CalibrateMonSpd
>Psychtoolbox>PsychCal
Calling script for monitor calibration. Assumes
you have CMCheckInit/MeasSpd functions that initialize
measurement hardware and return a measured spectral
power distribution respectively.
NOTE 30-June-2023: I tried to fix some of this mess, to make the functions
more compatible with the way current operating systems, graphics cards and
display devices work. Specifically the massive changes in how hardware gamma
tables work on modern gpu’s. I also added support for more Photo Research
colormeters. This seems to work ok with the simulated meterType 0, but I don’t
have actual measurement hardware to verify if things work as expected. I guess
it is less broken now, and maybe even working fine, but who knows?
NOTE (dhb, 8/19/12). This code is a bit dusty, as it is not
being actively maintained. In particular, the PTB display
control has evolved since this was last looked at carefully.
In general, you want to calibrate through the same set of
display calls you’ll use in your experiment. Below is some
prose written by Mario Kleiner that describes how the PTB
currently wants you to control and restore your clut. In
an ideal world, this routine would be updated to match. But
more generally, if you use this code you may want to modify
to make sure it is displaying colors the same way you will in
your experiments. The actual work is done in CalibrateMonDrvr
and CalibrateAmbDrvr so that is where you would look.
We have two functions for this. [LoadIdentityClut](LoadIdentityClut)() for loading an
identity clut and configuring the GPU for identity pixel passthrough,
and [RestoreCluts](RestoreCluts), which restores to the state before
[LoadIdentityClut](LoadIdentityClut)(). I think "[sca](sca)" calls that, as well as showing the
cursor and other cleanup actions.
[LoadIdentityClut](LoadIdentityClut) is also automatically called by [PsychImaging](PsychImaging) etc. if
the image processing pipeline is used for
Bits++/Datapixx/Viewpixx,video attenuators etc.
It is important to always use [LoadIdentityClut](LoadIdentityClut)() instead of self-made
code. Many (most?) graphics cards on most operating systems have
graphics driver bugs and hardware quirks which cause the self-made
identity clut to actually not turn out to be an identity clut in the
hardware. [LoadIdentityClut](LoadIdentityClut)() has heuristics to detect os/gpu combo
and select an appropriately fudged lut to actually get an identity
mapping. In case of new hardware with new quirks we should update
that files detection logic to cope. Additionally there is
[SaveIdentityClut](SaveIdentityClut)() to save a known good lut for automatic use with
[LoadIdentityClut](LoadIdentityClut), overriding its choice. And there is
[BitsPlusIdentityClutTest](BitsPlusIdentityClutTest) to test a Bits+ or Datapixx device
thoroughly for problems and aid in fixing them. It is easy to get
fooled into thinking you got it right, even when using a photometer,
because some of the problems are subtle and not detectable without
use of dynamic test patterns like in [BitsPlusIdentityClutTest](BitsPlusIdentityClutTest) with
Mono++ mode, or special debug functionality as in the
[DataPixx](DataPixx)/[ViewPixx](ViewPixx) devices.
I use a [DataPixx](DataPixx) to test such stuff and PTB has builtin diagnostic to
utilize that hardware to make sure that the video signal is really
untampered.
[LoadIdentityClut](LoadIdentityClut) also calls special functions in [Screen](Screen) that try to
workaround or fix other hardware interference. E.g., in addition to
cluts messing with pixel passthrough, there is also display dithering
on any digital video output, and some new pre- and post-gamma
corrections in latest generation hardware -- [Screen](Screen) can fix this for
some cards on some operating systems, but only when used through
[LoadIdentityClut](LoadIdentityClut).
E.g., almost all AMD cards will cause trouble on OSX when the
[PsychtoolboxKernelDriver](PsychtoolboxKernelDriver) and [LoadIdentityClut](LoadIdentityClut) etc. is not used,
almost all [NVidia](NVidia) cards on OSX will cause trouble unless you use some
special [NVidia](NVidia) kernel driver which is somewhere referenced on the
Bits+ pages on our wiki. We have/need similar hacks on Windows, e.g.,
[PsychGPUControl](PsychGPUControl)() for AMD cards. On Linux either PTB's low-level
fixes apply or they are not neccessary.
So the [CalibrateMonSpd](CalibrateMonSpd) etc. should be fixed to use
[LoadIdentityClut](LoadIdentityClut)/[RestoreCluts](RestoreCluts), everything else is just begging for
trouble.
7/7/98 dhb Wrote from generic.
dhb dacsize/driver filled in by hand, Screen fails to return it.
4/7/99 dhb NINEBIT -> NBITS.
dhb Wrote version for Radius 10 bit cards.
4/23/99 dhb Change wavelength sampling to 380 4 101, PR-650 native.
9/22/99 dhb, mdr Define boxSize.
8/11/00 dhb Save mon in rawdata.
8/18/00 dhb More descriptive information saved.
8/20/00 dhb Automatic check for RADIUS and number of DAC bits.
9/10/00 pbe Added option to blank another screen while measuring.
2/27/02 dhb Various small fixes, including Radeon support.
dhb Change noMeterAvail to whichMeterType.
11/08/06 cgb, dhb OS/X.
9/27/08 dhb Default primary bases is 1 now. Use RefitCalLinMod to change later if desired.
8/19/12 mk Ask user for choice of display output device.
6/30/23 mk Use new clut mapping to fix this mess on standard gpus. Also allow
choice of different supported colormeters, not just PR-650.
8/31/23 mk Assign cal.describe.dacsize also if g_usebitspp is already set.
Psychtoolbox/PsychCal/CalibrateMonSpd.m