[Gllug] Hardware requirements for video editing?

Chris Bell chrisbell at overview.demon.co.uk
Fri Apr 11 20:14:36 UTC 2003


On Fri 11 Apr, Rich Walker wrote:
> 

> 
> Right; so if the camera *is* producing 30fps, you'll need to filter the
> output to get 25 before you start.
> 
   The staff at a local Jessops told me today that they think it is most
likely to be 10 to 15 pictures per second. I expect that they use a long
persistence viewfinder to obscure any flicker effect. Professional cameras
are very sensitive, so can often use electronic shutter control to reduce
blur and image distortion effects associated with subject and camera motion
without running out of light.

> 
> We did some tests a while ago with photo's and JPEG. Since Richard
> Greenhill runs a photolibrary, we took a photo and JPEG'd it ~20
> times. The before and after were printed, and handed around. *Everyone*
> picked the original - as the one that had been repeatedly JPEG'd.
> 
> It's true that if you use a high degree of compression, the artifacts
> will rapidly become intolerable. However, jpeg'ing the frames (or using
> some other codec) to a high quality produces such a great reduction in
> data size for little cost that it's almost always worth doing. 
> 
> > equipment is designed to cope with very high data rates. There are systems
> > specifically designed as instant replay recorders using a huge box of hard
> > disc drives, able to store 20 minutes of sound and video, as well as slow
> 
> Yes, I'm aware that in the past this kind of thing was necessary. Modern
> hard drive thruput and sizes suggest that small RAID-0 or RAID-5 arrays
> with 5400rpm drives can attain good performance, and when backed up with
> motherboards with a couple of gig of ram on board, *great* performance.
> 
   They do not use standard computers to control them, and the data is
stored and recovered raw, together with edit point markers, for minimum
delay, maximum quality. There is an operator assigned to each unit who
watches as the video is recorded, placing markers for every possible edit
point so that the replay is almost instantaneous on demand without needing
to preview the recording. The same technique is used on standard videotape,
with the same edit point system, and the tape can be automatically spooled
back to the previous edit point ready for replay, although this is not
instantaneous.

> > motion replay systems using modified cameras and videotape recorders running
> > at 3 times the normal rate, so that the replay can be slowed without too
> > much jitter, but unfortunately they are not within the available
> > budget.
> 
> I can imagine!
> 
   I used to be able to show people round, but there are too many security
implications now. If you can imagine a maximum length artic, with sides that
can be wound out when parked making it almost the same width as its length,
filled with kit. One vehicle has two independant editing suites, plus a
number of general use recorders, making a total of around 30 machines, all
capable of record / replay / variable speed replay. Standard recorders are
about 35,000 each, plus monitors, control, sound, switching and routing
equipment, communications, etc.


-- 
Chris Bell


-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug




More information about the GLLUG mailing list