My Experiences with Video in E-Prime - freezing and jumping issues

[Benny Liebold]

After having read at lot in this group and the forum I feel the need
to share the experiences I made in the past few days regarding two
video issues.

For my trial I wanted to use 180 video stimuli (each about 5 s) that
were presented in a pre-randomized order. After intense testing of my
trial I experienced two major issues regarding the presentation of the
video stimuli. (1) Some test machines crashed during the trial with a
frozen picture and an audio loop or simply displayed an error that
should not be related to the design of the trial. (2) Some videos were
aborted after about 700ms of playtime and the script jumped to the
next one. I will refer to the latter one as “video jumping”.

Those two issues gave quite me a headache … especially because the
machine I used to design the trial did not produce any of these errors
at all. But at that point I realized, that I used quite a potent
machine for the trial design, being a 27” iMac with a 3.3 GHz Intel
Core i7, 8GB Ram and a Radeon 4870 with 1 GB of video memory running
Windows 7 x64 (fully updated). The test machines were not slow at all
(AMD 64 X2 with 2.7 GHz, 4 GB Ram and a onboard Video Card with 256 MB
of video memory running Windows 7 x86, fully updated), but the
difference is quite significant. So this really had to be the cause
for both issues I mentioned above. Consequently I had to deal with the
video codec I used as I already used the latest build of E-Prime 2.0,
the DivX codec packet and the K-Lite codec packet. The movie display
did work, but it was quite unstable as mentioned above.

At the beginning I used 720p-material (1280x720, 29,97 fps) and the h.
264 codec as it is know for its superb compression abilities. The
other side of the coin is the high CPU usage it produces when decoding
the videos. Speaking of CPU-usage: I realized that E-Prime only uses
one CPU-core! This really is a problem when you have quite potent
multi-core CPUs but with a poor single-core performance. In fact the
iMac’s Core i7 should be at least twice as fast as the AMDs in the
test machines when using a single core. This is due to the higher CPU-
clock, the i7’s ability to hyperthread and it’s Turbo mode which
boosts the clock to 3.6 GHz when using only one core. The architecture
is way ahead of the AMDs as well. So this really was on the one hand
the Problem of my trial and on the other hand the weakness of the
current E-Prime build. Why not using the full capabilities of current
CPUs?

Back to my trial: Consequently I had to lower the CPU-usage during
video playback. So I re-encoded my files into MPEG1 as stated by many
forum threads and built a small trial only running the stimuli in a
randomized order. Additionally I logged the Start and Finish Times of
the slides. After some testing with various video resolutions and bit
rates I came to the following conclusions.

1. the most important: take some time for intense testing and maxing
out the stimulus quality (if necessary as in my case)
2. and probably the second most important: use very fast machines/CPUs
with high single-core capabilities (i.e. Intel Core i5 and i7; an
older 3 GHz Intel Quadcore worked flawless as well) for the trial to
avoid performance issues and video jumping
3. preparation: deactivate any unnecessary services to maximize the
machine performance (Windows-Key+A; type “msconfig”; go to services;
mark “hide windows services” and deactivate all unnecessary services;
go to system start and deactivate any unnecessary programs that would
run in the background otherwise); also deinstall your virus scanning
program and deactivate the Windows Defender (and its real time
scanning ability); deactivate Windows 7 Aero; pluck out your network
cable to avoid viruses ☺
4. use MPEG1: this codec really IS CPU-friendly
5. in Codec Config I finally used the standard MPEG-codecs; only for
audio I used ffdshow-Audio as provided by the K-lite codec packet
6. do NOT – under any circumstances – use the “stretch video” function
if you experience issues related to poor performance; instead aim for
higher resolutions in the initial conversion process and display the
videos 1:1
7. set “stop after” to “no” if you don’t need the function (don’t know
if that really helped though …)
8. the freezing-script-issue seemed to be related to the available
video ram; every time the overall file size reached about 230 MB there
was a 50% chance a system would freeze; so keep your file sizes small!
Alternative: Use dedicated video cards with at least 512 MB of video
ram
9. try to experiment with the bit rate; for me bit rates between 800
kbps and 1300 kbps worked quite well; every thing higher would lead to
a freezing script; every thing lower lead to poor video quality; aim
for a bitrate as high as possible (in fact the iMacs Core i7 could
handle 6000 kbps in h.264 easily without any mistake; I settled with
1024x576 at 1000 fps for the AMDs, this led to 0,625 jumped videos in
average per trial, nothing the final data could not handle)
10. video jumping can be observed by calculating the difference
between the finish time and the start time of a slide; values lower
than a threshold you have to define indicate jumped videos
11. run theses tests on all of your test systems at the same time to
compare the results and write down any crashes and video jumps per
system to compare individual system stability as well
12. finally, important to avoid losing cases: split your Experiment
into the smallest possible count of parts; this way crashed parts can
be re-initiated without starting allover again or loosing the
participants data

I really hope this helps! So good luck and interesting data for future
experiments.

---

Benny Liebold, B.A.
Academic Assistant

Chair of Media Psychology
Institute of Media Research

Faculty of Humanities
Chemnitz University of Technology
Thüringer Weg 11, 09126 Chemnitz, Germany

-- 
You received this message because you are subscribed to the Google Groups "E-Prime" group.
To post to this group, send email to e-prime at googlegroups.com.
To unsubscribe from this group, send email to e-prime+unsubscribe at googlegroups.com.
For more options, visit this group at http://groups.google.com/group/e-prime?hl=en.