Original URL: http://www.theregister.co.uk/2012/09/21/jupiter_cantab_jupiter_ace_is_30_years_old/

The Jupiter Ace is 30

Two Spectrum stars go Forth

By Tony Smith

Posted in Hardware, 21st September 2012 07:00 GMT

Archaeologic The Basic programming language, although present in many different dialects, was the lingua franca of early 1980s home computers. One machine dared to be different: Jupiter Cantab's Jupiter Ace, a small unit that spoke Forth. It first went on sale 30 years tomorrow.

Forth was conceived by Charles Moore, a computer scientist employed by the US National Radio Astronomy Observatory. Moore's work on the NRAO's computer system led him to devise the language for programming telescope tracking control systems.

Perceiving a wider benefit, Moore and fellow NRAO employee Elizabeth Rather took Forth to a bigger audience by launching it as a commercial product in 1973. Forth was soon ported to a wide array of processor platforms, including Zilog's Z80 family, the basis for many an early 1980s home computer, including Sinclair Research's ZX80, ZX81 and ZX Spectrum.

Jupiter Ace

By Jupiter: the Ace
Source: Wikimedia

Enter two of the minds behind the ZX series. Steve Vickers was a software writer with a mathematics PhD. In the early 1980s, he was employed by Nine Tiles Information Handling, a company contracted by Sinclair to develop the ZX80's implementation of Basic. Vickers had been brought on board to update Sinclair Basic for the ZX81, and to write the machine's Basic programming manual.

In due course, when work commenced on the Spectrum, Vickers was tasked with bringing Sinclair Basic into the colour era. However, his desire to rewrite the language interpreter code afresh was frustrated by the tight development schedule Sinclair had imposed.

Vickers' work with Sinclair naturally brought him into close contact with Richard Altwasser, the hardware engineer designing the Spectrum's motherboard. A Sinclair Research staffer, Altwasser had joined the company in September 1980 and had been loosely involved in the development of the ZX81. The Spectrum, however, was his baby.

Source: Jupiter Ace Resource Site

Today, Altwasser remembers the two discussing the concept of setting up on their own while they still involved with Sinclair Research. "But during the design stage for the Spectrum we were incredibly busy. There was no chance at all to do any homework - certainly no serious work - about bringing a new product to market.

"But there was a time after the product design was finished, when [the Spectrum] went into manufacturing, where our work was less intensive than it had been. There was an opportunity then to spend some evenings thinking a little bit more seriously about other projects," he says.

Steve Vickers remembers Altwasser working on the basic board layout for a new micro around this time. This extra-curricular project led to the hardware engineer to ask the software engineer to join him in the development of a product of their own. The software guy readily accepted.

"The thought was, Clive Sinclair was making all this money out of us, why shouldn't we just make our own computer and make the money for ourselves," says Vickers.

The Spectrum was launched in April 1982. Not long after, Altwasser and Vickers quit their respective employers to form their new venture, initially called Rainbow Computing.

Steve Vickers and Richard Altwasser in 1982

Cantab creators: Steve Vickers (left) and Richard Altwasser in 1982

"We felt there was an opportunity, a gap in the market for a more programmer-oriented kind of product - a more technical, niche product," recalls Altwasser. "We saw the advantage in using the Forth programming language, and we thought we could differentiate ourselves this way."

Rapid development

Forth was brought to Vickers' attention by an acquaintance, Chris Paterson, who worked at the PA Technology Centre, located just outside Cambridge and, coincidentally, where the first Dragon 32 hardware had been developed during the latter part of 1981. Pattison convinced Vickers of Forth's advantages - near assembly level speed, custom commands and a low memory footprint - and Vickers, after pondering whether Forth could be used as the native language of a home computer, pitched the technology to Altwasser.

Altwasser and Vickers worked quickly, Altwasser draughting the motherboard layout by hand, Vickers coding the machine's Bios, and Forth editor and compiler in Z80 assembly language. Other CPUs were very briefly considered and rapidly rejected, says Altwasser. Since Vickers already had Z80A assembly programming off pat, sticking with the Zilog chip meant no time lost while he learned a new architecture.

Jupiter Ace issue 1 schematic

The Ace Issue 1's schematic
Source: Jupiter Ace Resource Site

Complex custom chips were dismissed too as these would have increased the machine's cost and hindered the developers' desire to get the product to market fast.

"It also meant we could build and test prototypes very quickly," recalls Altwasser. "The hardware design was pretty straightforward. Much more time was needed for the software design and writing the manual, which were Steve's areas."

Developing Forth for a small home micro was not just a matter of writing an editor and a compiler, recalls Vickers. It also required a little deviation from the normal Forth methodology.

"Forth as it existed then had an edit-compile cycle. You edit a program, store it on disk, read it in, compile it and run it. That wasn't quite going to fit the model we had in mind where for something like the ZX81 you had just a tape [for storage] and you want to it as little as possible, and you want to make the process as interactive as possible - users key in instructions and they see the results on the screen immediately - like you did on a ZX81 or Spectrum.

"So one of things I thought was going to be necessary was an in-place editor, which is different from the standard Forth model. We also wanted floating-point arithmetic, something a lot of Forth programmers felt was unnecessary - they're wrong, if you ask me. I knew how to implement that from Spectrum, so that wasn't a problem to do, and we stuck that in as well," he adds. "There are novel features in my Forth."

By August 1982, Vickers and Altwasser were ready to announce the product, called the Ace in honour of the Pilot Ace - 'Ace' for 'Automatic Computing Engine'. The Pilot Ace was developed in the 1950s by the UK's National Physical Laboratory, where Vickers' physicist father worked and to which Vickers junior, as a young boy, would be taken to marvel at the equipment it had at its disposal.

Forth code for the Ace by David F Corner

Sample code: part of Forth program Antarctica by David F Corner, from Personal Computer World August 1983

Vickers stated his debt to the Pilot Ace developers in his Forth programming manual, bundled with the new computer: "We [Jupiter] are simply the beneficiaries of 32 years of development that invented the printed circuit board, the transistor and then methods of packing thousands of transistors onto one small silicon chip; and in the process transformed computers [like the £30,000-plus, full-room Pilot ACE-based English Electric DEUCE (Digital Electronic Universal Computing Engine)] into machines for everyone."

To further cement this notion of standing on the shoulder of giants, the Rainbow Computing name was dropped in favour of Jupiter Cantab, to fly the flag for the company's home town, Cambridge.

Vickers recalls his work was essentially just a matter of writing the software needed to control the already defined hardware. "There weren't many things that still need to be tested on the hardware. I remember the cassette deck interface caused some problems and that needed some fine tuning during the development process. We had some embarrassing moments in exhibitions where it just didn't work properly."

While Vickers completed the software, Altwasser turned to establishing the manufacturing operation: choosing component suppliers, getting the parts they shipped to the firm handling the assembly.

Source: The Jupiter Ace Resource Site

"We were able to negotiate with one major supplier of a lot of the components a deal to supply as and when we needed it," he recalls. Custom parts had to be ordered in bulk, but assembly was largely handled on an 'as required' basis, avoiding the need to stockpile too many unsold units, he remembers.

Altwasser and Vickers were not over-ambitious. They forecast initial sales would be satisfied by a production run of 1000 units a month, to be sold through mail order, as the ZX80 had been.

"It will be popular in the educational field because Forth is an easier language for children to learn," Altwasser told Popular Computing Weekly at the time of the Ace's launch.

ZX81 killer in a post-Spectrum age?

Journalists unsurprisingly homed in on the new computer's unorthodox language. In response, Altwasser told them: "We chose Forth in preference to Basic because of its speed and flexibility. In Forth you can define your own functions and tailor your program exactly to your needs - that's why it is so exciting. Forth is also so much faster - it's ten times faster than Basic."

Ten times faster despite running on the same Z80A processor that was found in the Spectrum and being clocked to 3.25MHz, eight per cent lower than the Spectrum's 3.5GHz. Benchmark tests carried out by reviewers indeed showed a significant speed lead.

The Ace had just 3KB of Ram - more than the ZX81 but a lot less than the Spectrum. But then it didn't need so much. Forth is a language that makes for compact code - functions are compiled at runtime - and can happily operate in a lot less space than Basic can.

Jupiter Ace advert

Promoting the Ace
Click for full-size image

Still, that didn't help the Ace stand out in a market where 16-32KB of memory had become the norm. Neither did its use of monochrome graphics - an unusual design decision given the developers' early keenness, recalled by Vickers, to pitch partly the Ace as a games machine. As it turned out, although they had written games in the past, neither Vickers nor Altwasser found they had the time to produce the number of titles needed to get such a market rolling.

Unlike the ZX81, the Ace could generate sounds and present lower-case characters. It also had, albeit a crude one, a keyboard with moving keys in place of the ZX81's touch-sensitive deck, though as Personal Computer World reviewer Mike Curtis noted: "Each key needs a firm push in the centre otherwise it is liable to tip sideways and not make contact."

It's tempting to look at the Ace's specifications and see them very much as exceeding those of the ZX81 but not of Sinclair's follow-up product. Should Altwasser have given the Ace colour graphics, souped up sound and such?

"A lot of people looked at what we did and said 'you should have gone with Basic, you should have gone with high-resolution colour graphics', but the with those kind of design choices we would just have been making a 'me too' Sinclair product and that wasn't our vision," says Altwasser today. "That wasn't what we wanted to do at all."

"There were doubts," says Vickers of the initial design specification, "but we didn't know what to do about them so we said, 'look it's a good machine, let's make it'."

But they couldn't ignore the rest of the market. At launch, Altwasser and Vickers were touting a 48KB expansion pack, an add-on colour board and a plug-in printer port. The memory and the printer link, it was promised in August 1982, would be on sale by Christmas.

"We are not trying to make everything for three farthings," Altwasser told Your Computer at the time, despite the fact that the machine was packed in a low-cost, vacuum-formed plastic casing - the Ace's "worst feature", wrote Your Computer reviewer Bill Bennett.

Inside the Jupiter Ace

Inside the Jupiter Ace
Source: Jupiter Ace Resource Site

But with the need to control costs and ensure a quick time to market, an injection-moulded case would have to wait. "An ABS case would have looked much better but would have have taken a considerably longer period to tool up for," says Altwasser now.

The Jupiter Ace formally went on sale on 22 September 1982, priced at £89.95 including sales tax and shipping. The firm operated out of Altwasser's suburban house just off the Huntingdon Road in Cambridge, initially in a spare room, though when that was required for Altwasser's new-born son, the Jupiter operation relocated to the garage. Quickly outgrowing the garage, they soon moved to an office building in Cambridge's Bateman Street, just behind the University's botanic gardens.

Forth was central to the Jupiter Ace's design and it was no less key to the company's sales pitch.

"Comparing Forth to Basic is like comparing a Gothic cathedral to a mud hut," Vickers said at the time. "It's about time someone got away from Basic."

He told Micro Computer Weekly: "Forth is entirely interactive. You type in one word and the system automatically looks it up in a dictionary of definitions and then executes it. The user can define new ones and programming is like writing subroutines."

Jupiter Ace interior

Source: Jupiter Ace Resource Site

Many folk were tempted by the new language and by Jupiter's new machine. By December 1982, the company had signed retail deals - a big step forward into the unknown for a small company managed and operated by two techies - with chains Laskys and Debenhams to put the Ace into some of their high street shops, with distributor Micro Marketing managing the process. Christmas 1982 was boom time for the UK microcomputer industry.

It was promised that production would be doubled in January 1983 to ensure shops got sufficient stocks. Michael Scott, Micro Marketing's MD boasted: "We are looking to take 25 per cent of the market for micros costing under £100."

Ramping up for 1983

Micro Marketing would become one of the few sellers of add-ons for the Ace. The promised Ram pack failed to materialise, likewise the games and applications software that, Jupiter suggested, would hit the market in early 1983. Micro Marketing put out a 16KB memory extender in March of that year, and third-party software house were beginning to release apps around the same time.

Vickers and Altwasser, meanwhile, focused on developing the Ace range, initially revising the original model and then developing a new version, the Ace 4000.

The 4000 would sport a better-quality, injection-moulded case of the kind favoured by its competitors. Despite a new motherboard, the 4000 matched the spec of the original, ensuring "full software compatibility", but added a "Rom-deselect feature, enabling the internal Rom to be switched out and external Rom modules to be plugged in". It also gained sound output through the connected TV monitor. 16KB and 48KB Ram packs would be offered too to augment the 3KB built in.

Selling the Jupiter Ace

By Spring 1983, third-party add-ons were appearing

But Brits wouldn't get to see them - the 4000 was intended solely to support Jupiter's entry into the US market, company documents reveal. The 4000 was made in the UK, with Stateside sales and logistics being handled by Oxford, Pennsylvania-based Computer Distribution Associates, "not to be associated with Jupiter Systems of Berkeley, California", the 4000's box warned in big letters.

"There was a definite period when our long-term forecasts were of selling lots of machines in the States," says Vickers. "We had lots of transatlantic phone calls where they were telling us how many zillions of Aces they were going to sell, but unfortunately there were always delays - among then genuine delays around getting kit into America and getting approval for the electronics - but basically the US market just never happened."

No more than 800 units are thought to have been shipped over the Pond. In the UK, fans were offered the Ace 16+, a regular Ace bundled with a 16KB Ram pack. It was announced in June 1983 as a replacement for the original machine. The priced remained £89.95.

But it wasn't enough to encourage further sales. Both Altwasser and Vickers recall big orders coming in from a dealer only to find the reseller couldn't then be persuaded to take the units they'd agreed to take.

By the end of October 1983, Jupiter Cantab had formally ceased trading. The writing had been on the wall since the summer. In July, Altwasser decided he not longer wanted to be involved in the venture and quit - "in order to spend more time with his family", press reports put it at the time.

Selling off the Jupiter Ace

Selling off the Ace: Boldfield was offering cut-price machines through the mid-1980s

Looking back, Altwasser says he was simply less convinced about the company's ongoing fortunes than the more optimistic Vickers. "I think it's fair to say Steve had very bullish views and estimates for where the company and the product could go, and I though it was best not to hold Steve back with my more conservative, less optimistic views," he says.

For his part, Vickers clearly understands the pressure Altwasser, as the head of a new, very young family, was under to seek a more stable form of employment than "seeing out a failing company".

Jupiter desperately needed a more powerful follow-up machine to replace the first Ace, which was behind the hardware curve when it was launched let alone a year on. But without the company's hardware guru, Richard Altwasser, out of the picture, where would the 'Ace 2' come from?

Collapse

With an ageing spec and a paucity of third-party software, sales of the Ace slumped, despite a claim from newly installed marketing chief Geoffrey Walker - a friend of Vickers since his undergraduate days, brought in to fill Altwasser's administrative role - that 5000 Aces had been sold to date. Vickers' recollection is of a much smaller number.

"Geoff was doing a PhD in music but he was beginning to realise he wasn't going to finish it and he wanted to do something more business-like, so he was looking for an entry into the non-academic world, and I desperately needed someone to take over Richard's admin work." Bringing on board others better suited to running a growing business was considered too, but Jupiter simply couldn't then afford to employ likely candidates.

Vickers attempted to shift the focus of the company away from computing enthusiasts and on the education market, but kids wanted to learn Basic, not Forth with is peculiar "Reverse Polish notation" syntax which placed operands first, operators after.

Steve Vickers and Richard Altwasser

Happier days: Jupiter publicity shot of its co-founders taken before Altwasser's departure

"Richard and I are both engineers, and we thought Forth was the right solution for what we were trying to make. And an engineering terms, we achieve that. But from a marketing standpoint, I think we were probably a bit naïve. In retrospect, we sold a lot of machines to the few hundred Forth enthusiasts and that was it."

Beyond the Forth fans, Vickers and Walker also briefly explored pitching the machine as a Forth development kit for programming low-level industrial controllers. "We had some initial discussions with the railways, for instance, but they really came too late," says Vickers.

With no prospect of reversing the downward sales trend, either with new, more exciting hardware or fresh interest from app developers, Vickers and Walker opted to pull the plug.

Jupiter Cantab was wound up in November 1983, just 14 months after its two founder announced their first machine. The company's employees - the headcount had never risen above five people - were made redundant and Jupiter Cantab left £140,000 in debts, administrator Dennis Cross of accountants Chater & Myhill, appointed on 8 November 1983, revealed at the time. Some 1600 Aces were left to be sold, he added.

Brighton-based Remsoft was one potential buyer. It considered nabbing the remaining stocks from one of Jupiter's creditors, manufacturer Downsway Electronics. The few Ace 4000s left over could be sold to members of the Ace User Group, which Remsoft ran. The last old-style Aces were bought by Cambridge-based Boldfield Computing to be sold by mail order for £26 plus VAT a throw. No further units were ever manufactured.

Altwasser recovered from the stress of running his own computer company and, after a period as a consultant, joined Amstrad in 1986 as its engineering manager. He later became a company Director, and stayed with the firm through its acquisition of his former employer, Sinclair Research. He left in 1992 and has since occupied senior positions with RM (formerly Research Machines), Xitex and Icera. He most recently worked for educational computing firm Promethean.

Steve Vickers and Richard Altwasser

Ace pilots: Richard Altwasser (left) and Steve Vickers today

"Looking back I don't think we did badly," says Altwasser. "It was a great privilege to work with Steve, and it gave me a taste for being in the technology and entrepreneurial world, which is where I've been ever since."

After Jupiter, Steve Vickers returned to contract programming and to technical writing, but his love of mathematics would eventually lead him to an academic career, taking posts at Imperial College, London, in 1985; later at the Open University and, his current role, Senior Lecturer at the University of Birmingham, where he teaches both maths and computing. ®

The author would like to thank the many retro-tech fans for archiving home computing adverts, photos and documentation from the 1980s, without whom this feature would not be possible. Very special thanks go to Richard Altwasser and Steve Vickers. Thanks also to Steve Parry-Thomas at The Jupiter Ace Resource Site for his invaluable help sourcing pictures and scans.