[argyllcms] Re: Displaying sRGB graphics on wide gamut monitors - gamma problem?
- From: Anders Torger <torger@xxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Wed, 27 Oct 2010 21:11:52 +0200
On Wednesday 27 October 2010, Gerhard Fuernkranz wrote:
> Am 27.10.2010 17:50, schrieb Anders Torger:
> > However, I need to study more how the data in these tables are
> > actually used. The source of all this was that I expected vcgt
> > values at 0,0,0 to be 0,0,0 with the -k 0.0 parameter to dispcal
> > ...
>
> Well, do you stringently need that as long it is granted that the RGB
> numbers > 0 sent to the monitor still display the same black color on
> the screen?
No, you're right. I've looked on my monitor (a Dell 2408-WFP), and
uncalibrated black level 4-5 is the first that is visible to the eye,
3-4 with some leaning towards the side. Looking in the vcgt, the
starting 16 bit values are <1559, 1374, 1336> which indeed is the
numbers that get into XF86VidModeSetGammaRamp() call for X11, I checked
that to make sure I've not misunderstood how vcgt is applied. Just to be
even more sure I also tested to change the first value to something
insane (15000) just to see that it affected black <0,0,0>, and it did.
Since I use DVI (8 bit) and the LCD-TFT element in the Dell 2408-WFP
also only supports 8 bit these 16 bit numbers are eventually truncated
to 8 bit <6, 5, 5>. Thus there is only a very very slight increase in
brightness, which is not detectable by the instrument (Colormunki) using
spotread.
So I probably should not worry so much about it. I was just a bit
surprised that -k 0.0 does not cause calibration curves to start at
0,0,0. But looking into the dispcal source code I found the answer, it
seems like the parameter is used in the measurement stage rather than
after, so you get the instrument errors on top. I thought the parameter
was applied after measurement so I could use this as a way to hide dark
colour precision problems of my instrument.
For my model of pretending-that-the-monitor-is-perfect-near-black-to-
avoid-possible-worsening-by-applying-poor-precision-measurement to work,
the profiling step must also ignore this, and of course it doesn't.
It seems like hand editing is the only way to do what I planned
initially, but since its not a feature of the program I guess it is
generally not needed. I don't really know exactly how poor this low cost
spectrometer is at dark colors, perhaps not as bad as I think. There
does seem to be quite much of noise though.
/Anders
Other related posts:
