Sinclair ZX Spectrum Troubleshooting Guide

Sinclair's Issue 3 circuit diagram, the reference for the designators and pin numbers below.

Component-level fault finding for the original rubber-key Sinclair ZX Spectrum (16K and 48K), board Issues 1 to 6A. Sinclair published a full service manual for this machine with its own fault-diagnosis procedure, and that procedure — not a reconstruction of it — is what this page follows.

For the later machines, which use different boards and different RAM, see the Spectrum+, Spectrum 128 and Spectrum +2 guides.

Read this first. Three things are commonly got wrong about this machine:

• The input is 9 V, not 12 V, and there is no 7812 on the board. The +12 V and −5 V rails are generated on the board itself by the TR4/TR5 inverter from that 9 V input. A "12 V supply rail" at the regulator input does not exist.
• The ROM is 16 KB, not 8 KB. IC5 is a single 28-pin device covering 0000–3FFF.
• IC1 pin 40 is ground. The ULA's Vcc is pin 14. Pin 13 is a second supply, Vcc2, which sits at about +3 V.

Diagnostic technique

Sinclair's own advice is worth repeating: in a machine like this, where everything depends on everything else, the most reliable method is to compare waveforms and levels against the same point on a known-good board. A careful visual examination first often finds the fault outright — burnt discretes and overheated tracks show up immediately, "as do the attentions of an enthusiastic amateur".

Look for solder splatter, not just dry joints. A hairline short from splatter can measure several ohms and produce convincingly misleading symptoms.

Thermal checks

After a minute or two, feel the ICs. One package noticeably hotter than its neighbours is a strong suspect. Freeze spray is useful in reverse: if behaviour changes as a chip cools, change that chip. Sinclair's own servicing addendum suggests testing at elevated temperature to bring out marginal picture quality and RAM faults.

Substitution and the test clip

IC1 is socketed on every board, so the ULA is easy to swap. For the soldered-in devices Sinclair describes making a test IC: clip a known-good part of the right type (4116 or 4532) onto an IC test clip and bridge it across each suspect device in turn. It is not guaranteed to work, but it saves a great deal of unnecessary desoldering.

Step 1: the supply

Always verify the supply before suspecting the board.

The external pack is unregulated 9 V DC through a 2.1 mm coaxial socket, centre negative. Details, including why a healthy pack reads well above 9 V off-load, are in Sinclair ZX Spectrum General Maintenance.

Sinclair's basic checks, in order of priority

Function	Test point	Expect
Regulator input	Positive side of C50	+9 V DC ±2.0 V. Below +7 V the regulator will not function correctly
Regulator output	Positive side of C34	+5 V DC ±0.25 V, no discernible ripple
+5 V	IC6 pin 9	+5 V DC ±0.25 V, no discernible ripple
−5 V	IC6 pin 1	−5 V DC (setting-up tolerance −5.5 V to −4 V)
+12 V	IC6 pin 8	+12 V DC (setting-up tolerance ±1.2 V)
+12 VA, to IC14 only	LT end of C52	+12 V DC (setting-up tolerance ±1.2 V)
Clock pulses	IC1 pin 32, TR3 base and collector, IC2 pin 6	Square wave at +5 V amplitude
Address and data lines	Right-hand side of R17–R23 and R1–R8	Waveform of 3.5 V amplitude

If those all pass, check IC1 pin 14, IC2 pin 11 and IC5 pin 28 for +5 V, and IC1 pin 40, IC2 pin 29 and IC5 pin 14 for ground. If everything is still good and IC1 is the plug-in type, replace IC1. If that does not cure it, check the address and data lines of IC1, IC2 and IC5 for active data.

Note an inconsistency in the manual. Section 5.3.1 describes IC1 pin 32 as a 14 MHz square wave, while section 6.3.2 gives pin 39 as the 14 MHz point and pin 32 as the CPU clock. Section 6.3.2 is the self-consistent reading: X1 connects at pin 39, and pin 32 drives TR3 and then IC2 pin 6.

Control lines

Where a control line is suspect, compare origin and destination against a good board:

• IC2 (Z80A): pin 16 INT, pin 17 NMI, pin 19 MREQ, pin 20 IOREQ, pin 21 RD, pin 22 WR, pin 24 WAIT, pin 25 BUSRQ, pin 26 RESET.
• At a RAM device: pin 4 RAS, pin 3 WRITE, pin 15 CAS.

Edge connector

+5 V is 3A, +9 V is 4A, 0 V is 6A/7A/14A, −5 V is 20B, +12 V is 22B and −12 V is 23B. RESET is 20A, ROMCS is 25A, VIDEO is 15A.

Dead, with smoke on switch-on

This is the classic Spectrum failure and it has a specific cause. Failure of the internal power supply transistor TR4 has been found to be a significant cause of problems — Sinclair's own words — and it takes the lower RAM bank with it often enough that the DC-DC converter modification is recommended for any unit in which either TR4 or the 16K DRAM has blown.

Sinclair's procedure:

1. TR4 shorted base to collector — remove TR4.
2. Short still present with TR4 removed — check TR5 and C44, check the track visually, then check TR4 itself.
3. The replacement TR4 blows again — change TR4 and TR5, even though TR5 tests good, and confirm the TR5 circuit components are to the latest specification.

After a TR4 failure, check the lower RAM before assuming the machine is fixed.

Do not follow the DC-DC converter modification as printed in section 9.2 of the manual. The hosted edition carries an editor's note stating that those instructions are incorrect; they are reproduced there only for fidelity to the original. The section 9.13 parts list is the reliable record of what each later board actually carries.

Will not initialise

A working machine shows a clear screen with the copyright message in the lower left. If it does not:

1. Work the basic-checks table above, in order.
2. Check the supply pins and grounds of IC1, IC2 and IC5.
3. If IC1 is socketed, replace it.
4. On a 48K machine, isolate the 32K expansion RAM. Where the ICs are soldered, Sinclair's suggestion is to remove the +5 V from IC25 by very carefully cutting the track to IC25 pin 16. If that clears the fault, the expansion RAM is at fault and must be narrowed down further.
5. If the fault survives that, start changing ICs in the order IC13 down to IC6 (lower RAM), then IC1 (ULA), then IC2 (Z80A), powering up after each change.

RAM faults

The lower 16K is eight 4116 devices, IC6–IC13, each contributing one bit. They need +5 V, −5 V and +12 V — which is why a rail fault and a dead RAM look identical, and why the rails must be checked first.

The upper 32K on a 48K machine is eight 32k-bit devices, IC15–IC22 (TI 4532-3, TI 4532-4 or OKI MSM3732), with IC23–IC26 providing the address multiplexing and read/write control.

Sinclair's memory check

Where the machine initialises but RAM is suspect, this finds the faulty device.

Type:

PRINT PEEK 23732+PEEK 23733*256

The answer is the last valid memory location, and should be 65535 on a 48K machine or 32767 on a 16K machine. If a different value n is returned, the faulty location is n+1. A value below 32767 puts the fault in the original 16K.

Then, using that address — the manual's worked example uses 43201:

POKE 43201,85: PRINT PEEK 43201     (answer A)
POKE 43201,170: PRINT PEEK 43201    (answer B)

If answer A is not 85, or answer B is not 170, look the returned value up in the table.

RAM bit-to-IC fault table

Data 85 returns	Data 170 returns	Size of error	Error bit	Faulty IC below 32768	Faulty IC above 32767
84	171	1	0	IC6	IC15
87	168	2	1	IC7	IC16
81	174	4	2	IC8	IC17
93	162	8	3	IC9	IC18
69	186	16	4	IC10	IC19
117	138	32	5	IC11	IC20
21	234	64	6	IC12	IC21
213	42	128	7	IC13	IC22

Where there is more than one faulty location, the first fault has to be repaired before you can proceed.

No video

Sinclair's sequence:

• No video at all — change IC1; check TR1 and TR2.
• No video after the ULA was changed — check IC1 and its socket visually; check for a short between C65 and R53; repeat the power supply check.
• No +5 V input — check the voltage drop across R64; check the +5 V regulator output and the +9 V regulator input; check the PWR socket is not shorted; check the power pack.
• No video input at IC14 pin 12 — check the waveform at IC14 pin 13, and if that is correct check TR1 and TR2. Check the waveforms at IC1 pins 15, 16 and 17. On Issue 1 and Issue 2 boards check VR1 and VR2. Then change IC1. If the +5 V and the video input are both correct, change the modulator.

The modulator is replaced as a complete unit; it is not a serviceable assembly.

Video present but wrong

Sinclair's video fault list

Symptom	Action
Dark screen	Check the tuning; change the modulator
Dim "Sinclair" logo	Check +12 V; check C65 for high impedance; check R52; check +5 V on IC14 pin 3
Wavy lines across the screen	Press down on each electrolytic in turn — if the lines disappear, change that capacitor. Then check crystal X2, then TR1 and TR2, then IC14
Self-resetting	Check TR6
Random dynamic squares	Change C54
Paper area too low	Change IC1
Diagonal lines in the border area	Change R47
Distorted video	Check for +3 V at IC1 pin 13; change C30
Corrupt paper after a ULA change	Check the IC1 socket; check the PWR socket is not intermittent; check IC1 pin 14 for +5 V with no discernible ripple
No clock at IC2 pin 6	If the amplitude is low, check R25. Check TR3. Check IC1 pin 32, then IC1 pin 39 for 14 MHz, then crystal X1, then IC1 pin 13 for +3 V

Colour faults

Colour tuning is adjustable only on Issue 1 and Issue 2 boards, which carry the presets VR1/VR2 and the trimmers TC1/TC2. Issue 3 made the colour circuitry self-tuning and removed them, so a colour fault on an Issue 3 or later board is a component fault, not a maladjustment.

Colour fault list

Symptom	Action
Pin-cushioned or unstable "pin screen"	Change crystal X2
Blue haze around the logo	If untunable on Issue 1 or 2, change IC14
Screen too yellow	Change IC14
Intermittent colour	Change TC2 (Issue 1 and 2); change IC14
Red, blue or green logo	Change IC1
Washed-out colours	Check TR2
Alternate dark lines	Change IC14
VR1 or VR2 will not adjust the IC14 pin 13 waveform	Monitor IC14 pin 2 while trimming VR2 and IC14 pin 4 while trimming VR1; change IC14
No colour after trimming TC2, IC14 pin 17 not oscillating	Check crystal X2; change IC14; change X2
No colour after trimming TC2, IC14 pin 17 oscillating but will not tune	Check for a broken track; change TR2; change IC14

On Issue 1 and Issue 2 boards, VR1 and VR2 null the voltages on IC14 (LM1889) pins 4 and 2 relative to pin 3. The factory settings are 130 ±20 mV on pin 4 and 75 ±20 mV on pin 2; the optimum service values are +50 mV and −50 mV respectively, with an overall usable range of +45 to +150 mV and −45 to −100 mV. The sub-carrier oscillator should be 4.433619 MHz ±50 Hz.

IC14 (LM1889) pin signals

Taken at switch-on with the copyright logo displayed, no keys pressed, all DC levels positive with respect to 0 V.

Pin	Function	Signal
1	Chroma lead	Sine wave, 500 mV, 0.2 µs, 9.5 V DC
3	Chroma bias	4.8 V DC
5	Ground	0 V
12	Video input	12 V DC unregulated
15	Sound tank	Linked, 12 V DC
17	Chroma oscillator output	Square wave, 0.2 µs, 4 V peak to peak, 8.0 V DC
18	Chroma lag	Sine wave, 500 mV p-p, 2 µs, 9.5 V DC
6–11	—	Not connected

Keyboard

The keyboard is an 8 × 5 matrix: eight blocks of five keys horizontally, five blocks of eight vertically. That geometry gives an immediate diagnosis:

• A block of five keys dead — the fault is in the KB2 circuitry or the 8-way membrane.
• A block of eight keys dead — the fault is in the KB1 circuitry or the 5-way membrane.

Sinclair's fault list:

• KB1 fault — check for a short across two or more of the five tracks; change the membrane.
• KB2 fault — check for a short across two or more of the eight tracks; change the membrane.
• Dead keyboard, wrong response, or response too fast — change IC1. Check for a track short first: a short holds one key permanently on and disables the rest.

Membranes fail by cracking where they bend. Withdraw the tails straight and refit them fully without kinking. See Sinclair ZX Spectrum Keyboard Membrane Repair.

Sound

No speaker output: check the load input waveforms, then TR7, then D9, then change the speaker. The speaker is a 40 Ω TV type driven by TR7 (ZTX450) through D9.

Tape

SAVE or LOAD failing — Sinclair's servicing addendum is to inspect the EAR and MIC jack sockets for deformation of the contact springs before looking at the circuit.

The ULA's MIC/TAPE output is pin 28.

Expansion

Interface 1 not recognised — inspect the edge connector for damaged pins, then check the M1 signal from the CPU. M1 is used only with Interface 1, so a failure in it is completely invisible when the Spectrum is used on its own.

The on-board ROM is disabled by an external device pulling ROMCS (expansion port pin 25A) high; that is the mechanism Interface 1 uses to substitute its own ROM.

Self-inflicted faults

Things that break a working machine if done in the wrong order, or with the wrong part:

• TR4 replacement type KTC2236A-Y has a different pinout from every other approved replacement — pin 1 is the emitter, pin 2 the collector and pin 3 the base, so base and collector are reversed. Fitting one as if it were a ZTX650 puts the inverter straight into the fault it was replaced for.
• If TR3 is a KSC839, R24 must be 15 K, not the 1 K value used with other types.
• IC3, IC4, IC25 and IC26 must not be National Semiconductor 74LS157s. Sinclair marks this explicitly in every parts list.
• All eight expansion RAM devices must be the same type, and the board links must match. On an Issue 2 board, link 3 is fitted for 4532-3 parts and link 4 for 4532-4; on Issue 3 two links in the grid between the edge connector and the MIC socket select manufacturer and enable polarity.
• From Issue 4A the ULA must be 6C001-7 or later, because two spare gates of IC24 were brought into the RASL path to improve timing. An earlier ULA in that board will not behave.
• If you change the ROM to a different manufacturer, move the H/N links. They are between the ROM and the heatsink on Issue 1, under the raised part of the heatsink on Issue 2, and next to the loudspeaker on Issue 3.
• Refitting the heatsink carelessly. On some Issue 3 boards the heatsink hole allows enough play to foul the edge connector. Fit it away from the connector, and use thermal grease behind the regulator body.
• Cutting the track to IC25 pin 16 to isolate the expansion RAM is Sinclair's own suggestion, but it is destructive — plan the repair before you make the cut.

Things that are not faults

• The regulator coil buzzing faintly. Only an excessively noisy coil is a fault, and the remedy is to change it.
• The picture shifted one character position to the left on a machine fitted with a 6C001 ULA. That timing change was deliberate, made to improve performance with certain television sets.
• A warm regulator and heatsink.
• An original power pack reading well above 9 V with nothing connected. It is unregulated.

System test

Sinclair's production test used the ZXTP test tape, loaded conventionally with the keyboard connected and the machine running from its own supply. It exercises everything except SAVE, which is tested by typing in a short program, saving it and verifying it.

Related pages

• Sinclair ZX Spectrum
• Sinclair ZX Spectrum General Maintenance
• Sinclair ZX Spectrum Capacitor Replacement Guide
• Sinclair ZX Spectrum Keyboard Membrane Repair
• Sinclair ZX Spectrum Service Manual
• Sinclair ZX81 Troubleshooting Guide
• Recommended Tools