Sinclair ZX Spectrum 128 Troubleshooting Guide

Fault diagnosis for the Sinclair ZX Spectrum 128 (the "toastrack") starts from the checks in Sinclair's service manual. Sinclair never completed the manual's symptom table: section 4.2.4 is printed as "TBD", with a note that it would follow once in-service history was known. The symptom advice below therefore comes from the 48K service manual, the T.I.S. field-service notes and repairers' published work on the 128 and the grey +2, which shares its design. The Sinclair ZX Spectrum+ Troubleshooting Guide covers the 48K faults that carry over.

Before you start

• The computer runs from an external 9 V pack, so there is no mains voltage inside the case. The highest rail on the board is +12 V.
• Never plug or unplug anything on the expansion edge connector with the machine switched on. Ian Worsley attributes at least 80–90 % of the 48K faults he repaired to interfaces plugged in live or with the keyway peg missing, which lets the plug slip sideways and short the outputs together.
• Identify the board version (2, 2-1, 6K, 6U or 9G). Two manufacturing defects are version-specific (see below).
• Look over the board before measuring anything. Sinclair warns that hairline solder splatter can have a resistance of several ohms and cause misleading symptoms.
• Sinclair's digital and analogue schematics for the 128 are in section 6 of the service manual (pp. 44 and 45), as low-resolution scans.

Sinclair ZX Spectrum 128.

Defects and behaviour that are not faults

Rule these out first

Symptom	Boards	Cause and fix
Numeric keypad does not work	Version 6K	R137 has the wrong value, so the −Vcc supply to the MC1488 driver (IC33), which serves the keypad and serial ports, is out of specification. Replace R137 (47 Ω in Sinclair's parts list) with 15 Ω. Corrected on the 6U.
A peripheral that needs the CPU clock (a DivMMC, for example) does not work	Versions 2 and 2-1 (Investronica, Spain)	/CLK on the edge connector is not connected to the CPU clock. Add a wire. Correct from version 6K onwards.
Beeper sound or sampled speech is distorted in some software	All	The 128 and grey +2 contend RAM banks 1, 3, 5 and 7. The +2A and +3 contend banks 4–7, as the original 128 specification did, so software written on one behaves differently on the other.
AY music much quieter than beeper effects	All 128-family machines	The mix of the ULA and AY outputs is unbalanced by design.

Supply checks

Check power first. The on-board supply derives every rail from the 9 V input:

Rails, sources and loads

Rail	Derived from	Feeds
+5 V	7805 regulator	Logic, the ULA, the sound and UHF modulators
+12 V	TR4 (ZTX650) and TR5 (ZTX213) inverter, rectified by D15 and smoothed by C44	RS-232 driver IC33 and the keypad
−12 V	Charge pump C111, C112, D28, D29 on the same square wave	RS-232 driver IC33
−5 V	The −12 V rail through zener D19	Expansion port

Sinclair's basic checks (section 4.2.3)

Function	Measure at	Expected
Regulator input	+ve side of C50	+9 V ±2.0 V; below +7 V the regulator may not work correctly
Regulator output	+ve side of C34, IC15 pin 8	+5 V ±0.25 V, no discernible ripple
+12 V	IC33 pin 14	+12 V ±0.5 V
−12 V	IC34 pin 1	−12 V ±3 V
−5 V	D19 anode	−5 V

The external pack should give a trough of at least 7.0 V at 1.4 A with 215 V AC mains, and a peak of no more than 13 V at 600 mA with 265 V AC mains.

Sinclair's and Timex's circuit drawings in the Ferranti Archives 2026 show the supply in more detail. The 9 V input passes through a toroidal common-mode choke (with link positions L1 and L2 to bypass it) to the 7805. The +12 V inverter uses TR4 (ZTX650) and TR5 (ZTX213) with feedback through R59 (1.8 kΩ). The negative rail comes from a BA157 diode pump (D28, D29, C111, C112) on the inverter's 12 V AC output, and zener D19 (BZY88C) with R79 (2.2 kΩ) gives the −5 V on edge connector pin 20B.

Inverter faults

On the 128 the inverter rails feed only the RS-232 and keypad circuits and the −5 V on the expansion port. The RAM runs from +5 V and the video from +5 V, so a dead inverter gives a dead serial port, a dead keypad and no −5 V on the edge connector while the machine otherwise works.

The inverter uses the same TR4 and TR5 types as the 48K. T.I.S.'s 48K notes say that TR4 is very prone to damage and that, before replacing it, TR5 and the coil should be checked too, because all three sometimes fail together and a new TR4 can fail again at once if the others are faulty. Worsley warns against fitting anything but a ZTX213 at TR5 on the 48K, although Sinclair's later 48K parts list names BC213 among the alternatives. The 128's own parts list gives ZTX650 for TR4 and ZTX213 for TR5 with no alternatives. Tynemouth found a TR4 replaced by a lower-rated part on one +2 board and fitted a ZTX651, the higher-voltage version of the ZTX650.

Power socket

Crashes or resets when the machine is moved are usually the power plug or socket. Replace a loose socket; Worsley advises against trying to tighten it.

Clock checks

Clock and carrier checks (section 4.2.3)

Signal	Test point	Expected
Master crystal	IC37 pin 6	17.73447 MHz, no jitter
Z80 clock	IC2 pin 6	3.54689 MHz, no jitter
Colour encoder	IC1 pins 46, 47	8.8 MHz, no jitter
Sound carrier	IC38 pin 4	Within 2 kHz of 6.0 MHz (UK) or 5.5 MHz (other European versions)

Sinclair's crystal note specifies a series-resonant 17.734475 MHz part accurate to 10 ppm. The manual says to adjust the sound carrier "as per the Setting Up Instructions" if it is out of tolerance, but its section 3.1, Setting Up Instructions, is printed as "TBD".

Master oscillator and crystal

The 17.7345 MHz oscillator is X1 with IC37, a 74S04; its output at IC37 pin 6 is the clock input on ULA pin 1. The parts round it differ between documents:

Oscillator parts by source

Part	Sinclair drawing (Nov 1985) and service manual Table 5.4	Timex production drawing (March 1986)
R101, R102	820 Ω	180 Ω
C120	100 pF	Kony crystal: 150 pF. C.S.I. crystal: 10 nF (10 %)
C124	20 pF 2 % (provision for a 2 to 22 pF trimmer)	Kony crystal: 20 pF. C.S.I. crystal: 22 pF (5 %)

If the crystal has been replaced, or the oscillator will not start or is unstable, check C120 and C124 against the make of crystal fitted. Ferranti's ULA specification asks for a clock high time of at least 0.45 of the 56.39 ns period at 25 °C (0.48 at 50 °C).

CPU clock

The ULA makes the Z80 clock and stops it while the display is reading the contended RAM. The clock leaves ULA pin 39 (PHICPU), which also goes to edge connector pin 8A with a 1 kΩ pull-up, R73. It reaches the base of TR3 (ZTX313) through R24 (1 kΩ) and C67 (100 pF), and TR3, with R25 (180 Ω) as its collector load, gives the Z80 a full 5 V clock. C129 (100 pF, base to ground) was added in Sinclair's second modification of September 1985. Sinclair's parts list notes that if TR3 is a KSC839, R24 should be 15 kΩ. If the ULA's pin 39 is switching but there is no clean 3.54689 MHz clock at IC2 pin 6, TR3 and these parts are in the path.

Timex's provisional drawing of March 1986 has R73 and C129 circled and marked "removed" in another hand, while Sinclair's parts list still includes both, so boards may differ here. The ULA's own output is specified at only 1.0 V minimum high at 1 mA, so judge the clock's level after TR3. TR6, the other ZTX313 nearby, is not in the clock path: it gates the ULA's IORQ input with A0.

If it does not initialise

Sinclair treats a clear screen with the copyright line in the lower left as the sign that most of the system is working. If that does not appear and the rails and clocks are correct, the manual's sequence is:

1. Check the +5 V and ground distribution to the ROM, Z80, ULA and RAM.
2. Check /RD, /WR, /MREQ, D0–D7 and A0–A15 from the Z80. All should be active immediately after a reset.
3. Check the RAS and CAS lines to the uncontended RAM (IC15–IC22). They should be active immediately after a reset.
4. Check the RAS and CAS lines to the contended RAM (IC6–IC13) against the oscillograms. Press RESET to get a clear trace.
5. Check that the ROM (IC5) is enabled by an active-low signal at pin 20.
6. Check that the bank register loads correctly: immediately after reset, its pins 2, 5, 7, 10, 12 and 15 should be low. The text calls this chip IC32, but Table 5.4 lists the 74LS174 as IC31 and the AY-3-8912A as IC32.
7. Check the outputs on the RGB connector.
8. Check the picture on a television and listen for a key click on each ENTER press. Then check LUM0 at IC36 pin 7, the sound carrier at IC38 pin 4, and the drive into the modulator at the TR11 emitter, where the DC level at the bottom of the waveform is typically 185 mV.

ULA faults

Ferranti's sketches of the picture from returned ULA7K010E5 parts, November 1985.

The ULA (IC1) is a Ferranti 7K010, listed by Sinclair as 7C001; it is described on the Sinclair ZX Spectrum 128 ULA page. On most 128s it sits in a 48-way DIL socket, so a known good ULA is the quickest test. Spectrum for Everyone reports that some late Version 6U boards have a soldered Amstrad 40056 instead.

Check its supplies before condemning it. VCC1 is +5 V on pin 14 and ground is pin 48. VCC2, pin 13, is fed from +5 V through R34 (15 Ω, 0.5 W; the parts list allows 0.5 W or 1 W), with C30 (22 nF) at the pin and C75 (100 nF) on VCC1. Ferranti's specification gives VCC2 as 2 V nominal and typical currents of 45 mA on VCC1 and 85 mA on VCC2; the archive documents give no voltage to expect at pin 13 on a working board. Then check the 17.7 MHz clock on pin 1 and the 8.8 MHz pair on pins 46 and 47 (see clock checks).

In late 1985 Ferranti tested 19 ULAs that Sinclair had returned from early production (marked ULA7K010E5, date codes 8536 and 8540), all of which had passed Ferranti's own production test. The symptoms it recorded in a test machine are a useful list of what a failing 7K010 can do:

• no reset at power-up, sometimes with horizontal black and white bands, a grid or a blank picture on screen
• resets, but vertical bars appear in the border after the first key press
• picture and sound correct, then a crash after a short time or about a minute
• noise over the picture and failure after a few seconds ("clock failure (counter chain)"), or wavy vertical lines
• the BORDER command gives the wrong colours
• the picture flashes continually, with black and white lines between the colours
• colours correct but some lines and characters black, with random squares of colour
• colour noise, where one of the 8.8 MHz outputs was dead, or no colour at all
• a white picture

The parts that crashed after running for a while also failed Ferranti's speed test once warm: their on-chip resistors were high, which lowers the chip's supply current and its maximum clock frequency. These were faults in the early production parts, which Ferranti then screened for; they are listed here as symptoms, not as a fault every 128 has.

RAM faults

The 128 has sixteen 4164 DRAMs: IC6–IC13 (contended) and IC15–IC22 (uncontended). Sinclair specifies 150 ns access time and 128-row refresh. The contended bank hangs directly on the ULA's address, RAS, CAS and write-enable pins, with its data reaching the CPU bus through R1 to R8 (470 Ω). The uncontended bank is addressed by IC27, a ZX8401, through 470 Ω resistors, and its CAS timing comes from an RC delay built round IC28 (74LS04) inverters with R89 (8.2 kΩ), R88 (1 kΩ) and C105 (180 pF). The drawings also show C103 (25 pF) in that delay; Sinclair's parts list marks C103 "not used". If only the upper bank misbehaves and its chips test good, these delay parts are in its CAS path.

Tynemouth Software's repairs to five grey +2 boards turned up 34 faulty RAM chips, all Samsung KM4164B from mid-1986. His method:

• Power the board from a current-limited bench supply. Tynemouth set 1.5 A; a working +2 draws 800 mA to 1.2 A depending on the RAM. A board that goes straight into current limiting has shorted chips, which get hot. Remove the hottest first; several may be bad.
• Use a diagnostic ROM in the ROM socket. Brendan Alford's ZX Diagnostics reports failing data bits with beeps and coloured bars even when the screen RAM itself is faulty; Paul Farrow's 128K RAM tester then checks every bank.
• Where most of a bank has failed, replace and socket the whole bank.

A RAM chip that gets hot is shorted.

Video faults

Vertical jailbars on RGB

Some 128s show faint vertical banding on plain, light colours on the RGB output that does not appear over RF or composite. Spectrum for Everyone's fix is to replace C28 (22 µF, to the right of the 5-way keyboard connector) with a 47 µF axial capacitor and C7 and C8 (next to IC12 and IC13) with 1 µF ceramics. This reduces the banding a good deal without removing it entirely. On a grey +2 the capacitor to change is C52.

Other picture faults

Sinclair's own video checks are those in the last step of the initialisation sequence above: the RGB outputs, LUM0 from the TEA2000 encoder at IC36 pin 7, the MC1376 sound carrier at IC38 pin 4, and the modulator drive.

• No colour: check the 8.8 MHz colour-encoder clock at IC1 pins 46 and 47, then the TEA2000 (IC36). The 8.8 MHz pair reaches the TEA2000's XTAL1 and XTAL2 pins through C118 and C119 (1 nF) and R99 and R104 (470 Ω), with R90 and R100 (1.5 kΩ) to ground. Ferranti specifies the pair as a small signal, 0.4 to 0.7 V peak to peak about a mean of about 3.6 V, so expect a sub-volt waveform, not a logic-level square wave. Ferranti found returned ULAs with one 8.8 MHz output dead, which gave colour noise or no colour.
• Bright not working: BRIGHT reaches the low bit of each colour input of the TEA2000 (R0, G0, B0) through diodes D20 to D25, with pull-ups R96 to R98 (10 kΩ).
• RGB monitor will not lock vertically: the vertical sync for the RGB socket is made by TR13 (BC239) with R145 (15 kΩ), C132 (2.2 nF) and R146 (1 kΩ) and an IC28 inverter, from the ULA's sync output. These parts were added at Sinclair's third modification in November 1985 and R145, R146 and C132 are not in Sinclair's parts list, so check what the board has. Composite video for the RGB socket comes from TR14 (ZTX313) through R134 (75 Ω).
• For checking the colour outputs, hold BREAK while pressing reset: the 128 shows a colour test of vertical bands in the eight paper colours, bright and normal, with the eight ink colours on them.
• RGB picture correct but no RF picture or sound: check the carrier at IC38 pin 4 and the drive at the TR11 emitter. Sinclair says to replace the modulator (an Astec UM1233 on UK machines) as a complete unit.
• Wavy lines across the screen: Sinclair's 48K manual says to press down on each electrolytic in turn and change the one that makes the lines disappear.
• Corrupt or striped display: RAM. Run a diagnostic ROM.

Sound faults

The 128 has no internal speaker. The ULA's sound (through C123, 1 µF, and R112, 68 kΩ) and the AY's (through C127, 1 µF, and R132) are summed into IC38, an MC1376, which with its tank coil L4 generates an FM sound carrier that is added to the video for the TV. With a monitor, the sound is available from the MIC socket. R132 is 18 kΩ on Sinclair's and Timex's drawings and 39 kΩ in Sinclair's parts list. On boards built to Sinclair's third modification the three AY outputs are mixed through R139, R140 and R141 (1.8 kΩ each) into TR15 (BC239) before reaching the sound line; these parts are not in the service manual's parts list.

• Picture good but no sound on the TV: IC38, L4 and the coupling into the video (R131, C126), then the modulator.

• No sound at all: on the 48K, faulty or missing sound usually means the ULA.
• Beeper works but no AY sound: the AY-3-8912A (IC32). Test with a 128 BASIC PLAY command, which uses the AY, rather than BEEP, which uses the ULA.
• AY audible but much quieter than the beeper: normal for the 128 family.

Tape loading and saving

• EAR: the socket feeds ULA pin 34 through R36 (680 Ω) and C32 (100 nF), with D13 (1N4148) and C35 (10 nF) from pin 34 to ground and R37 (1 kΩ) from the socket side to ground. Inside the ULA the EAR input is a Schmitt trigger with typical thresholds of 0.7 V and 0.85 V, and the CPU reads it on data bit D6.
• MIC: ULA pin 35 drives the MIC socket through C31 (100 nF) and R35 (10 kΩ). The same pin carries the beeper. Ferranti's specification gives the pin three output levels, and Sinclair's manual explains that the save signal is small while BEEP (bit 4 of port 254) drives the pin harder so that the tone is heard on the TV.
• The start-up menu has a Tape Tester option, which shows a bar for setting the playback volume.
• If the external parts check good and the EAR or MIC is still dead, the ULA's cells for those pins are the next suspect.

Keyboard faults

The 128 uses the Spectrum+ membrane, connected by tails at KB1 (5-way) and KB2 (8-way).

• Sinclair's 48K manual: a dead block of five keys points to KB2 circuitry or the 8-way tail; a dead block of eight keys points to KB1 circuitry or the 5-way tail.
• Most keyboard faults are the membrane: worn pads under the keys or a cracked tail.
• Bottom-row keys that do not work can sometimes be cured by removing the tail clamps, cleaning between the layers of each tail around the clamped area, and adding packing under the clamps.
• A key that repeats means the reaction plate is too tight. Slacken the screw nearest that key by a quarter turn.
• Make sure both tails are fully inserted and not kinked.

The full keyboard procedure is on the Sinclair ZX Spectrum 128 General Maintenance page.

Serial port and keypad faults

The RS-232 port and the keypad both use the 1488 driver (IC33) and 1489 receiver (IC34). IC33 needs both the +12 V and −12 V rails, and the keypad is powered from +12 V.

1. If both the serial port and the keypad are dead, measure +12 V at IC33 pin 14 and −12 V at IC34 pin 1. If either is missing, see inverter faults and check the charge pump (C111, C112, D28, D29).
2. If only the keypad is dead on a 6K board, fit 15 Ω at R137.
3. Otherwise check the 6-way BT-type sockets, then substitute IC33 and IC34.

Faults that appear when warm

• Check that the regulator is firmly fixed to the heatsink with thermal grease behind it.
• T.I.S.'s 48K notes say to suspect the heatsink first and then the ULA for faults that appear only when warm, and to look for hairline cracks radiating from the heatsink screws when chasing intermittent faults.
• Tynemouth set aside two +2 boards whose faults developed after warm-up even with every relevant chip swapped for a known good one, and suspected hairline cracks or a connection opening with thermal expansion.

Contacts and random crashes

T.I.S.'s general check-over for the 48K lists loose or overtightened heatsink screws, dry joints, cracked boards, badly fitted ICs, tarnished contacts in IC holders and on the edge connector, and damaged keyboard tails. For intermittent faults it adds checking every holder for tarnish and replacing faulty holders completely. On the 128:

• Reseat socketed ICs and replace any holder with poor contacts.
• Clean the edge connector fingers. Tynemouth found dirty edge connectors on two boards that stopped a DivMMC working.
• Re-tension the EAR and MIC socket contacts if loading is intermittent: push the spring contact down and inwards until it touches the case.

Quick reference

Symptom	Likely cause	See
Keypad dead, everything else works, 6K board	R137 value	Defects
DivMMC or clock-dependent interface fails, version 2 or 2-1 board	/CLK not connected	Defects
Serial port and keypad both dead, no −5 V on the edge connector	Inverter or charge pump	Inverter faults
Board goes into current limit, chips hot	Shorted RAM	RAM faults
Crashes, corrupt display	RAM	RAM faults
Jailbars on RGB only	C28, C7, C8	Jailbars
No colour	8.8 MHz clock, TEA2000	Picture faults
Will not reset, bands, grid or blank picture, supplies and clocks good	ULA	ULA faults
Picture good, no sound on the TV	IC38 MC1376, L4	Sound faults
Upper 64K fails with good RAM chips	CAS delay round IC28	RAM faults
No Z80 clock, ULA pin 39 switching	TR3 and its parts	CPU clock
Oscillator unstable after a crystal change	C120, C124 for the crystal make	Oscillator
No sound at all	ULA	Sound faults
AY silent, beeper works	AY-3-8912A	Sound faults
Block of five or eight keys dead	Membrane tail or its connector	Keyboard faults
Faults only when warm	Heatsink fixing, ULA, cracks	Warm faults
Resets when moved	Power plug or socket	Power socket

Related pages

• Sinclair ZX Spectrum 128
• Sinclair ZX Spectrum 128 General Maintenance
• Sinclair ZX Spectrum 128 Capacitor Replacement Guide
• Sinclair ZX Spectrum 128 Service Manual
• Sinclair ZX Spectrum+ Troubleshooting Guide
• Sinclair ZX Spectrum 128 ULA
• Sinclair ZX Spectrum +2 Troubleshooting Guide
• Sinclair ZX Spectrum Keyboard Membrane Repair