Sinclair ZX Spectrum General Maintenance

Sinclair ZX Spectrum 48K main board, Issue 3B, with the heatsink removed to expose the regulator and the DC-DC converter area at the upper right.

Routine care for the original rubber-key Sinclair ZX Spectrum (16K and 48K), board Issues 1 to 6A. Sinclair's ZX Spectrum Service Manual and the Issue 3 circuit diagram give the test points, part numbers and procedures.

For the later machines see Spectrum+, Spectrum 128 and Spectrum +2.

Sinclair asks for normal anti-static precautions when handling ICs and a suitably earthed, low-power soldering iron.

What is on the board

ZX Spectrum 16K/48K main integrated circuits

Ref	Part	Notes
IC1	Ferranti ULA, 40-pin	Video, keyboard scanning, tape I/O, chip selects and RAM multiplexing. Fitted in a socket
IC2	Z80A or µPD780	The CPU
IC3, IC4	74LS157	Address multiplexers for the lower RAM. Not National Semiconductor parts
IC5	Spectrum ROM, 28-pin	A single 16 KB ROM at 0000–3FFF, from NEC or Hitachi. Links marked H and N on the board select which
IC6–IC13	4116, 150 ns	Lower 16K of RAM. Eight 16k-bit DRAMs
IC14	LM1889	Colour encoder
IC15–IC22	TI 4532 or OKI MSM3732	Upper 32K of RAM, 48K machines only. Eight 32k-bit DRAMs
IC23, IC24	74LS32, 74LS00	48K machines only; not used on Issue 5 and 6A
IC25, IC26	74LS157	48K machines only, not National parts; not used on Issue 5 and 6A
IC27	Mullard ZX 84/01	Issue 5 and 6A only: a second ULA replacing six 74LS packages
IC28	74LS04	Issue 5 and 6A only
REG	7805	5 V regulator, bolted to a heatsink with a 4BA screw, shakeproof washer and nut
MOD	Astec UM1233	UHF modulator, channel E36 vision carrier (UK)
TR4 / TR5	ZTX650 or TIPP31 / ZTX213	The inverter that generates the −5 V and +12 V rails
X1 / X2	14.000000 MHz / 4.433619 MHz	Master and chrominance sub-carrier crystals
KB1 / KB2	5-way / 8-way	Keyboard tail connectors, Molex 5229
Loudspeaker	40 Ω, TV type	Driven by TR7 (ZTX450) through D9

Which ULA

The 16K/48K Spectrum was fitted with several ULA revisions, and they are not freely interchangeable between board issues.

16K/48K ULA revisions

ULA	Boards	Notes
5C102E	Issue 1 and a few Issue 2	Timing fault cured by a 74LS00 on a small board with flying leads (the "cockroach" modification). Fitted to about 40,000 units
5C112E, -2, -3	Issue 2	Needs the transistor modification (TR6, ZTX313) or the diode-and-resistor modification. TR6 is part of the board layout from Issue 3
6C001E-5	Late Issue 2 only	Rare
6C001E-6	Late Issue 2, most Issue 3 and 3B
6C001E-7	All issues; required from Issue 4A	The final 48K revision and the only one documented to work on every 48K board

The 7K010E is not a 16K/48K part. It belongs to the Spectrum 128 and the grey +2, which output RGB where the 48K ULAs output YUV.

Several resistor values change with the ULA fitted: with a 6C001 in an Issue 2 board, R47 becomes 1 K, R49 10 K, and R56 and R63 470 R.

Ferranti's records of the ULA types

Ferranti made the ULA, and its files from 1981 to 1984 are in the Ferranti Archives 2026; the chip is covered in Sinclair ZX Spectrum ULA. Ferranti's own lineage chart of the Spectrum ULA runs:

• UK (PAL): 5C102; 5C112 (a logic change); then the 6C001 Mk3, Mk5 ("customer mods"), Mk6 ("new R64 osc") and Mk7 ("customer mod delays").
• USA (NTSC): 5C112; 5C114 (a logic change); then the 6C011, starting at Mk2 (derived from the 6C001 Mk3 with a logic change and modifications), followed by Mk3 ("new R64 osc") and Mk4 ("delays"). In May 1984 Ferranti was ordered to bring the 6C011's timing up to that of the 6C001 Mk7.

On the chips and in Ferranti's test notes the mark is a suffix: 6C001-5, 6C001-6 and 6C001-7, written "6C-001E-7" on Timex's Issue 6A drawing. Ferranti's 1983 notes also distinguish the 5C112-1 (date code 8234) from the 5C112-2 (date code 8301), which had higher internal resistor values and drove the RAS line less strongly, and its timing photographs include a 5C112-3 marked 8320. Sinclair's parts list for Issues 4A and 4B gives the ULA as "6C001-7 (6C011)".

The Mk7 change, agreed between Ferranti and Sinclair on 29 September 1983, added four inverters to the CAS path (about 22 ns) and two to the address-select path (about 11 ns), after fast 6C001-6 parts had failed when cold; see the troubleshooting guide.

Sinclair's manual says the 6C001 "alters the timing of the colour burst signal" and shifts the picture one character to the left. In January 1983 Sinclair asked Ferranti for the change "concerning moving the colour burst signal" made on the 6C001 PAL ULA to be repeated on the NTSC ULA.

Which ULA each board drawing calls for

Period board drawings in the Ferranti archive

Drawing	Date	ULA	Other parts named
Sinclair circuit diagram, unnumbered, with pencil modifications	© 1982; annotated to 15 October 1982	Labelled "G. Ex."; modification 1 marked "5C102 only"	Lower RAM NEC µPD416C-3 (4116, 150 ns); ROM Hitachi HN613128P
Sinclair "ZX Spectrum Circuit Diagram - Issue Two", SRC 102	16 August 1982	"ULA", type not given	Lower RAM 4116, 150 ns; "optional 32K RAM" of 4532, 200 ns
Timex "ZX Spectrum Circuit Diag. Issue 3", 356-201891-02	May 1983; change 02 of 10 June 1983 added TR7	"5C112E or 5C112E-2 or 5C112E-3 (see mod. shts)"	Lower RAM 4116 (ITT or NEC, 150 ns) or National MM5290-2; upper RAM TMS4532-20NL3 or -NL4; ROM HN613128P or NEC D231286, "250nS"
Timex "ZX Spectrum Circuit Diag. Issue 6A", Ferranti version, 356-280106-01	25 September 1984	6C001E-7	ZX8401 (IC27) in place of the 74LS157s; 74LS04 (IC28); UM1622 modulator as an option

Sinclair's parts list gives the 5C102 for Issue 2, the 5C112 as the Issue 2 modification and the 6C001 for Issue 3, while Timex's Issue 3 drawing of June 1983 calls for the 5C112E family. Sinclair's telex of May 1983 (below) gives resistor values for a 6C001 on both Issue 2 and Issue 3 boards.

Factory modifications on early boards

Sinclair's manual says the first ULA, the 5C102, "has a timing fault which was cured by connecting a 74LS00 IC mounted on a miniature board and spider", fitted to about 40,000 units, and that the 5C112 which replaced it has "either a diode or resistor or transistor connected to it" instead. The Ferranti archive has a 1982 Sinclair print on which these changes are pencilled as numbered modifications, and a separate sketch of the 74LS00 patch:

• Mod 1, marked "5C102 only": the 74LS00 patch that collectors call the "dead cockroach". The sketch numbers its connections 4, 5, 20 and 11, which are the Z80's A14, A15, IORQ and +5 V pins, and 36 and 37, the ULA's A14 and A15 inputs. As drawn, the gates pass A14 and A15 to the ULA unchanged except during an IORQ cycle, when the ULA sees A14 high and A15 low.
• Mod 2: a ZTX312 transistor with its base on the Z80's A0 (IC2 pin 30), its collector on IC2 pin 11 (+5 V) and its emitter on the ULA's IORQ input (IC1 pin 33). Sinclair's manual gives the same connections for the 5C112 "transistor mod", with TR6 as a ZTX313. Its alternative, the "diode/resistor mod", is a 4.7 kΩ resistor from +12 V to IC2 pin 30 and a 1N4148 from IC2 pin 30 to IC1 pin 33.
• Mod 3, dated 15 October 1982: the CPU clock from IC1 pin 32 through 1 kΩ with 100 pF in parallel to the base of TR3 (ZTX313), with a 1 kΩ pull-up, TR3 driving the Z80's clock at IC2 pin 6. Sinclair's manual lists the matching Issue 1 change (D14 replaced by C67, 100 pF; R24 from 3k3 to 1k; R27 from 680R to 470R; R73, 1k, added from IC1 pin 32 to +5 V) and says it was built in for all 5C112E-2 and later ULAs.

Sinclair's list of Issue 1 modifications also includes a 100 pF capacitor from RAS to ground, "necessary only when using ULA 5C102". A 74LS00 on a small board, an extra transistor or diode near the Z80, or these clock-circuit values on an Issue 1 or early Issue 2 board are Sinclair's own modifications, fitted in production or by service engineers, not later bodges.

The same 1982 print notes the NTSC version of the colour encoder: X2 changed to 3.579545 MHz, 4.7 pF added across C37 and 10 pF across C38.

Fitting a 6C001 to an Issue 2 board

Sinclair formalised the resistor changes in a telex to Ferranti on 26 May 1983, amendment 1 to Engineering Instruction/Spectrum 309, which "formalises the position concerning the use of 6C001 ULA in Issue 2 PCB's":

R47, R49, R56 and R63 by board issue and ULA

Board issue	2	2	3
ULA	5C112	6C001	6C001
R47	220 Ω (printed "2020R" in the telex)	1 kΩ	1 kΩ
R49	8.2 kΩ	10 kΩ	10 kΩ
R56	220 Ω	470 Ω	1 kΩ
R63	220 Ω	470 Ω	1 kΩ

The telex prints the 5C112 value of R47 as "2020R", which is not a preferred value; Sinclair's manual gives 220 Ω in the same table. The telex adds that an Issue 2 board with a 5C112 "will operate with resistors as for Issue 2 with 6C001, but this may affect setting up procedure".

All four resistors are in the video circuit. On Sinclair's Issue Two and Issue 3 circuit diagrams, R47 runs from the ULA's Y-bar (luminance) output, pin 17, to 0 V; R56 and R63 run from its U and V outputs, pins 15 and 16, to 0 V (on Issue 2 the colour presets VR1 and VR2 also connect at these points); and R49, with R50, sets the bias on the base of TR1, the first video transistor, which takes the luminance signal through R48.

The colour clamp circuit

In January 1983 Sinclair sent Ferranti "the clamp circuit we wish to use on the Spectrum to remove the tuning pots": two BC184 transistors clamping the U and V feeds to the LM1889 (IC14) on the sync pulses of the Y output, on a small board "over the area where the 2 x 2K2 pots are located". Its fitting notes remove VR1, VR2, R38, R39, R40, R41 and C36, set C52 and C53 to 150 pF, and add 470 Ω from LM1889 pin 3 to +5 V.

Sinclair's parts lists show VR1, VR2, R38 to R41 and C36 fitted on Issue 2 and absent on Issue 3, which adds TR8 and TR9 (BC184 or BC549B). The Issue 3 circuit diagram has the clamp on the main board: TR8 and TR9 (BC184) with R74 and R75 (10 kΩ) and R76 and R77 (1 kΩ) on the U and V feeds to IC14 pins 4 and 2, R78 (470 Ω) from +5 V to pin 3, and 150 pF at C52 and C53. This is the colour circuit that, in the words of Sinclair's manual, Issue 3 introduced to remove "the need for potentiometers and trimmers and the associated adjustment".

Upper 32K and the RAM select links

On Issue 1 boards the upper 32K is a plug-in board on special DIL connectors. The Ferranti archive has drawings, photocopies and a circuit of one, Sinclair's "RAM 32 Issue 1" (© 1982): eight TMS4532-20NL4 RAMs, two 74LS157 multiplexers, a 74LS32 and a 74LS00, two RC delays (220 Ω with 100 pF and 220 Ω with 47 pF) in the control logic, and a 47 nF decoupler at each RAM. Its two 16-way sockets carry only A0–A15, D0–D7, MREQ, RD, WR, +5 V and 0 V, and one multiplexer input is tied to +5 V. Sinclair's manual gives the RAM select level as 0 V for 4532-3 parts and +5 V for 4532-4.

Sinclair's Issue Two drawing puts the same circuit on the main board as "optional 32K RAM", with a "RAM type select" link to 0 V or +5 V. From Issue 3 the select links form a small grid marked H, L, 3, 4, TI and OKI, between the edge connector and the MIC socket; Timex's Issue 3 drawing shows it "as on PCB edge, bottom view".

Regular cleaning

Case and keyboard

• Wipe the case with a damp cloth and mild detergent.
• The rubber key mat can be removed and washed in warm soapy water. Dry it completely before refitting.
• The keyboard membrane is the fragile part. Do not flex or fold it, and withdraw the tails straight out of KB1 and KB2. When refitting, Sinclair's instruction is that the ribbon connectors must be fully inserted and not kinked.
• If the keyboard template lifts at the top edge, Sinclair's fix was a 6 mm strip of double-sided tape along that edge, which early boards lacked.

See Sinclair ZX Spectrum Keyboard Membrane Repair for membrane work.

Board

• Disconnect the power and all leads first.
• Remove dust with a soft antistatic brush or dry air.
• Reseat the socketed ICs. IC1 is socketed on every board, and on a 16K machine the expansion-RAM positions are sockets too.
• Clean the edge-connector fingers and check for pitting or bent contacts.

Power supply

Inside a ZX Spectrum power pack: the mains transformer, with two electrolytic capacitors on the small board below it.

The Spectrum takes an unregulated 9 V DC supply through a 2.1 mm coaxial socket. The ZX81 uses a 3.5 mm jack; the prototype Spectrum did too, and pads for a 3.5 mm socket remained on the board up to Issue 4B.

External supply

Property	Value	Source
Connector	2.1 mm coaxial socket
Polarity	Centre negative, outer positive: the opposite of most modern adaptors. Getting it wrong can seriously damage the machine
Type	DC, unregulated. Mains transformer, full-wave rectifier and capacitive smoothing, with a thermal fuse at the transformer input
Original model	Sinclair UK1400, rated 1.4 A. Reads about 12.5 V unloaded, because it is unregulated
Sinclair's load specification	At 215 V AC mains input and 1.4 A output, the trough should not fall below 7.0 V; at 265 V AC and 600 mA, the peak should be less than 14 V
Measured at the board	+9 V ±2.0 V at the positive side of C50. Below +7 V the regulator will not function correctly

An unloaded original pack reading about 12.5 V is normal for an unregulated supply. Judge it by the +9 V ±2.0 V measured at C50 with the machine running.

The UK1400 is not suitable for a Spectrum 128 or a +2, even though the plug fits: those machines draw more current.

The pack is repairable. The primary side is at mains potential and the smoothing capacitor holds charge after unplugging, so unplug it, wait, and confirm the capacitor is discharged before putting a hand or a probe inside. Replace a damaged lead completely. A Television magazine repair column reported several packs whose leads had gone short-circuit and whose rectifier diodes had caught fire and burnt the panel.

The rails the machine makes for itself

The board derives four supplies from the 9 V input:

• +5 V from the 7805, for the logic, the ULA and the modulator.
• −5 V and +12 V for the 4116 lower RAM, generated by the TR4/TR5 inverter, which lifts the 9 V input above 12 V; D15 and C44 rectify and smooth it.
• +12 VA for the colour circuits, taken through R62 and C45 so that it is free of noise from the RAM, and also present on the expansion port.

Sinclair calls failure of TR4 "a significant cause of problems", and recommends the DC-DC converter modification for units in which TR4 or the 16K DRAM has blown.

Supply, clock and test points

Voltage and signal checkpoints

Point	Expect	Source
IC6 pin 9 (lower RAM)	+5 V, ±0.25 V, no discernible ripple
IC6 pin 1	−5 V (acceptable range −5.5 V to −4 V)
IC6 pin 8	+12 V, ±1.2 V
LT side of C52	+12 VA, ±1.2 V
Positive side of C50	+9 V ±2.0 V (regulator input)
Positive side of C34	+5 V ±0.25 V (regulator output)
IC1 (ULA) pin 14	+5 V (Vcc)
IC1 (ULA) pin 40	Ground, not a supply pin
IC1 (ULA) pin 13	Vcc2, about +3 V, fed through R34 (15 Ω, 0.5 or 1 W) and decoupled by C30. Not a 5 V pin
IC2 (Z80A) pin 11 / pin 29	+5 V / ground
IC5 (ROM) pin 28 / pin 14	+5 V / ground
IC1 (ULA) pin 39	14 MHz, from crystal X1
IC1 pin 32, TR3 base and collector, IC2 pin 6	The CPU clock, a square wave at +5 V amplitude
Right-hand side of R1–R8 and R17–R23	Data and address waveforms of 3.5 V amplitude

Section 5.3.1 of the manual labels 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 consistent reading, because pin 39 is where crystal X1 connects and pin 32 feeds TR3 and then the CPU clock input at IC2 pin 6. The manual's fault list refers to the CPU clock loosely as "3 MHz"; X1 is specified as 14.000000 MHz.

Edge connector supplies

From Sinclair's Issue 3 diagram: 3A is +5 V, 4A is +9 V, 6A, 7A and 14A are 0 V, 20B is −5 V, 22B is +12 V and 23B is −12 V. RESET is 20A, ROMCS is 25A and VIDEO is 15A.

Full contact list

Sinclair's Issue Two, Issue 3 and Issue 6A diagrams letter the underside A and the component side B. The edge-connector tables on Timex's Issue 3 and Issue 6A drawings letter them the other way round, underside B and component side A, so Timex's +5 V contact is "3B". Timex's Issue 3 drawing still labels its +12 V and −5 V rails 22B and 20B, in Sinclair's lettering, so read each drawing against its own table. The contact numbers and signals are the same on all of them, and Ferranti's 1983 test-box wiring (A3 +5 V, A20 RESET, B17 IORQ) and Sinclair's manual (ROMCS on 25A) use Sinclair's lettering. The table uses Sinclair's.

Edge connector (Sinclair lettering)

Contact	A (underside)	B (component side)
1	A14	A15
2	A12	A13
3	+5 V	D7
4	+9 V	Not used
5	Slot	Slot
6	0 V	D0
7	0 V	D1
8	CPU clock	D2
9	A0	D6
10	A1	D5
11	A2	D3
12	A3	D4
13	IORQULA	INT
14	0 V	NMI
15	VIDEO	HALT
16	Y	MREQ
17	V	IORQ
18	U	RD
19	BUSREQ	WR
20	RESET	−5 V
21	A7	WAIT
22	A6	+12 V
23	A5	−12 V (Issue 2 and Issue 3 diagrams); 12 V AC (Issue 6A, and Timex's Issue 3)
24	A4	M1
25	ROMCS	RFSH
26	BUSACK	A8
27	A9	A10
28	A11	Not used

On Sinclair's Issue 3 diagram the −12 V line to 23B is taken from the inverter before the rectifiers; Sinclair's Issue 6A diagram labels the same contact 12 V AC.

Capacitor health

Sinclair gives no replacement interval for the capacitors. Its test for a suspect electrolytic: with wavy lines on the picture, press each electrolytic in turn and change the one that makes the lines go.

Which capacitors are fitted depends on the board issue; the full per-issue list is in the Sinclair ZX Spectrum Capacitor Replacement Guide. See also Capacitor Failure Symptoms.

Known failure-prone components

Failure-prone parts

Component	Symptom	First check
TR4/TR5 inverter	Dead with smoke on switch-on; no −5 V or +12 V; lower RAM damaged	Sinclair names TR4 failure as a significant cause of problems. If TR4 is short base-to-collector, remove it; if the short persists, check TR5 and C44. If a replacement TR4 blows again, change TR4 and TR5 together and check that the TR5 circuit components are to the latest specification
ULA (IC1)	No video, wrong colours, dead or wrong keyboard response	It is socketed, so substitution is easy. Check the socket itself and pin 14 for +5 V with no ripple before condemning the chip
Lower RAM (IC6–IC13, 4116)	Will not initialise; corrupt display	These chips need all three rails. Check +5 V, −5 V and +12 V at IC6 before changing RAM
Upper RAM (IC15–IC22)	48K machine behaves as 16K, or crashes	Can be isolated for testing. All eight must be the same type, and the board links must match
Keyboard membrane	Whole rows or columns dead	The keys are wired in eight blocks of five and five blocks of eight. A dead block of five points to KB2 or the 8-way membrane; a dead block of eight points to KB1 or the 5-way membrane
7805 regulator	Faults that appear when warm	Check it is making good contact with its heatsink and that thermal grease was used. On some Issue 3 boards the heatsink hole allows enough play to foul the edge connector
Regulator coil	Audible whine	A faint buzz is normal. Sinclair's remedy for an excessively noisy coil is to change it
Jack sockets	SAVE or LOAD fails	Sinclair's servicing addendum: inspect the jack sockets for deformation of the contact springs
Edge connector	Interface 1 not recognised	Inspect for damaged pins, and check the M1 signal. Only Interface 1 uses it, so a failure goes unnoticed when the Spectrum is used alone
Speaker	No sound	Driven by TR7 through D9. Check the input waveforms, then TR7, then D9, then the speaker itself
ULA (IC1), 6C001-7	A block of eight keys dead, or wrong characters, especially with a Spectrum+ keyboard; crashes only when cold or only when warm	Swap the ULA and see whether the fault moves with it. Measure +5 V at IC1 pin 14 itself: Ferranti found 6C001 ULAs in 1984 that worked at 5 V and above and failed below. See the troubleshooting guide

Preventive maintenance

1. Measure the supply and confirm that it is centre-negative before plugging it in.
2. Remove the five self-tapping screws from the base, then the single board screw.
3. Inspect the board for corrosion, cracked tracks, burnt components and dry joints, particularly around the regulator, the power socket and the edge connector.
4. Look for solder splatter as well as dry joints. Sinclair warns that hairline splatter can measure several ohms and produce very misleading symptoms.
5. Inspect the electrolytics for bulging or leakage.
6. Reseat IC1 and any other socketed devices. Use the correct removal and insertion tools and avoid handling the pins.
7. Clean the edge-connector fingers.
8. Check the regulator is still firmly bolted to its heatsink, and that the heatsink cannot touch the edge connector.
9. Refit the keyboard tails fully and without kinking them.
10. Run the machine and confirm the rails against the table above.

Things that are not faults

• The ULA runs quite warm; the other chips run cold.
• The inverter coil buzzes faintly when the DC-DC converter is working. T.I.S. treats a missing buzz as the fault; Sinclair treats only an excessively noisy coil as one.
• An original power pack reads well above 9 V off-load, because it is unregulated.
• On a 16K machine the expansion-RAM positions are empty sockets from the factory, ready for the upgrade.
• A picture shifted one character to the left on a machine with a 6C001 ULA. Sinclair made the timing change to improve compatibility with certain television sets.
• A seal with a four-digit code over the central screw hole at the rear of the base. Sinclair fitted it from August 1984.

Tools

See Recommended Tools. Sinclair's minimum list for this machine is an oscilloscope with a ×10 probe (rise time 0.02 µs/cm), a variable 0–30 V bench supply, a general-purpose multimeter, a mono cassette recorder with record and playback, a colour television, a ZX Printer, the ZXTP test tape, a blank tape and double-sided tape for the keyboard template. On a 9 V bench supply a 16K Spectrum draws about 500–700 mA and a 48K about 700–900 mA.

Common faults and first checks

Symptom	Where to look first
Dead, smoke on switch-on	TR4 shorted base to collector; then TR5 and C44
Does not initialise	Work Sinclair's basic-check table in order: regulator input at C50, regulator output at C34, the three rails at IC6, the clock, then the address and data lines
No video	Change IC1; check TR1 and TR2; check the modulator's +5 V and the video input at IC14 pin 12
Wavy lines	Press each electrolytic in turn; then crystal X2, TR1/TR2, IC14
Dim copyright logo	Check +12 V, then C65 for high impedance, then R52 and IC14 pin 3
Self-resetting	Check TR6
Random dynamic squares	Change C54 (470 pF). Sinclair's parts lists show C54 on Issue 2 boards only
Dead keyboard or wrong keyboard response	Change IC1. For a fault on one tail, check for a short between two or more of its tracks, then change the membrane
Corrupt display, suspected RAM	Use Sinclair's memory check and the bit-to-IC table in the Sinclair ZX Spectrum Troubleshooting Guide

For the full symptom-by-symptom treatment, see the Sinclair ZX Spectrum Troubleshooting Guide.

Related pages

• Sinclair ZX Spectrum
• Sinclair ZX Spectrum Capacitor Replacement Guide
• Sinclair ZX Spectrum Troubleshooting Guide
• Sinclair ZX Spectrum Keyboard Membrane Repair
• Sinclair ZX Spectrum Service Manual
• Capacitor Failure Symptoms
• Sinclair ZX Spectrum ULA
• Ferranti Archives 2026: Ferranti's ULA files of 1981 to 1984, with the period board drawings cited here
• Recommended Tools