Re: Spectrophotometer accuracy and IT8

"samourai" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]
> hi
>
> your remarks about ccd are quite disappointing. I am not a specialist of
ccd
> or atronomical
> equipment, but as far as i know ccd is only a collection of silicon
> photodiode arranged as matrix.
> Then I do not see why they should have lower performance than single
> photodiode structure
> I think most of telescope use ccd, cooled much below 0°C to reduce noise
>
CCD of course means charge coupled device.  CCDs work by coupling charges
into a "well" and then draining them off by turning on a photodiode. CCD
measures the accumulated charge and a photodiode measures the current flow.
Cooling does reduce noise in both CCD and photodiode.  But it is more
critical in a CCD because noise accumulates charge in the well.  Thus a CCD
can "saturate" while running in total darkness but photodiodes exhibit
constant thermal noise of a few nanoamps.  CCD's back back illumination and
must have their backs, "thinned" to allow this light to pass efficiently.
They operate very differently.  It has only been recently that CCD arrays
were able to exhibit a room temperature dynamic range of more than 1000:1.

> I think the intrinsic the number of bits of ADC is also a useless
parameter
> because the adc
> is generally integrated in a dynamic control loop in a way that its signal
> input range is permanently
> adjusted to input signal.
> A good example of this is sigma-delta converters which is a finally a "1
> bit" adc !
>
Successive Approx & dual slope integrating converters do not function this
way.  If  one uses a precision voltage source on them then a 0 - 1 volt
signal is divided into either 4096 (12bit), 16384(14bit) or 65536(16bit)
steps.  If the ADC is -1 to +1 range then the 0 to 1volt is 32768(15bit).
If my dark current is around 768 counts then my dynamic range is 0 to 32000
counts or 0 to 320% with a linearity and resolution of ±1 lsb.

> regards
>
>
> "Danny Rich" <[EMAIL PROTECTED]> a écrit dans le message news:
> [EMAIL PROTECTED]
> >
> > "Timo Autiokari" <[EMAIL PROTECTED]> wrote in message
> > news:[EMAIL PROTECTED]
> > > Hello Danny,
> > >
> > > thank you for the information. However I tried to ask about the system
> > > error or the accuracy of the spectrophotometers.
> > >
> > > I do know that the manufacturers specify the range like 0% to 300% and
> > > I know that they use 10 bit (or even better) A/D conversion that boils
> > > down to the 0.003% digital resolution for the above range.
> > >
> > Minimum requirements for most spectrocolorimeters is 16bit A/D
> >
> > > But what I'm asking is about the system error or accuracy. I believe
> > > the spectrophotometer suffer somewhat from the noise just like the
> > > digital cameras and scanners do, even if the spectrophotometer do
> > > sample averaging over e.g 16 samples it still is not a huge
> > > improvement over the single sample.
> > >
> > Uncertainty in values of diffuse reflectance factor as supplied by most
> > national standards laboratories is 0.3% to 0.5%.  On uniform, solid
> patches
> > of process ink that more than 1.0 CIELAB unit.  Inter-instrument
agreement
> > between two instruments of the same brand and model will be less than
0.5
> > CIELAB unit.  I am just finishing a study for an ASTM report on
precision
> > and bias where we tested 16 instruments (45:0) against the NPL values on
> 12
> > ceramic tiles, each read 3 times.  There the average difference is about
> > 0.67 CIELAB units.  This is very similar to a study reported in 1995 by
> the
> > NPL for their spectrophotometry & colorimetry club where the average
(for
> > d:8) was 0.5 CIELAB units.  The worst instruments had maximum values
above
> > 2.0.  The best results that I have ever seen reported was a study that
> > included 4 instruments on 3 continents where the maximum difference was
> 0.15
> > CIELAB unit and the average was 0.09 CIELAB units.  Portable (handheld)
> > units alway perform worse than benchtop units.
> >
> > > For CCD based system I believe full scale non-linearity of +/-5% is
> > > quite a good value. Then there is the question of the noise level. If
> > > we assume that at room temperature the best CCDs can deliver some 10
> > > stops range then dark end raw accuracy of a 0% to 300%
> > > spectrophotometer could be something like +/- 50% and when 16 such
> > > samples are averaged it would reduce down to about +/- 6% and together
> > > with the nonlinearity this would result something like +/-11 accuracy
> > > for the dark end. Does this reasoning make any sense? If this is in
> > > the right ballpark then an ICC profiler software really should take
> > > the spectrophotometer error into account somehow.
> > >
> > Your logic follows from your assumptions.  But CCDs are lousy light
> detector
> > and no serious instrument uses them.  silicon photodiodes have been
shown
> to
> > be absolutely lines (within the uncertainty of the measurements - about
> > 0.05%) over at least 11 decades of light levels and some lab have
reported
> > observing 13 decades.
> >
> > The noise in a photometer is almost always at the high end of the scale.
> > Dark noise is generally at or below the resolution of the A/D.  There
have
> > been some exceptions to this in portable instruments where the
instrument
> > packaging contributed thermal noise to the detectors.  CCDs and
> > photomultipliler tubes are inherently noisy devices and special care
must
> be
> > taken to make them stable.  Still, RCA and Photometrics makes camera
chips
> > that can reliably detect single photons and collect them continuously
for
> > hours without saturation in astronomical applications.
> >
> > Danny
> > > Timo Autiokari
> > >
> > > On Wed, 26 Nov 2003 00:36:45 GMT, "Danny Rich"
> > > <[EMAIL PROTECTED]> wrote:
> > >
> > > >I do not know what is being used in the Q60 lab but Kodak has one of
> the
> > > >best optical metrology labs in the world.  Their master instruments
> have
> > the
> > > >equivalent uncertainty of any national standards laboratory.  The
> dynamic
> > > >range of a laboratory grade spectrocolorimeter is 0% to 300% with a
> > > >resolution of 0.003%.  Lower cost portable instruments are not quite
as
> > > >good - probably 0.1% to 150% with a resolution of 0.025%.
> > > >
> > > >There are some diffuse reflectance spectrophotometers available for
> much
> > > >money which have 8 or 9 decades of range but are not used in the
> graphic
> > > >arts.
> > > >
> > > >Danny Rich
> > >
> >
> >
>
>