Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 Apollo 11/21/85; site apollo.uucp
Path: utzoo!watmath!clyde!cbosgd!ihnp4!qantel!lll-crg!seismo!harvard!think!mit-eddie!genrad!decvax!wanginst!apollo!johnf
From: johnf@apollo.uucp (John Francis)
Newsgroups: net.graphics
Subject: Re: Film Recorders (Matrix/Oxberry, Dunn)
Message-ID: <2a5a2ffc.917@apollo.uucp>
Date: Mon, 25-Nov-85 15:57:15 EST
Article-I.D.: apollo.2a5a2ffc.917
Posted: Mon Nov 25 15:57:15 1985
Date-Received: Fri, 29-Nov-85 07:57:39 EST
Organization: Apollo Computer, Chelmsford, Mass.
Lines: 65

Here at Apollo we have a Matrix QCR D4/2 - Oxberry animation camera combination.
For those of you unfamiliar with this hardware, the Matrix is a dual resolution
device - either 2K or 4K resolution (software programmable). The amount of user
control on the Matrix is good - you can load compensation tables of your choice
as well as the "standard" sets supplied for Ektachrome, as well as setting the
beam intensity for each exposure pass. (The Matrix is a constant beam intensity
device, with the exposure being varied by lingering the beam on each pixel). You
can also rotate or reflect images to deal with horizontal or vertical formats. It
also has internal calibration logic, and you can tell it to perform a calibration
scan whenever you want (between frames). It is a raster-scan only device, but it
does have a form of run-length encoding to reduce the amount of data to be sent.
You can get several different camera modules as well as the Oxberry camera - we
also have the standard 35mm slide camera. Changing modules is simple - all you
do is pull one module out and plug in the other one. The system can even tell
the software what camera module is loaded (in fact you can read back just about
everything you could ever want to know, and then some!) We have just had some of
our images of parts of the Mandelbrot set enlarged to 20x30 (inches). These were
exposed at 2K resolution, but only computed at 1K resolution, so we had to do
software pixel (and scanline) replication. We were using standard Kodak 100 ASA
negative film, and the results were impressive! There are really square pixel
blobs visible on the prints, so after a few days of CPU time we are now ready to
try a 4K resolution exposure, with every belief that the camera is really able
to achieve this degree of resolution. For what it is this device is also fairly
cheap - the Matrix is about $20K, and the Oxberry about the same. I do not know
enough about animation cameras to compare the Oxberry to other possibilities,
but it is the standard Oxberry pin-register animation camera adapted for use
with the Matrix (presumably the adaptation is simply to allow computer control
of the shutter). Enough of the sales pitch - now for the bad news:
The Matrix camera really only has 64 different pixel intensities. This leads to
very obvious contouring artifacts when the image is a computer-generated picture
(such as a ray-traced image) that has smooth variation of intensities. This can
be fixed by adding random low-order noise to the image, but of course this means
that run-length encoding goes out of the window! As the interface to the Matrix
is over an IEEE-488 (GPIB, or HP instrumentation) bus, and the maximum data rate
the camera can accept is about 50K bytes/second, a 4K by 2.7K RGB image takes
about 12 minutes. Even in 2K resolution mode the time taken is 3 minutes/frame.
This is fine for the occasional slide, but trying to shoot animation sequences
(at 2K by 1.5K resolution) requires 24 HOURS of exposure time for ten SECONDS
of running time. There are rumours that Matrix are working on a faster interface,
and/or a lower resolution mode (512 by 384 is often adequate for animation), but
at the moment I have heard nothing more than rumours.
When changing camera modules, dust has a nasty way of creeping in and settling
on the CRT. This is exacerbated by the fact that the Matrix has to be mounted
vertically for use with the Oxberry (that thing is HEAVY), so dust settles on
the screen by gravitational as well as by electrostatic attraction. This means
that after a week of continuous shooting you may well find everything ruined
by a black blob of dust right in the middle of your image! Mounting the Matrix
vertically also requires a modification to the filter-wheel assembly, since
normally the drive capstan for the filter wheel is held in place by gravity.
When we first got our unit from Oxberry this modification had NOT been made.
(Still at least the Matrix DID tell us that it could not rotate the filters,
rather than blindly charging ahead).
The version of microcode in the Matrix we got with our Oxberry camera does not
assert the "not ready for data" signal during the RESET sequence, and so we
have to put in a software delay to allow the RESET to complete. Although this
only occurs at the start of an exposure sequence (rather than for every frame)
the delay needs to be about 40 seconds to allow time for initial calibration!
An older Matrix that we borrowed as an evaluation unit did not show this problem,
but that had an old version of the microcode that did not know about the extra
commands added for the Oxberry.

And finally :  Dunn Instruments have just announced a new system - the DFR 8000.
This allows for resolutions from 512x512 to 8Kx8K (including run-length encoding),
several different interfaces (including offline use from 2400' 800/1600 bpi tape),
and can expose an RGB 8Kx8K image in about 20 seconds! It does, however, cost $115K.