>> >>Hi Jeremy,dear george
>> >>
>> >>This is just something you might want to look into, concerning timing
>> >>with video cards in the new G4s. The software we use (Vision Shell),
>> >>does not give accurate timing with the video cards that are built
>> >>into the new G4s. I have also had difficulty getting my timing to be
>> >>accurate with MATLAB and the psychophysics toolbox. I know the
>> >>website says everything is compatible, blah blah. But Just checking
>> >>my timing with this simple timing test I attached, I find that my G4
>> >>is much less accurate with timing than Helga's G3. Maybe its related
>> >>to the same video card problem. For example, waiting for 25 ticks, my
>> >>computer was off by anywhere from 1 frame to 3 or 4 frames, whereas
>> >>Helga's had absolutely perfect timing (i.e., it waited as long as I
>> >>told it to, down to the millisecond). I think you are using newer
>> >>computers for testing, so it might be worth checking the timing on
>> >>those.
>> >>
>> >>George
>> >>Attachment converted: Geldard:simple_timing_test.m (TEXT/MSIE) (0004BFF1)
>> >
>> >
>> >dear jeremy & george
>> >
>> >We pay a lot of attention to timing issues, and in fact have identified
>> >and solved timing problems in the new ATI video drivers that accompany
>> >the G4 processors. At this time i'm not aware of any timing problems, and
>> >our tests are quite rigorous. But we always welcome user reports that can
>> >document a problem, as we all use different configurations that can
>> >exhibit unexpected differences.
>> >
>> >by the way, it is very unlikely that the processor difference itself (G3
>> >vs G4) matters. We have yet to find any circumstance where that mattered
>> >(except, of course, for the welcome doubling of floating point speed in
>> >the G4). What does matter, always, is the video driver, and they keep
>> >changing, model to model, as apple and ATI fiddle. However, despite
>> >problems we had to work around, I'm happy with the ATI drivers. They're
>> >fast, and they're getting better.
>> >
>> >I didn't receive the attachment, but if George is simply measuring
>> >how long it takes for 25 ticks to elapse,
>> >and finding that it's slow, then the answer probably has nothing to do
>> >with his video card. Macs aren't great at servicing interrrupts. They
>> >turn off interrupts within interrupt service routines. So if your machine
>> >is really busy servicing interrupts, it may miss a few. Thus if George's
>> >machine has a really busy internet connection, or an empty zip drive (yes,
>> >ironically, it issues more interrupts when empty), his machine may miss
>> >a few ticks. He can test this assertion by turning off
>> >interrupt-demanding services and noticing whether his timing becomes
>> >accurate again.
>> >
>> >In the PsychToolbox, we offer a timing service based on an internal
>> >counter in the PPC processor that's immune to interrupts (unlike all the
>> >other timers offered by the Mac OS). So GetSecs is always accurate.
>> >
>> >best
>> >
>> >denis
>> >
>
>
>Hi Professor Pelli,
>
>Thank you for your thorough response. You were right on the nose by
>the way. I was getting interruptions from our server. It was a silly
>mistake, but I just overlooked this possibility. Thanks again for
>helping us figure this out.
>
>George
oh good! I could only guess, based on my experiences. Glad we got that
under control.
I've cc-ed to the psychtoolbox forum, so that others might learn from
this experience.
best
denis