Editing ZX81 (section)

=== Development and manufacture ===
[[File:ZX81 Leiterkarte.jpg|thumb|The ZX81 motherboard, Issue One version. The Ferranti ULA is on the left of the image and the NEC Z80 processor is in the centre. The TV output modulator is on the top left. At <!--in, on, at ... wondering which is best-->the bottom right is a ribbon cable connecting to the membrane keyboard.|alt=Top-down view of the ZX81 motherboard showing the layout of the components. Four chips are prominent, along with a TV modulator on the top left and a ribbon cable on the bottom right.]]
[[File:Sinclair ZX81 PCB Top.JPG|thumb|Revised Sinclair ZX81 PCB Revision 3 Top Side]]
[[File:Sinclair ZX81 PCB Revision 3 Bottom Side.JPG|thumb|Sinclair ZX81 PCB Revision 3 Bottom Side]]
The development of the ZX81 got under way even before the ZX80 had been launched. Sinclair's chief engineer, [[Jim Westwood]], was given the task of improving the ZX80's hardware to reduce the number of components and thus bring down the cost.<ref>{{Cite web |title=Sinclair ZX81 |url=https://www.timexsinclair.com/computers/sinclair-zx81/ |access-date=2023-03-15 |website=timexsinclair.com |language=en-US}}</ref> He also sought to fix some of the more annoying problems with the ZX80. Westwood and his colleagues found that the component count could be reduced greatly by combining eighteen of the ZX80's chips into a single [[Gate array|uncommitted logic array (ULA)]], a type of general-purpose chip full of logic gates that were connected up as the customer required during chip manufacture. This short-lived technology of the day was cheaper and quicker than the design of a customised logic chip, which typically required very high volumes to recoup its development cost.{{sfn|Adamson|Kennedy|1986|p=103}} [[Ferranti]] produced the new chip for Sinclair,{{sfn|Adamson|Kennedy|1986|p=104}} who hailed Westwood's design as a triumph of innovation: "The ZX81 had four chips when our nearest competitor in this respect, the TRS-80, had 44."<ref>[[#NS260686|Sinclair (26 June 1986)]]</ref> Only 70% of the [[logic gates]] on the ULA were supposed to be used, but Sinclair decided to use them all to squeeze more functions in. This resulted in the machine becoming uncomfortably warm during usage.<ref>[[#TSS-CRASH|''CRASH'' (May 1985)]]</ref> Computing folklore held that the ZX81 had to be refrigerated by balancing a carton of cold milk on top of the case.<ref>[[#Goodwin|Goodwin (Jan 1988)]]</ref><ref>[[#cw|Gillies-Jones (22 June 2000)]]</ref>

The ZX81's ROM was doubled to 8&nbsp;KB, from the ZX80's 4&nbsp;KB ROM. This enabled a fuller implementation of a version of [[BASIC#Standards|ANSI Minimal BASIC]] (termed [[Sinclair BASIC]] by the company). Clive Sinclair re-commissioned a company called Nine Tiles, which had produced the ZX80 ROM, to develop the new ROM software for the ZX81.{{sfn|Adamson|Kennedy|1986|p=104}} The code was written by John Grant, the owner of Nine Tiles, and [[Steve Vickers (computer scientist)|Steve Vickers]], who had joined the company in January 1980. Grant concentrated on the software that drove the ZX81's hardware, while Vickers developed the new BASIC and the accompanying manual. Sinclair's brief to the pair was fairly non-specific but primarily concerned remedying a key defect of the ZX80 so that the new machine could be used for practical programming and calculations. Vickers later recalled:

{{blockquote|As far as Clive was concerned, it wasn't a question of what the machine ought to be able to do, but more what could be crammed into the machine given the component budget he'd set his mind on. The only firm brief for the '81 was that the '80s math package must be improved.{{sfn|Adamson|Kennedy|1986|pp=89–90}}}}

The new ROM incorporated [[trigonometry|trigonometric]] and [[floating-point arithmetic|floating-point]] functions, which its predecessor had lacked – the ZX80 could only deal with whole numbers. Grant came up with one of the ZX81's more novel features, a syntax checker that indicated errors in BASIC code as soon as it was entered (rather than, as was standard at the time, only disclosing coding errors when a program was run).{{sfn|Adamson|Kennedy|1986|p=105}} Unfortunately for Vickers, he introduced a briefly notorious error – the so-called "square-root bug" that caused the square root of 0.25 to be returned erroneously as 1.3591409 – as a result of problems with integrating the [[ZX Printer]] code into the ROM. Although it was eventually fixed, the bug became the subject of controversy and Sinclair was forced to replace some of the ZX81s sold to early customers.{{sfn|Adamson|Kennedy|1986|pp=107–8}} On a more positive note, Vickers' work on the manual was received favourably, being described in 1983 as "one of the classic texts on BASIC".{{sfn|Adamson|Kennedy|1986|p=106}} Max Phillips commented in a ''What Micro?'' retrospective:

{{blockquote|It does a reasonable job and sensibly provides lots for the reader to do. It's quite honest about the [ZX81]'s shortcomings and provides hints and tips for ways round them ... Best of all, the manual is complete and comprehensive. There's some fairly advanced and often undisclosed information in there. The beginner won't understand it for a long time but if he or she learns some more advanced ideas, the manual is ready for them.<ref>[[#Phillips|Phillips (April 1983)]]</ref>}}

The task of designing the ZX81's case again fell to Rick Dickinson, who produced an updated version of the ZX80's wedge-shaped case. This time round, the design team were able to use [[injection moulding]],{{sfn|Laing|2004|p=61}} which enabled them to deliver a higher-quality case. Dickinson originally envisaged the ZX81 as "an expandable range of boxes following a vaguely modular approach with a common width", though this approach was eventually dropped.<ref>[[#Dickinson|Dickinson (24 August 2007)]]</ref> From start to finish, the design process took about six months.<ref>[[#Cooke|Cooke (August 1982)]]</ref>

[[File:ZX81 concept design.jpg|left|thumb|One of Rick Dickinson's original concept drawings for the ZX81, envisaging the machine as part of "an expandable range of boxes following a vaguely modular approach"|alt=Sketch of two concept designs for the ZX81, showing the computer with a series of similarly shaped boxes stacked behind it in a row.]]
The ZX81 was launched on 5 March 1981 in two versions (though with identical components) – a pre-assembled machine or a cheaper kit version, which the user could assemble themself. Both versions were manufactured in [[Dundee]], [[Scotland]] by [[Timex Corporation]]<ref name="FT 6 Mar 1981" /> at the company's [[Dryburgh (Dundee district)|Dryburgh]] factory.<ref name="McManus">[[#McManus|McManus Galleries]]</ref> Timex had not been an obvious choice of manufacturing subcontractor, as the company had little previous experience in assembling electronics. It was a well-established manufacturer of mechanical watches but was facing a crisis at the beginning of the 1980s. Profits had dwindled to virtually zero as the market for mechanical watches stagnated in the face of competition from the digital and quartz watches. Recognising the trend, Timex's director, [[Fredrik Olsen|Fred Olsen]], determined that the company would diversify into other areas of business.{{sfn|Adamson|Kennedy|1986|p=94}}

This shift by Timex came at an ideal time for Sinclair. The ZX80 had proved more popular than expected and Sinclair's existing manufacturer, a small electronics company in [[St Ives, Cambridgeshire|St Ives]], lacked the resources to deal with the demand. Timex took over production of the ZX80 late in 1980.{{sfn|Adamson|Kennedy|1986|p=94}} The arrangement worked well for both companies and Timex took on the manufacture of the ZX81, aided by capital investment in its Dundee plant.<ref name="Church">[[#SU05-82|Church (May 1982)]]</ref> Sinclair initially planned to produce 10,000 ZX81s a month, rising to 30,000 a month within a year.<ref name="FT 6 Mar 1981" /> However, Timex initially had significant problems in producing enough ZX81s to satisfy demand. As a consequence, it took up to nine weeks for ZX81s to be delivered by mail order. It was not until September 1981, five months after the launch of the ZX81, that the delivery times finally came down to the promised twenty-eight days.{{sfn|Adamson|Kennedy|1986|p=111}} Those who already owned or had recently ordered the ZX80 were not excluded; anyone who had ordered a ZX80 in the two weeks before the ZX81's launch would receive the newer machine, while existing owners were able to upgrade their ZX80s by plugging an extra £20 ROM chip into the circuit board.<ref name="FT 6 Mar 1981" />

The reliability of the ZX81 was controversial. W.H. Smith, one of the machine's key distributors, had a company policy of ordering a third more ZX81s than were actually required for sale, so that it would have enough replacements for faulty machines.{{sfn|Adamson|Kennedy|1986|p=112}} Similar problems were reported in the US market, where contemporary reports suggested that only a third of the ZX81s shipped actually worked.{{sfn|Adamson|Kennedy|1986|p=133}} However, figures released by Sinclair claimed that only 2.4 per cent of pre-assembled machines were returned, although 13 per cent of kits were returned.<ref name="Church" /> Clive Sinclair strongly denied any problem with reliability:

{{blockquote|We have a lower rate of failure on our computers than anybody else in the world, and the reason for that is that we do everything to keep the quality right. The ZX81 production line is a miracle of efficiency; after all, one is made every 10 seconds. They go through the most amazing quality control. Also we have a far lower component count than anyone else. We have only four chips where everyone else has 40.<ref name="Hayman">[[#Hayman|Hayman (June 1982)]]</ref>}}

Sinclair attributed the higher failure rate of the kits to customers breaking the components by inserting or soldering them the wrong way, though the company admitted that there was a persistent problem with power supplies that affected both kits and pre-assembled ZX81s.<ref name="Church" /> The bigger problem was perhaps Sinclair's lack of after-sales service, which Robin Clarke of ''New Scientist'' described as "one of the worst after-sales performance records of any company ever established."<ref name="Clarke" /> The ''Financial Times'' observed that "Clive Sinclair's offices are filled with returned computers which can take months to be repaired."<ref name="FT 20 Mar 1982" /> The company's slowness in replacing returns and delivering freshly ordered machines meant that Sinclair Research gained a reputation for poor customer service.{{sfn|Adamson|Kennedy|1986|p=108}}