>Dear Denis,dear chris
>
>I haven't made much progress with measuring the voltage as yet. The
>scope I've been using really doesn't have the resolution necessary
>for the job. I'm borrowing a 400 Mhz scope next week that can give
>us 20 mv resolution per division, which should be sufficient.
>
>In the meantime, I've done some luminance testing by hand with a
>Minolta photometer. I went through the first 128 10-bit CLUT values
>where the slope of the gamma function is shallow and found
>consistent incremental increases in luminance.
>
>Today I carefully sampled the red and green 10-bit CLUT values near
>512. In our detection threshold tasks, we flicker the target on a
>gray background [512 512 512], so this range of CLUT values
>determines our threshold contrasts. Again, I found consistent
>luminance increments for each CLUT value for both the red and green
>guns (presumably blue would be the same).
>
>I'm fairly confident we are getting good resolution in this range of
>the 10-bit CLUT table and can produce cone- and luminance contrasts
>below 1%.
>
>If we get the scope successfully displaying VGA voltage, we will
>make samples in these same regions of the 10-bit CLUT table. Are
>there other suggestions that you might have about where to sample?
>I certainly have seen nothing to suggest 8-bit resolution (e.g.,
>incremental luminance for every fourth 10-bit CLUT value).
>
>Yours,
>
>Chris
>-----------------------------------------------
>Chris Chase, Ph.D
>Associate Professor of Psychology
>Claremont McKenna College
>850 Columbia Ave.
>Claremont, CA 91711
>909.607.3668
>chris.chase@...
your tests are (or will be) definitive as far as ruling out mere
8-bit performance, and as far as demonstrating that there really is a
10-bit DAC present. An 8-bit DAC can only produce steps of 1/256 V =
4 mV, so measuring 1 mV steps proves that you've got a 10-bit DAC and
that all the bits are in play. However, given our past experience
with the anomalies in transformation of values in the driver
software, one worries that even if most values are passed through
correctly, there still might be some values that are not. It takes
only one anomaly to produce an artifact that might wholey determine
the visibility of the signal.
Thus it would be highly desirable to test EVERY value, from 0 to
1023, to make sure they are ALL ok. Then we could certify 10-bit
accuracy.
It would be ok to only test one color (e.g. green) since it is very
hard to imagine that the three channels would not be identical.
i'm cc-ing to the forum, as others may be interested and have useful
suggestions.
best
denis