Original URL: http://www.theregister.co.uk/2014/02/19/archaeologic_sinclair_research_might_have_beens/

Loki, LC3 and Pandora: The great Sinclair might-have-beens

The ‘Super-Spectrum Entertainment Engine’ and more

By Tony Smith

Posted in Vintage, 19th February 2014 10:02 GMT

Archaeologic We all know about the ZX80, the ZX81, the ZX Spectrum and its successors, and the QL. But these weren’t the only microcomputers Sinclair Research worked on during its brief life between 1979, when it emerged as the renamed Science of Cambridge, and 1986, the year its brand and products were bought by Amstrad and it was shut down for good.

Clive Sinclair had prompted the development of the ZX80, which was largely an evolution of Science of Cambridge’s MK14 microprocessor learning kit, a board computer based on the National Semiconductor SC/MP 2 chip. The ZX80 was a essentially a MK14 based on a Z80A processor and with a fitted cassette port and TV modulator. The ZX81 was an logical upgrade of the ZX80.

At the time of the ZX81’s release, the Spectrum was in development as the ZX82 and was again an upgraded version of its predecessor, this time with colour graphics and an updated Basic. Even then, though, there were many ideas being bandied about as to what kind of micros the ZX83 and the inevitable ZX84 might be.

QL+

The QL revamped for the wafer-scale era
Source: Rick Dickinson

The ZX83 would eventually become the QL, but not before the codename had been attributed to a number of possible machines that Sinclair managers and engineers thought they might bring to market in 1983. One was a mobile micro, very likely inspired by the March 1981 launch of the Osborne 1 portable computer, and this notion of a portable computer for business users would tax various Sinclair engineers for the rest of the company’s life. It was briefly attached to the ZX84 codename.

Clive Sinclair’s desire to build a mobile computer that, crucially, could operate as the user’s sole machine and not as an adjunct to a desktop would eventually be realised through a product named "Pandora", and development work was proceeding on the device when Alan Sugar formally took over in April 1986.

Sugar killed it, but in doing so allowed Sinclair engineers soon employed elsewhere to move the project on without being hindered by the baggage that working at Sinclair had involved: specifically the need to incorporate the company’s flat-tube screen technology. The result was the LCD-based Cambridge Computing Z88, a much better product than Pandora ever came close to becoming. The Z88 was “one of [Clive Sinclair’s] best products”, says Perran Newman, who led the Pandora team after working on the C5 electric vehicle.

Sinclair’s failure to build a working mobile micro ensured Sir Clive was never credited with developing Britain’s first laptop computer. That honour goes to Thorn EMI, which built and released the battery-powered Liberator in the year between the summers of 1984 and 1985, while Sinclair Research was completing the QL and, later, developing the Spectrum 128.

Sinclair also missed out on producing the first British games console, as we’d understand the term today, when a project dubbed the Low-Cost Colour Computer - aka LC3 - fizzled out when its two key developers got working on other, more pressing projects.

The QL itself might have been the means by which Sinclair released the first British 16-bit home computer, if the business machine’s development team had been given extra time to adapt their hardware to become a platform for future Spectrum-branded machines. They tried, but under the tight development deadline they were working to, they failed and the idea was dropped.

The QL was never the great success Sinclair hoped it might be, but it might have been had more time been granted to its development and features considered for a later, second-generation model, codenamed "Tyche". These included a 3.5-inch floppy drive in place of the Microdrives, a full 68000 processor instead of the QL’s cut-down 68008 and a mouse-driven graphical user interface. Tyche would have provided all of those had it got further than the drawing board. Its GUI would have been Digital Research’s Graphical Environment Manager, better known as Gem.

Gem was ported to the 68000 for Atari’s ST during the last quarter of 1984 and the gaming power of the ST but more particularly of its 68000-based rival, Commodore’s Amiga, prompted two Sinclair engineers to propose an Sinclair alternative, codenamed "Loki". Like Pandora and Tyche, it never survived the Amstrad takeover. Ditto projects like Janus and Proteus, or even left-field options, such as "Dylan", conceived as a low-cost PC clone that might succeed where the QL had failed.

This, then, is the story of some of these great Sinclair "might have beens".

LC3: Low-Cost Colour Computer

Martin Brennan joined Sinclair Research in 1982 specifically to investigate and develop new artificial intelligence algorithms, but he soon became wrapped up in broader efforts centred on bringing Sinclair’s promised Microdrives to market. Brennan eventually engineered the ZX Interface 1 through which the Microdrives would connect to the Spectrum.

But Sinclair Research wasn’t the kind of place where engineers were told to work on specific projects to the exclusion of all else. The company hired young, brainy people and gave them space for their natural creativity to flourish.

One of Martin’s ideas: a really cheap machine that could bring punters in through its gaming capabilities but provide scope for expansion so they might build it up into a full computer in due course. Designed to be inexpensive, it would perhaps replace the monochrome ZX81 and certainly protect the Spectrum from Far Eastern rivals feared to be on a mission to undercut Britain’s market leader.

There was a template for such an expandable system: the Mattel Intellivision. Launched in America in 1979 but not made available in the UK until 1982, the Intellivision had been developed as a games console to rival Atari’s massively popular VCS 2600. Unlike the Atari box, the Intellivision was created with the idea that its console hub could, with the addition of a bionic keyboard add-on, become a true personal computer.

The Intellivision Keyboard Component, as the add-on was called, provided extra memory, a cassette interface and a 6502 processor. Essentially it was a computer in its own right, using the console as a TV-output peripheral. It proved hard to engineer, and was late to market.

Sinclair Microdrive

"Simple black box with a ROM slot" - the inspiration for the LC3’s look?

LC3 would avoid these problems by starting out as a computer. The notion was to fit it with a Z80 processor, a little bit of memory and enough ROM for the operating system, which would load games in off ROM cartridges. It was envisaged as a small black box with a ROM slot in the front, says Stephen Berry, the Sinclair software engineer Martin Brennan got working on the LC3’s OS and Basic interpreter while he worked up a prototype motherboard. It’s not hard to imagine the visual inspiration: the ZX Microdrive itself.

Not that LC3 ever got as far as an industrial design. Breadboards were built with TTL to provide Berry with hardware on which to run his code, but the project doesn’t appear to have progressed any further than that. Berry says he was shifted to more pressing matters: wrangling the documentation for the QL, for one, a task that took him many months of full days and weekends of work to complete.

There was no time for the LC3 until all that was done, and by the time the QL manuals had been completed, around June or July of 1984, the LC3’s time had passed.

“I wasn’t at Sinclair for the LC3 but was for Loki, and LC3 only ever got mentioned in passing,” recalled Quondam Sinclair programmer and occasional Register journalist Rupert Goodwins in 2002. “I remember seeing a ULA design for it, but it was only pulled out of a cupboard for nostalgia reasons.”

More than 30 years on from those days, details of the proposed specification, beyond the basic notion of implementing very little memory in order to keep the price as low as possible, remain vague. Stephen does recall some of the work he did on the software. The operating system, he says, was built out of a “little kernel” capable of multi-tasking. He may not have completed support for windowing - really just segmenting the screen into arbitrary independent areas - but he certainly built in the hooks for the mechanism behind such a system to be added later. “The design had to overcome the twin restrictions of low processor power and limited amounts of memory,” he says.

“The LC3 OS was indeed written, in skeletal form,” said Goodwins in 2002. “[It had] multitasking, windowing and was potentially very fast - a complete context switch in around a hundred T states, ISTR. I don’t remember it being resurrected for Loki, but some of it may have been unearthed for Pandora.” Very possible, that, since Stephen Berry later designed the Pandora laptop’s system software.

One way of reducing the cost - and dealing with a low memory capacity - was to fake up a bitmapped screen by making use of a character-based display, Berry says. Text VDUs store a small grid of Ascii values which are used to look up appropriate entries in a character matrix table accessed by the display processor. Change those character matrices - essentially treat them all as variable user-definable graphics - and you can replicate a bitmap screen and use less memory. Or so the theory went.

Could the LC3 ever have come to market? That would have hinged on the Spectrum: would it have become cheap enough to produce for the price to be cut to around £50 or less, what Sinclair was selling the ZX81 for. Sinclair Research’s managing director, Nigel Searle, was, according to several ex-Sinclair staff, entirely unwilling to risk the company’s cash cow. That, plus the fact that an attempt to sell ROM cartridges for the Spectrum was a woeful failure - why buy £20 cartridges when tapes could be had for less than a fiver? - show that the economics would have to be very advantageous if the LC3 was to succeed.

Whether Brennan and Berry got an official go-ahead to begin development of the LC3 is unclear. Quite apart from the freedom Sinclair Research engineers enjoyed, the company was never one for formal approval processes and sign-offs. A nod from Clive Sinclair, Nigel Searle or even Jim Westwood would have been considered sufficient. However, unlike some projects, the LC3 did get beyond the proposal stage, and was even considered by management when the company’s leaders met in November 1983 to ponder the firm’s activities in 1984 and beyond. Since no serious Asian threats had emerged, and Spectrum production was getting cheaper, a line was drawn through the LC3.

“I still think the LC3 was the best approach,” said Sir Clive in a 1986 post-sale interview with Sinclair User. “It was a complete colour computer, entirely on two chips - very much like the Japanese Nintendo machine but many years beforehand.”

Loki

If it had ever been produced, Loki - just a codename, not a go-to-market handle - could have been Sinclair’s post-Spectrum 128, next-generation home computer. But it was never to be: it was an early casualty of Amstrad’s acquisition of the Sinclair brand and Sinclair Research’s IP. “Loki didn't ever really get anywhere,” said former Sinclair engineer Martin Brennan a few years ago. “It had a name but there weren’t any drawings as such. It was a concept.”

His collaborator on the project, such as it was, was fellow Sinclair engineer John Mathieson, who said separately: “The idea was never more than a paper design.”

The team - fellow Sinclair techie Ben Cheese made a third member - began thinking about a new home machine after completing work on the Spectrum 128, a version of Sinclair’s stalwart that had been created for Spanish distributor Investrónica. In 1985, Spanish government decided to tax computers with 64KB or less memory, so Investrónica asked Sinclair to come up with a 128KB version. It was the same tax regime that prompted Amstrad to devise the CPC 472, a slightly more capacious version of the CPC 464 for the Spanish market.

Brennan took the opportunity to revamp the machine, adding not just more memory but also the popular General Instruments AY-3-8912 chip for superior, three-channel sound routed out through the TV speaker, plus MIDI compatibility, a proper RS-232 serial port and RGB monitor output. The Basic was overhauled and given a new editor.

Loki proposal

Loki proposed

The Spectrum 128 was first demo’d in public in September 1985 on its release in Spain, but didn’t go on sale in the UK until January 1986. By this point, Brennan, Cheese and Mathieson were wondering what to do next.

“There remained a core group of people within Sinclair who wanted to continue selling cheap home computers with dual educational and game playing capabilities - machines that the kids would want and that parents would be willing to pay for. We were impressed when the Amiga came out [in 1985] because as a computer it seemed to fit that role perfectly,” said Mathieson. “We believed that the Amiga was over engineered. It could animate with sprites and with a blitter, for example, but why have both? We figured that if you could implement a system that could animate the screen then you only needed one set of hardware to do it. So we proposed a computer called ‘Loki’ as a Spectrum successor.”

Ironically, the QL was at one point mooted as the foundation for a second-generation Spectrum. It would be based on the same Motorola 68008 processor as the QL, and use the QL’s two custom chips, which were modified during development to accommodate this new role. Yet the notion fell out of favour, and ultimately Sinclair would choose instead to give the existing Spectrum a new, QL-style case and call it the Spectrum Plus.

Had the Loki specifications been defined to any extent? Sinclair User magazine claimed in June 1986 that Loki would feature a 7MHz Z80H processor, 128KB of RAM and two custom chips: one for sound, the other for graphics. The latter would deliver three different modes: 512 x 256 pixels with 16 colours, 256 x 212 with 256 colours, or 256 x 212 with 64 colours. The video chip, dubbed “Rasterop” by the magazine, would have sprite animation and collision hardware built in.

Backward compatibility would be provided for the original 48KB Spectrum thanks to a copy of its ROM, and the Z80H able to flip from 7MHz to 3.5MHz - presumably why this chip, normally rated at 8MHz, was set to run at the lower speed.

Amiga in its sights

Sinclair User described the machine as Amstrad’s next home computer, but the only facts it got correct were the memory size and the presence of CPC 464-style integrated cassette deck. The Spectrum Plus 2, which Amstrad released in 1986 - according to Lord Sugar’s 2010 autobiography, he had his people working on the design before he’d even agreed to buy Sinclair - was a Spectrum 128 with a built in tape player. The magazine’s report now appears as a mix of a glance at the Loki proposal write-up, extrapolations made from it and assumptions based on what Amstrad had already done with the CPC series.

Brennan, Cheese and Mathieson fleshed out their notions with a broad "wish list" specification - something Sinclair Research could achieve if the project got the go-ahead - and this is proposal is what Sinclair User saw and the basis for Sir Clive’s recollection. But Loki never made the next step to become a formal design endeavour, a process which might well have changed this back-of-an-envelope sketch further still.

The proposal describes Loki as a “Super-Spectrum Entertainment Engine” that will replace the Spectrum 128 and sell for “less than £200” yet deliver Amiga-level sound and graphics performance - “the capabilities of a commercial sythesizer... this facility includes a MIDI interface”; “cartoon quality graphics”; “support the addition of CD or LaserVision to upgrade these effects”; “a facility to allow the machine to be locked to an external video source... this will allow the machine to be used as a tool for home video use (captions, etc)” - though the Commodore machine isn’t named directly.

Other features: an unspecified Z80 running at 7MHz, 128KB of Ram, expandable to 1MB and split 50:50 between program and system memory, and the video buffer; display resolutions of 256 x 212 with eight bits per pixel, or 512 x 212 with 4 bits per pixel; “Rasterop hardware performing screen animation, scrolling and line draw”; Spectrum 48/128 software compatibility mode - but “128 sound may be a problem”, the front page warns.

Loki preliminary sketch

Loki appears to have at least made it to sketch form
Source: Rick Dickinson

It’s a "what we’d like to try to build" specification rather than a description of a computer that can be built. At this stage in the machine’s development, it would taken 12 months to produce a working prototype at least. Yet it got so far as to end up on Rick Dickinson’s desk, from which he worked up a number of sketches. “The Loki was directed absolutely at the games market - unlike any other Sinclair computer,” says Rick.

“The main box housed a massive CPU. Peripherals, including RAM increases could be added until they dropped off the desk - everything you could ever need could just be plugged on. Low entry cost and utterly expandable.” That actually sounds more like the LC3, so it’s entirely possible elements of the older idea were revived for Loki, especially since both came from the fertile mind of Martin Brennan.

If that makes it seem that Loki was more than the ‘only on paper’ notion that Brennan and Mathieson later said it was, perhaps there’s something in that view.

In 1986, after the Amstrad acquisition of Sinclair tech, Clive Sinclair claimed that “Loki... chip designs were under way but there was still quite a lot of work to be done on [detailed] specifications. Basically, we thought if we could do an Amiga for under £200 with decent graphics there’d be a big market. I wasn’t actually convinced of that. There was a danger of running into the same trap as the QL of not giving people what they actually needed”.

Other erstwhile Sinclair staff say the same. “Lots of Loki got breadboarded before Sinclair’s computer bits got flogged to Amstrad, but there wasn’t anything there that could count as an ongoing, coherent project,” remembered Rupert Goodwins a few years ago. “I don’t think such issues as palette switching were resolved, for example. Brennan and Mathieson were notoriously opinionated, and declared that palette switching wasn’t needed as ‘it was only ever good for demos’. Everyone else demurred... But I did see some video capture stuff, the disk drive hardware and some other miscellaneous graphics breadboards.”

Loki never got the green light. After Sinclair shut its doors, Brennan, Cheese and Mathieson set up Flare Technology to turn their concept into a real design, the Flare One. The new machine wasn’t Loki. It didn’t - and couldn’t - contain any Amstrad IP, such as Sinclair Basic or the QL’s SuperBasic.

Rather, Flare One was a machine broadly descended from Loki in the sense that its developers’ heads were at that point full of ideas about low-cost Amiga rivals. The participants’ entirely reasonable desire to ensure no one ever came to conflate the Flare One and Loki would undoubtedly have encouraged them to play down the Sinclair proposal as just a “paper design”.

Pandora

Sir Clive Sinclair had wanted to build a portable computer for years, and Sinclair Research had engineers looking into the notion of an untethered machine as far back as 1981. It even steered his thinking about the Sinclair Radionics micro, which (eventually) became the NewBrain. Sinclair would eventually release a portable, Cambridge Computer’s Z88, in 1988 - essentially his take on one of the earliest mobile computers, the slab-like Epson HX-20.

The Pandora was entirely different: a clamshell machine with not an LCD screen like the HX-20 or the Z88 but one of Sinclair’s two-inch flat cathode ray tubes. Based on the Z80 processor, Pandora would be compatible with the Spectrum, early rumours leaking out of the company indicated, and would include a Microdrive. It would, like the QL, come bundled with productivity software, stored on ROM.

It would, in Clive Sinclair’s mind, entirely replace the owner’s current computer: “We have to come up with a portable which people will be happy to use as their only machine - so that they won’t have need of any other,” he said in an interview with Popular Computing Weekly in February 1985. “Swapping files from one machine to another is just not on - the data has to be there all the time.”

It would, he hoped, be a true first for Sinclair Research: a computer that was not only genuinely portable – as the HX-20 was but the Osborne 1 certainly was not – but also usable. It had a decent-sized display, which, again, machines like the Epson and the Osborne lacked. It would ensure that the Sinclair reputation for innovation, bruised by the QL, would be seen to be as healthy and as active as it had once been.

Pandora

One of the first Pandora mock-ups, built to demonstrate the optics, not as a working computer
Source: Rick Dickinson

“The resultant virtual image has a residual curve which the eye interprets as a 3D surface reminiscent of a Cinemascope screen,” Perran Newman is said to have written in a Pandora progress report in January 1986.

That sounds impressive. The Pandora’s display was a Sinclair flat tube fitted into the lid. It was, of course, far too small to make a usable text display on its own, so Sinclair engineers contrived an arrangement of mirrors and lenses which folded out as the computer’s lid was lifted. The notion, says Stephen Berry, who worked on the computer’s software, was that, the tube’s image was reflected and magnified into a “floating” image placed where a modern laptop’s screen would be.

Stephen says one problem was that in order to see the virtual image, the user needed to focus his or her eyes at infinity, even though the screen appeared close up. It was like trying to see those 1980s ‘magic’ pictures, which embedded pictures inside repeating patterns. Only by trying to focus your eyes ‘behind’ the picture could you see the hidden object. It wasn’t easy to do; distracted, your eyes didn’t automatically focus on the image but rather on the poster itself. The Pandora screen was much the same.

Rupert Goodwins said of it: “I did the screen driving software and designed some fonts – the screen was so weird that if you didn’t have fonts explicitly matched to it, the chances of reading stuff was minimal – all tested on an old Zenith green monitor I’d jimmied to match the aspect ratio of the final screen.”

Pandora’s industrial design when through a number of iterations, first as sketches made by Sinclair’s designer, Rick Dickinson, which led to a final design that was modelled as a physical reproduction. This was subsequently revised, at least once if not twice.

Box clever

“Only a couple of prototypes got made,” recalled Goodwins. “The circuit never got further than a collection of breadboards, and the models were there to test the optics and screen details. People were understandably nervous about things like the extra-hight tension supply to the overdriven tube, and even more nervous about whether the whole thing would either work or be manufacturable.

“I can’t remember too much about the video modes, although I think there was one that had 64 x 24 characters. That was going to be the main one, because Pandora was seen as a productivity tool rather than a games box. It also had a 6 or 8MHz Z80 - I think it was the Hitachi variant with some extra instructions - and everything else bar the memory in one ULA.

“I also seem to remember that it wasn’t going to have Microdrives: the flat screen was seen as quite enough nonsense to be getting on with, thank you very much, Clive.

“Journalist Guy Kewney asked Alan Sugar whether he’d bought the rights to Pandora. ‘Have you seen it?’ asked Sugar. ‘Yes,’ said Guy. ‘Well then.’ said Sugar.”

Amstrad would, just a few years, release the Notepad NC100, a rather less sophisticated Z88-alike.

Pandora

A later Pandora design revision
Source: Rick Dickinson

Clive Sinclair never lost his vision for a workable portable computer, and almost immediately assembled a team to work on a machine without Spectrum compatibility. Goodwins was asked to participate, but chose to work for Amstrad instead; he was one of very few Sinclair people to be hired by Sugar’s firm. No Spectrum support, then for the Z88, and no flat tube, either. Sir Clive was persuaded that perhaps LCD - a technology he detested, says Stephen Berry, because of the of carcinogenic chemicals used in its production - had to be adopted after all.

The architecture remained as it had in Pandora: Z80 CPU plus ROM plus RAM plus one ULA. Ditto the use of disposable rather than rechargeable batteries, Sinclair long having averred that you could always buy fresh batteries but rarely find a power outlet to feed a charger.

The ROM contained a new OS, OZ, plus PipeDream, an integrated word processing, spreadsheet and database program. It also contained a version of... BBC Basic. At long last, Sir Clive had laid a very specific ghost to rest.

QL 2.0: Enigma, Tyche, Proteus and Janus

During the latter part of 1985, Sinclair Research seems to have been awash with ideas about how the QL line could be extended - or, given the QL’s blemished reputation, caused by announcing the original far too early - another brand but one also aimed at a more sophisticated audience than the Spectrum.

"Enigma" emerged in the Autumn of 1985: a system with one or two 3.5-inch floppy disks in a system unit which would come with a bundled colour monitor, printer and a separate keyboard. It would, claimed Your Computer, feature 1MB of RAM and bundle versions of the Psion-produced QL productivity applications in a “full Window, Icon, Mouse environment”, perhaps Digital Research’s Gem, which had been released in the UK the previous April.

Enigma in Your Computer

Your Computer lauds ‘Enigma‘

In many ways, Enigma can be seen as Sinclair’s response to the PCW 8256, Amstrad’s Z80-based "word processor" launched in September 1985 and quickly becoming hugely popular. The QL’s hardware designer, David Karlin, had wanted the QL to be a machine along the lines of the QL. Perhaps Enigma was his suggestion for a way for Sinclair to release not only a PCW rival but a machine closer to the one he’d wanted to make in the first place.

Enigma was rumoured to be set to ship toward the end of the first half of 1986. Was it ever a runner? It was rumoured that the creator of the Archimedes casing had designed the box for Sinclair but sold the design to Acorn when Sinclair changed its mind. If not Enigma, then the case might have found a home as an IBM PC clone, which company executives were also pondering as the money began to run out during the latter part of 1985 and the situation at Sinclair Research began to look increasingly untenable. Stephen Berry says the PC clone was "Dylan".

"Tyche" is another QL successor and perhaps has more weight than Enigma because it eventually made it outside Sinclair, at least in the form of an updated ROM image developed by Jonathan Oakley and made public many years later. Thanks to the Amstrad takeover, the Tyche code became Sinclair’s final version of QDOS, though it was never released at the time.

It is said to have been prepared not for the QL but for a successor that has been described by QDOS creator Tony Tebby - though he had been long gone from Sinclair by then - in terms that sound a lot like the Enigma description.

“There was a program internally to do a QL-type machine, that’s to say based on the same chips, but with disc drives,” said Sir Clive in 1986. “It wasn’t really a QL, though, and it was a much more expensive machine.” That sounds a lot like Enigma/Tyche.

QL+

How a second- or third-generation QL might have looked
Source: Rick Dickinson

Janus remains something of a mystery. The god Janus had two faces, to look two ways at once, and it’s tempting to see that reflected in a concept machine able to succeed in both the business and the games markets - perhaps the way it had once been thought that the QL might go. Equally, and perhaps more likely, it reflects looking both forward and backward: the latter in its support for QL technology, but forward to a new world of computing based on Sinclair’s wafer-scale integration ideas.

Certainly designer Rick Dickinson sketched and produced mock-ups of a number of wafer-scale machines, some in upright towers, others in squat, domed units. Many are marked “QL+”. Was Janus and the QL+ one and the same thing, and if so were they ever more than a concept for a future? We shall probably never know. ®

The author invites anyone with anecdotal or documentary evidence of these and/or other Sinclair "might have beens" to drop him a line.