[argyllcms] Re: CCT of a Speedlight
- From: ternaryd <ternaryd@xxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Sat, 10 Jul 2021 22:02:56 +0200
On Sat, 10 Jul 2021 10:46:02 -0700
Ben Goren <ben@xxxxxxxxxxxxxxxx> wrote:
I’ve found CCT to not be all that useful in
photography. Its best use is in marketing of
consumer household lightbulbs, where it gives
an easy-to-understand expression of what sort
of “look" / “mood” the light will create.
I thought to have understood that CCT is only
half of one story, the "tint" being the other
half. And that the color rendering of an
illumination would be a second story. D50 is a
computed spectrum, not a CCT, I know, but it is
identified by a CCT (5003), and xenon discharge
light is said to be quite close to daylight.
Also argyll computes the Duv to the closest
D-spectrum. Or did I get this wrong?
Actually, I'm trying to make a new camera
profile which is "well behaved" (in the sense
of Elle Stone). I started with a flash lit
photograph of the target; I believe it should
be good enough, although I know most recommend
sunlight. Using the supplied reference data I
got a dE of almost 30. I know I shouldn't chase
the dE numbers, but the manual says above 5 I
_should_ worry. So I measured the target
myself, stuck the numbers into the reference
file and tried again. And voilà, dE went down
below 2dE.
But it isn't "well behaved" yet. Elle Stone
adds a white and a black point. I didn't repeat
the shot yet with a black trap and a teflon
tape, but just stuck in two additional lines
into the ti3 file. Now the curves said, the
profile is "well behaved", but dE went up to
90. So I guess the white balance must be the
key. And while preparing those two items, I was
thinking about the white balance of a compact
flash as mine, and how different it might
actually be from D50. Hence the question. I
guess, I won't need the CCT to do this, but
still feel curious about it. Most flash
advertising claim the flash has 5600°K, and I
ask myself if it might actually be true that my
flash is that much cooler. Being the only one.
Two light sources with radically different
emissive spectra can have the same CCT. A fun
experiment would be to find (or construct)
such a pair and set them up in side-by-side
viewing booths. The neutral patches of a
ColorChecker would look “close enough” the
same as likely wouldn’t make a practical
difference, but the rest of the patches would
be radically different. You could have some
fun, covering and then revealing the colored
patches. I imagine an exhibit like this can
probably be found in a science museum
somewhere.
Sure, the CCT doesn't say anything about CRI,
nor about any other system with the same goal
but less critics. A layman as myself can only
suppose that a flash produces light with a
spectrum that is working reasonably well to
produce reflective colors, as there is little
practical alternative. In the end, when I
bought that flash, it was Nikon's flag ship
model. It shouldn't be complete trash, and
incompatible with any other product, right?
But do you mean by this, the best choice of a
target's illumination for a camera profile would
be an incandescent bulb? After all, they say it
has the most "complete" spectrum of all. But
it's not an illumination that is easy to find
in any natural or posed scene.
1) Colorimetrically accurate representation,
such as with fine art reproduction.
This is your case, isn't it?
You need a fairly smooth illuminant with no
significant spikes nor dips in its spectra;
... and probably a more professional instrument
than the x-rite i1pro3, which gives only a
sample every 10nm, skipping all those narrow
spikes.
then, the math works well to translate (using
ICC profiling) everything to D50.
Well, colprof also doesn't have lots of
choices; it'll have to be happy with any
numbers the users deliver. As colprof can't
know about the precision of those numbers, the
applied math remains the same. Does this mean
that colprof is useless unless top-of-the-line
instruments are used? I just want to get close
enough to trigger my color memory, even a long
time after I took the shot.
Good studio flashes are moderately spiky, but
still smooth enough for this to work well.
Well, Nikon's SB-910 isn't a studio flash, but
it's not a no-name thing from no where. So I
thought, it should produce some reasonable
spectra. Photographers much, much better than
me use it quite successfully.
“Hot” lights (plain ol’ tungsten bulbs) are
superlative, provided you can get enough light
on the subject without starting a fire.
Just to light a profiling target, I guess I
could manage without a fire extinguisher. But
in this neighborhood, it's hard to get an
incandescent bulb nowadays.
2) Reportage. A photographer at a concert
wants to capture the mood of the light as a
faithful rendition of the experience of being
there. In this setting, white balance becomes
all but meaningless. If the performer is
wearing a giant 18% gray card but standing
under a purple spotlight, you technically
“should” balance the color such that the
performer’s outfit is gray. But nobody at the
concert saw the outfit as gray; everybody saw
it as purple.
That purple light is so far from any type of
"white", that our eyes would never adapt to
that as a white point, even if we gave them
enough time (which isn't the case in a concert
anyways). If the goal was to reproduce
faithfully the colors actually seen by the
audience, someone would have had lit that gray
card with a D50-spot.
For more common situations, where the
illumination is close enough to white so our
eyes can adapt, all depends on the purpose. If
I want to see the colors as they are under D50
light, I probably would take the camera's
guess. If I want to capture the mood of that
illumination as seen by someone not yet
adapted, I would move the white balance as many
mireds into the other directions of D50 as the
camera's guess is away from that.
3) “Pleasing” color. Many, many photographers
aim for “pleasing” color. There’s some
overlap, but a lot of this comes down to style
and personal preference.
...or what ever I'm trying to express. In both
cases above, I've got a starting point more or
less close to reality, allowing me to
exaggerate whatever part I feel an urge to. And
that in contrast to a somewhat random outcome,
having to decide which part should match my
color memory, and which one I choose to alter.
Note that, if the light is fully controlled
such that the scene as viewed is as desired, a
colorimetric workflow might work best — but
it’s typically only skilled and experienced
product and fashion photographers who can
control the light this well.
Oh, well. As a friend of mine used to say, Miss
Perfection lives upstairs, in apartment 5B. But
even not perfect, at least I can try to come
somewhat closer, right?
So… you probably don’t need to worry about
the CCT value of your flash. (For that
matter… unless it’s a studio flash on the
level of an Alien Bees Einstein or above, the
CCT will change with output power.) If it ever
becomes significant…you’ll know that it has
and why.
Some convincing sounding comments I've read
recently searching for the SB-910's CCT, claim
that every xenon flash, by laws of physics, will
vary CCT with the change of output power, but
not to a degree of like 1500°K. Until recently I
haven't tried measuring it myself, but this
kind of difference, I'm sure I would have
noticed. But I didn't. And the thing is, I did
try it between 1/1 and 1/64 and I couldn't
reproduce it: All CCTs reported by argyll where
in the same neighborhood; no tendency.
As a bottom line: Does the color temperature
(what ever it might be) matter for the creation
of a camera profile? If I take tungsten or the
most perfect studio flash or sunlight (at any
time of the day or weather), the colprof would
always yield the same profile (due of the
chromatic adaptation code in colprof)? Is this
the meaning of "don't worry about CCT"?
After reading Bruce Lindbloom's essay about
chromatic adaptation, I know that this is also
just an attempt to get closer. But for the sake
of that argument…
Thanks for your thoughts!
Other related posts:
