Category: Other Computers

This machine is not easy to manoeuvre; it takes two people to lift it. And this is with a switch mode power supply.

The screen looked to have deteriorated.  There were what looks like mould spots all over it.  I thought that the CRT may have been broken, but casting light on it showed the spots were not on the phosphorous but just below the surface of the glass.

It looked like it may have been under a protective layer.  A check on the web confirmed that this is a case of CRT cataracts where the adhesive holding protection screen on deteriorates.  It’s not terminal; it can be repaired.

The owner showed me how to take the top cover off – just two screws at the sides that need to be loosened.

Inside there is a long baseboard with a terminal block at one end and a passive backplane at the other. It was not clear to me whether each of the 7 cards had to be in a particular location.

The primary power supply is located on the right and the secondary low voltage supply, providing regulation of various voltages, is on the left.

The backplane did not appear to correspond to any bus that i had seen or read about.

The CRT is at the front left and is driven by the board at the far left of the card cage. The drives are mounted in a frame at the right.

The primary supply is a switching unit. It provides several unregulated outputs to the secondary power supply to create all the required voltages for the backplane and the drives. I could not see any of the infamous Rifa line filter capacitors. All other caps looked fine.

The floppy drives were removed, cleaned, lubricated and refitted. The thumb screws up the top retain the drives.

The main processor is an 8085. The card doesn’t have a lot on it but includes the printer interface, the keyboard interface, and the boot ROM. I don’t have any schematics or technical documents, so exactly how this card works is a bit of a mystery.

The date codes include 1981 so i’m thinking this was quite a late unit. There seems to be a lot of kynar for a design that i’d expect to be a few years old.

The DRAM card provides 96kB of memory so it must have some form of windowing going on. It uses 16kb DRAM chips which worried me a bit because they are often unreliable.

There are two floppy disk controller cards, which is unusual; normally a controller card would look after multiple drives. This was either very inefficient, or there was some scheme that allowed concurrent reads and writes.

There is a switch on each board near the 34 pin connector that sets the controller for Drive 0 or Drive 1.

There’s a character ROM at the top of Video Card 2.  Character memory on the right of it: 6x256x4bit SRAM. That doesn’t seem like enough for a 24 x 80 display, so something else must be going on.

The only I/O that i could see was the printer port. No serial port is sad.

Both the printer and the computer have 110V rating plates but the seller assured me that they had always run both units from 240V and the accompanying power leads certainly supported that proposition. 

I thought that the supplies had almost certainly been converted, but I didn’t want to leave it to chance. I checked each unit and went looking for some evidence of the conversions.  There is a wire on the power supply which has been moved from 100V to 200V.

The keyboard is secured by two screws underneath the unit, and the board is screwed to the surround. It uses a D8748 microcontroller.

All up, the machine struck me as being in pretty good shape. Sure, it’s just a word processor, but underneath it’s a computer looking for an operating system.

I wasn’t particularly vested in this machine but, i wanted to see if there was life and if so how much life.

I turned on the machine using the circuit breakers at the rear of the machine. These felt a bit clunky but gave a satisfying click and the fans came to life with quite a lot of noise.

The drive lights flashed regularly, as if beckoning me to pop in the boot disk. I answered that call, but there was no attempt to read the disk.

I still had the unit open, and I could see an LED on the power supply flashing with the drive LEDs, so I figured it was probably a power fault.  The CRT wouldn’t switch on.

I pulled all the cards from the rack (except the CRT supply which was a bit tricky) and the supply came up fine – solid LED.  This prompted me to measure all the power supply lines and work out which supplies go where, including the computer bus. 

The CRT still wouldn’t switch on, but I thought that it might need one of the cards that I’d removed.  (The front panel switch is wired via the keyboard, as it turns out.)

Replacing the cards and connectors progressively went well until I connected the keyboard.

I checked which “pins” on the processor board edge connector provided power, –5V and 15V, and found that the keyboard did indeed have a short from 15V to ground.  I cut a track to narrow it down.  There was a tantalum on the remaining piece so, I took a wild guess that it would the cause.  It was.  I replaced it and made the cut track good.

After that, the unit fired up.  The screen turned on, and the unit requested a boot disk.  I gave it the only one that I had, and it booted.

Unfortunately, it wasn’t picking up anything that I typed. That led me back to the keyboard.

The keyboard is a Cortron unit.  The company still exists and still makes keyboards, but I can’t see them lending much support.

The unit is almost certainly a custom unit made for lexitron.  I was able to find some documentation for another computer that described how the cortron keyboards worked.  I also had a good poke around the board. The working description for the other keyboard seemed to accord with what I could see on this unit.

The keys use contactless magnetic cores.  When they are pressed, a pulse can be transmitted from one winding to another. 

It’s all controlled from a microcontroller.  The switches seem to be arranged in a 10 x 10 grid.  The microcontroller addresses the array via two decimal decoders.  I expect there must be a sense line back – I need to find that. The micro controller presents a byte to the host processor reflecting the key that’s been pressed or released.  I imagine there’s some debouncing, and they must have done something about “shift”.

The keyboard also presents the state of the front panel switches.   A mux routes either the keystroke or the switch state to the 8085 board depending on a select (address line) from the 8085 board.  I suspect that the processor sends a “come and get it” strobe (maybe an interrupt) when a key press is detected.

The keyboard also has status LEDs (eg caps lock) and those are written by the 8085 board to a register on the keyboard.  Ie the keyboard seems to just be a peripheral device on the 8085 bus rather than be accessed via, say, a serial port.

The bad news is that I could see no strobes.  The micro controller looks dead to the world, although it is happily clocking.  It is almost as if it has lost its program.  That’s a real problem because I doubt that I’m going to be able to find that ROM binary.  It is a UV erasable part rather than a mask program – there must be some kind of finite life.

This is going to be a bit of a problem to remediate.  I don’t fancy writing that code using the original processor – it was probably in assembler. 

At this point the sensible thing to do was to say, “well i tried” and then grab the drives and the RAM and chuck the rest into recycling. I put it down for a while before deciding that i could make a replacement based on a teensy board using the Arduino development environment. This board is vastly more powerful than the microcontroller, so i’d be able to work in c.

There are two pieces to be solved:

• the protocol and character coding (maybe ASCII) to the 8085 board
• how to detect keystrokes

This was a genuine case of suck it and see, and i knew it would be time-consuming.  I decided to park the project until i could find a way to duplicate disks. In 2019, i don’t think greaseweazle had entered my conciousness.

The system came with a box of what looked to be well-worn disks. I had found no other disks online so as far as i knew, these disks, or copies of them, were the only disks that would boot up this machine.

I have one boot disk and a few data disks. They also seem to work in the machine.

I cleaned up the drives, and they seemed to work ok except for one quirk: they will only read data disks from the drive that is used to boot the machine.  I’ll have to think about what that means – perhaps an addressing issue – or perhaps there’s a command required to change the current drive.

I’m pretty sure I’ve tried both drives with both controllers and with both controllers as drive 1.  It’s another mystery, but I can’t mess with the disks too much until they are copied.

As far as I can tell, the machine makes no provision for copying disks.  A bit of an oversight, I think.  Surely there is, but without a keyboard I have no way of exploring anyway.

The disk drives are Shugart SA-400 units.  They are single sided and have only 35 tracks.  We’re peering back to a 1976 design here. 

As I understand it from the manuals, they are single density disks (which turned out to be wrong) with 10 sectors per track.  Unfortunately, single density really means that the coding is different.  FM instead of MFM used for double density.  The drives don’t care about the coding – the decoding is done on the floppy disk controller.

The disk controller cards use western digital FD1791B disk formatter/controller ICs, which is a fairly common family.  Between the industry popularity of the shugarts, and the western digital disk controllers, i expected the formats to have a lot in common with other formats. I was dreaming.

There is a reasonable chance that ImageDisk could read the disks on an IBM machine if I was using a disk controller that supports single density.  I don’t have one.  It’s worth pursuing though because some other old machines also use FM anyway.

I purchased a relatively cheap Goldstar Prime 3b card which was rumoured to support FM, but i had no success with the card and Imagedisk.

Next was an Adaptec SCSI card which is also a floppy drive controller and is known to support FM.   Fail. I tried another Adaptec card well known for supporting single density: Adaptec AHA-1542B. Fail. I concluded that it wasn’t FM coded.

I thought perhaps it might use some other variation. The Lexiton floppy disk cards are, after all, custom cards.

That really left me with very few options.  Perhaps some custom hardware that might do the job – a catweazle card or a kyroflux – but they are hard to come by, expensive, and there was no guarantee of success.   Either greaseweazle (and fluxengine) was not available in 2018 or it had escaped my attention. I put the job on hold. 

My floppy disk skills have come a long way since 2018.  These are double density disks but with a unique coding.  The coding is built into the floppy disk controller cards.  The disks can be read and written with greasweazle, but they cannot (at present) be decoded into recognizable sectors.

In early 2019, i noticed an isa card on a popular auction site, with a vague description, that was called a copy card.  Being priced moderately, being located in Australia, and noting that it had what looked like a floppy disk edge connector, I thought it was worth making an enquiry.  The seller was kind enough to send me some scanned pages from the manual, and it did sound remarkably like it could do a low level floppy disk copy.   It sounded like it may have been a clone of the old CopyIIPC hardware.

It works by taking over the floppy drive data signal.  The floppy drive cable is routed from the controller to the copy card and then on to the drives.   It has quite a good little manual (no doubt inspired by the original) and the copy program has a few options to play with.  It’s really made for copying tricky games disks but claimed to also be able to copy other formats such as apple and Atari. The irony of a copy card and its software being cloned was not lost on me.

I popped it into my PC XT. Initially i had no success with the lexitron disks, but I tried an apple ii disk and although it wouldn’t boot it did give a catalogue listing back on the Lingo Apple II clone which was encouraging.

I tried all the options, and found that with track length matching, and weak sectors turned on, I could get a Lexitron disk that would sometimes boot. I made a few more attempts and got some disks that seemed to boot every time.  That doesn’t mean that they’re error free, but at least I could work on the Lexitron without putting further wear on that original disk.