Keith Winterburn imported and supported Ontel systems in the UK until the late 1990's. Here are some of his reminiscences, related to me in July 2005:
Ontel: An Engineer's View
In the later part of the 1970's, I was a consumer electronics engineer. One day, I was asked to go to look at a faulty machine. It was a Ontel word processor, an OP1/64 with a 8" diskette drive. It had a display fault that turned out to be a transistor on the frame circuit, which I fixed. I was then asked to look at a roomful of equipment--all faulty. It turned out that the service company in the UK had been using the new stock for installation to maintain customer's machines until the stock ran out! Obviously they needed an engineer in house.
The machinery was a mixture of OP1/64s, Diablo 1355wp printers, Okidata SL160 line printers, dual floppy drives, and CDC 9427h Hawk drives and file controller 1's. Over the next few months, I repaired what was faulty and did onsite repairs for the customers, some of which were significant sized businesses. I had to learn on the job. There was a promise of a training course, but that never happened. Neither did the promise of an additional engineer.
Over the years the systems progressed with multiple disk drive machines, which made backing up interesting. We used a few file controller 1's, which had 4 ports linked to one or two CDC Hawk drives, and a larger number of file controller 11's: 8 ports linked again to CDC Hawk drives. Some of the systems got to six or seven processors. We migrated the larger systems to the 96 Meg CDC Phoenix CMD drives which used the file controller 111's with 8 ports, and again got up to seven processors.
The file controller 1 (FC1)--as opposed to FC11 or FC111--was a disk driver / multiplexer which interfaced between the essentially serial disk drive and the CPUs. (The disks were only fast tapes when all said and done.) The FC1 had Diablo connectors to the disk drive and you could daisy chain four drives. It was really a dedicated parallel to serial converter which selected the sector required by the CPUs and managed the data transfers. It had four ports which were connected by 25 pin /pin dee connectors to the OP1 series CPUs. It was really a shared resource manager and you could not transfer data from one port to another without it being written via the disk.
The FC11 was a eight port with Diablo connectors. The FC111 was an 8 port with a ribbon cable connector 60 pin to daisy chain 2 CDC Phoenix drives. The DC111 disk controller had a jumper plug which altered the logic to work with all file controllers. It also had a 26 pin cable to each drive. In extreme cases you could fit two hard disk controllers and run up to two FC111's, making a maximum of 384 megs online. But you still had to run with a single file split over more than one disk platter if it was bigger than 16 megs.
With up to 100 yards of 25 core cable between CPU and file controller, and the link running at bus speeds, timing was always an issue. The spec was 200 ft of Belden 25 core cable per pair of ports but we used more. Sometimes had to add a 10mm sq (8awg ?) earth to get error free transfer. If you had a mains splat you had a major problem as the file controller was then sometimes totally scrap with bits blown off all over the boards and a smoking power supply. The FC111 power supply used the OP1 power transformer and when it cooked it smoked dirty black smoke all over the machines in a sort of black slug coming from the power supply about a half inch high carbon mark and it stunk.
The basic machine was a nine slot machine: display, CPU, RAM 32k , another RAM 32k , I/O controller, and four device slots: comms controller, hard disk , diskette ( could be programmable diskette controller which could support industry standard IBM drives) and a printer controller or a byte string controller (whatever that was-- we only saw part populated ones used for the printer controller function). [Editor: the Byte String Controller did high-speed memory moves and comparisons independently from the CPU.]
Some of the machines were extremely reliable but some were not. The usual fault was the corruption of sector 19h on the hard disk. This was the free space (freelist record) and consisted of the start address, end address and sector count. There was also a lock byte which was used to lock the system whilst adding to a file. There was no automatic means of recovery; if you rebooted the machine it just sat locked up, requiring a program called HZAP to correct the problem. Eventually a program called AUTOFIX was available to automate the recovery from freelist corruption, which was inevitable as the links were lost. Another problem was the single byte offset of a sector caused by removing a disk controller 111 board from a terminal. The disk controller transceivers were powered down the line from the file controller and sometimes you got a sector whose contents were shifted by one byte. This corrupted the current address bytes and you got an error 8A: address error. This was a consequence of the format with each sector being started with three two byte links: From Address, Current Address, and To Address.
We also had some OP1/70's, which were compressed backplane machines, with a single CPU/RAM board instead of 3 boards. The OP1/70 lost the two RAM (orange slots), so the purple CPU slot was then CPU/RAM, making a seven slot machine. The OP1/50 was the same except it only had two device slots, which were programmed by jumper plugs. (Usually they were diskette and printer, so it was tied to being a diskette based system only.) The slots were essentially a S100 slot but not quite a standard S100 layout.
The OP1/R was available but we rarely installed any as the maximum RAM size of 32K was insufficient to run word processing, which was our main sale. We did have some OP1/R terminals used for enquiries in an estate agent. They used Gandalf line drivers in a 9600 baud multidrop configuration. They ran a diskette configuration with compiled OPL programs, but it required a OP1 to run the terminals, which could not run anything at the same time I believe
There was a Winchester disk option with 5 or 10 meg ST506 interface. The Winchester and 5.25 floppy were built into a dual diskette drive case which looked like the op1/15 but smaller. But we never had any: by then (about 1983) the PC's had arrived and the writing was on the wall.
The last machines were the OP1/15, which was a LSI-based machine with a part of the circuitry on a massive 64 pin LSI. The first units had a linear power supply with the usual faults of burning connectors, etc. They later had a switching mode supply which was reliable.
Other faults were motor bearings in the CDC drives--I must have changed hundreds of sets of spindle motor and fan motor bearings on site. Head crashes were great fun on a Phoenix; they weren't bad on a Hawk. Relay contacts on drives power amplifier transistors on phoenix mj11015 11016 were the main problems.
The OP1/70 had a six layer printed circuit board which was prone to bad connections inside the board, needing a twist in the board every once in a while.
The only really poor point of the machine was the snaking of the picture with stray magnetic fields inside the machines. Very thorough screening on the tube neck with a Mu metal sheet reduced the problem but we could not shift it all. (Mu metal is a magnetic shielding metal like thin tinplate and softer than steel.)
Another feature of Ontel was the printers, usually the Diablo 1355wp. It was a slave. It had no intelligence of its own: it received its commands from the Ontel in byte form along the Centronics interface to a Ontel board in the printer which converted it to the 50 pin Diablo interface in three bytes (I think): commanding column required and to move and then character required and strike. This was done on the fly, requiring the OP1 to know where the printer carriage was and what letter was in front of the hammer. It also allowed backwards printing, good justification, and proportional type, which was difficult. The printers did skip counts, which had the text slowly walking one way or another. Ontel shoved in a occasional restore when convenient to reset the counters if you had a missed couple of counts. It hissed at the end of travel as it covered an opto sensor.
I was maintaining some customers machines until the late 1990's, with the last of our machines--a OP1/70--in use printing address labels and shipping documents when I left in 2002. There was a customers machine still in use, but not maintained by me, in Gateshead in 2003 operating a legacy accounts program written in 1979 in OP/l.
All in all a very fine machine series with world beating facilities and the best word processor I ever used, bearing in mind the very limited CPU (0.8uS cycle time), 64 k RAM, and 10 meg of disk (if you are lucky). To run a multi million pound business on such a machine beggars belief these days!
keith winterburn line7 at engineer dot com July 2005