[argyllcms] Re: Calibrate a HCFR hardware
- From: Graeme Gill <graeme@xxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Mon, 14 Jan 2008 11:45:36 +1100
Gerhard Fuernkranz wrote:
I guess that the CRT/LCD switch could have other side effects too, like
syncing to the display's refresh rate (CRT) or omitting the
synchronization (LCD), and not just to select one of two different
calibration sets? Graeme, can you please tell, how this is implemented?
Btw, does the HCFR instrument actually sync to the dispay's refresh rate
too?
Some instruments don't have the facility to synchronize to
the CRT refresh rate. The HCFR and Eye-One Display 1 are such
instruments. In these cases, the CRT/LCD switch only changes
the calibration matrix. The repeatability error created
can be minimized by trying to make the integration time
reasonably long, but it will never be as good as having
it a multiple of the refresh period.
Maybe dispcal and friends should support instrument specific options,
which are transparently passed through to the driver and interpreted
only by the driver? (and for the HCFR driver, a file with individual
calibration data to use could be such an option).
Colorimeter & display specific correction matrices are certainly of
interest in tackling the problems associated with some of the LED displays,
and could be applied to the HCFR as well.
It's a matter of how to put this infra-structure into place.
One of the issues I see with correction matrices is that they
are a combinatorial management problem. To actually separate the
information needed would be to need spectral calibration information
for the instrument and displays.
Btw (maybe a little bit OT), when I look at the TCS230 datasheet then
I'm wondering, whether the HCFR instrument is equipped with an IR filter
in front of the sensor chips?
No it is not by default. I had a bit of a look around, and it's
not at all easy to get hold of such a filter either. It's one
reason I think that if there is ever an HCFR mark 2 instrument,
that they should switch to the Mazet sensor.
Otherwise the presence of IR may
significantly tamper the readings, since all four photo diodes have a
significant sensitivity at IR wavelengths up to 1100nm (IMO colorimeter
sensors should not respond to radiation outside the human visible
spectral band).
I suspect this happens in practice.
I'm also wondering, whether possibly the use of all 4 diodes (R, G, B
and clear) in conjunction with a 4x3 transformation matrix to XYZ (i.e.
treating it as a 4-band instrument) might be helpful to improve some
aspects of the gadget (e.g. same calibration valid for a larger set of
displays - this requires of course linear independence of the four
spectral photodiode responses, and just by looking at the diagram in the
datasheet I don't feel able to assess to which extent this is granted).
I'm not sure how the math would be used for this. I guess if a bunch
of readings were obtained for representative displays with different
spectral characteristics plus spectrometer reference data, that a linear
regression could be used to derive the matrix. It's not an easy
task to collect the data though, and the end result may not be
much better than using the RGB channels only.
Graeme Gill.
Other related posts:
