Sinclair and the 'Sunrise' Technology: The Deconstruction of a Myth (24 page)

The ZX83 design process outlined above never got much of a chance to proceed smoothly. The delays in the design of the Spectrum Interfaces held up the development of the upgraded microdrives for the QL. Having finalized the actual tape-cartridge design, Dave Southward, Sinclair Research’s technical director, had to try to gain faster data transfer and capacity by modifying the drive mechanisms and electronics. Sir Clive is inclined to blame delays on this aspect of the design:

The problems centred principally on the microdrives, which worked a treat in the Spectrum version, but in the QL were re-engineered in what looked like subtle and fairly minor ways, but in fact turned out to have a lot of problems. (Interview, 6 November 1985.)

Since portability had vanished, there was no longer any particular need for the low power consumption of the microdrives, and not as much cost benefit as might be imagined. Standard 3-inch floppy-disc drives bought in from Japan would be nearly as cheap, and have both faster performance and higher storage capacity. There was no likelihood that Sinclair would take this route, however, since it had succeeded in producing an ‘innovative’, if inefficient and idiosyncratic, storage device, and would proceed with the design despite good arguments to the contrary. Of course there was little else that was distinctive about the machine, and ICL presumably wanted some innovative Sinclair technology in exchange for its investment.

The need to stick with the Sinclair approach is also apparent in the keyboard for the QL. This revised membrane design, although a distinct improvement over previous Sinclair designs, is not the ‘professional keyboard’ it pretends to be. It was assessed as the better of two designs produced for the QL but, symptomatically, the sample keyboards produced by a Japanese company to Sinclair’s requirements, of typewriter standard, were not included in the assessment. The obsessive Sinclair Research mentality says, apparently, that it has to look like a Sinclair design even if it doesn’t work as well as it might. Again, there is no significant cost saving that acts as a justification. We were informed that the sample keyboards costed out at the same price as the Sinclair design chosen! The movement of the yen might have altered this by a pound or so, but most purchasers of a ‘business’ computer would probably happily have paid such a surcharge for something their secretary could type on.

It’s unlikely that any decisions could have been made to go against the Sinclair style, since the management couldn’t even manage to decide which of the sample printers offered by manufacturers they should choose. (The QL has a serial printer interface, rather than the industry standard parallel interface.) Since the sample printers were available in 1983, it is somewhat peculiar that only in late 1985 was one announced. One explanation for the lack is of course that a useful business system must have a printer. Since cheapness would be a selling point, reminding the potential purchaser that another large investment would be required would be counter-productive.

The same argument went for the VDU monitor that is essential for extended use. Again, there was a very good argument for Sinclair to provide a monitor, because the QL system drives an international standard display, which will not work effectively with some of the British monitors available. Marketing-wise, to trumpet a price breakthrough of £400 for a business system, and then admit that you needed a printer (about £250) and a colour monitor (about the same) to put it to use might remind potential purchasers that they were looking at a total cost of nearer £1000. At £400, the QL could be made to look like a bargain, but at £1000 there were other options the customer might look at. This train of thought, or incapacity to commit to a competitive assessment - however you look at it - led to two decisions, or rather one non-decision and a fudge. The first was not to provide a monitor, which was one less commercial task, and the second was the addition of an output suitable for a domestic television. Again, this was retrogressive in terms of the original concept, but comforting in the sense of reverting to Sinclair style — all previous machines had only plugged into televisions. The fact that the display wraps off the corners of a lot of television screens and produces flicker was discounted in favour of increasing the market; the specifications changed again, and another job was added to the design task.

Back to the hardware that was supposed to use the peripherals. The heart of the design was the custom gate array. At the point when the TTL prototypes should have been built, the design process had slipped so far that the electronic design for them was not completed. Since the turnaround time on a custom chip was only three weeks (and about £10,000), the decision was to go straight for the custom chip, and then incorporate this in a prototype, cutting out the TTL version. Predictably, the chips had problems, and with mounting pressure it appears that as soon as one problem had been identified and corrected the modification was incorporated in another reiteration of the chip. Another three weeks, in the course of which more flaws would come to light, and the process would then be repeated.

Two consequences are apparent from this illusorily time-saving approach. The first is that the integrated modem that David Karlin was to produce, as part of one of the chips, was never designed. Another feature vanishes from the vision. The second was the incorporation in the design of what the ad people turned into a virtue, an Intel 8049 microprocessor chip. This ‘second processor’, says the blurb, ‘controls the keyboard, generates the sound, and acts as an RS232 receiver. None of the power of the 68008 processor is wasted on these functions.’ Well, yes, but a lot of the potential processing power of the 8049 is wasted on doing three minor jobs, none of them very well. It wasn’t supposed to be there at all, since the functions were supposed to be performed by another custom chip.

This was a sensible choice, as an initial design decision, because it would replace the three standard chips that could do the job, be cheaper, and simplify the circuit. Unfortunately, they didn’t have time to design this, so the 8049 was pressed into service. The beeping noises it makes are more variable than the Spectrum’s, but just as useless. It handles the keyboard encoding, but this could as easily be done in software, and it handles RS232 serial communications (signals in and out of the machine) in a peculiarly inefficient way, operating one channel but multiplexing two channels into this. The result is various problems with the RS232 facilities, one being that both serial ports provided can be set only to the same speed of transmission. If there was a cost saving there might be some argument for it, but since the 8049 is more expensive than the custom chip, and less efficient than three dedicated chips (together not significantly more expensive), it gives a measure of the confusion of the design process. It is not as if the 8049 merely dropped into the position reserved for the custom chip, either, since it didn’t have enough pins (connections) and additional custom-chip work was required to overcome this problem.

In the light of this catalogue of circuit changes, Sir Clive’s explanation of design problems as due to lack of control over the engineers rings a little hollow:

The project started off in a totally different fashion, and then diverged from what I originally wanted because the engineers who worked on it wanted something very different. Engineers always do, they want something that they would like, and you’ve either got to pull them back, or you’ve got to persuade them, or you’ve got to switch engineers. In this case, they weren’t having to persuade me, really, they were persuading Nigel, and he bought the package. (Interview, 6 November 1985.)

What the engineers wanted was not a better machine, but more time, better co-ordination, a consistent specification and things like that. Given the way the QL turned out, to blame the engineers for a common trait of their profession, the quest for excellence, is perverse. The engineers were in fact doing their best, but if you can’t put all the bits together to test them you are bound to have problems with the overall system.

Since the interactions are supposed to be controlled by the software, the writers of the operating system also have problems. It is quite feasible to write the bulk of the software from the specification of the hardware and its interrelationship, but revisions in the design must be reflected in changes to at least the lower level of software, that directly controlling chips and other devices. The operating system (OS) was commissioned from a software company called GST. It worked from the low-level OS drivers produced by Tony Tebby, and in contrast to the modest fees paid to Nine Tiles, GST was to have done fairly well out of their involvement. In addition to time and material payments that ran into six figures, GST was to receive a royalty. Tony Tebby, inside Sinclair Research, was also developing an operating system ‘as a backup’.

Psion, a software company that had done very well out of games software for home micros, had forged links with Sinclair Research on earlier software ventures. When Sinclair put the word out in early 1983 that a suite of business programs was needed, Psion had already started to develop just such a package of integrated software for IBM micros and other MS-DOS machines. It put its proposals in and they were accepted. Since all the development work was done on a VAX minicomputer system the lack of hardware didn’t matter so much in the first place. It would later be customized to the QL operating system. The only hardware available in mid-1983 was mock-ups of the main board, minus microdrives, serial ports and the like. Somewhat later some hand-built versions were produced, with a single Microdrive, and the serial ports, but since the Microdrive couldn’t be used in conjunction with the serial port, and the logical faults were still there, their appearance, although better than nothing, could not have given anybody a sense of rampant progress. We now leave the QL saga for a moment to consider another development taking place around this time.

There were persistent rumours that out of the Far East would emerge a cheap (£50 or so) colour computer as the Orient moved into the market. Responding to this potential challenge to the Spectrum’s pre-eminence, the Low Cost Colour Computer (LC3) project was started. The hardware was virtually a one-man project for Martin Brennan, who designed and produced it in a matter of a month as a TTL prototype. With a Z80 chip, and designed to use ROM software cartridges, and with data storage on battery-backed RAM packs, it was a nice concept. Steve Berry produced an operating system, complete with the full overlapping ‘windows’ that the QL doesn’t possess. (True windowing, as on the Macintosh, allows a separate screen portion to overlay whatever was originally there, and then be removed, while preserving the contents of the original screen.) This cheap and powerful machine, with superior display handling to that of the QL, was one of the topics discussed at a planning meeting in November 1983 held, for some reason, in the Lake District.

Sir Clive, the technical members of the board, and various members of the technical staff forgathered. The LC3 project was chopped, on the grounds that the competition had not appeared, and there was no reason to introduce a cheaper computer unilaterally, especially with the lower absolute profit margins it would produce. So much for cheap computing for the masses! Further development of the LC3 would be costly, and the view was being sustained at this time that the QL was almost ready for production. The capacity for self-delusion that this implies is explicable only in terms of a lack of communication. While Nigel Searle puts it like this:

We are a very bureaucratic company and don’t spend a lot of time in formal communication, written or otherwise. (Quoted in International Management, November 1984.)

The technical staff found themselves wondering why, in a fairly small company, communication was worse than that found in the major companies they had worked for. When Searle decided to go for a QL launch date, the stresses became acute. Again, the implication was that the engineers were a problem:

At some point in a project that has been going on for 18 months, you have to put a stake in the ground and say you are launching the product on such and such a date. If you wait for the guys who are working on the product to tell you when it will be finished, you will wait for ever, (ibid.)

Waiting until there’s a working prototype, however, would seem to be sensible. The decision to go for a launch was imposed rather than negotiated, just as Searle implies, but ill-advised. Tony Tebby recalled the state of affairs:

Communications were deliberately distorted. If I talked to marketing, they would describe to me a product I’d never heard of. They said, ‘Well, give us the finished product in a couple of weeks’ time and we’ll review our position.’ I said, ‘But it’s not going to be working for six months!’ They say, ‘But we’re starting the ad campaign in two weeks’ time, placing the ads.’ (Interview, 14 October 1985.)

The case design was frozen, which didn’t allow time for the design of the ‘feet’ to prop up the case to a good typing angle to be completed. The resulting bodge occasioned comment from reviewers:

If you do a lot of typing you might find the keyboard lies a bit flat. To overcome this, Sinclair has supplied three funny little plastic feet which are supposed to fit into rubber pads under the keyboard. I found that these fell out regularly and in the end I dispensed with them and got used to a new typing position. (Personal Computer World, June 1984.)

Another reviewer suggested you also dispense with the feet, but rest this 'professional’ micro on a book! Another result of freezing the case design meant that positions of sockets for external connections were also frozen, and the PCB (printed circuit board) tracks had to lead, however inefficiently, to those locations. Rick Dickinson, the case designer, and John Williams, the draughtsman, or the others involved (effectively everybody) can’t be blamed or even disparaged for any of these problems. The lack of co-ordination and communication, and the pressure for an unrealistic launch date must bear the blame. The last prevarication and specification change indulged in by the management of the time illustrates the point.