormunki), which should be around 13.4 volts when looking at the spec sheets, and hope that the lamp doesn't burn out too fast.
« Last Edit: February 26, 2015, 09:19:22 am by torger »
Logged

AlterEgo

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1995
Re: Daylight simulators for camera profiling
« Reply #4 on: February 26, 2015, 09:43:37 am »

Any tips, ideas, experience?
Iliah Borg (of, for example, profiles for RPP) once suggested a simple halogen worklight /a pair/ (power stabilized) + gel to adjust the temperature up, and their light reflected off brushed aluminium surface towards the target (= no direct illumination)
Logged

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Daylight simulators for camera profiling
« Reply #5 on: February 26, 2015, 10:01:43 am »

I think I'll attempt with direct light without diffuser first. I know there will be uniformity issues, but I think a flatfield can solve that. If it doesn't work out I'll add diffusers.

I guess the exact light temperature is not too important really, if it's 4700K or 5000K shouldn't matter but I'd like to have adjustment anyway for the sake of experimentation.

I doubt that filtering up the temperature works, if the intention is to have as good peak-free daylight spectrum match as possible. The ushio whitestar have builtin filters and have a 6500K version, but I have failed to find a D65 spectrum comparison, while Solox 4700K in overdrive does match D50 very well seen in independent tests. The Solox is filtered too of course (it's not a 4700K blackbody...), but it seems like their thing is that they've got a great daylight match with carefully devised filtering.

It seems to me like the whitestar filter is more simplistic, just to get the XY coordinate right without caring about D65 spectrum matching. I guess the same would happen if I tried to match with generic gels.
« Last Edit: February 26, 2015, 10:05:55 am by torger »
Logged

Redcrown

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 507
Re: Daylight simulators for camera profiling
« Reply #6 on: February 26, 2015, 12:03:09 pm »

What do you think about Andrew Rodney's (Digitaldog) claim and demo that the color temp of daylight does not make any difference when generating camera profiles from a CC chart?

https://www.youtube.com/watch?v=_fikTm8XIt4

It would seem to make your effort to match the spectrum of daylight at a specific temperature not necessary.
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: Daylight simulators for camera profiling
« Reply #7 on: February 26, 2015, 12:18:55 pm »

What do you think about Andrew Rodney's (Digitaldog) claim and demo that the color temp of daylight does not make any difference when generating camera profiles from a CC chart?

https://www.youtube.com/watch?v=_fikTm8XIt4

It would seem to make your effort to match the spectrum of daylight at a specific temperature not necessary.

Though I agree with your point, torger states in his original post he's wanting to do some camera profiling experiments so we really have no idea of the purpose behind what he's wanting to accomplish.
Logged

AlterEgo

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1995
Re: Daylight simulators for camera profiling
« Reply #8 on: February 26, 2015, 12:25:21 pm »

What do you think about Andrew Rodney's (Digitaldog) claim and demo that the color temp of daylight does not make any difference when generating camera profiles from a CC chart?

when daylight cease to be a daylight - consider tungsten... it is also smooth, not spiky and full spectrum... so at some point you might still want to use filter to reallocate intensity within spectrum, right ?... that was, i guess, the point behind gelling halogen worklights from any walmart/lowes/homedepot out there
Logged

AlterEgo

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1995
Re: Daylight simulators for camera profiling
« Reply #9 on: February 26, 2015, 12:29:18 pm »

Though I agree with your point, torger states in his original post he's wanting to do some camera profiling experiments so we really have no idea of the purpose behind what he's wanting to accomplish.

he wants to create a "camera" profile (icc/icm or dcp) and asking about a suitable source of artificial illumination (to match a daylight spectrum properties - hence one might assume a generic/outdoors profile) to do it indoors instead of outdoors
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Daylight simulators for camera profiling
« Reply #10 on: February 26, 2015, 02:17:41 pm »

What do you think about Andrew Rodney's (Digitaldog) claim and demo that the color temp of daylight does not make any difference when generating camera profiles from a CC chart?
Not originally my claim but that of someone who knows far, far more about this process than I do: Eric Chan of Adobe. Of course, trust but verify. Sure enough, it's the illuminant's SPD not the specific color temp of that illuminant (within a broad range according to Eric) that requires the use of differing custom DNG camera profiles. IF we had to build DNG profiles based on color temp, it wouldn’t work anywhere as well as it does now where the two are not joined at the hip so to speak.

Not sure this helps the OP.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Daylight simulators for camera profiling
« Reply #11 on: February 27, 2015, 03:43:20 am »

Yes it's probably true that daylight temperature is not that important for the profile result, I've used free outside daylight so far for my profiles and it seems to work well. It's something I want to test in my experiments though, I do this both for making profiles to use, but also learn something about how camera profiling works and maybe write a little article about my findings.

However I think even if the daylight temp is not that important, it could be the case that it's important that the spectrum is not spiky, I'll probably test it too becuase I have a flourescent daylight lamp already that I can compare with.

It would be supercool to have one of those multichannel programmable spectrum leds to play with (could then experiment also with twilight spectrums etc), but that €6000 price tag is a bit hefty for a hobbyist project :)
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Daylight simulators for camera profiling
« Reply #12 on: February 27, 2015, 04:57:59 am »

However I think even if the daylight temp is not that important, it could be the case that it's important that the spectrum is not spiky, I'll probably test it too because I have a fluorescent daylight lamp already that I can compare with.

Hi Anders,

I also think that a continuous spectrum (Spectral Power Distribution, SPD) is very important for a predictable profile generation and behavior when we start manipulating it by deviating from the spectrum's Correlated Color Temperature (CCT). Even Daylight isn't completely continuous, due to absorption of sunlight in the atmosphere (Fraunhoffer lines), but those absorptions are so extremely narrow band that we can still work under the assumption of daylight having a continous spectrum.

Of course, in practice, the SPD will deviate from a Blackbody emission (e.g. due to locally reflected color, which is why the base of clouds have a green cast over meadows and foliage), and our sensors are also anything but continuous in their spectral sensitivity. So that's where knowing how it deviates from the continuous model will give us at least a fighting chance to calibrate for the differences. But even then, it's probably exposure level dependent due to our color models. So there will remain to be a lot of art after the first science has done it's job.

Quote
It would be supercool to have one of those multichannel programmable spectrum leds to play with (could then experiment also with twilight spectrums etc), but that €6000 price tag is a bit hefty for a hobbyist project :)

All it takes is to win the lottery ;)

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: Daylight simulators for camera profiling
« Reply #13 on: February 27, 2015, 01:51:07 pm »

Yes it's probably true that daylight temperature is not that important for the profile result, I've used free outside daylight so far for my profiles and it seems to work well. It's something I want to test in my experiments though, I do this both for making profiles to use, but also learn something about how camera profiling works and maybe write a little article about my findings.

Since I now have three process versions of Adobe imaging apps CS3, CS5 & LR4.4 and their subsequent updating to their DNG converter and Adobe DNG PE AND what others say about Base Profiles as starting points such as Adobe Standard for DNG camera profiling, I've now noticed the differences applying the old profiles generated using the same daylight lit CCchart in CS3 vs LR4.4 but only to a wide range of legacy images.

The differences vary depending on the image scene some slight and some more pronounced especially the LR4.4/Adobe Standard Base Profile created from a CCchart lit by overcast clouds vs direct sun. But I didn't start noticing these differences at first because the LR4.4 with current DNG PE made the CCchart target image look less saturated (more accurate by the numbers) than the old CS3 profiles that I took that as an improvement. Now I'm finding the hue/sat differences on some images don't produce desirable results and keep reverting back to using either ACR4.4, Adobe Standard or the old CS3 custom single and dual illuminant profiles.

This discovery between these profile versions was an unfolding process so I'ld have to conduct a rigid test to see what was influencing what due to all the variables in order to track down whether the changes are due to the newer updated software, Base Profile in the DNG or illuminant CCchart.

I know for my camera Adobe Standard neutralizes the entire image to appear less greenish yellow which does make the CCchart gray patches look reddish blue but I don't know if the DNG PE CCchart Wizard is removing the Base Profile's influence on WB of the gray patches before it automatically corrects the cast and builds the new custom tables. I do notice this particular Adobe Standard/LR4.4 sourced DNG profile makes images look different than the old CS3 made profiles.
« Last Edit: February 27, 2015, 01:52:53 pm by Tim Lookingbill »
Logged

torger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3267
Re: Daylight simulators for camera profiling
« Reply #14 on: February 27, 2015, 04:05:18 pm »

I think camera profiling has been somewhat romanticised by the raw converter makers. They produce "looks" much like oversaturated printer profiles we see from paper makers, ie made to impress the casual user, they are noway accurate but very subjective indeed. I don’t buy into that you need to be a color expert at Phase One to get good color from the camera, especially in landscape photography.

What I want is a neutral starting point and produce my own look with available color tools in post-processing software, and custom profiling is necessary to put the photographer in the front seat.

Surprisingly little is written about camera profiling though and software for profiling is rare and often limited. So it's an interesting subject to experiment with.
« Last Edit: February 27, 2015, 04:29:16 pm by torger »
Logged

Tim Lookingbill

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2436
Re: Daylight simulators for camera profiling
« Reply #15 on: February 27, 2015, 06:02:28 pm »

Quote
They produce "looks" much like oversaturated printer profiles we see from paper makers, ie made to impress the casual user, they are noway accurate but very subjective indeed. I don’t buy into that you need to be a color expert at Phase One to get good color from the camera, especially in landscape photography.

What would define a color expert? Are there tests? I found these statements from your linked "Nist.gov" on CCT issues with solid state lighting quite telling about whether we all know enough about light's affect on color...  http://www.nist.gov/pml/div685/grp03/vision_lighting.cfm ...

Quote
As the spectra of LED sources are dissimilar to traditional incandescent and discharge lamps, some of the existing standards and measurement methods are insufficient or deficient when applied to LEDs. For example, a new metric for color rendering of light sources is being developed to address the problems of the color rendering index (CRI) for LED light sources. Such studies require well-designed vision experiments. Because solid-state lighting sources have much greater flexibility of spectral design than traditional lamps, manufacturers have more freedom in the selection of correlated color temperature (CCT) and color quality of solid-state lighting products for various lighting applications. However, the interrelated effects of chromaticity, color rendering, and other aspects of spectra on lighting are still not well understood. Thus, the introduction of solid-state lighting necessitates re-visiting many of these questions regarding the effects of spectra on lighting.

Not well understood?! Huh?!

From all that's been discussed online using spectral plots, graphs and 3D modeling analysis to prove one product is better at color reproduction over another since I got into digital imaging back in '98, I was convinced color scientists especially with the government had it worked out and understood just fine. What other possible "Standard" could they come up with over how we're defining the effects of spectra on lighting?
« Last Edit: February 27, 2015, 06:05:15 pm by Tim Lookingbill »
Logged

AlterEgo

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1995
Re: Daylight simulators for camera profiling
« Reply #16 on: February 27, 2015, 06:15:43 pm »

What would define a color expert?
you need to decide what will be the criteria... dE or "I like it"
Logged

digitaldog

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 20630
  • Andrew Rodney
    • http://www.digitaldog.net/
Re: Daylight simulators for camera profiling
« Reply #17 on: February 27, 2015, 06:28:31 pm »

From all that's been discussed online using spectral plots, graphs and 3D modeling analysis to prove one product is better at color reproduction over another since I got into digital imaging back in '98, I was convinced color scientists especially with the government had it worked out and understood just fine.
None of those metric's (spectral plots, graphs and 3D modeling analysis) prove one product is better at color reproduction over another per se. It is a bit like suggesting that the proof of the best car to buy is the one who's MPG rating by the Government is the highest. A gamut plot can tell you a lot about a color space, not all that much about the subjectively of the final output. Got a huge gamut but piss-poor gray balance? You'll not see that attribute (gray balance) unless you have a gamut plot of the neutral axis and know to look there. If you want to poke holes in data, data used by marketing departments to imply they've proved their devices are superior, how about those who tell us their displays produce "Billions of colors"?  Or that their displays have 14-bits instead of 12-bits internally since more has to be better.
Logged
http://www.digitaldog.net/
Author "Color Management for Photographers".
Pages: [1]   Go Up