#PAR or FAST
27-Jul-95 23:18:50
Sb: #182361-#PAR or FAST
Fm: Martin Doudoroff/KUB 74664,2144
To: KEVIN SANDERSON 74002,35
>> What kind of compression settings have you tried? How much time do you
get on what size hard-drive before you run into problems? <<
PMFJI, but here are some of my Perception experiences:
I'm using the Perception (w/ capture card) and a Micropolis 3243AV Gold drive.
The computer is a DEC XL90 — I originally installed the PVR in a P66, but the
PVR didn't like the bus speed, so I had to swap computers (contact DPS on this
issue before purchasing!).
Other than the bus speed issue, installing the PVR was a no-brainer.
The 3243 drive, a 4Gb Fast SCSI-2 interface device, has MAX/MIN advertised
sustained data rates of 7.2/4.0 Mb/second and an average seek time of 8.9 msec.
Actual performance will vary across the surface of the drive, decreasing as you
move to the inside tracks; disk analysis through the PVR suggests I am getting
about 6.2 Mb/second on the outside tracks and about 3.75 Mb/second in the
innermost tracks.
The PVR uses adaptive compression at the frame level (it may be at the field
level, but the documentation says frame) to maximize compression efficiency.
The main quality factor, as in the PAR, is block limit (the maximum amount of
space provided for each frame). There appears to be approximately 68 DPS
"blocks" per Mb per second. DPS includes four default "record quality" settings
(you can create your own) which are as follows:
Name Block limit MB/Sec
——————————————————
Offline 100 1.5
Industrial 270 4
Broadcast 340 5
Digital 410 6
For perspective, the best block limit I could ever get away with on the PAR was
210. (There are various explanations for this: my IDE drive was slower, the PAR
has less-sophisticated compression, etc.)
There's no point in comparing the PAR and PVR from an equipment standpoint:
they are completely different devices. It is worthy to note that the PVR is
component throughout–an important issue for broadcast applications. From a
subjective standpoint, there is definitely NO comparison: the PVR is produces
(and digitizes) vastly superior video than the PAR. Speaking simply in terms of
JPEG artifacting, the PAR at its best falls a little below the "Industrial"
setting on the PVR. At "Broadcast" and better, you cannot really discern
artifacting, even on a BVM-1910. The signal quality (rendering) leaving the PVR
is top-notch (to be expected from DPS). My animation looks, well… like it
came off an Abekas.
Everything we've done with the PVR has been at the "Digital" level, out of
sheer principle (and glee). I have digitized as many as ten minutes of video
and I've output many minutes of 3D Studio frames, and I have yet to see the PVR
glitch–all at "6 Mb/second". (Because of the adaptive compression, I'm finding
also that the video rarely gets anywhere near the block limit threshold, so the
PVR isn't really even pushing 6 Mb/sec.)
Editing performance is, so far, very good with the 3243: we've done some simple
cuts-only output from the PVR with flawless results. I havent had the
opportunity yet to test the PVR with Speed Razor for editing and generation
degradation: that will be the acid test.
I expect to get about 15 minutes of video out of the 3243 at the 6Mb/sec
setting.
A special note about the capture card on the PVR: its ability to capture image
detail from video is astonishingly good. We digitized several minutes of
ultra-detailed Betacam imagery and the PVR retained almost everything. I hope
to run some tests on this, now that my curiosity is piqued.