Hardware Feature Extra
Name   Source
Micro Clinic   Television magazine 1986

Micro Clinic
Hints for repairing Sinclair Spectrums

Sinclair Spectrum & Opus Disc Drive - Not recognising the Opus.
Although this machine seemed to be working normally, when an Opus disc drive was purchased the computer ignored it. The cause was traced to pin 27 of the CPU chip being stuck low. Replacing the chip provided a cure - pin 27 is the /Ml line and was telling the disc drive that something else was being loaded. As the edge connectors appeared to have had plenty of use I was left wondering what had happened to cause the fault. As luck would have it another of these machines came along - with a blown ULA chip because the customer had pushed the computer into the disc drive and then switched on without bolting the two together. Now the slot in the Spectrum's edge connector can become slightly enlarged, and the Opus's key is rather thin. It's possible for the connectors to short together if the two are left free. After changing the ULA I checked for any other damage - and found that the CPU's pin 27 was stuck low. Problem solved!

Sinclair Spectrum - Solder warning
The following tales show how careful you have to be when working on micros. The computer had a faulty ROM and after removing it and fitting a holder for the replacement, I found I had a dead keyboard. After checking the ULA chip I started to look for solder splashes, using an eyeglass. Not being able to find one, out came the trusty scope. After much running around the circuit it finally dawned on me that the /INT line was not going low enough for the CPU to scan the keyboard. There was a minute solder splash under the CPU's holder. It was removed by sliding a piece of paper under the holder. The /INT line was going down to about 2V. A lot of wasted time and slapped wrists! The next machine had been 'got at' in some way by a small boy, but we didn't know the exact details. Transistor Tr4, the ROM and the CPU were all faulty, but the machine still wouldn't initialise when these had been replaced. Again out with the scope: data line 5 was found to be shorted to chassis. This once more meant an eyeglass search for something conductive. The cause was found under the ULA's holdier: it looked like silver paper. Presumably this had got in during the said small boy's investigation, i.e. he took it apart while eating something wrapped in silver paper! Perhaps something easy now?

Sinclair Interface 1/Microdrive
The ULA in the interface can suffer if the interface moves about or is suddenly disconnected. The usual result is a "microdrive not present" message when a microdrive cartridge is loaded, and sometimes the Spectrum won't initialise due to a grounded data line. Just occasionally a ULA chip will overheat spectacularly, causing a crater in the case just above it. As the ULA's are expensive it pays to bolt the two machines together. It's surprising how many odd faults will go away if the main board edge connectors and the microdrive/interface connectors are given a good clean. I cannot stress this point enough. In the long run it pays to examine each machine carefully after repair.

Sinclair Spectrum
Several of these machines have been brought in with a permanent black raster and white border. The cause is a faulty ULA chip.

Sinclair Spectrum Plus
A common problem with these machines is no colour due to a faulty encoder chip. We find that it usually happens when the "SN" equivalent of the LM1889 is fitted.

Sinclair Spectrum
A blank, black raster - slightly snowy on two occasions - has been traced to a faulty 6C001E ULA chip. A thin liquid had been tipped inside one of these machines that was brought to us - it had run considerably. The customer accepted a large estimate so I set about putting it right. First I replaced the usual faulty transistors (ZTX650 and ZTX2I3) in the 5V regulator circuit, also the coil. All the 4116 RAMs had internal short-circuits (shown up by the fact that they got very hot very quickly). After replacing these the machine still didn't initialise. The CPU. ULA and ROM chips were all found to be faulty. Sinclair power supplies give a lot of trouble. Leads and plug problems are cured by replacing the whole lead - a five minute job and the leads cost less than a pound. Several of these machines have come in where the lead has gone short circuit and the diodes have caught fire, burnt the panel and the only economical answer has been a new unit - they are clearly not adequately protected. Don't forget the thermal fuse on the primary of the transformer when servicing these!

Sinclair Spectrum
It's been said before that the first check with these machines should be on Tr4 and the -5V line. The machine can initalise and appear to be ok. (until the keyboard is used) with the -5v line missing, so to save time and heart heartache remember to make voltage checks first. Colour problems with later Spectrums and Spectrum Pluses are generally due to the SN94459N chip - replace it with an LM1889.

Sinclair Spectrum Plus
We sold a batch of "new" Spectrum Pluses to a local school and congratulated ourselves on beating Dixons etc. for the order. That was a mistake! We've had a few back with odd problems which we've had to repair ourselves under guarantee. The common link is that they've all received attention before. One machine had had reset problems since we'd supplied it. The reset had been slow and not always complete & Also the logo had occasionally been accompanied by the flashing cursor. On investigation the machine turned out to be a Spectrum issue 3B that had been converted to a Spectrum Plus. The reset arrangement was novel - instead of taking the CPU reset pin to chassis it took the 5V line to chassis! No apparent damage had been caused by this brutal treatment and the machine now works normally. Following this we've had a number of other Spectrum Pluses with old boards inside. One was a real heartache. It came in with Tr4 short-circuit and a number of chips damaged. We did a memory check before boxing it up and found that it worked only as a 16K machine. To shorten a long story, it would appear that Sinclair sold a number of 48K machines that were working only with 16K - labelling them as 16K of course. The extra 32K of RAM was soldered in (no holders), ready to spring any poor unsuspecting soul about to upgrade them a surprise. This particular board had been recycled in a Spectrum Plus case and had been sold again (in a large Liverpool store), still only as a 16K machine. No one appears to have checked it at any stage during this procedure. Just to add a little spice to the fun, the membrane was very intermittent on extend mode, delete and symbol shift. As we're a long way from Liverpool we repaired all this and levied a nominal charge, hoping to get the loss covered by future business- Way back in BC (before computers) we made it our policy always to check TV sets before delivering them. We continued this policy AD (after digital?) and it has paid dividends over and over again. We refuse to sell an item still boxed unless the customer is adamant. Even so we recently missed an Amstrad printer whose ribbon was twisted.

Sinclair Spectrum
In this Spectrum both Tr4 arid the 5-1 V zener diode had gone short-circuit, thus preventing operation of the power supply.

Sinclair Spectrum Plus
This machine came in with a simple looking fault that proved to be one of the more elusive ones. At power up the screen area was completely black and the logo didn't appear. It had apparently gone part way through the initialisation process, so there couldn't be much wrong. Testing the signals at the CPU and ROM pins produced different results each time checks were made  this should have been a clue that an unusual fault was present. After removing both the CPU and the ROM I found slightly low resistances on the data lines. The cause was faults in the 32K extension memories - five of the TMS4532 memory chips were found to be faulty! These can be checked by measuring the resistance between the data lines and 0V. With the CPU and ROM removed a reading of 10k ohm is satisfactory: a normal working computer gives a reading of 5-6k ohm. In any case a quick comparison between all the data lines is a useful exercise. Connect the red (negative) lead to 0V if you're using a standard analogue meter. Eliminating the extension memory by removing IC25/6 or cutting the control tracks to these ic.s had no effect on this fault. Only removing the faulty chips or cutting the data tracks to them cleared it.

Spectrum Plus
We've had similar Spectrum Plus faults to those described by Ken Taylor in the January Micro Clinic. When one machine was powered up the screen showed a changing brick pattern but no Sinclair logo. The current consumption was excessive at over 1A - it should be nominally 680mA. The ROM, CPU and ULA chips were checked by fitting them in a known working machine. The ROM and CPU were ok but the ULA proved to be defective. Unfortunately the same fault was present when the ROM and CPU chips plus a new ULA were fitted in the defective machine. A quick prod around with the logic probe then showed that the CPU had crashed (/halt pin 18 =1). The ROM. CPU and ULA were again removed and power applied. The logic probe was then used to check the data bus - all lines should have been high due to resistors R9-R16. Lines D1,D2, D4, D5 and D6 were found to be low however. A resistance check showed a direct short-circuit and we next found that pin 14 (data out) of 1C16/17/19/20/21 was shored to ground. Lifting pin 14 of these i.c's removed the excessive current drain. IC25 and IC26 were at this stage removed to disable the top 32K. A scope probe check then revealed, that the /RAS control signal for the lower 16K RAM was missing at pin 4. It was present at pin 35 of the ULA. A resistance check showed a contact resistance of 14ohm between the ULA and its socket. Fitting a new socket re-established the signal - but still didn't clear the fault. Address line A6 was found to be present at one side of R20 but not the other side. A resistance check showed that R20 had infinite resistance. Replacing R20 cured the remaining faults, allowing the machine to initialise correctly - but only as a 16K machine (IC25+26 removed) Phone calls to various distributors revealed that the TMS4532 and the MSM3732 have been obsolete for a couple of years, the replacement being the 4164 series. Which links do you use with these devices? I assume the Texas link is used and either link 3 or link 4 depending on whether the upper or lower 32K is used.

Taken from www.8bit.ht.st