Chapter 11
Power amplifiers
Power amplifiers are uneventful devices. They are usually big and heavy, take up a lot of rack space, and feature very little (or sometimes nothing) beyond input and output sockets. Because one tends to ignore them, it is all the more important that they are chosen and used with due care. Coming in a variety of shapes, sizes and ‘generations’, they are all required to do the ostensibly simple job of providing voltage amplification – converting line levels of up to a volt or so into several tens of volts, with output currents in the ampere range to develop the necessary power across the loudspeaker terminals. Given these few requirements, it is perhaps surprising how many designs there are on the market.
Domestic power amplifiers
The domestic power amplifier, at its best, is designed for maximum fidelity in the true sense of that word, and this will usually mean that other considerations such as long-term overload protection and complete stability into any type of speaker load are not always given the type of priority which is essential in the professional field. A professional power amp may well be asked to drive a pair of 6 ohm speakers in parallel on the other end of 30 metres of cable, at near to maximum output level for hours on end if used in a rock PA rig. This demands large power supplies and heavy transformers, with plenty of heat sink area (the black fins usually found on the outer casing) to keep it from overheating. Cooling fans are frequently employed which will often run at different speeds depending on the temperature of the amplifier.
The domestic amplifier is unlikely to be operated at high output levels for a significant length of time, and the power supplies are often therefore designed to deliver high currents for short periods to take care of short, loud passages. A power supply big enough to supply high currents for lengthy periods is probably wasted in a domestic amplifier. Also, the thermal inertia of the transformer and the heat sinks means that unacceptable rises in temperature are unlikely. Although there are one or two domestic speakers which are notoriously difficult to drive due to various combinations of low impedance, low efficiency (leading to high power demand), and wide phase swings (current and voltage being out of step with each other due to crossover components and driver behaviour in a particular speaker enclosure), the majority of domestic hi-fi speakers are a comfortable load for an amplifier, and usually the speaker leads will be less than 10 metres in length.
It is unlikely that the amplifier will be driven into a short-circuit due to faulty speaker lines for any length of time (silence gives an immediate warning), which is not the case with a professional amplifier which may well be one of many, driving a whole array of speakers. A short-circuit developing soon after a show has begun may cause the amplifier to be driven hard into this condition for the whole evening. Protection circuitry needs to be incorporated into the design to allow the professional amplifier to cope with this without overheating or catastrophically failing which can affect other amplifiers in the same part of the rig.
Several ‘classes’ of amplifier design have appeared over the years, these being labels identifying the type of output stage topology employed to drive the speaker. These are outlined in Fact File 11.1.
Fact file 11.1 Amplifier classes
Class A
The output stage draws a constant high current from the power supply regardless of whether there is an audio signal present or not. Lowcurrent class A stages are used widely in audio circuits. The steady bias current as it is known is employed because transistors are non-linear devices, particularly when operated at very low currents. A steady current is therefore passed through them which biases them into the area of their working range at which they are most linear.
The constant bias current makes class A amplification inefficient due to heat generation, but there is the advantage that the output transistors are at a constant steady temperature. Class A is capable of very high sound quality, and several highly specified up-market domestic class A power amplifiers exist.
Class B
No current flows through the output transistors when no audio signal is present. The driving signal itself biases the transistors into conduction to drive the speakers. The technique is therefore extremely efficient because the current drawn from the power supply is entirely dependent upon the level of drive signal. Class B is therefore particularly attractive in batteryoperated equipment. The disadvantage is that at low signal levels the output transistors operate in a non-linear region. It is usual for pairs (or multiples) of transistors to provide the output current of a power amplifier. Each of the pair handles opposite halves of the output waveform (positive and negative with respect to zero) and therefore as the output swings through zero from positive to negative and vice versa the signal suffers so-called ‘crossover distortion’. The result is relatively low sound quality, but class B can be used in applications which do not require high sound quality such as telephone systems, handheld security transceivers, paging systems and the like.
Class A–B
In this design a relatively low constant bias current flows through the output transistors to give a low-power class A amplifier. As the input drive signal is increased, the output transistors are biased into appropriately higher-current conduction in order to deliver higher power to the speakers. This part of the operation is the class B part, i.e.: it depends on input drive signal level. But the low-level class A component keeps the transistors biased into a linear part of their operating range so that crossover distortion is largely avoided. The majority of high-quality amplifiers operate on this principle.
Other classes
Class C drives a narrow band of frequencies into a resonant load, and is appropriate to radio-frequency (RF) work where an amplifier is required to drive a single frequency into an appropriately tuned aerial. Class D is ‘pulse width modulation’ in which an ultrasonic frequency, modulated by the audio signal, is used to drive the output transistors. A low-pass filter is employed after the output stage. This technique has been revived in one or two designs in the late 1980s. A variation on class D called class T (from the Tripath company) has recently been seen. Here, the ultrasonic frequency is continuously varied in accordance with the amplitude of the audio signal. The frequency is about 1.2 MHz at low signal levels, falling to around 200 KHz for very high signal levels; a greater overall efficiency is claimed as a result. Classes E and F were concerned with increasing efficiency, and currently no commercial models conform to these particular categories. Class G incorporates several different voltage rails which progressively come into action as the drive signal voltage is increased. This technique can give very good efficiency because for much of the time only the lowervoltage, low-current supplies are in operation. Such designs can be rather smaller than their conventional class A–B counterparts of comparable output power rating. Class H is a variation on class G in that the power supply voltage rails are made to track the input signal continuously, maintaining just enough headroom to accommodate the amplifier’s requirements for the necessary output voltage swing.
Since the early 1980s the MOSFET (Metal Oxide Semiconductor Field-Effect Transistor) has been widely employed for the output stages of power amplifiers. MOSFET techniques claim lower distortion, better thermal tracking (i.e.: good linearity over a wide range of operating temperatures), simpler output stage design, and greater tolerance of adverse loudspeaker loads without the need for elaborate protection circuitry.
Professional amplifier facilities
The most straightforward power amplifiers have input sockets and output terminals, and nothing else. Single-channel models are frequently encountered, and in the professional field these are often desirable because if one channel of a stereo power amplifier develops a fault then the other channel also has to be shut down, thus losing a perfectly good circuit. The single-channel power amplifier is thus a good idea when multi-speaker arrays are in use such as in rock PA systems and theatre sound.
Other facilities found on power amplifiers include input level controls, output level meters, overload indicators, thermal shutdown (the mains feed is automatically disconnected if the amplifier rises above a certain temperature), earth-lift facility to circumvent earth loops, and ‘bridging’ switch. This last facility, applicable to a stereo power amplifier, is a facility sometimes provided whereby the two channels of the amp can be bridged together to form a single-channel higher-powered one, the speaker(s) now being connected across the two positive output terminals with the negative terminals left unused. Only one of the input sockets is now used to drive it.
Cooling fans are often incorporated into an amplifier design. Such a force-cooled design can be physically smaller than its convection-cooled counterpart, but fans tend to be noisy. Anything other than a genuinely silent fan is unacceptable in a studio or broadcast control room, or indeed in theatre work, and such models will need to be housed in a separate well-ventilated room. Ventilation of course needs to be a consideration with all power amplifiers.
Specifications
Power amplifier specifications include sensitivity, maximum output power into a given load, power bandwidth, frequency response, slew rate, distortion, crosstalk between channels, signal-to-noise ratio, input impedance, output impedance, damping factor, phase response, and DC offset. Quite surprising differences in sound quality can be heard between certain models, and steady-state measurements do not, unfortunately, always tell a user what he or she can expect to hear.
Sensitivity
Sensitivity is a measurement of how much voltage input is required to produce the amplifier’s maximum rated output. For example, a model may be specified ‘150 watts into 8 ohms, input sensitivity 775 mV ≡ 0 dBu’. This means that an input voltage of 775 mV will cause the amplifier to deliver 150 watts into an 8 ohm load. Speakers exhibit impedances which vary considerably with frequency, so this is always a nominal specification when real speakers are being driven. Consideration of sensitivity is important because the equipment which is to drive the amp must not be allowed to deliver a greater voltage to the amplifier than its specification states, otherwise the amplifier will be overloaded causing ‘clipping’ of the output waveform (a squaring-off of the tops and bottoms of the waveform resulting in severe distortion). This manifests itself as a ‘breaking-up’ of the sound on musical peaks, and will often quickly damage tweeters and high-frequency horns.
Many amplifiers have input level controls so that if, for instance, the peak output level of the mixer which drives the amplifier is normally say ‘PPM 6’ – about 2 volts – then the amp’s input levels can be turned down to prevent overload. In the given example, 2 volts is 8 dB higher than 775 mV (PPM 4 ≡ 0 dBu) and so the input level control should be reduced by 8 dB to allow for this. If a dB calibration is not provided on the level control, and many are not particularly accurate anyway, a reasonable guide is that, compared with its maximum position of about ‘5 o’clock’, reducing the level to about 2 o’clock will reduce the sensitivity by about 10 dB, or by a factor of three. In this position, the power amplifier with an input sensitivity of 775 mV will now require 0.775 × 3, or about 2 volts, to develop its full output.
If input level controls are not provided, one can build a simple resistive attenuator which reduces the voltage being fed to the amplifier’s input. Two examples are shown in Figure 11.1. It is best to place such attenuators close to the power amp input in order to keep signal levels high while they are travelling down the connecting leads. In both cases the 3k3 resistor which is in parallel with the amplifier’s input can be increased in value for less attenuation, and decreased in value for greater attenuation. With care, the resistors can be built into connecting plugs, the latter then needing to be clearly labelled.
Figure 11.1 (a) An unbalanced resistive attenuator. (b) A balanced resistive attenuator
Power output
A manufacturer will state the maximum power a particular model can provide into a given load, e.g.: ‘200 watts into 8 ohms’, often with ‘both channels driven’ written after it. This last means that both channels of a stereo amplifier can deliver this simultaneously. When one channel only is being driven, the maximum output is often a bit higher, say 225 watts, because the power supply is less heavily taxed. Thus 200 watts into 8 ohms means that the amplifier is capable of delivering 40 volts into this load, with a current of 5 amps. If the load is now reduced to 4 ohms then the same amplifier should produce 400 watts. A theoretically perfect amplifier should then double its output when the impedance it drives is halved. In practice, this is beyond the great majority of power amplifiers and the 4 ohm specification of the above example may be more like 320 watts, but this only around 1 dB below the theoretically perfect value. A 2 ohm load is very punishing for an amplifier, and should be avoided even though a manufacturer sometimes claims a model is capable of, say, 800 watts of short-term peaks into 2 ohms. This at least tells us that the amp should be able to drive 4 ohm loads without any trouble.
Because 200 watts is only 3 dB higher than 100 watts, then, other things being equal, the exact wattage of an amplifier is less important than factors such as its ability to drive difficult reactive loads for long periods. Often, ‘RMS’ will be seen after the wattage rating. This stands for root-mean-square, and defines the raw ‘heating’ power of an amplifier, rather than its peak output. All amplifiers should be specified RMS so that they can easily be compared. The RMS value is 0.707 times the instantaneous peak capability, and it is unlikely that one would encounter a professional amplifier with just a peak power rating.
Fact file 11.2 Power bandwidth
Power bandwidth is a definition of the frequency response limits within which an amplifier can sustain its specified output. Specifically, a 3 dB drop of output power is allowed in defining a particular amplifier’s power bandwidth. For example, a 200 watt amplifier may have a power bandwidth of 10 Hz to 30 kHz, meaning that it can supply 200 watts – 3 dB (= 100 watts) at 10 Hz and 30 kHz, compared with the full 200 watts at mid frequencies. Such an amplifier would be expected to deliver the full 200 watts at all frequencies between about 30 Hz and 20 kHz, and this should also be looked for in the specification. Often, though, the power rating of an amplifier is much more impressive when measured using single sine-wave tones than with broad-band signals, since the amplifier may be more efficient at a single frequency.
Power bandwidth can indicate whether a given amplifier is capable of driving a subwoofer at high levels in a PA rig, as it will be called upon to deliver much of its power at frequencies below 100 Hz or so. The driving of high-frequency horns also needs good high-frequency power bandwidth so that the amplifier never clips the high frequencies, which easily damages horns as has been said.
Power bandwidth is not the same as power rating, as discussed in Fact File 11.2.
Frequency response
Frequency response, unlike power bandwidth, is simply a measure of the limits within which an amplifier responds equally to all frequencies when delivering a very low power. The frequency response is usually measured with the amplifier delivering 1 watt into 8 ohms. A specification such as ‘20 Hz−20 kHz ±0.5 dB’ should be looked for, meaning that the response is virtually flat across the whole of the audible band. Additionally, the −3 dB points are usually also stated, e.g.: ‘−3 dB at 12 Hz and 40 kHz’, indicating that the response falls away smoothly below and above the audio range. This is desirable as it gives a degree of protection for the amp and speakers against subsonic disturbances and RF interference.
Distortion
Distortion should be 0.1 per cent THD (see ‘Harmonic distortion – technical’, Appendix 1) or less across the audio band, even close to maximum-rated output. It often rises slightly at very high frequencies, but this is of no consequence. Transient distortion, or transient intermodulation distortion (TID), is also a useful specification. It is usually assessed by feeding both a 19 kHz and a 20 kHz sine wave into the amplifier and measuring the relative level of 1 kHz difference tone.
The 1 kHz level should be at least 70 dB down, indicating a well-behaved amplifier in this respect. The test should be carried out with the amplifier delivering at least two-thirds of its rated power into 8 ohms. Slew rate distortion is also important (see Fact File 11.3).
Crosstalk
Crosstalk figures of around −70 dB at mid frequencies should be a reasonable minimum, degrading to around −50 dB at 20 kHz, and by perhaps the same amount at 25 Hz or so. ‘Dynamic crosstalk’ is sometimes specified, this manifesting itself mainly at low frequencies because the power supply works hardest when it is called upon to deliver high currents during high-level, low-frequency drive. Current demand by one channel can modulate the power supply voltage rails, which gets into the other channel. A number of amplifiers have completely separate power supplies for each channel, which eliminates such crosstalk, or at least separate secondary windings on the mains transformer plus two sets of rectifiers and reservoir capacitors which is almost as good.
Fact file 11.3 Slew rate
Slew rate is a measure of the ability of an amplifier to respond accurately to high-level transients. For instance, the leading edge of a transient may demand that the output of an amplifier swings from 0 to 120 watts in a fraction of a millisecond. The slew rate is defined in V μs−1 (volts per microsecond) and a power amplifier which is capable of 200 watts output will usually have a slew rate of at least 30 V μs−1. Higher-powered models require a greater slew rate simply because their maximum output voltage swing is greater. A 400 watt model might be required to swing 57 volts into 8 ohms as compared with the 200 watt model’s 40, so its slew rate needs to be at least:
30 × (57 ÷ 40) = 43 V μs−1
In practice, modern power amplifiers achieve slew rates comfortably above these figures.
An absolute minimum can be estimated by considering the highest frequency of interest, 20 kHz, then doubling it for safety, 40 kHz, and considering how fast a given amplifier must respond to reproduce this accurately at full output. A sine wave of 40 kHz reaches its positive-going peak in 6.25 μs, as shown in the diagram. A 200 watt model delivers a peak voltage swing of 56.56 volts peak to peak (1.414 times the RMS voltage). It may seem then that it could therefore be required to swing from 0 V to + 28.28 V in 6.25 μs, thus requiring a slew rate of 28.28 ÷ 6.25, or 4.35 V μs−1. But the actual slew rate requirement is rather higher because the initial portion of the sine wave rises steeply, tailing off towards its maximum level.
Musical waveforms come in all shapes and sizes of course, including near-square waves with their almost vertical leading edges, so a minimum slew rate of around eight times this (i.e.: 30 V μs−1) might be considered as necessary. It should be remembered, though, that the harmonics of an HF square wave are well outside the audible spectrum, and thus slew rate distortion of such waves at HF is unlikely to be audible. Extremely high slew rates of several hundred volts per microsecond are sometimes encountered. These are achieved in part by a wide frequency response and ‘fast’ output transistors, which are not always as stable into difficult speaker loads as are their ‘ordinary’ counterparts. Excessive slew rates are therefore to be viewed with scepticism.
Signal-to-noise ratio
Signal-to-noise ratio is a measure of the output residual noise voltage expressed as a decibel ratio between that and the maximum output voltage, when the input is short-circuited. Noise should never be a problem with a modern power amplifier and signal-to-noise ratios of at least 100 dB are common. High-powered models (200 watts upwards) should have signal-to-noise ratios correspondingly greater (e.g.: 110 dB or so) in order that the output residual noise remains below audibility.
Impedance
The input impedance of an amplifier ought to be at least 10 kΩ, so that if a mixer is required to drive, say, ten amplifiers in parallel, as is often the case with PA rigs, the total load will be 10 k ÷ 10, or 1 k, which is still a comfortable load for the mixer. Because speakers are of very low impedance, and because their impedance varies greatly with frequency, the amplifier’s output impedance must not be greater than a fraction of an ohm, and a value of 0.1 ohms or less is needed. A power amplifier needs to be a virtually perfect ‘voltage source’, its output voltage remaining substantially constant with different load impedances.
The output impedance does, however, rise a little at frequency extremes. At LF, the output impedance of the power supply rises and therefore so does the amplifier’s. It is common practice to place a low-valued inductor of a couple of microhenrys in series with a power amp’s output which raises its output impedance a little at HF, this being to protect the amp against particularly reactive speakers or excessively capacitive cables, which can provoke HF oscillation.
Damping factor
Damping factor is a numerical indication of how well an amplifier can ‘control’ a speaker. There is a tendency for speaker cones and diaphragms to go on vibrating a little after the driving signal has stopped, and a very low output impedance virtually short-circuits the speaker terminals which ‘damps’ this. Damping factor is the ratio between the amplifier’s output impedance and the speaker’s rated impedance, so a damping factor of ‘100 into 8 ohms’ means that the output impedance of the amplifier is 8 ÷ 100 ohms, or 0.08 ohms. One hundred is quite a good figure (the higher the better, but a number greater than 200 could imply that the amplifier is insufficiently well protected from reactive loads and the like), but it is better if a frequency is given. Damping factor is most useful at low frequencies because it is the bass cones which vibrate with greatest excursion, requiring the tightest control. A damping factor of ‘100 at 40 Hz’ is therefore a more useful specification than ‘100 at 1 kHz’.
Phase response
Phase response is a measurement of how well the frequency extremes keep in step with mid frequencies. At very low and very high frequencies, 15° phase leads or phase lags are common, meaning that in the case of phase lag, there is a small delay of the signal compared with mid frequencies, and phase lead means the opposite. At 20 Hz and 20 kHz, the phase lag or phase lead should not be greater than 15°, otherwise this may imply a degree of instability when difficult loads are being driven, particularly if HF phase errors are present.
The absolute phase of a power amplifier is simply a statement of whether the output is in phase with the input. The amplifier should be non-phase-inverting overall. One or two models do phase invert, and this causes difficulties when such models are mixed with non-inverting ones in multi-speaker arrays when phase cancellations between adjacent speakers, and incorrect phase relationships between stereo pairs and the like, crop up. The cause of these problems is not usually apparent and can waste much time.
Coupling
The vast majority of power amplifier output stages are ‘direct coupled’, that is the output power transistors are connected to the speakers with nothing in between beyond perhaps a very low-valued resistor and a small inductor. The DC voltage operating points of the circuit must therefore be chosen such that no DC voltage appears across the output terminals of the amplifier. In practice this is achieved by using ‘split’ voltage rails of opposite polarity (e.g.: ± 46 volts DC) between which the symmetrical output stage ‘hangs’, the output being the midpoint of the voltage rails (i.e.: 0 V). Small errors are always present, and so ‘DC offsets’ are produced which means that several millivolts of DC voltage will always be present across the output terminals. This DC flows through the speaker, causing its cone to deflect either forwards or backwards a little from its rest position. As low a DC offset as possible must therefore be achieved, and a value of ± 40 mV is an acceptable maximum. Values of 15 mV or less are quite common.