Dear Denis and All,
I got my photometer back today and conducted Doug's luminance test.
MyWin = Screen(1,'OpenWindow',[128],[],[]);
clut8 = repmat([0:255]',1,3);
clut10 = clut8 + 384;
%Screen(myWin, 'Gamma', clut8,8);
Screen(myWin, 'Gamma', clut10,10);
For i = 0:20
GetClicks
Screen(myWin, 'FillRect', [128+i], [512-128,384-128,512+128,384+128])
end
GetClicks
Screen(myWin,'Gamma',clut8,8);
Screen(myWin,'close');
The results appear to be consistent with 10-bit resolution. When I ran
Doug's test for 8-bit gamma, the photometer read about a 0.5 cd/m2 increment
in luminance for each step. When I changed to 10-bit, I recorded about a
0.12-0.13 cd/m2 increment in luminance for each step. Under 10-bit
resolution, I was getting about 4 times the luminance precision, exactly
what we'd expect. If we only had 9-bits to work with, I think I should see
about a 0.25 cd/m2 increment for each step, but the steps clearly were much
smaller.
I sampled across a broader range of CLUT values by varying some of Doug's
parameters, and the steps looked smooth, although stepping up luminance by
hand it is difficult to see jumps. I'd have to know where to look in the
range of CLUT values from 0:1023 to be able to find jumps of 2- or 4-fold
increases that would indicate places in the CLUT where we slip back to 9- or
8-bit resolution.
At this point, I'm not sure how to integrate the information gathered over
the last several weeks about the Mobility 7500. To summarize, the card:
1. Passes Ben Singer's 10-bit test and reports using the upper 8-bits.
2. Fails the 10-bit cluttest, with an error pattern that suggested the upper
8 bits are correct but the lower 2 bits are a repeat of the top 2 bits.
3. Passes Bob Dougherty's uniform field-flicker test.
4. Passes three of Denis Pelli's TwoBitFlickerTests, but fails one of them.
5. Passes Doug Taylor's luminance measure test for the range of CLUT values
he sampled (which were 512-532 for the 10-bit CLUT) and for some other
samples I made throughout the 10-bit CLUT range.
These results suggest a bug in the driver that affects getClut as described
by the error pattern of the cluttest, but may not affect other screen
functions. Any thoughts?
Best,
Chris
--
Chris Chase, Ph.D.
Associate Professor of Psychology
Claremont McKenna College
850 Columbia Ave.
Claremont, CA 91711
909.607.3668 (office)
909.621.8419 (FAX)
chris.chase@...
I got my photometer back today and conducted Doug's luminance test.
MyWin = Screen(1,'OpenWindow',[128],[],[]);
clut8 = repmat([0:255]',1,3);
clut10 = clut8 + 384;
%Screen(myWin, 'Gamma', clut8,8);
Screen(myWin, 'Gamma', clut10,10);
For i = 0:20
GetClicks
Screen(myWin, 'FillRect', [128+i], [512-128,384-128,512+128,384+128])
end
GetClicks
Screen(myWin,'Gamma',clut8,8);
Screen(myWin,'close');
The results appear to be consistent with 10-bit resolution. When I ran
Doug's test for 8-bit gamma, the photometer read about a 0.5 cd/m2 increment
in luminance for each step. When I changed to 10-bit, I recorded about a
0.12-0.13 cd/m2 increment in luminance for each step. Under 10-bit
resolution, I was getting about 4 times the luminance precision, exactly
what we'd expect. If we only had 9-bits to work with, I think I should see
about a 0.25 cd/m2 increment for each step, but the steps clearly were much
smaller.
I sampled across a broader range of CLUT values by varying some of Doug's
parameters, and the steps looked smooth, although stepping up luminance by
hand it is difficult to see jumps. I'd have to know where to look in the
range of CLUT values from 0:1023 to be able to find jumps of 2- or 4-fold
increases that would indicate places in the CLUT where we slip back to 9- or
8-bit resolution.
At this point, I'm not sure how to integrate the information gathered over
the last several weeks about the Mobility 7500. To summarize, the card:
1. Passes Ben Singer's 10-bit test and reports using the upper 8-bits.
2. Fails the 10-bit cluttest, with an error pattern that suggested the upper
8 bits are correct but the lower 2 bits are a repeat of the top 2 bits.
3. Passes Bob Dougherty's uniform field-flicker test.
4. Passes three of Denis Pelli's TwoBitFlickerTests, but fails one of them.
5. Passes Doug Taylor's luminance measure test for the range of CLUT values
he sampled (which were 512-532 for the 10-bit CLUT) and for some other
samples I made throughout the 10-bit CLUT range.
These results suggest a bug in the driver that affects getClut as described
by the error pattern of the cluttest, but may not affect other screen
functions. Any thoughts?
Best,
Chris
--
Chris Chase, Ph.D.
Associate Professor of Psychology
Claremont McKenna College
850 Columbia Ave.
Claremont, CA 91711
909.607.3668 (office)
909.621.8419 (FAX)
chris.chase@...