[argyllcms] Re: Slightly wrong black levels on projector

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Mon, 16 Nov 2015 11:15:29 +1100

Antonio Marcheselli wrote:

I'll take the lack of interest as a "get a life" reply? :)

Hi,

Typically no reply means that you've asked a questions to which
it's difficult to frame a useful response.

Please find below a picture I've just taken with my camera. I have set the
exposure so
it kind of looks as it should. Level 18 is barely flashing, you really need
to look for
it and go by the screen. Level 17 is completely dark.

https://www.dropbox.com/s/ukclizi1ngi6m7l/IMG_9087.jpg?dl=0

I have tried adjusting the levels before doing a calibration. Makes sense to
me, but
when adjusting brightness and contrast I am also changing the white balance
on the
Panasonic - which I can further adjust but losing some of the light. In fact
adjusting
the levels before doing a calibration generated a better result in terms of
black levels.

What is the level of fussiness here in your opinion? I feel that the perfect
black level
is crucial for a good calibration.

Sorry - what exactly are you asking ? - it's not clear.
i.e. what are you expecting here that's different from what you are getting ?

Note that there is nothing magical about a step wedge -
i.e. there is no reason to think that every step of some
wedge should be visible when correctly calibrated, since
visibility depends on the step size and the target curve shape.

A sufficiently high pure gamma curve on top of a high black offset + 256 step
wedge
will be expected to have several steps that are not visibly different.

See
<http://www.avsforum.com/forum/139-display-calibration/1471169-madvr-argyllcms-135.html#post38694313>
for example of a case where the first steps are expected
to be invisible.

So the best way would be to measure your step levels and compare to the expected
value, to assure yourself that the calibration is sufficiently accurate
(assuming
your instrument is capable of the necessary accuracy). Easiest approach would
probably be to run a verification chart with the step wedge values as the
target.
[ You could do this one step at a time using ColorMeter too. ]

Graeme Gill.


Other related posts: