Re: Accuray of EyeOnePro
Re: Accuray of EyeOnePro
- Subject: Re: Accuray of EyeOnePro
- From: Robin Myers <email@hidden>
- Date: Tue, 30 Oct 2007 15:03:57 -0700
OK, I will try to explain why companies are probably not going to give
you absolute performance metrics.
Measuring color is in the same class as measuring temperature, or
heat, with heat having much longer wavelengths. Measuring temperature
is the most inaccurate and lowest in precision of all our
measurements. Time, distance, velocity, volume, everything else can be
measured to incredibly small values, but not temperature. A quick
search on the Web yielded only one reference to measuring as
accurately as to 0.0004 degrees C. Yet we can measure time to below
femtoseconds (that's 10^-15 seconds), distance to at least as small as
picometers (10^-12), and so on, but temperature to only 10^-4. So,
measuring color is not all that precise to begin with.
Then let's look at how we get the spectrum to measure. For an emissive
subject the light goes through a slit, is split into the different
wavelengths by a diffraction grating, then it interacts with the
silicon sensor elements to produce an electrical signal which is
measured and reported back to the user. For reflective subjects light
from a source is reflected from the subject, then proceeds through the
system like an emissive subject. I realize this is an incredibly
simplistic view of a spectrometer, but the major components are there.
Starting with the slit, the intensity of the light coming through the
slit is non-uniform, introducing our first source for measurement
inaccuracy. This light is sent to a diffraction grating in a way which
minimizes this non-uniformity. The diffraction grating splits the
light up into a spectrum. The diffraction grating does not produce a
linearly uniform spectrum, depending on the number of rulings, angle
to the light, shape, surface, materials, etcetera, making it the
second potential inaccuracy source. When the spectrum hits the sensor,
the fixed spacing of the silicon sensor elements will result in
measuring varying width spectral regions per sensor element. The
sensor also has a wavelength dependent response, responding better to
long wavelengths than short ones. In addition, the sensor can produce
a signal without any light hitting it, depending on its temperature,
making the sensor the third source for measurement inaccuracies.
Then there is the random nature of light itself. This produces a
variation in the light before it even enters the spectrometer. The
light source can be quite non-uniform and variable in its spectral
output in addition to its intensity variability. This is a fourth
source for inaccuracy.
Think of your spectrometer as a digital camera with a diffraction
grating in place of the lens. All the noise and inaccuracies you see
in your digital images are the same types of problems the spectrometer
sensor has, plus all the rest of the spectrum producing system.
The upshot of all this is that the measurement accuracy of a
spectrometer varies with wavelength, illumination, the type of sensor,
the temperature of the sensor, the size of the slit, the type and
shape of diffraction grating, the optics used to manipulate the light
in the device, and a myriad of other factors. This makes it incredibly
difficult for a manufacturer to specify their devices in the way
request by Mr. Breton. Reporting the measurement error as an average
and maximum delta-E value is about as simply as the error can be
reduced.
Even the values reported by Minolta for their emissive device varied
by intensity, were reported as a variance range and were only reported
for a single small region of the color space.
Robin Myers
On Oct 30, 2007, at 12:44 PM, Klaus Karcher wrote:
Marco Ugolini wrote:
In other words, if you buy a Toyota, don't expect to win the Indy 500
with it... <g>
That's not the issue IMHO.
If you want to buy a car, you can compare some elementary technical
data like maximum speed, mileage or CO2 emission of different
models. There are standardized, traceable test procedures and
common, objective criteria to compare a Toyota and a Ferrari. They
make it possible to decide whether one of them meets your
requirements or not.
This seems not to be the case for measuring devices.
Klaus
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden