Bringing Back A Spectrum’s Rails

The Sinclair ZX Spectrum was to most Brits the computer to own in the early 1980s, it might not have had all the hardware features of its more expensive competitors but it had the software library that they lacked. Games came out for the Spectrum first, and then other platforms got them later. If you didn’t have a rubber keyboard and a Sinclair logo, you were nothing in the playground circa 1984. That low price though meant that in true Sinclair tradition a number of corners had been cut in the little micro’s design. Most notably in its power supply, all the various rails required by the memory chips came from a rather insubstantial single-transistor oscillator that is probably the most common point of failure for these classic machines.

[Tynemouth Software] had an Issue 2 Spectrum with a missing -5V rail, and has detailed both the power supply circuit used on these machines and the process of faultfinding and repairing this one. A single transistor oscillator drives a little ferrite-spool transformer from which the various supplies are rectified and filtered. Similar circuits appear in multiple generations of Sinclair hardware, where we might nowadays use a little switching regulator chip.

We’re taken through the various stages of faultfinding this particular circuit, and the culprit is found to be a faulty Zener diode. It’s certainly not the last dead Spectrum that will cross an enthusiast’s bench, but at least in this case, the fault was less obtuse than they sometimes can be in this much-loved but sometimes frustrating machine.

Sinclair enthusiasts might also appreciate the great man’s earliest work.

10 thoughts on “Bringing Back A Spectrum’s Rails

  1. Indiegogo has said that it is willing to extend the deadline it gave to a project attempting to make a handheld version of a classic British computer..But the crowdfunding site says that the team behind the ZX Spectrum Vega+ has yet to meet its conditions.In February, Indiegogo threatened to appoint debt collectors if the campaign had not fulfilled its commitments by the end of May. The project’s chief told the BBC he was “still determined to deliver”.

    Dr David Levy added that he believed many backers “still are fully supportive of our finishing the project”.The campaign originally pledged to send out the console in the summer of 2016. The company he chairs, Retro Computers Ltd (RCL), has issued an update to backers saying it now intends to deliver the first consoles by 15 June.

  2. My interest in things Sinclair/Timex is current, as I still have a large box full of Sinclair computers, parts, and magazines (boxed for potential removal; too many other projects); so I would definitely like clarification, vis-a-vis the date(s) shown here–

    “….The campaign originally pledged to send out the console in the summer of 2016. The company he chairs, Retro Computers Ltd (RCL), has issued an update to backers saying it now intends to deliver the first consoles by 15 June.”

    …and the posting date is 2 June 2018.

  3. I’ve got to remember this nifty little power supply when I want to incorporate planned obsolescence!
    In all seriousness, I’d love to get my hands on a ZX here in the states!

    1. Planned obsolescence, 36 years later, a good couple of decades since the last Spectrum was sold commercially. We could do with more obsolescence like that.

      Other good thing is, it’s a well known weak point. So it’s something you’d check for first, according to well-known symptoms, and then fix really easily and cheaply with a new transistor or zener. Which, being Sinclair, are absolutely the cheapest ones ever sold.

      Lesser companies would’ve put extra windings on the PSU transformer, and a DIN plug to deliver multiple actual voltages from the actual power supply. Eurgh! Not for Sinclair. If it barely works, and can be gently encouraged towards reliability in the manufacturing process, then that’s what the customers get. And they love it!

      If it weren’t for the silly bugger’s obsession with building a car that absolutely nobody with any sense of style would buy, we’d be reading all of this on Sinclairs now. Actually the C5 was where he failed by being un-Sinclairish. Instead of taking advantage of cheap use of well-proven technology, sneaking extra functionality out of stuff where it was possible, even if not intended, he went the other way.

      The C5 was an attempt to bend technology and functionality to fit his ideal product, rather than shaping the product around what was available and cheap. Lead-acid batteries, bicycle laws, road safety and all the rest just weren’t conducive to a one-man electric cart. It wasn’t crying out to be made, it was Sir Clive trying to impose his will on the world. Whereas the good stuff he made, the computers, were all clever uses of what was cheap to produce something that was more than you’d expect for the money.

      Also the QL had a keyboard nobody could touch-type on. That’s a bit of a SNAFU, business-wise.

  4. I never looked at the Spectrum schematic before (old C= fan here :^) so it baffles me why they used that circuit to generate that voltage when tapping into the CPU clock using one of the available nand gates and use it to drive a transistor with single output inductor could have spared the other transistor and a dual coil inductor which probably weren’t cheap at the time. A zener diode feeding back the output V to one of the nand gates could have been used to make the output stable.
    Disclaimer: I just made up the schematic in my head, no idea if it works, but it should.

    1. The clock to the CPU is not consistent as the ULA stops it when it is using the shared RAM, but it should be possible to tap the ULA clock or even the colour burst frequency and use that as you say. It also seems like they could have used the same RAM chips as the upper RAM, these are 5V only. They are reject 64K RAM chips, with a good upper or lower half, used as 32K RAM chips. With cost savings in quantity, using the same chip in both banks and getting rid of the switching regulator would seem a better option that the way they chose. (and losing the voltages on the edge connector, they weren’t very consistent over the lifespan of the machine anyway).

  5. All things considered, not a bad design. Not a great design. A serviceable power supply. A few too many components (for example) for the very undemanding -5v. memory substrate bias ( a shunt regulator, for goodness’ sake!). And “high-quality, low-cost zener diode” has always been an oxymoron.
    Two well-worn old saws from the hardware design business: “Anybody can design a power supply”, and “The power supply design can wait ’til the last (moment)”. Just like here.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.