Simplistic question: The spectrum is useful obviously in showing spikes and gaps and guiding filtration, but that has issues as has been pointed out, how do you, from that spectrum, then set the single figure white balance the camera allows and which is only in RAW data anyway so has no effect on what is captured, although we are repeatedly told to get it near enough before going to the converter because it does have some effect?
Simplistic answer:
I usually just shoot on daylight or tungsten preset- it seems to get close enough as a starting point for RAW processing and it means I know where I am starting from. I know that I need to dial in some plus green to match daylight or tungsten on the RED, but I do that on the lights rather than on the camera.
If I think it is very tricky I set using grey card in most "representative" bit of the scene, this can help dial out gross green and magenta shifts.
Flash is so much more powerful and just as portable, and has something much more like a true black-body thermal emission spectrum, so I still use flash 90% of the time for stills.
Where I have found the LED panels particularly useful is a little extra warm fill and catchlights in model's eyes when shooting available light; here I definitely rely on daylight or tungsten preset depending on what the main light source is, and consult the back of the camera to get some handle on green/magenta shift, then dial that up on down on the panel.
How does the Sekonic meter chooses its overall "Average" colour for a very lumpy spectrum? I guess only the engineers at Sekonic can answer that. Maybe they just do a one-parameter fit to the spectrum and accept the answer, however poor the fit actually is?
But that's not any help when setting stuff on the camera, as the camera's colour temperature settings aren't for continuous spectra either. They are to mix relative amounts of R, G and B channels, as that's the only thing the camera CAN do. The only data it records are the RGB intensities inside the filter passbands.
This sort of thing is what I mean:
(from
http://www.samirkharusi.net/spectrograph.html it was the best passband graph I could find with a quick Google).
As you can see, the R G and B passbands overlap, and the red filter leaks at the extreme violet end. So the camera's response will be the convolution of each filter response with the input spectrum.
Note that the dip in the cyan in the LED spectrum is pretty much where the blue and green filter passbands cross over. This will make the exact response of the camera highly dependent on exactly where the dip falls relative to those two pass bands.
Ironically this might be why my Hasselblad (which generally seems to have very clean colours and low sensitivity, suggesting they may have used stricter passband filters for the RGB pixels) reacts so much worse to the LED spiky spectrum than something like Canons or REDs, where the dyes are chosen with half an eye on low-light performance rather than going all out for colour purity.
If a camera from a particular manufacturer happens to have the filter cross-over more cleanly up in the blue, whereas another has it more down in the green, you're going to get very different camera responses from the same light. One might register the light as very blue, another camera as very green. As this is an artefact of the camera and the light convolved together, it is impossible to see by eye: you can only see it once you've taken a shot. And in neither case does it say anything about how the intensity of the red channel will vary in parallel with the recorded intensity in the blue vs the green. So in general there will not be a "correct" colour temperature setting anyway- that models the response of RGB channels assuming a continuous thermal spectrum, which isn't what's present.
Note that the leakage in the red passband up into the blue/violet happens to coincide roughly with the peak of the blue emission from the LED, too- so the contamination of red in the blues might give a much stronger red contamination than the RAW processor is expecting, resulting in red or magenta shifts in the blues.
There's also a general lack in the far reds both in filter and light, so objects that appear red to the eye might end up looking very dark on camera.
Adding in extra passbands only makes the situation more complex:
(From
http://www.leefilters.com/lighting/colour-details.html#247&filter=tf for regular minus-green).
Look at what CTO does:
To get the response of the camera to a spiky spectrum light filtered with that is beyond my capability to visualise in my head! (I think you multiply the light spectrum by the fiilter response, but then have to convolve with the camera dye passbands to get the R, G, B responses). Whatever you get, it probably won't be what you naively expected, and you'll very likely be throwing away a lot of intensity, too.
This is why I worry when I see manufacturers shipping daylight balanced LED panels with slot in CTO filters. What comes out will likely bear little resemblance to a genuine 3200K black body thermal spectrum!
I note that Lee are now doing a range of filters specifically for LED, which have cleaner passbands than their regular range:
http://www.leefilters.com/lighting/colour-details.html#CL180&filter=ldbut at the moment those are more like party gel colours matching to Tungsten, not the subtle effects of a +1/8th CTO or -1/4 green.
Cheers, Hywel
