Along with loudspeakers, transistor power amplifiers are the prime cause of spuriously extinguished music; but their erratic reliability isn't, as in the case of speakers something to be mitigated by sensitive husbandry. On the contrary, the vast majority of power amplifiers using bipolar transistors fail quite randomly, regardless of design sophistication or elaborate 'protection' networks. This calls to mind a large and libellous body of information about the suicidal predilections of allegedly 'professional' power amplifiers featured in the dynamic adverts of other magazines.

In this article though, we are primarily concerned with a down-to-earth method of repairing power amplifiers, and a discussion of design shortcomings is an idea for a future article. Indeed, the DIY approach described here hinges upon the assumption that the amplifier design is inherently a going concern, even if it has faults which cause occasional failures, and clearly any mass-produced amplifier will meet this criterion.

Initial Observations

As the vast majority of power amplifier failures are catastrophic, we'll begin with the assumption that the amplifier is initially dead. Begin by disconnecting the mains and removing the cover(s). Then, using a small hand torch, make a careful physical inspection looking for loose wires or screws, bad connections, dry joints or damaged PCB tracks. If you encounter blown fuses or burnt/discoloured components, then serious failure can be assumed. Otherwise, the fault is quite possibly something 'subtle', such as a wire hanging from an input socket or a blown mains fuse. So, if nothing horrendous appears to have taken place internally, it's sensible to make good the loose screw, wire or blown fuse and then go on to test the amplifier, with due precautions, as enumerated shortly.

Returning to 'well cooked' components, our first task is to remove the PCB. At this point, it's very important to make notes and sketch plans of any wiring which has to be desoldered. Alternatively, if such an invaluable record is bound to be mislaid, use a water-based felt tip pen or apply self-adhesive labels to identify terminations in situ. Also take care to retain nuts and bolts removed during dismantling; placing them in a resealable plastic bag to be firmly gaffer taped to the lid is a neat anti-disappearance precaution. Once the PCB is removed, the amplifier chassis can be put to one side and surgery can commence. Before proceeding, it's worthwhile laying your hands on a good desoldering tool, to ease the task of component removal and avoid serious damage to the PCB, particularly when an awkward component has to be replaced several times.

Our first setback is the anonymous value of badly charred components, although some of these may turn out to be merely covered with a layer of soot. Revelation begins with a generous squirt of aerosol foam cleanser followed by vigorous scrubbing with a toothbrush and a rinse in hot, soapy water. This will remove most of the carbon deposits from the board and components alike, though care should be taken, as over-enthusiastic scrubbing may obliterate crucial markings. Usually, we're then left with a handful of charred resistor and transistor skeletons.

If the amplifier is a stereo model, values or type numbers can obviously be ascertained by reference to the other channel. Failing this, there are other avenues. First, try phoning the manufacturer or importer (written communications rarely work here) asking for someone who can answer a 'technical query'. Engineers are generally quite happy to discuss component values quite openly in response to sensible and lucid descriptions and you may even be able to obtain a circuit diagram, particularly if you patronise the company in purchasing replacement semiconductors from them. Or you may discover a colleague owns identical amplifiers in working condition, whereupon a casual glance inside will bring the information you require.

When in desperation, it's possible with a practical knowledge of power amplifier topology to make an inspired guess as to the mystery component, particularly if you have a number of circuit diagrams to draw order-of-magnitude values from. Last of all, don't ignore the possibility of measuring a badly charred resistor to discover that it still reads within 10% of its original value!

The next task is to replace resistors and capacitors lying in the vicinity of the recent fireworks display, particularly those which bear visible burn or scald marks or were coated with vaporized copper or soot prior to cleaning the board. Such components may measure satisfactorily, but intense heat won't enhance their long term reliability and as their cost is usually small, replacement outright is a sensible means of avoiding needless failures in future. As regards replacement values, if you can't copy the originals exactly, err on the side of generosity, whilst maintaining the intrinsic nature of the component. For instance, we can replace ¼ watt carbon resistors with ½ or even 2 watt versions, but a 1 watt wirewound resistor would be inadvisable, because its construction may introduce inductance problems. Similarly, replacing a 10 volt 22uF PC electrolytic with a tantalum capacitor of identical value would render the circuit susceptible to misbehaviour whenever small reverse voltages appeared across the capacitor and this inversion might be an entirely normal aspect of the circuit's behaviour. Hence the original choice of electrolytic, which wouldn't be damaged by small reverse voltages. On the other hand, replacement with a 63 volt 33uF electrolytic would be unlikely to give rise to problems.

Generally, the only truly critical passive components in power amplifiers are small ceramic or polystyrene capacitors with values in the range 10 to 820pF, which govern the high frequency power response and compensate the negative feedback so as to restrain it from becoming positive feedback at high frequencies. Thus values deviating greatly in either direction from the original may give rise to HF oscillation.

After the reinstatement of damaged or suspect passive components, we proceed with the assumption that all the semiconductors on the board are faulty and go on to test every one. Although pessimistic, this assumption is reasonable in respect of transistor power amplifiers, because their direct-coupled configuration implies mutual dependence, and a blown transistor at the output, say, will usually precipitate a chain of destruction throughout the amplifier.

Semiconductor Assessment

To test a transistor, use an analogue multimeter switched to 'Ohms x 10 or x 100'. (Many digital multimeters lack the biasing voltage necessary to make definite 'go/no go' judgements and making sense of the readout takes longer in any case). Remove each transistor in turn and check for the pattern depicted in Table 1. Although this data implies a knowledge of the transistors terminal configuration, in practice this information isn't necessary, for devices showing the appropriate pattern are invariably okay. This confidence is particularly welcome when a board bristles with T092 encapsulated devices, where of course the leadout configuration is entirely spurious.

If any readings differ in any respect from the general pattern, for instance the collector-emitter resistance isn't near infinite, assume the device to be guilty and make a note of its position and orientation on the PCB. Naturally, devices which measure satisfactorily are replaced on the board, taking care to solder them the right way round!

Aside from transistors, other semiconductors which may require testing include diodes, zeners and op-amps. The latter can't readily be tested without specialist jigs and substitution is usually the most sensible course of action. The other two devices are invariably okay if they have a high — but not necessarily infinite - resistance in one direction and give a low reading in the opposite, and whilst this measurement obviously can't take place with the components in circuit, in contrast to transistor checks, it's only necessary to pull one leg from the PCB.

The next stage is to obtain spare transistors, and if you can't obtain exact replacements through normal distribution channels, it's worthwhile phoning the amplifier manufacturer or importer for a quotation. Failing this, check semiconductor manufacturer's data books or a general purpose equivalents book such as 'Tower's International Semiconductor Selector' to arrive at nominal equivalents. For small signal devices, the salient parameters are V_CEO (normally 10 to 30% higher than the total supply volts), I_C, P_tot (aim for similar order of values here), and the current gain H_FE, which need only be within 50%, but must be quoted at the same order of bias current. For instance, if the original device had an H_FE of 250 @ 3mA, then an equivalent with an H_FE of 350 @ 2mA would be a sensible choice, whilst another with an H_FE of 250 @ 500mA would probably be unsuitable.

Two other parameters to consider are V_CBO (look for a device where V_CBO is similar to the original or higher), and f_t, which indicates the transistor's frequency response. This figure ought to be of the same order as the original, but a lower f_t will probably only be of audible significance if the amplifier is expected to reproduce frequencies above 10kHz at high levels (viz: treble amplifiers in active systems). At the other extreme, excessive f_t might at worst lead to HF instability, so it's wise to aim for f_t values similar to the original unless you have access to an oscilloscope in order to confirm that no instability has arisen. Table 2 lists high voltage small signal transistors which will suit the majority of transistor power amplifiers, whenever numerical replacements are unavailable.

Ad-lib power and driver transistor replacement is altogether a more dicey art because subtle parameters such as f_t become more significant and crucial data which isn't readily available such as SOA (safe operating area) and H_FE curves must be taken into consideration. However, Table 3 lists some general purpose audio power devices having ample ratings for run-of-the mill applications.

One area to be discussed concerns components which are chassis mounted. Of course, the power transistors are tested in the same manner as the PCB mounted devices, and here again, complete removal is the only way to be sure you're measuring a faulty transistor and not a hidden, parallel resistor or a short circuit across the mica insulating washer.

Before replacing power devices, clean down both mounting bases and apply fresh heatsink compound. Then inspect the mica washer for cracks and replace it if it appears at all weary. Check also the nylon insulating bushes for abrasion or crushing. Then, once the devices are repositioned and tightened up, use an ohmeter to check for short circuits between the heatsink (assuming a mica washer is used) and the transistor's collectors.

If a Zobel network is fitted across the output socket, check the series resistor for continuity. Finally, the power rail voltages should be checked, making sure that the bare ends of those wires which lead to the PCB are insulated or at least unable to touch each other. If the power supply won't work, and you're certain that blown fuses aren't responsible, look for broken transformer terminations or a faulty mains switch. Alternatively, persistent fuse blowing at the instant the mains is applied suggests a short circuit that's most likely to be traced to the bridge rectifier. However, also check the transformer primary, as it's not unknown for tappings to be changed to 110 volts "to get more power" — a famous quotation usually attributed to fervent guitarists!

The Meticulous Reinstatement

At this point, we have a chassis bearing serviceable power devices and a working power supply. And next to it, a PCB with all damaged and suspect components replaced. Of course the natural temptation is to dive in, replacing the PCB and switching on, rashly expecting the return of loud music. But aside from a nerve-shattering 'bang', music is unlikely to return if you approach the reconstruction phase with impatience. Unless you're possessed with inner calm and great confidence, the best way to ensure your labours aren't wasted is to take a break from the amplifier session for an hour or two, or at least make a cup of tea. When you return, check over your work with a refreshed and, above all, critical eye. Bear in mind that 'everything depends on everything else' in direct coupled power amplifiers; one error, however minor or stupid, can readily precipitate the chain destruction of your beloved amplifier once again. When you're satisfied, replace the PCB, double checking the connections noted earlier and also ensure that water used to wash the board isn't lurking in droplets under large components.

Next, wire 470 ohm 7 watt wirewound resistors in series with the positive and negative supply leads (or use a 1k resistor if the amplifier has a single supply rail relative to 0V), and reconnect the input and output leads. Connect a meter across the positive and negative terminals on the PCB side of the resistors as in Figure 1, and after making final checks, turn on. If the resistors heat up and the voltage across the rails at the PCB falls to a value well below the nominal, e.g., only 20 volts across +45 volt rails, then a fault is in evidence and further checks are called for. However, if there's a preset potentiometer on the board, this is probably intended to set the quiescent current, so watch for confirmation when its adjustment to one extreme causes the voltage across the PCB to rise to a near normal value.

Assuming satisfactory results, switch off and discharge the PSU reservoir capacitors. Then insert smaller value resistors (say 100R, 7 watt — the exact value isn't critical) and place a milliammeter in the positive supply lead. Switch on again and if there's a quiescent current preset, adjust the current after a few minutes of warm up time to circa 30mA (again, this is not critical). If no preset is present, assume that quiescent current lying somewhere between 10mA and 60mA spells a healthy amplifier. Switch off again, discharge the PSU reservoir capacitors and place the meter across the output to read volts. Then, after shorting the input, turn the gain control to minimum and checking the positive supply lead is reconnected minus the meter, switch on and examine the residual DC voltage at the output. If this lies below 1V, it's unlikely anything is seriously amiss, though under 100mV is a more satisfactory indication. If the DC offset exceeds 1V, a fault is likely, but note that some amplifiers boast a preset pot enabling any residual DC to be trimmed to zero. If so, forestall this adjustment until the 100 ohm current limiting resistors have been removed. If the amplifier survives, at this crucial point, it will probably live. But before pressing it back into service, apply a music signal peaking at half power (-3dB) and soak test it for a few hours into a suitable speaker — or, less taxingly (especially if you're restricted to BBC Radio 2 as a continuous soak testing source!) into a load resistor.

Finally, listen to the sound quality and remember that any freshly developed harshness or distortion probably indicates high frequency instability, or more subtle faults. These call for investigation with more elaborate test gear, and using only a test meter, this is about as far as we can go. Nevertheless, this technique will enable some 75% of direct-coupled bi-polar transistor amplifier catastrophies to be rectified at a fraction of the usual cost. In subsequent articles, we'll look at MOSFET and valve amplifier repairs, possibly also with a detailed examination of sophisticated test and tuning procedures for musicians' amplification.