Measuring the Color of White

Gary Eickmeier writes:
OK, so I got this Ott Light, a portable desk lamp with a daylight
balanced fluorescent bulb, in order to check my prints and compare them
to the monitor.

But the light seems a little green to my sore eyes, so I am wondering
how I can measure the color temp of it to be sure of what I am looking
at.

There is no good answer to this question. Incandescent sources give off
light with a continuous spectrum, but the actual shape of the spectrum
can be biased towards blue or red depending on the source temperature.
Most light sources that are not incandescent (including fluorescent
tubes) have spiky spectra that don't look anything like an incandescent
source.

In the first place, most colours do not have a "colour temperature". If
you look at any version of the CIE colour diagram, which shows all
possible colours while ignoring brightness, you'll see that it is a 2D
space. The set of colours that can be produced by an incandescent
object (e.g. the sun, an incandescent lamp filament) forms a somewhat
curved line on this chart. The position along the line can be specified
by a single measurement (a colour temperature) because the line is 1D.

Colour temperature meters work by measuring the ratio of red to blue
light, or some equivalent comparison, *assuming* that the light is
actually from a incandescent source.

So *if* you have an incandescent source, its actual colour is somewhere
along that line, and a colour temperature value is all that's needed to
fully describe the apparent colour you get, and the shape of the light
source spectrum. If you have an incandescent source, a colour
temperature meter will tell you what the colour temperature is.

But for a light source with a weird spectrum, its apparent colour is
most likely not equivalent to *any* point on the incandescent source
line, and its colour temperature is not defined.

Even if you found a fluorescent lamp that exactly matched the colour of
(for example) 5500K incandescent lamp, it *still* probably wouldn't
look the same for viewing prints. Matching the colour of the light
sources just means that a piece of matte white stuff (that reflects all
wavelengths equally) illuminated by those two sources will look the
same.

But the apparent colour of something like a print is determined by the
effect of the cyan, magenta, and yellow dyes in the print on the
incoming light. And *this* is determined by wavelength-by-wavelength
multiplication of the light in the lamp spectrum by the absorption of
the dye. In general, you're going to get different results if the two
light sources have different spectra *even if the sources look exactly
the same when illuminating a white object*. To some extent, this is
measured by the Colour Rendering Index (CRI) of the fluorescent lamp.
This tells you how much colours will be distorted by the fact that the
fluorescent lamp's spectrum is different from an incandescent source of
the same apparent colour.

The Ott fluorescent tubes are designed to have a higher CRI than most
fluorescents, but they're still not equivalent to a good incandescent
source. If you want to see what prints will look like under sunlight or
incandescent lamps, use incandescents for viewing.

Dave

"ZONED!" writes:

I use a color temperature meter (sorry could not resist) some photo
rental places rent these.

The problem is that the output of the meter is meaningless if the source
is not an incandescent lamp or the sun. Colour temperature meters don't
measure the whole spectrum of the light; they just measure the ratio of
red to blue light *on the assumption that the source has a continuous
spectrum*.

Dave

"Dave Martindale" posted:
"...
There is no good answer to this question.
...."

Wrong. It is a VERY well researched subject. That is, if you are talking
about what illumination to use to "check ... prints" for proper color and
appearance - which I believe is the OP's actual question. In fact, IIRC
here are ISO Standards covering the subject, and the literature published
by the old US NBS (National Bureau of Standards ... now known as NIST, or
National Institute of Standards) goes back to well before the 1930s.

"RSD99" wrote in message
news:ZCVCd.16702$Y57.12969@trnddc08...
"Dave Martindale" posted:
"...
There is no good answer to this question.
..."

Wrong. It is a VERY well researched subject. That is, if you are
talking
about what illumination to use to "check ... prints" for proper
color and
appearance - which I believe is the OP's actual question.

I think the OP wanted to know how to measure the color temperature of
his lightsource. For that he would need a spectrophoto meter, a bit
too expensive for a single measurement I would guess. The Gretag
MacBeth EyeOne Photo would be very useful for that purpose.

Of course pleasant looking (daylight) color also depends on the
illumination level. At lower luminance levels a shift (of up to 65nm
for nightvision) in the human eye's peak sensitivity takes place
(Purkinje effect), requiring a lower Kelvin temperature to match the
intended color balance.

Bart

"RSD99" writes:
"Dave Martindale" posted:
"...
There is no good answer to this question.
..."

Wrong. It is a VERY well researched subject. That is, if you are talking
about what illumination to use to "check ... prints" for proper color and
appearance - which I believe is the OP's actual question.

You're guessing what he meant to ask, and then answering that question.
That may certainly be useful, but it does not make what I said wrong.
I answered the specific narrower question he did ask, and quoted that
question to make it clear what I was answering. I don't see how that's
"wrong".

His actual question was written as:

But the light seems a little green to my sore eyes, so I am wondering
how I can measure the color temp of it to be sure of what I am looking
at.

Dave

"RSD99" writes:
"Dave Martindale" posted:
"...
There is no good answer to this question.
..."

Wrong. It is a VERY well researched subject. That is, if you are talking
about what illumination to use to "check ... prints" for proper color and
appearance - which I believe is the OP's actual question.

You're guessing what he meant to ask, and then answering that question.
That may certainly be useful, but it does not make what I said wrong.
I answered the specific narrower question he did ask, and quoted that
question to make it clear what I was answering. I don't see how that's
"wrong".

His actual question was written as:

But the light seems a little green to my sore eyes, so I am wondering
how I can measure the color temp of it to be sure of what I am looking
at.

Dave

For a more precise explanation of what colour temperature actually
means, I'll quote a bit from Hunt's "The Reproduction of Colour" (5th
ed, p214):

If the relative spectral power distribution of a source is
exactly the same as that of a Planckian radiator, then the
temperature of the latter is referred to as the _distribution
temperature_ of the source. Most sources, however, do not
duplicate the relative power distribution of a Planckian
radiator exactly, but many have the same chromaticity as that of
a Planckian radiator; in this case the temperature of the latter
is referred to as the _colour temperature_. It is common with
other sources of whitish light to quote their _correlated colour
temperature_: this is defined as the temperature of the
Planckian radiator that produces light most closely matching the
particular source. These correlated colour temperatures then
provide a useful indication of the relative bluishness or
yellowishness of the sources.

In other words, if a source has a distribution temperature, it really
has a spectrum like a black body, and objects viewed under it will look
the same as they do under any other blackbody source with the same
distribution temperature.

If a source has a colour temperature, then it *looks* the same as some
true blackbody source, but the spectrum generally won't match, and the
colour rendering won't be the same if the spectrum doesn't match.

If a source only has a "correlated colour temperature", it doesn't look
exactly like any blackbody source in colour, but it's close to one.
Again, the spectrum may be very different from a blackbody.

Dave

For a more precise explanation of what colour temperature actually
means, I'll quote a bit from Hunt's "The Reproduction of Colour" (5th
ed, p214):

If the relative spectral power distribution of a source is
exactly the same as that of a Planckian radiator, then the
temperature of the latter is referred to as the _distribution
temperature_ of the source. Most sources, however, do not
duplicate the relative power distribution of a Planckian
radiator exactly, but many have the same chromaticity as that of
a Planckian radiator; in this case the temperature of the latter
is referred to as the _colour temperature_. It is common with
other sources of whitish light to quote their _correlated colour
temperature_: this is defined as the temperature of the
Planckian radiator that produces light most closely matching the
particular source. These correlated colour temperatures then
provide a useful indication of the relative bluishness or
yellowishness of the sources.

In other words, if a source has a distribution temperature, it really
has a spectrum like a black body, and objects viewed under it will look
the same as they do under any other blackbody source with the same
distribution temperature.

If a source has a colour temperature, then it *looks* the same as some
true blackbody source, but the spectrum generally won't match, and the
colour rendering won't be the same if the spectrum doesn't match.

If a source only has a "correlated colour temperature", it doesn't look
exactly like any blackbody source in colour, but it's close to one.
Again, the spectrum may be very different from a blackbody.

Dave

Bob Williams wrote:

You can probably tell if you have the CRI 83 or the CRI 93 by taking a
picture of a Kodak Gray card illuminated ONLY by the Ott Light.
Look at the image in Photoshop.
If the RGB values are pretty close to each other (+/- 5 units), you have
the CRI 93. If the G value is way out of line with R&B, you have the CRI 83
Bob Williams

But that's the crux of my question, Bob. I can't just "take" a picture
of a grey or a white card. It has to be done with some WB setting. The
White Balance that I use will be the main factor in the RGB values I
would pick out in Photoshop. For example, if I balance for that subject,
then obviously I will get a great RGB reading in Photoshop.

So I'm wondering if I should set a 6500 or a 5000k manual setting, and
see if some setting or other gives me some equal RGB readings, then I
know that is the color of the Ott light.

I think I just answered my own question. Shoot the white card with every
WB setting that is near daylight, and see which one matches white
closest. What I have available are 3000, 3700, 4000, 4500, 5500, 6500,
and 7500. I could probably eliminate anything below 4500.

Gary Eickmeier

Bob Williams wrote:

You can probably tell if you have the CRI 83 or the CRI 93 by taking a
picture of a Kodak Gray card illuminated ONLY by the Ott Light.
Look at the image in Photoshop.
If the RGB values are pretty close to each other (+/- 5 units), you have
the CRI 93. If the G value is way out of line with R&B, you have the CRI 83
Bob Williams

But that's the crux of my question, Bob. I can't just "take" a picture
of a grey or a white card. It has to be done with some WB setting. The
White Balance that I use will be the main factor in the RGB values I
would pick out in Photoshop. For example, if I balance for that subject,
then obviously I will get a great RGB reading in Photoshop.

So I'm wondering if I should set a 6500 or a 5000k manual setting, and
see if some setting or other gives me some equal RGB readings, then I
know that is the color of the Ott light.

I think I just answered my own question. Shoot the white card with every
WB setting that is near daylight, and see which one matches white
closest. What I have available are 3000, 3700, 4000, 4500, 5500, 6500,
and 7500. I could probably eliminate anything below 4500.

Gary Eickmeier