Would anyone be interested in making some of these DIY LED Panels?
I have seen many builds and they often use aluminum cookie sheets and Warm or Cool White LED strips. Patrick has made a flexible version with both so you can adjust the color temp. The only limitation I see is his design can’t drive the LED to full power due to heat. Still seems very bright.
Let’s build some for either Digital Media or to use in the eLab as work lights. Probably be about $75 per panel. If you want the gimbal arm he suggest it would be another $18. Take a look at the DMS Lighting Wish List on Amazon.
Just a general FYI, but be REALLY careful when you select LEDs to use for photography. You can very easily pick one that “looks good” but does not transmit a full-enough spectrum to make colors look right.
In my experience, about half of the pop you get out of DSLR-style photography is the really deep color – which you could easily lose with poorly selected lighting sources.
If color matters to you, don’t go for a CRI of less than 90.
I would use just a couple to light a test panel before building whole assembly to see how color balnce comes out and to view view RGB graphs for any major gaps. But that’s just me.
This COULD work, as long as your test panel is appropriate.
But imagine this situation. Three monochromatic sources, on a perfectly white background. You get a perfect RGB histogram out of the camera. Now take a sheet of something that is monochromatic brown, and it will be black.
That is an oversimplification, but it illustrates the problem. Pigments, inks, and dyes do not reflect on exact RGB spectra.
I do not see what your concern is. It is standard practice for ccd based photography to use shots of color charts to calibrate lighting to achieve accurate color renditions. The only real concern is color consistancy, not temperature.
Even with a light source that has one color component much stronger then the others you can still obtain accurate color renditions by using raw files and calibrating against a standard color chart like this:
Objects in general don’t display a trichromatic response. In other words, while you can mix three pure lights to obtain a visual equivalent to any shade, the process is not bidirectional. Therefore, if you have significant gaps in the spectrum of your light source, you can get reasonable-looking results when looking at a chart, but not be able to get reasonable rendering of photographic subjects.
There’s not any difference between video & stills taken with the same sensor, in any case.
Not a difference in the sensor, but a difference in the ability to post process for color. With still, you need the raw ccd data, I dont think video captures this raw information.
Thank you for the response, I was not aware that a ‘white’ led could be produced that had a complete abcense of a part of the color spectrum and still remain ‘white’
Do you know of a specific part number that exhibits this problem? I would like to play around with one.
No, your chart doesn’t show what you previously claimed. That a white LED had a complete absense of part of the spectrum. While 480nm produces far less power, it still emits light at that frequency and the color response curve of the ccd can therefore be calibrated to produce an accurate color rendition.
The calibration, and color adjustment procedures I referenced above often involve a great deal more the simply calculating a color temperature to adjust the image. They can, and for cases where really accurate color rendition is required, do include calibrations based upon a light curve just like you show.
Heck, I have seen accurate color photos produced with tri-phospher flourescents that produce similar response curves.
If you try to apply that kind of differential correction, with about a 15-to-1 variation across the spectrum (orange curve), you will get quite bad results.
I have seen amazingly accurate photos taken with a diy led light using cheap and presumably very bad, especially since they were early models, of white leds. She used tha calibration techniques I referenced above.
There are problems with using CRI to evaluate these lights.
The Yuji leds look promising but are probably not cost effective for this project. Cheap leds could be used w/color correction done in post. Shoot a Macbeth color card, create a set of color correction curves, apply accordingly to rest of shoot.
Don’t mix leds of CRIs. Why? Over all coverage of subject may not be of a consistent CRI making final color correction difficult if not impossible.
