iiinova — where your digital self keeps evolving

PS: I have (small) Three Rivers Computer Corporation stickers to hand out!

RSRE Flex - made by MoD researchers and one of the weirdest OSs I have ever experienced

Proof of life: my PERQ 2T2 running RSRE Flex. Flex never had a fancy graphical demo: this is just the login screen.

POS - a simple DOS-like OS that was more popular in the US

PNX: a System III Unix by ICL with a pre-X11 GUI, the UK's most commonly-used PERQ OS

Proof of life: my PERQ 2T2 running the PNX graphical demo. A high-resolution monochrome video display shows a set of overlapping windows with line and raster graphics and with sales text extolling the features of the PERQ 2.

Tom Stepleton @stepleton.bsky.social · 20/05/2026

Please join me in praying that nothing else fails before the show. RetroFest 2026, Swindon, 30-31 May. You *might* see a working PERQ 2T2 there.

A bare window on a spare 1980s graphical interface bearing the message "hello retrofest 2026"

040

Tom Stepleton @stepleton.bsky.social · 20/05/2026

EIO board U48 CPU board U134 🎯

Two DIP IC packages sitting on a blue anti-static mat. The 20-pin device is a 74S225 16x5 FIFO and the 16-pin device is an AM25S10 4-bit shifter. Both have big Xs written on them in yellow paint pen.

Tom Stepleton @stepleton.bsky.social · 14/04/2026

45 days until #RetroFest 2026 in Swindon. #PERQ status: - Flaky Z80-based I/O subsystem: sometimes the #Z80 jumps off the tracks and executes nonsense code - Disk boots fail: the microcode attempts to load blocks far beyond the end of the disk Trying today: DIY diagnostic microcode

A still from a video on the Technology Connections YouTube channel. The presenter Alec is sitting at his usual desk with a device that makes the bell noise heard at American railway crossings. The closed caption visible on screen says "But almost anything can be fixed if you're patient enough."

050

Tom Stepleton @stepleton.bsky.social · 11/04/2026

POV: #PERQ hardware debugging, as usual. The #Z80 (serving as an I/O coprocessor) is behaving strangely: fetching instruction bytes without toggling the M1 pin, jumping to random locations at around 47uS after coming out of reset; what's going on here? RetroFest in six weeks or so...

A hardware debugging scene: from left to right, a Tektronix TLA714 logic analyser (a big squat box with an LCD computer display), a PERQ 2T2 workstation with the covers off and the EIO board jutting out from the front (I'm using bus extenders for easier access to the ICs), and an oscilloscope. Wires snake everywhere to connect the test equipment to the guts of the ailing PERQ.

081

Tom Stepleton @stepleton.bsky.social · 04/04/2026

Developing technique for "disk baking". The box is a filament dryer for 3-D printing, and the disk "bakes" inside at 55 °C for six hours in an attempt to stabilise the degraded binder that holds the magnetic oxide to the PET substrate.

An 8" floppy disk rides "hammock-style" on a pair of wires slung inside of a smoked perspex box with a backlit indicator panel.

Tom Stepleton @stepleton.bsky.social · 16/03/2026

You want: a regular #Ethernet MAU to plug into an AUI port You have: a MAU that only works with Apple's weird AAUI port No problem! Here's an adaptor: codeberg.org/stepleton/aa...

Two Farallon EtherMac 10BASE-T MAUs with their AAUI connectors plugged into small circuit boards with AUI plug connectors. These circuit boards are the AAUI-to-AUI adaptors described by the project linked in this post.

021

Tom Stepleton @stepleton.bsky.social · 21/02/2026

Do you have 5.25" and 8" floppy disks that are spare, low-quality, degrading, or just generally in bad shape? I'd like to take my floppy disk data recovery game to the next level. I have equipment, but there's no substitute for experience.

A trio of floppy disks: a big 8" disk, a fairly ordinary 5.25" disk for PCs, and an Apple FileWare (aka Twiggy) disk.

695

Tom Stepleton @stepleton.bsky.social · 17/12/2025

Has anyone ever heard of a (UK?) journal called University Computing? Know where I could find archives? Long shot, but this paper could contain info that helps us open-source a remarkable operating system developed by the UK MoD in the 1970s and 1980s. Boosts appreciated!

Picture of a citation at the end of a paper. The citation says:

15. "Flex and the ICL Perq2" M.Stanley. University Computing Vol 8, No 2, Summer 1986.

152

Tom Stepleton @stepleton.bsky.social · 18/11/2025

@breakintoprogram.co.uk so the problem with my Spectrum+ was the Z80, it turns out. The address bus was stone dead, +0V on all sixteen lines. Subbing in the CPU from my Jupiter Ace got it going; a replacement is now on order. Thanks for the chat on Sunday!

The display from a broken Sinclair ZX Spectrum+ computer: inside a thick green border, eight alternating black and white vertical bands with a few small, coloured squares arranged on the left-hand side.

240

Tom Stepleton @stepleton.bsky.social · 29/09/2025

Just in time for #PERQtober ! An adapter to replace your missing or broken #PERQ T2 landscape monitor with a VGA display: codeberg.org/stepleton/pe... Thanks to @skeezicsb@oldbytes.space for early advice.

A circuit board with two DIP ICs and some passive components connects a VGA cable and two DIN-5 cables (top) to a DIN-4 cable, a DA-15-terminated ribbon cable, and a 75-ohm coaxial cable (bottom). The bottom cables go to a PERQ T2 workstation; the top cables go to mouse, keyboard, and display.

Tom Stepleton @stepleton.bsky.social · 26/08/2025

The Olivetti Programma P203 (1967) uses a circuit construction technique known as "transistor jail" (or "cordwood" [1]. Seen at [2].) [1] en.wikipedia.org/wiki/Printed... [2] www.thecomputerchurch.org

030

Tom Stepleton @stepleton.bsky.social · 20/06/2025

Wheeling Airglow (en.wikipedia.org/wiki/Airglow... ) out to the runway at the British Human Powered Flying Club's 2025 Icarus Cup event, Lasham Airfield, Hampshire.

A group of people walk a slender aircraft with long wings down a tarmac taxiway, all silhouetted by the rising sun ahead of them. A shallow mist rises off of the airfield's grass surface.

Tom Stepleton @stepleton.bsky.social · 20/05/2025

#Smalltalk -80 on #PERQ --- Mario Wolczko's implementation (www.wolczko.com/st80/) running with maybe some gremlins still to sort out.

A monochrome graphical display from an old computer workstation. In a large window at the centre of the display, a Smalltalk-80 session can be seen, with a browser window partially occluded by a System Transcript and a System Workspace window.

041

Tom Stepleton @stepleton.bsky.social · 23/04/2025

Here are some images from ICL's looping demo that runs on the #PERQ 2T2 under PNX 5.

A black and white screen image of a graphical desktop with several windows. One window shows a clock, another a bar graph with the bars bouncing around, another a graphical diagram of a heterogeneous computer network with many PERQs on it, and a final slide describing various attributes of PERQ Networking, viz.

PERQ Networking

* 10 MBit Local Area Network
- ECMA 72 Transport protocol
- File Transfer
- Remote Printing/Plotting

* Wide Area Networks
- IBM 2780/3870 emulation
- ICL-C03 protocl
- Asynchronous Terminal Emulation
- uucp

A black and white screen image of a graphical desktop with several windows. One window shows a clock, another a world map, another a pie chart of world population by region, another the word "PERQ" in a stylised ICL font of the time, and finally a "text slide" describing various attributes of PNX, viz.

PNX Operating System

* Derived from Unix System III
* Microcode implementation
* 32 Bit virtual address
* Paged virtual store
* Single (32 Bit) & Double (64 Bit) precision arithmetic
* Window Manager
* Fortran 77, Pascal and "C"

Tom Stepleton @stepleton.bsky.social · 23/04/2025

OK, so the #PERQ 2T2's disk system will work if: * you let it warm up for a bit * you have the logic analyser attached to U9 on the EIO board * ...and running

Splayed out on a wooden floor: a covers off, partially-disembowled PERQ 2T2 under-desk workstation computer. Wires snaking out of the PERQ attach to a solderless breadboard which is then attached by a sheaf of probe leads to an old Tektronix logic analyser nearby. It looks a mess but atop it all is a 19" monitor showing video coming out of the PERQ. The machine is running and shows the diagnostic code 255 on a 7-segment display.

Tom Stepleton @stepleton.bsky.social · 25/02/2025

This is pretty funny to, like, maybe three people on Earth.

680

Tom Stepleton @stepleton.bsky.social · 27/12/2024

After learning the true meaning of "we're not doing presents this Christmas", it was off to the makerspace to hastily laser-engrave and laser-cut jigsaw puzzles for everyone.

A parody clone of the cover of a Ravensburger puzzle box, advertising a "Laserburger Puzzle" with "up to 570" pieces and "Crummy Laserburger Qualtiy [sic]". The puzzle claims to have "Arduousclick Technology", referencing the "Softclick Technology" of actual Ravensburger puzzles. The puzzle itself appears to be a halftone image of a bunch of potatoes, and helpfully the original potato image appears at the bottom of the box art alongside the text "it is potates".

1232

Tom Stepleton @stepleton.bsky.social · 22/12/2024

I would have mentioned the 4051 if you hadn't, of course! Running the 6800 at a mighty 800 kHz. They were working on this 4051 at the vintageTEK museum last year, and as nice as the 6800 was, check out these gorgeous 6820 PIAs...

A circuit board with a collection of DIP ICs from the early- to mid-1970s, with characteristic "wavy" hand-drawn circuit traces. Closest to us is a ceramic-package MC6800 with a gold cover, but behind it is a rank of five MC6800 PIAs made by AMI in white ceramic packages with grey "spider leg" traces extending from the chip die to the DIP pins. If only microchips were packaged like this today...

150

With U141 replaced, U159 starts pulsing. I fixed the mouse. I have a working PERQ at last! I can run manufaturer demos that are honestly pretty good for the 1980s! The honeymoon lasts a week, and then some chip in the hard drive system fails. That's what I've been fixing ever since. (fin)

A brief check with the logic analyser and we had our smoking gun. You can see data going into one of the 74S225 and then one of the bits missing when it comes out. U141 is the culprit.

The screen of a logic analyser showing numerous signal traces. A circled area on the left shows the hexadecimal digit C going into the 74S225 memory, while a circled region on the right shows the hexadecimal digit 4 coming out of it. The binary pattern for C is 1100, while 4 is 0100. A 1 bit has been lost.

We've learnt to be suspicious of 74S225 chips: it seems they don't age very well. It's been necessary to replace two of them in the PERQ already, since they made the display look like this. (That's a whole different story.) Good thing a PERQ 2 system uses about 30 of the things...

Close-up of a computer display showing text-filled areas and blank areas with graphical corruption. The nature of the corruption is lots of black vertical bars as well as vertically-aligned dots. For the curious: the PERQ has special hardware for moving data around in memory called a "blitter". It's especially useful for drawing text to the screen, since you can store characters in other areas of memory and then copy them to the display each time you need to show a new letter. But the blitter temporarily stores the data in a 74S225 as it does its copying, and if the 74S225 doesn't store data correctly, you get corruption like this.

100

Anyway, why is this happening. Well, what does the Z80 code encounter on its journey between the PERQ CPU and the I/O-handling Z80? The schematic shows that it filters through these 74S225 memory chips, which basically handle the traffic jam of a two-lane highway for data merging down into one lane.

Excerpt from the PERQ EIO board schematic showing four 74S225 16-by-5-bit FIFO ICs that connect the PERQ memory data bus to the Z80 data bus. Those are a 16-bit and an 8-bit bus respectively, so the analogy of a two-lane road becoming a one-lane road is pretty apt. The bytes must take turns, and the 74S225s gives bytes a place to wait as they zipper-merge onto the data bus. There are four 74S225s here, but that's because each one is only five-bits wide: you need two of them to store a byte. (Each one has an unused bit.)

100

Maybe it only applies if you're lucky. I think my PERQ's Z80 was experiencing some truly Buster Keaton-level good luck. BTW @ctrlaltrees.bsky.social , this bug is why that keyboard glitch you saw was happening..

In a modified clip from Sherlock Jr (1924), Buster Keaton rides a driverless motorcycle on its handlebars and narrowly avoids getting run over by a train (repeatedly).

100

It's especially important because the instruction the Z80 runs to send configuration to U159 is D3, the "OUT" instruction. 53 does something entirely different. Suddenly it's clear why U159 wasn't getting set up properly. The Z80 was running corrupted code!

Just more text describing the difference between the bytes D3 and 53, as far as a Z80 is concerned. D3, which is 11010011 in binary, is written as "OUT (n),A" in Z80 assembly language and means (simplifying) "emit the A register's contents on I/O port n". OUT instructions are what the Z80 uses to send configuration commands to U159. 53 meanwhile is "LD D,E" and means "move contents of the E register into the D register", which is some data-shuffling internal to the Z80 that doesn't involve U159 or any other external device in any way.

100

The scene: A and I at the bench. A reads eavesdropped bytes from the logic analyser. I follow along in the code. We start finding differences. What should be C3 becomes 43 in the PERQ. 96 is 16. D3 is 53. Something is turning some high-bit binary 1s into 0s! *dramatic music, fade to commercial*

Double-spaced lines of text showing a hexadecimal representation of the program code that the Z80 should be running. In the space beneath the line, corrupted digits are noted. Extra info for you, the ALT-tag reader: there are also pipe characters that happen to mark "seams". The PERQ uploads program data into the Z80 in 31-byte chunks, and the pipe characters show the boundaries between those chunks.

100

But then at the end of the show, there was a breakthrough. To understand, it helps to know that (a) in the PERQ, I/O is handled by a second processor, a Z80 (b) the main processor has to download code to it to use the mouse, and (c) we know what that code should be.

The PERQ's EIO board, which handles all I/O for the PERQ except for the video display. It's a large circuit board with over 200 separate DIP ICs on it. Arrows point to the board's Z80 microprocessor as well as U159, the 8254 programmable interval timer that generates the clock pulses for the mouse.

100

I'm so grateful to @binarydinosaurs.co.uk for the chance to exhibit, but regret being so absorbed in the troubleshooting! I didn't socialise much and I feel a bit bad about it. Sorry to fellow exhibitors; I do really like seeing you all there and wish I had made more time to chat!

A closeup of my electronics workbench exhibit at the Centre for Computing History's Retro Computer Festival in November, 2024. The logic analyser is displaying a schematic (it's basically a Windows 2000 machine, after all). The oscilloscope is on but idle. The PERQ has its CPU board extending outward from its installed position, showing many dozens of chips. Directly in front, on the anti-static mat, illuminated by an overhead lamp, the "solder side" of the EIO board, the large circuit board that handles nearly all of the PERQ's I/O. Some thin, barely-visible "flying leads" have been tacked to locations on the board, providing electrical access to signals that will soon be measured on the logic analyser.

There was a whole exciting festival going on around me when I snapped this photo, but it was easy sometimes for just this scene to fill my entire field of attention.

Extra for ALT tag readers: at this stage we are pursuing a red herring, investigating whether incorrect address decoding logic is preventing the Z80 from sending signals to U159. This was not the problem...

340

BTW, this troubleshooting is happening in public at the wonderful Retro Computer Festival at the Centre for Computing History in Cambridge. Only have a broken computer? Exhibit yourself trying to fix it as a kind of live diorama, a bit like a blacksmith pounding on hot iron in a "living museum".

A person in a blue plaid shirt and beanie gazes at my electronics workbench exhibit at the Centre for Computing History's Retro Computer Festival in November, 2024. Arrayed on a table atop a blue anti-static mat: a logic analyser with mouse and keyboard, a multimeter, an oscilloscope, and of course the PERQ itself with most of its covers off. Atop the PERQ, a screen and keyboard plus a whiteboard proclaiming "Tom vs. PERQ" and the goal of fixing the mouse.

130

So let's find that pulse-maker. I've probably got the #1 spot in 2024 for Time Spent Looking at PERQ Schematics, so it's not hard to trace the signal back to p8-15 of www.bitsavers.org/pdf/perq/per... and U159 in particular, an Intel 8254 programmable interval timer...

An excerpted page from the PERQ schematics, with highlights pointing out U159, the Intel 8254 programmable interval timer, and the TABLET CLK clock signal that it generates. The skew away from the vertical is present in the scanned schematics document; it's not artistic license, in other words...

The mouse isn't a normal mouse, it's more like a graphics tablet. Its dedicated "mouse pad" is a circuit board that detects the mouse's location. To work, the mouse needs to receive steady signal pulses from the computer, but it wasn't getting those. Our problem lies elsewhere...

The PERQ's "mouse pad": a fairly empty-looking, square circuit board with a few clusters of DIP ICs at its upper edge. Elsewhere, a grid of circuit traces spreads across the otherwise-bare board. Current flowing through these traces generates magnetic fields that are detected by a coil inside the mouse.

100

Tom Stepleton @stepleton.bsky.social · 12/12/2024

About a month ago I fixed the mouse on my circa-1983 #PERQ 2T2 computer: a key feature for an early graphical workstation. It was a wild troubleshooting journey --- in the end, there was nothing wrong with the mouse itself. You might not believe what it was...

An electronics workbench with a PERQ 2T2 computer resting atop it: its system unit is not quite as large as the kind of refrigerator that fits beneath a desk. Most of its covers are removed. Highlighted in the image is the PERQ's three-button mouse. Text on the image says, "Issue: mouse not functioning, other peripherals seem to work ok".

180

I'm running low on tractor-feed paper. There's barely enough to make this year's greeting cards!

Close-up of an Apple Lisa computer's display as the LisaDraw program sends a festive card design with a Christmas tree to a dot-matrix printer. The Lisa's print dialogue is verbose. It tells the user that print spooling is available: that is, the user can dismiss the dialogue and do other work while printing continues. But it also points out that if the user does this and then at some point "hear[s] a whistle", they should reopen the print dialogue. (The Lisa's "whistle" is a rising bloopy sound that the Lisa makes if printer trouble occurs, like running out of paper. I only know early versions of the Lisa Office System to make this sound.)

0391

Tom Stepleton @stepleton.bsky.social · 25/11/2024

Fixing 40-year-old #PERQ computers eats notebooks! It was working, now it won't talk to the hard drive. After probing three interface boards on the bench, I now think the problem lies elsewhere...

A notebook with dense, tabular notes in fine writing is propped up by an oscilloscope resting atop a PERQ computer system unit. At the bottom, the PERQ's diagnostic numeric display reads 014, which means the computer couldn't boot from the hard disk.

150

Tom Stepleton @stepleton.bsky.social · 11/11/2024

Greetings new followers, I'm glad you're here! As a periodic reminder and despite my day job, this account remains essentially dedicated to OLD COMPUTERS and OLD COMPUTER ACCESSORIES. Stay tuned for a thrilling post about dropout as it pertains inadvertently to executable code for a Z80 processor.

The cartoon character Hank Hill from the TV show King of the Hill gives a wary eye to a desktop PC with a CRT monitor perched atop.

040

Screenshots like these show that RasterOp still isn’t copying data perfectly. That’s enough to crash the OS! And that’s where things stand now this Halloween: troubleshoot and fix this #PERQ bug. Then the next one. Then the one after that. It’s a really fun puzzle! I’m hot on the trail. Stay tuned!

Screenshot detail of text from POS attempting to boot. Letters no longer have random vertical bars, but pixels and other parts are missing. The complaint about System.6.A.ZBoot is not unusual or relevant here.

Screenshot detail of text after the PERQ attempts to scroll the display image by one line. Each row of characters is a different letter with some of the same maladies as in the other image, but now entire vertical lines of the display are clearly missing.

100

I tracked the problem down to a FIFO chip, which sets up a queue for bit patterns waiting to be processed. Replacing it got POS to boot to 999! But having booted, POS and PNX both are still buggy or frozen. There’s still trouble…

A fancy chip testing gizmo, looking like a stack of circuit boards, tests a 74S225 FIFO IC. The display says “78225 Test failed. Memory failed. Again?” The smoking gun…

But at least there is this. An oblong black panel a bit bigger than a deck of cards. There is a big red power switch at right. At bottom left is a black button with the word RESET printed above it; at top left, beneath the word DIAGNOSTIC, is a three-digit 7-segment LED display that reads 999.

100

It’s video!! But something is obviously wrong. The patterns suggested a flaw in the dedicated “RasterOp” hardware the #PERQ uses to move blocks of memory, both for video and for internal OS stuff. And at 231, POS is using RasterOp!

A modern(ish) LCD display shows a white screen with text on it from the initial boot of PNX announcing version numbers and sundry other information. Blank areas of the screen have vertical “pinstripes” made of tiny dots, and letters are corrupted with things like random vertical lines.

Close-up of a matrix of alphabet letters on the display. Many letters are perturbed by vertical lines, missing pixels, and other kinds of corruption.

100

My #PERQ monitor needs repair for a part that emits 20000 volts of DC power. I’m not ready to work on that yet! So I went on a new side quest: convert PERQ video to VGA. This adapter took some work, but it’s doing it. There’s a 70 MHz signal passing through that mess! Schematic eventually…

A pair of solderless breadboards holds a collection of jumper wires and passive components along with cable ends that extend out of shot and a pair of DIP ICs (a 74LS123 and a 74S241 for anyone curious).

100

The original #PERQ OS is “POS”, but in the UK, a Unix called “PNX” (yes, odd names) was preferred. When PNX boots, it only counts up to 255. Imagine my astonishment when it got that far on the first try: my PERQ had booted an operating system! And my oscilloscope found something new: a video signal…

A yellow trace on an oscilloscope display has the hallmark sign of a monochrome video signal: an oscillation between mostly high and occasionally low signals with barcode-like spikes on the high part. The signal has lots of overshoot which is just due to my poor probe grounding; not really relevant here.

100