[argyllcms] Re: normal monitor calibration vs. applycal
- From: Vallo Kallaste <kalts@xxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 17 Jan 2020 15:22:36 +0200
On Thu, Dec 19, 2019 at 12:19:31AM +0200, Vallo Kallaste <kalts@xxxxxxxxx>
wrote:
On Tue, Dec 17, 2019 at 06:09:23PM +0100, Florian Höch
<lists+argyllcms@xxxxxxxxx> wrote:
as always, "it depends". Technical factors that can contribute:
- Does the video card apply the videoLUT in high bit depth and then
dithers down the result (or outputs a high bitdepth signal to a
display device supporting high bitdepth input, which usually means
dithering then still happens in the display itself) or does it use
integer truncating to 8 bits per pixel per component?
- Does a color managed application do its internal color-related
calculations in high precision and then dithers down for output (or
again, outputs high bitdepth directly if the video card and display
device supports it)?
In summary, the real-world contributing factors can be:
- the operating system (and version) [under Linux, this includes the
display server and window manager]
- the video card driver (and version)
- the video card output settings (if configurable)
- the specific color managed application(s) used and their settings
(e.g. hardware accelerated vs software rendering, if hardware
accelerated then the video card driver comes into play again)
- the display device, its capabilities and (hardware) settings as well
as the physical connection to the video card (e.g. DVI is limited to 8
bits per pixel per component)
For the record I changed old HD 3650 to used Radeon HD 7850 (Pitcairn
PRO). I was very reluctant because all such cards have active coolers
and old card was chosen specifically for passive cooling.
Anyway it seems to be the best 15€ I've ever spent PC HW vise.
It _was_ noisy as expected and didn't have PWM-controlled fans (2) which
could be connected to MB socket(s) for rpm control. But fans still
started and worked well after dropping few series diodes to fan power
circuit. Almost inaudible at idle now and almost always idle given my
usage. One day I will rip open this old 30" Dell of mine and find the
whining component which now drowns all the other sounds...
This card has 2 mini-DP ports, HDMI and dual-link DVI-D. I've got two
monitors connected to DP-0 and DVI-0, both have their own 10-bit LUT and
dithering can be controlled per output via xrandr.
The effect of dithering is most noticeable by turning it off and on
after calibrating to different white point with dithering turned on.
With old GPU after calibration grey gradient looked very uneven and
coloured. With new GPU grey gradient looks even without any coloration.
Turn dithering off and gradient looks awful, turn on and all is well
again, so it seems to work as expected. Anyway, I can now calibrate
to D58 without visible posterisation or having weird green-purple skies
on B&W pictures.
I've also noticed that calibration is much more repeatable even on the
high (dispcal -qh) level, before anything higher than 'low' was
meaningless waste of time. That's even before the current ArgyllCMS
2.1.2 version which has relaxed calibration targets in the darkest tones.
I still see mach banding for most of the grey gradient when stretched
over the screen width. Dithering works because gradient has now some
parts where it is smooth to my eye. Most of it still has mach banding,
perhaps I'm very sensitive. 10-bit LUT + dithering in newer Radeon cards
is not panacea but better than before for sure.
Other related posts:
