User`s guide

E-Prime User’s Guide
Appendix A: Timing Test Results
Page A-4
code and less than 50% with using the vertical blank simulation code
19
. This allows the software
to resynchronize, on average, once every 2 refreshes. However, this might cause the top 10% of
the display to not be updated on a missed refresh event. For experiments in which you are
altering the very top of the screen on single refreshes, the rate of missed refreshes should be less
than 0.1%. Figure 3 below illustrates the use of the vertical blank simulation code, which
simulates a refresh event. The third function of the Refresh Test is to report the refresh rate and
its stability.
Figure 3. Illustration of the vertical blank simulation code, which estimates the
time when the next refresh should occur.
If the Refresh Clock Test displays only green “+”s in the top row (1% Rule), this indicates that the
machine is capable of providing millisecond timing. If a red “X” appears only on the first run, the
timing errors were likely due to operating system “stabilizing/tuning” the memory space after
loading the program. If there is a red “X” on the first run and green “+”s after the first test, the
program is likely to run well after a short practice block. However, if many “X”s appear, the
system is running programs that are suspending E-Run and disrupting critical timing. Common
causes are:
1. Failure to exit all programs (other than E-Run) before the test
2. Virus-scan programs and background programs (e.g., MS Office Assistant
®
)
3. Network traffic or shared disk activity interrupting the computer
4. A slow machine (e.g., below 120MHz) that is unable to provide good timing
In order to achieve stable results, expose the running programs (by pressing Ctrl-Alt Del), shut
down unnecessary programs, and rerun the test. If need be, temporarily remove the network
cable and reboot the machine to check the timing. Short, one-minute tests may be run until
stable results are obtained. Next, run the test for 25 minutes. If that test succeeds, run the 6-
hour test (overnight).
19
The vertical blank simulation code is a special code, internal to E-Prime, using the microsecond clock
to cross check the expected time of a vertical blank event. The vertical blank simulation code estimates the
time that the next refresh should occur. If it was missed by a millisecond, it performs a simulated refresh
event. This works well as long as there are occasional hardware refresh events that occur and are detected
(e.g., if the refresh was detected ten refreshes previously, the E-Prime clock can predict the next refresh
with microsecond accuracy). Hence, a large RefreshMissedPercent (e.g., 50%) will not cause significant
timing errors due to the vertical blank simulation code cross checking. PST has never seen a card with a
RefreshMissedRate of over 50% so all machines operating with the vertical blank simulation code enabled
should not miss any refresh events. The vertical blank simulation code is enabled by default in E-Prime.