Sinclair ZX Spectrum 16K & 48K


After the success of his two previous computers, the ZX-80 and particularly the ZX-81, in April 1982 Sir Clive Sinclair presented the ZX-Spectrum.

It was originally to be called the ZX82 (as shown in these two design prototypes) , but Sinclair particularly wanted a name to reflect the new colour capabilities.  The Spectrum name came from a suggestion from division head Nigel Searle's girlfriend!

For the first time, a computer with high-resolution graphics, colour, sound and 48k RAM was sold for under £200. Once more Sinclair revolutionised the microcomputer industry with new standards.

A much improved keyboard over the ZX81, it was still a low cost, multi-function affair, but with a rubber key mat rather than a flat plate. It probably held the world record for the number of functions on one key - as many as six!

Again based on the Z80A CPU, two models were launched: one with 16K and one with 48K RAM.  The ROM size doubled again to 16K.
There were 8 revisions of the motherboard (revisions 3 and 6 also had 3A & 6A versions) with revision 1 machines much sought after now by collectors as only around 60,000 were sold.

An original issue 1 machine had light grey/fawn coloured keys. All the other versions had the familiar darkish blue/grey keys which made the letters/keywords easier to see under electric lighting.

The best feature of the Spectrum was its graphic possibilities. With 256 x 192 pixels, it was one of the highest resolutions for a microcomputer of its range in 1982.  Unfortunately, to minimise memory requirements, colours were limited to 1 ink colour and 1 paper colour which was applied to an entire per 8 x 8 pixel character block.

Although the addition of sound was a step forward after the silence of the ZX80 and ZX81, it was also to be a major disappointment - a poor quality (low ocst) speaker controlled by the BEEP statement – that was it!
To make matters worse, as there was no dedicated sound chip, the Spectrum could not do anything else whilst it was "beeping", so games could stutter badly (unless programmed cleverly in machine code) making background music virtually out of the question.

However, the Spectrum was an enormous success, becoming Britain's best-selling micro and spawning a huge market for software, peripherals and other add-ons, such as proper keyboards to replace the "dead flesh" standard rubber key affair. 

Technical Details

  • CPU: Z80A
  • Speed: 3.5MHz
  • RAM: 16KB or 48KB
  • Rubber Keyboard
  • Colour Graphics
  • Sound (beeper - 1 channel, 10 octaves)
  • Local price at launch: £125 - 16K version, £175 48K version

Problems and Issues

Issue 1 ULA Problems
Issue 1 Spectrums had problems with the ULA so that INKEY$ (gives the currently pressed key) only worked in 50% of the cases. This was solved by adding a small circuit board (the "spider", or "dead cockroach"), piggybacked on the ULA.  On some models, rather than the spider, there was just a mess of soldered wires and components, looking like a not very neat home repair.  I sold an Issue 1 spectrum some years ago, only for the buyer to return it to me because of my "shoddy repair work"!  This was genuinely how they shipped, so if you are lucky enough to come across an Issue 1, don't be alarmed with what you may find inside...

Issue 1 Spectrums were essentially a 16K machine with no space for expansion on the main board.  48K versions had an additional "daughter board" populated with the extra 32K of RAM fixed near the rear of the main board.

Only around 60,000 Issue 1 Spectrums were sold making it the rarest (and most valuable) of the early Spectrums.

Introduced late in 1982, Issue 2 boards had all the memory mounted on the main board in response to the success of the 48K version, effectively making the 48K machine the Spectrum of choice.  The internals were tidied up somewhat as the chip error which the spider fixed was sorted, so it was no longer required. There were still some problems with the ULA, but these were improved by adding a transistor on top of the CPU.
There was also a new colour for the keys which were now blue (rather than grey/brown) to improve clarity under electric lighting.

Around 500,000 Issue 2 Spectrums were sold in 1982-83.

Tuning & Colour
Both Issue 1 & 2 boards are famous for poor background colour and poor tuning.  Yellows which should be white, dot crawl and difficulties getting the colour to “lock” are very common. On some TV’s it is impossible to even achieve colour.
There are two pre-set capacitors and two variable resistors all in a row which
can be adjusted to help with these difficulties - this is covered in the REPAIRS section of the website.  Click here for the specific link.

Issue 3 boards were a complete redesign and have a much improved, lower power ULA which is self adjusting so the preset capacitors were not fitted.  There was also an uprated speaker (!) to make the BEEP louder.

This ULA upgrade caused significant compatibility problems as, to reduce power consumption, the value of some resistors was altered meaning a binary value on the EAR (mic in) socket previously present on Issue 1 & 2 Spectrums, was not present until the ULA had fully warmed up.  The keyboard input port reads in this value as well as the keyboard bits and lazy programmers were just checking for the value on the EAR socket.  This would not have been a problem if programmers had checked the whole keyboard input byte as was supposed to happen.

Production was switched to Taiwan from this model onwards and the Issue 3 version is the most widely available as over 3,000,000 were sold.

The Issue 3B was a minor redesign using slightly different components and circuitry, but there was very little practical difference.
THis is the most common board found in the Spectrum+.

In the Issue 4A and 4B boards, the only real new addition was the 6C001-7 version of the ULA which gave a much cleaner and brighter output to the screen.
For reasons unknown to me, this board version seems to give the most problems with add ons such as the Interface 1.  When connected along with a Sinclair Microdrive, the "CAT 1" command (or any other microdrive command) fails with the message "CAT ?1".  The Spectrum works perfectly in every other way, but I have never been able to fix this problem.
Also, if the Spectrum displays this problem, it will not work with the new divIDE interfaces which are becoming popular.

Issue 5 was a major tidying up exercise as six decoder/multiplexor chips were replaced with a Mullard ULA (ZX8401).  A 74LS04 hex inverter chip completed the redesign.

The final version of the Spectrum 48K was the Issue 6 which differed in only very minor ways - certain capacitors and resistors were changed and some models had a ULA provided by Saga rather than Ferranti.

Memory Issues
To keep the prices down Sinclair used faulty 64K chips (internally 2 X 32K). All the chips in the 32K bank of RAM had to have the same half of the 64K chips working. A link was fitted on the pcb in order to choose the first half or the second half.   It was possible with a few logic chips for an experimenter to have access to the faulty 32K bank with all sorts of odd results possible !

48K Machines working as 16K
I have seen this unusual problem several times now - I have had fully populated boards which should have been 48K spectrums only working as 16K machines.
Sinclair sold a number of 48K machines that had memory faults as 16K machines (labelling them 16K of course). The extra 32K of RAM was soldered in place however, even though it was not working.  At the time this caused a major problem for people who had bought a 16K machine because of cost constrainst and then saved enough dosh for an upgrade pack - imagine the surprise when they opened up their machine to add the extra chips!  Through the years, people have swapped motherboards around trying to get working machines, or the 16K labels have been lost or removed giving a nasty shock the moment anyone tries to load a 48K program and can't understand what is wrong.
As a check, enter the follow code directly:

PRINT PEEK 23733 and then [ENTER]

For a 48K machine the result will be 255
For a 16K machine the result will be 128

If your result is 127, then you have a 48K model only working as a 16K unit.

Flashing Coloured Blocks
One of the most common serious Spectrum faults.
The outputs of the voltage regulator circuitry are taken to the edge connector and an oscillator in the circuit is extremely sensitive to any extra load. This extra load is usually generated when a joystick or other peripheral is added or removed with the computer turned on.

The other culprit is one or more of the lower 16K memory chips shorting. These 4116 chips can still be found and faulty ones replaced if you are good at de-soldering, but finding the faulty memory can be very hit and miss unless you have a good record on the lottery or football pools. Sometimes you might be lucky and it will be obvious as a particular chip getting hot much more quickly than the rest.

Higher Memory Faults
These problems often go un-noticed until a certain memory address is accessed which results in a system crash or “Out of Memory” message.
To check that the entire memory is operational enter the following line:

PRINT PEEK 23732 + 256 * PEEK 23733 and then [ENTER]

The result should be 65535 for a 48K spectrum or 32767 for a 16K model. Any shortfall (lower number) indicates a memory fault.

Permanent Black Raster and White Border
The cause of this is a usually faulty
ULA chip these are sometimes available in the shop. On some models the chips are not soldered onto the boards and replacement is very easy. Carefully lever the chip out of its socket cradle with a flat screwdriver blade (or chip puller if you’re posh), levering evenly at each end to avoid bending the pins. Push the replacement evenly into the socket ensuring that it is the right way around (the notch on the ULA chip goes towards the back of the board).

Loose Cassette Connector Sockets
Another very common problem with all Spectrums using cassette leads are loose clamps inside the connector sockets.
This causes the leads to slip out of the socket, or move around affecting program loading or saving performance.
I often have spare sockets in the shop, but if your soldering skills aren't up to removing the old and fitting the new then a gentle push with a small flat blade screwdriver is often successful in re-tensioning the spring contact.

Keyboard problems
Groups of keys mysteriously stop working - probably the most common Spectrum faults and fortunately one of the easiest to put right.  This is almost always a worn out membrane, although splits in the keyboard mats can also cause similar problems.
keyboard membranes and rubber keyboard mats are usually available in my shop and fairly easy to fit provided you have strong fingernails and lots of patience.  The keyboard plate has to be removed and is usually stuck down pretty solidly on "rubber key"  Spectrums.

For instructions on how to replace a "rubber keyboard" membrane, see the REPAIRS section of the site, or click here.


© DataServe 2006 - 2011