Refurbishing a 128K ‘Toastrack’

Today I’m going to take you through the process of refurbishing a 128K Spectrum Toastrack machine.

A quick warning – this post is quite photo heavy, but is the best way of helping to illustrate what I do without relying on walls of text, and will hopefully make you feel like you’re more involved. As always, you can click on each image for a larger version.

Anyhow, onwards we go. I received this slightly anonymous package this morning:

Could have been packaged a little better, but upon opening it looks like the machine in question survived the spin across the Irish Sea intact. This is a Sinclair Spectrum 128K Toastrack, quite sought after by collectors as the last true incarnation of the Spectrum line before that Alan Sugar fellow got his greasy mitts on the brand. It’s quite crusty on the outside, but that’s to be expected after all these years.

These machines usually clean up quite nicely, being black in colour helps as this disguises any marks or yellowing seen on other systems.

The first thing I do with machines I buy is to crack them open and have a quick look inside, just to see what I’m facing, and take note of any issues I need to rectify. This machine is an Issue 6K, and can be differentiated from the Issue 6U 128 by having green solder mask on both sides of the board. Several prominent components (capacitors, coil etc) match those found in the Spectrum 48K Issue 4S, so it’s highly likely that Samsung produced this run of 128’s for Sinclair. They all have serial numbers beginning with 007, for those hoping to obtain their own.

Looking at the underside of the machine, it’s got a repair warranty label (violated without a thought) and an additional note – ‘5v rail 28/9/92’. Heartening to know that someone thought enough of this computer to have it fixed so late in the Spectrum’s commercial life.

Repair evidence is obvious immediately as there are two 100uf radial capacitors on the board, these would not have been factory original. (February 2020 – I was incorrect in this statement, it appears that a large number of 6K boards were manufactured using these radial 100uf caps. The footprints are present to replace using either radial or axial parts, and for complete historical correctness radial parts should be used where this type of fitment is found.)


The MIC socket was a little worse for wear, in that it was partly missing. A brief search located the remains, hiding in the recess between the regulator and the heatsink:


TR4 (a ZTX650) is also riding a little high, and examination of the PCB on the underside reveals flux residue. This and the two radial caps (C44 and C45) confirm the fix for the ‘5v rail’ fault – DC-DC converter failure caused by excessive ripple due to failed 100uf capacitors. This would actually have taken out the 12v rail, but left the main 5v rail intact.

(This circuit is almost identical to that in the 48K Spectrum range, and can fail in the same way. However in the 128 range there’s no 4116 RAM to kill, so usually all that happens is that video output is lost.)

So, not a huge amount to do, so let’s crack on. The machine was sold as working, and I verified this before starting the teardown, so in this case there’s no diagnosis to perform.

First fix is to replace the MIC socket with a good salvaged part. I also replace the 9v DC socket as it was looking a little worn. These are quite cheap so a no-brainer to do.


Next, I’ll perform a full recap. Laid out in front of the machine are all the axial capacitors I require for the job (3 x 100uf/25v, 3 x 47uf/25v, 4 x 22uf/25v, 1 x 4.7uf/25v and 3 x 1uf/63v):

Handily, Sinclair designed the PCB to allow fitting of radial capacitors even though they were never shipped in such a configuration. You can see this in the following set of images.


I also took the opportunity to change R115 (seen here just left of the ULA) for a 1K5 resistor, which helps better balance the AY and beeper levels. (The ideal level is 1.65K, so you can either piggyback two 3.3K resistors, or go for either 1K5 or 1K8 parts). I also reinstall TR4 to be lower on the board (reducing the risk of it bending and shorting during future maintenance).


I use composite video output for most of my machines, so didn’t perform the mods required to obtain a SCART-compliant signal from the RGB port. If you wish to do this, it’s a case of removing R134 and then replacing with a 680 ohm resistor to the left R134 pad, and the other end to the right end of C131 (47uf capacitor directly north of it). This gives you proper SCART blanking voltage on pin 1 of the RGB port but removes the composite output.

Anyway, recap complete, and the top of the PCB cleaned with a stiff brush. It was surprisingly clean given the state of the case, if it were in worse condition a bath in some IPA would be warranted at this point.

I do however clean the bottom solder side with IPA and a toothbrush, and then lay a piece of kitchen towel over the cleaned area and scrub with the toothbrush again (this soaks up all of the debris and flux residue into the towel, and stops it from becoming sticky when dry).

The edge connector was filthy, but is easily cleaned with a pencil eraser followed up with some more IPA.

Finally, the EPROM is missing its sticker, so I apply one to help preserve its contents. A quick diagnostics check shows all is well with the PCB after surgery.


Internals complete, now it’s time to turn our attention to the grubby exterior. I start by cleaning the heatsink with some baby wipes which easily get between the fins.

At the same time I take the opportunity to replace the original 7805 regulator with a L78S05CV – a Toastrack consumes 900mA or so without any peripherals which sails close to the 7805’s maximum output, and the L78S05CV is rated for 2A. I’ve replaced the thermal pad I found (which must have been added later) with a smear of thermal paste under the regulator when reinstalling.

I’d recommend regulator replacement in any Toastrack, simply because it’s the only Spectrum in which you can do so without the need for soldering (the PCB connector is a simple push fit back onto the regulator legs, but be careful to get the orientation correct).


Time to get stuck into the case then. I remove the keyboard backplate, membrane and bubble mat, followed by all keys and their posts. The old membrane is naturally discarded as it only works across a few rows. There’s not a huge amount of dirt underneath the keys, luckily – some right horrors can be encountered here depending on how much of a dribbler the previous owner was.

The upper and lower case halves get a scrub with some warm soapy water followed by a rinse in cold water (using a double sided sponge for the case, and a brush for the keyboard posts). The bubble mat gets a gentle clean at this point too.

I use baby wipes again to remove the last remnants of dirt from the case grooves and other nooks and crannies. The key is to not scrub the lettering too hard or use anything too abrasive, otherwise some of the red paint will be removed.

The next part is perhaps the most tedious of all but absolutely worth it – using baby wipes (this is becoming a common theme), get your fingernails into every upper recess of every key to remove the decades of grot. Reinstall as you go, just the keys and their posts at this point. If you’ve spent enough time, the machine should start looking quite nice at this point.

One last cosmetic detail on the case, unfortunately the metal rainbow badge is damaged. Replacements can be sourced from, however when I bought some of Djordje Mitic’s replacement faceplates last year, he included a load of blanks removed from the stamping process. Some of these have the rainbow logo on them and with a trim fit the 128 recess perfectly.


On the home straight now. Just a case of reassembling the computer and installing a replacement membrane to the case top. Obviously wait until everything is dry – if you’re impatient, careful use of a hairdryer can speed up this process.


And (drumroll…..) the finished machine, which has cleaned up quite nicely.

After taking the photos for this article, I did notice that the machine wasn’t working correctly with DivMMC devices. If the machine is otherwise working okay, this is almost always a fault with the CPU (specifically the /M1 line being missing or too weak for the DivMMC to use). The old CPU was extracted, a socket fitted in its place, and finally a new CPU added, and the DivMMC was happy.

A number of months later, I acquired a Spanish 128 Keypad. These do not work on Issue 6K Toastracks out of the box due to an incorrect resistor being fitted in one of the serial port power generation rails. R137 was changed for a 15R part, and this allowed the keypad to work perfectly with this machine. (It is speculated that this is the reason that UK keypads were not officially released – all 6K machines would have to have been recalled for this fix, at a time when Sinclair were under extreme financial pressure).

Hopefully this article has given you some insights into the process of rescuing a Spectrum from the brink, if you have any other handy top tips or advice I’d love to hear them!