Speak and Spell Power Board Repair

This article briefly describes the steps to investigate and repair a broken Texas Instruments Speak & Spell. This educational toy was first sold in 1978 using an innovative speech system based on linear predictive coding (LPC). A repaired Speak & Spell can make a fun retro gift and/or can be circuit bent for additional fun.

This 1980s vintage** toy was probably damaged by using an external power adapter with a higher voltage than the recommended 6V DC. One broken clip suggests there had been a previous attempt at a repair.

The repair involved desoldering and replacing some of the components on the power supply daughterboard.

** The version sold from 2019 onwards has a different display, different batteries and a tiny modern PCB.

1. One vintage, broken Speak & Spell.
2. A multimeter.
3. Soldering iron, solder and desoldering aids like (good quality) desoldering braid (solder wick) or a solder sucker.
4. Two BJT PNP transistors, 2N2907 were used.
5. One 6.8k resistor.
6. Optional
7. One 2.2nF ceramic capacitor.
8. One BJT NPN transistor (optional), 2N2222 were used.
9. A basic component tester (optional).
10. Four alkaline C batteries for testing.

There are no screws on this Speak & Spell, six plastic clips hold the two parts of the case together. These can be gently prised open with a flat-bladed screwdriver. Other revisions have two screws and four clips.

Inside there are two PCBs, the main board has a smaller daughterboard which converts the (nominal) 6V battery power to approximately -12V for the vacuum fluorescent display and a fairly stable -3.5V. The main board can be gently lifted up but it won't be completely free to move far due to the blue speaker wires, the red and black power wires and the flat ribbon cable from the membrane keyboard. The speaker can be unclipped from the surrounding plastic tabs and the ribbon cable can be pulled from the connector** to allow the board to be stood on its end.

These photographs were taken after the repair. The PCBs looked fine from the underside but there was and still is a distinct burn mark on the top of the power board. The original BJT PNP transistor in a TO-92 package above the burn mark had a visibly-cracked black plastic case.

If the wires are repeatedly flexed they may weaken the strands near the solder joints.

** This needs some practice to reinsert.

This schematic drawn by Furrtek** is for a different revision of the power board but it's close enough to understand how this simple switch mode power supply works.

There are two sections which were relevant to this repair.

1. The Q3 and Q4 BJT PNP transistors are connected in parallel***. These will be the same type perhaps matched to some degree. These share the load when switching the coil L2 on.
2. The STARTUP and POWERDOWN lines start the power supply and keep it on, respectively, via the Q1 BJT NPN transistor.

The photograph with just the top of the power board is the one that matches the schematic. The other photograph which has a flipped photograph of the underside alongside it is the one inside this Speak & Spell.

** D5 might be a normal diode rather than a zener.

*** This is also fairly obvious from a glance at the PCB traces.

One of the PNP transistors from the pair was cracked presumably from excessive heat caused either by current significantly beyond the maximum value or by its partner failing leaving it with a doubled current. The non-cracked one was also replaced as it was likely to either be broken or stressed by high currents. Two 2N2907 transistors were used to replace these.

After the replacement of the PNP transistor pair the Speak & Spell did not work on batteries (4 C cells with a recommendation for alkalines) but did work when powered by an external power supply. The external power supply set to 6V produced 6.4V DC possibly with some ripple. This was problematic as even fresh alkaline batteries only produce about 1.55V unloaded making 6.2V for the four batteries. Alkaline batteries decrease in voltage as they are used - the device also needs to be able to turn up with used batteries which still have some remaining charge.

With the Speak & Spell powered at 6.00V by an external power supply attached to the battery terminals the STARTUP line was checked with a multimeter by holding down the ON button. This action only raised it to 1.8V. The D1 diode drop reduces this to 1.2V and the potential divider reduces this further to 0.6V. This is going to be very close the Vbe of a (silicon) transistor and was the likely explanation for why the power supply did not turn on using battery power as it needed a slightly higher supply voltage.

The Falstad circuit simulation above with a lamp standing in the for the rest of the circuit shows this scenario. The push button raises the STARTUP voltage to 1.8V which isn't enough to turn on the transistor. The POWERDOWN line can easily turn on the transistor - a manual switch represents the microprocessor here. The 400 ohm resistor is a guess at the microprocessor's output pin impedance, the 13.5k resistor was chosen to represent the membrane keyboard based on the 1.8V measured value.

The membrane keyboard's connector was re-seated. This did not improve the functioning of the ON button. Ideally, understanding the nature of the high resistance of the ON button / membrane keyboard would be useful but in this case the easiest approach was to adjust the power board's potential divider to increase the voltage at the base of the resistor. The potential divider was quickly checked with in-circuit measurements using a multimeter showing reasonable values of 2.19k and 2.16k. The lower 2.2k resistor was replaced with a 6.8k resistor. This should be a safe change, free of side effects as it just increases the voltage at the transistor's base and there's still a 2.2k resistor to limit the current.

The NPN transistor was removed for testing. It tested okay but this provided an easy opportunity to replace it with a new transistor and one with a higher gain.

A 2.2nF capacitor was also replaced as it had a small crack near one of the leads. It tested okay at 34% above its nominal value. The component had been bent by 90 degrees - the crack may have been caused by mechanical damage.

The resistor is slightly fiddly to replace as the untrimmed leg of the original component is used to connect the daughterboard to the main PCB.

While the case was open all four of the battery connectors were cleaned with a contact cleaner strip.

The device was tested powered at 6.0V and again at 4.0V to ensure it would still start with partially discharged batteries. The photograph of the oscilloscope shows the the STARTUP on CH1 (cyan) and the POWERDOWN on CH2 (yellow) lines with the ON button being pressed with 6.0V power after the resistor has been replaced.

The components replaced were:

1. PNP transistor pair - one definitely broken - 2N2907.
2. Ceramic capacitor 2.2nf - tiny crack, but tested okay - 2.2nF ceramic.
3. NPN transistor - tested okay - 2N2222.
4. 2.2k resistor - tested okay but a higher value needed to compensate for ON button resistance - 8.6k.

The three electrolytic capacitors weren't changed. These can age/fail and are worth checking and perhaps pre-emptively replacing. They are outlined in yellow in the photograph above.

1. 10uF 35V (black) aluminium electrolytic capacitor.
2. 22uF 16V (orange) tantalum (electrolytic) capacitor.
3. 4.7uF (15V?) (grey) aluminium electrolytic capacitor.

The red and black power wires were resoldered as the movement of the main board during inspection and repair had caused the ends to be stressed causing one to break and one to fray.

The fixed Speak and Spell is powered in the video by a benchtop power supply at 6.0V. The oscilloscope running at 500ms per division shows the STARTUP line on CH1 (cyan) and the POWERDOWN on CH2 (yellow).

1. 0:07 ON button pressed - STARTUP is 4.7V for the duration of the press, the microprocessor raises the POWERDOWN line to 5.7V.
2. 0:13 OFF button pressed causing the POWERDOWN line to go to 0V.
3. 0:25 Turned back ON to check some buttons and speech synthesis.
4. 0:28 Edit in the video!
5. 0:30 Partial keyboard check.

The buttons were all checked for correct operation after this video and then checked again after the plastic case was closed.

The power board circuit is shown above in the Falstad Circuit Simulator with a layout close to the physical layout. This is an approximation as it's based on some guesswork about the coils, doesn't model all of the non-ideal characteristics of the components, doesn't include all the PCB layout-related parasitic elements and magnetic effects and was implemented by someone with no experience in the complex world of switch mode power supply design! It does oscillate as intended starting at 390kHz and dropping to 260kHz. It takes about 1ms to reach the target -3.5V and the zener diode then starts clamping the voltage to stay above -3.5V via the Q2 transistor. Two resistors are used as a substitute for the load of the Speak & Spell main board.

The 1982 photograph above features some early circuit bending.

Speak & Spell repairs and technical information:

1. fastmatt (YouTube)
2. Troubleshooting Speak & Spell - won’t turn on
3. Repair a glitchy Texas Instruments Speak & Spell Speak & Read Speak & Math - capacitor issue
4. Furrtek: Speak & Spell (french version) - reverse engineering and creation of a new ROM.
5. Circuit Bending Wiki: Speak & Spell
6. Casper Electronics Artistic Circuit Bending Schematic via MatrixSynth
7. THIS MUSEUM IS (NOT) OBSOLETE: Speak And Spell Breakout 1.2 A Look inside A Modular Circuit Bent Machine (YouTube) from This Museum Is (Not) Obsolete in Ramsgate, UK not too far from the Turner Contemporary Museum in Margate.
8. Ingmar's Retroblog: Speech output in the 80s – Speak and Spell (german version)
9. Datamath Calculator Museum:
10. Texas Instruments Speak & Spell
11. Texas Instruments Speak & Spell (British Voice Type 3)
12. TMC0271 - the 40 pin 4 bit (PMOS) microcontroller with VFD driver used in the Speak & Spell. The Vdd (pin 40), Vgg (pin 39) and Vss (pin 38) on the Speak and Spell are -3.5V, -12V, VBAT in the terminology used in this article and by Furrtek.

Electronics repairs:

1. Mend It Mark (YouTube) - no Speak & Spell repairs, but many other inspirational repairs by the talented Mark Maher of Perton Electronics.
2. Electronics Repair School (YouTube) - complex repairs on laptops, tablets, smartphone and more.

Speech synthesis including non-Texas Instruments products:

1. Vintage Computing & Gaming Interview: 30 Years Later, Richard Wiggins Talks Speak & Spell Development
2. IEEE Spectrum: The Consumer Electronics Hall of Fame: Texas Instruments’ Speak & Spell
3. The 8-Bit Guy (YouTube)
4. Speak and Spell - 1983 vs 2019 model!
5. How Speech Synthesizers Work
6. Island Drive Media Archive: Original Voice Recordings for Speak & Spell (Hank Carr, 1977) (YouTube)
7. LGR: The Iconic 80s Speech Synthesizer: DECtalk for PC - LGR Oddware (YouTube)
8. GadgetUK - Retro Gaming Repairs & Mods: Sinclair ZX Spectrum Currah Microspeech & Sound Output (YouTube)

Switch-mode power supply basics:

1. GreatScott! DIY Boost Converter || How to step up DC voltage efficiently (YouTube) - making a simple boost converter on a breadboard and perfboard.
2. element 14 presents: A Noise-Free DIY Switching Power Supply - How Hard Can It Be? (YouTube) - a practical look at how layout affects a basic switch-mode power supply based on the ON Semiconductor MC34167 staying well clear of a breadboard!
3. Rhode & Schwarz: Understanding Switching Mode Power Supplies (YouTube)
4. ElectrArc240: very Component of a Switch Mode Power Supply Explained (YouTube)

Some music featuring the Speak & Spell:

1. Kraftwerk - Computer World (Computerwelt): YouTube
2. LFO - LFO: YouTube
3. Jean Michel Jarre - Touch to Remember: YouTube