The F1 was originally built by Triple-I in hopes of getting a large contract with the Government for an Optical Character Recognition system. Its design was done by some graduate students (Stanford?) named Poole, Pettit, Holloway, and Dyer (?). As I understand it, their design became the DEC KL-10, but I could have that wrong. The CPU was built on five wire-wrap pages, that were machine wrapped.
The design was done using one of the first CAD systems I had seen, (SUDS?) which I believe ran on a PDP-15 clone that Triple-I used in its PFR-80, which was originally intended to be a film recorder. The PFR had a vector-tube graphics display, and the user could graphically connect various logic elements from libraries and have wire-wrap lists generated for machine wrapping. The PFR, incidentally, was the reason the F1 became a film-maker. Triple-I had figured out how to get a smaller, better controlled spot out of a CRT than anybody else at the time, and used that technology to print film for newspaper pages.
Author's Note: I have several of the high-rez CRTs used in the III film printers, if anybody is interested in them, please contact me, (dave (at) zfx.com).
The F1 had a bizzarre bus interface to the PFR that allowed it to dump high resolution imagery to the PFR and have it print to a standard 35mm camera. The images were send in Red, Blue, and Green records and exposed through Wratten filters onto Kodak 5247 color negative film. Another PFR note, someone at Triple-I had programmed a simple graphics game to run on the PDP-15 and the vector tube. You used front-panel dials and buttons to control acceleration and rotation vectors for spaceships that could shoot particles at each other. I'm not sure of the exact origin of this program, but I believe it was the first interactive vector graphics game, which ultimately gave rise to a group of vector-displayed videogames in the mid-eighties.
The F1 had enough in common with "standard" DEC-10 peripherals that it could use a disk channel controller( RH-10?). This was attached to three RP-04 (~50MB) disk drives, one for the system, and two striped together to hold a total of two high-res images. One would be computed and saved as separate RGB files, while the previous one was being printed on the PFR. The F1 had a huge Ampex core memory unit that I believe held 256K 36-bit words. That turned out to be barely enough to do high-resolution computer graphics, so Triple-I had built the MMM, Movie Machine Memory, which was built out of 4K static RAM chips. The memory was organized so that it could be displayed as an RGB image at 1024 x 1024 on a raster high-resolution monitor. In that mode, the MMM held 6 bits per color, 18-bits per pixel. It could also be used as a bulk 512K word memory for the TRANEW graphics rendering program, which used it to store hash tables. It was fascinating to "watch" the program execute on the MMM monitor.
I actually have one of the MMM memory boards.... somewhere in my junk pile.
The F1 required a DEC KA-10 as a "console processor". Since it was a development machine, the console was used to load and run the microcode. It was also an important debugging tool when (not if!) the system did something screwy. There was a modem attached to the KA-10, and in an emergency, Dave Poole could usually be persuaded to dial into it and run diagnostics. Of course, the console machine had its own disk controllers, disks, tape drives, even a line printer, and it ended up holding user accounts and acting as general non-job user disk storage. The KA-10 itself booted from punched paper tape, so a cold-start for the entire system was quite a journey up from ground zero!
When the F1 was up, it probably ran at something like 6 MIPS. That seems really slow by today's standard, but the code had been highly optimized. The original Fortran program, TRANEW had been written by Gary Demos with help from Jim Blinn, the JPL guy that did the pioneering computer-generated fly-by animations for NASA. They had optimized inner loops in assembly code, and had removed every last bit of "extraneous" code from the operating system, TOPS-10. It was really more like Swiss Cheese than an OS.
A second PFR unit acted as an "encoding station", and was attached to a huge Calcomp digitizing tablet. External vector displays had been added, as had another puck, so that the operator could digitize from engineering-type drawings to build a 3D database for the objects used in the animations. There were some fairly sophisticated programs for extruding solids, sweeping shapes through 3D paths, and other wonders, all executing on a PDP-15!
So when I arrived and saw this system in operation, my initial awe for the amount of pioneering work that had gone into making the system do all the things it had done quickly turned into a sense of dread given the tasks I had before me, which were to: (1) re-assemble enough of the original team-members that had used the system to be able to do feature film production, (2) learn how the system functioned so that it could be maintained and kept working on a grueling production schedule, (3) attempt to find areas where its performance or reliability could be improved, (4) MOVE the system to new facilities that were being built at Paramount Studios, and (5) stay sane!
Jim Rapley, who was one of the maintenance guys who knew the Foonly, and had worked with many of the people as it was developed at Triple-I came onboard first, followed soon by Art Durinski, a designer who had worked on TRON, Adam the Juggler, and many other Foonly projects. Rich Schroeppel, a consultant for Triple-I had done the TOPS-10 hacks and occasionally consulted, as did Jim Dungan, who knew TRANEW and a lot of the code that ran in the PFR.
Of course, Gary Demos and John Whitney, having gotten Triple-I started in the movie business, saw all the shortcomings of the Foonly. They had pushed for Triple-I to build the DFP, the first (that I know of) high-resolution digital film printer for motion pictures. This was the next generation PFR, using an 8" CRT which had fast-decaying phosphors so that it could be used for scanning in film (using photomultiplier tubes built into a special camera) as well as printing. The imagery was amazing. When they left Triple-I, they put together a deal where they got Control Data Corp. to buy a Cray supercomputer, rent the DFP from Triple-I, attache Ramtek displays and IMI and Evans & Sutherland vector tubes for encoding and motion. That was Digital Productions, which Omnibus eventually bought from CDC, but I'm getting ahead of myself.
The point is that Triple-I just wanted to unload the Foonly and all its peripherals and software as one big package, and Omnibus and John Pennie were the perfect sucker. So Jim Rapley knew a guy named Serge Polevitzki, who was at the Naval Ocean (?NOSC) down in San Diego. NOSC apparently had a bunch of stuff that needed to run on a TOPS-10 system and since DEC no longer made them, Dave Poole and Foonly, Inc. obliged them by coming up with the F4, a single rack-unit 36-bit computer that ran TOPS-10. I flew up to Berkeley to see Poole and discuss upgrades to the F1.
Poole was a character. He had long hair and a long beard and glasses. I met him at Foonly, Inc. which was a small warehouse that had been the assembly line for some thirty F4's. He wore ragged clothing and bluejeans, and had a pet parrot that rode on his shoulder most of the time. When the parrot wasn't on his shoulder, it was perched in one of the "unfinished" F4's. There was a precision shaped piece of cardboard duct-taped into this particular machine in such a way that the bird's droppings were deflected away from the internals of the machine. Poole's recommendation was that we have him build us an SMD disk controller so that we could use bigger, faster disks, like the Fujitsu Eagles that were popular SMD drives at the time. Unfortunately, his schedule to do this was going to take months, and our production schedule was already overfull. He did what he could on the microcode side, but the upgrade never happened.
Probably the worst moment of my life with the Foonly came after we had managed to get it moved to Paramount Studios and were cranking through the film for a long sequence for Flight of the Navigator, a Disney film. In those days, the idea that you could have a camera make a seemingly infinite move, untethered, fascinated film directors, so they all wanted long camera moves over an infinte plane, probably the hardest thing for rendering algorithms to manage, what with all that depth sorting. Since the Foonly only had enough disk storage to hold the frame being computed and the frame being printed, the numbers worked out like this: 30 seconds of film at 24 frames per second works out to 720 images each computed and printed at 6000 x 4000 pixels. At twenty minutes rendering time each frame, that meant it took ten days to do "the run". As we were five days into our final "run" for Navigator, unknown to us, a Paramount electrician was outside hot-wiring a new circuit in a power panel, when he dropped a wrench and shorted two of the three phase power busses together. Supplies in the KA and the F1 died, as did disk drives, interfaces, you name it! When we finally got the systems to power up, the KA system drive had crashed, and we had to get new heads in and then rebuild that drive. Remember with 1600 bpi tape drives, restoring a big system could take all day! Then the KA wouldn't load the Foonly microcode. I remember about a three day period when I would drive home and try to sleep for a few hours only to drive back and try to get running again. The really awful thing was that I kept seeing big billboard signs on the way in advertising Flight of the Navigator, saying "coming next week!". "We hope!" I would mutter to myself.
As Omnibus grew, we also were developing our own 3D software that ran on Unix and Silicon Graphics systems. Ultimately, the hope was to move in that direction and let the Foonly die a peaceful death. But it lived on because it worked (mostly) and cranked out better high-res film than anything else except the DFP. When Omnibus was contacted by CDC about buying Digital Productions, we ended up with the Cray/DFP combination, which was impressive as hell, but every bit as kludged together and one-of-a-kind as the Foonly.
Last I heard, (1999?) Dave Poole had sailed into some bay up near Anchorage, AK, never to
be heard from again. He was reportedly lost at sea. He had a 100' Swede sailboat that was legendary. I can just
about envision him sailing it, parrot on his shoulder, off into oblivion.
Here is a closeup view of the "Guts" of the F1.
There were four "Pages", each machine wire-wrapped ECL on Augat standard
PC forms.
Since it was all ECL, the signals all ran twisted pair,
and there was a theory that the machine failed frequently because of
"cut through" from the many wires going around sharp corners.
To the right you can see all the signal and power cables.
Note the prominent air-conditioning outlet immediately beneath the pages!
The only remaining F1 project engineer who would fool with the F1 was Dave Poole, who lived up near Berkeley, and would occasionally dial in to help us with problems. He had instituted an upgrade program that was underway when Omnibus bought Digital Productions. Needless to say, that purchase brought all further work on the F1 to a halt. I heard that Jim Rapley and Intergon had hauled the F1 away and made movies with it for a while, but I have not been able to reach Jim to find out if this is so. Jim worked with me at Omnibus, and had worked with the F1 at Triple-I, as well as the Cray at Digital Productions.
Last Updated 9/23/2011
Back to Dave's History Page
Please email me for more info: dave@zfx.com