Top of the class in amplification
We lightly toss out terms like Class A, Class D, Class A/B in respect of amplifiers. But what do they actually mean? And what difference do they make? Stephen Dawson explains.
Let’s decode the letters of amplifiers. What is Class A, Class D and so on? But before we do that, let’s first understand what it is that an amplifier does.
The job of the amplifier
Well, that seems kind of obvious. It ‘amplifies’, makes something bigger. In this case the electrical signal. But what about the electrical signal is it amplifying? A phono preamp is clearly increasing the voltage by something like a hundred times. But a component preamplifier, connected to a power amplifier is often reducing the voltage. The CD player plugged into it will typically output up to two volts. Feed two volts into a power amplifier and it will be running at full volume!
Sometimes preamplifiers are called ‘control amplifiers’, presumably for that reason. Nonetheless, a preamplifier can boost the output voltage when required.
Which brings us to power amplifiers. There are many circumstances in which the voltage output by a power amplifier is no higher, perhaps even lower, than the input voltage. Take that 2V peak CD output. Two volts out of an amplifier into an 8Ω load equals 0.5W output. Half a watt of power into an average high fidelity loudspeaker means a peak output of 86dB. That’s not thundering, but it’s certainly a decent volume.
Half a watt of power into a Klipschorn loudspeaker, though, means a peak of 102dB, which is thundering. So, could you plug a preamplifier’s output into Klipschorn and skip the whole power amplifier thing?
No. Preamplifiers and CDs and other ‘line level’ devices will only deliver their voltages into very high impedances. Typically 47,000Ω, and almost always at the same order of magnitude, or higher. So they’ll go up to 2V, but never supply more than the tiniest trickle of current (less than 0.05mA) and insignificant power levels (less than 0.1mW).
The job of a power amplifier is, as the name suggests, to supply power. Which means it has to be able to deliver large currents into low impedance loads – those magic numbers of 8Ω, or even 4Ω, which we all deal with all the time.
Over the decades clever engineers have come up with many, many ways of doing this. But the most common in use in the home entertainment space fall into five categories, or Classes. But in order to get things rolling and help in understanding some of the others, I’ll have to start with a sixth class which I’d guess never finds itself in a quality listening room.
The signal that has to be amplified consists of a continuously varying voltage. It spends roughly half the time as a positive voltage, half as a negative one, swinging around the zero voltage point. It’s like AC, but with much less regularity.
Amplifiers in the analogue domain work by increasing the voltage at each point proportionately, and allowing large currents to be drawn. Increasing the voltage means making the positive voltages more positive, so to speak, and the negative ones more negative. In order to do that, the two halves – the part above the zero line and the part below – have to be in effect separated, amplified, and re-united at the output stage to drive the speakers.
The amplification of each half is itself performed by using the input signal to control the passage of current created by the rail voltage of the amplifier though the output transistors.
The problem is with the re-uniting. This is always imperfect. Those imperfections are kind of regular, so the result is harmonic distortion. This can to some extent be addressed with negative feedback, but the high levels required can inflict other unfortunate audible damage to the signal.
So Class B as such is not used in high fidelity equipment.
But Class A is, fairly rarely. Rarely because it is almost entirely used for very expensive equipment and because it is horribly inefficient. The equipment is, in fact, expensive because of that inefficiency.
As we saw, the problems of Class B were to do with the point at which the signal switches from positive to negative and from negative to positive. What if that switch could just be eliminated? That’s what Class A does.
It does that by ‘biasing’ the signal – adding either a positive or negative voltage to the signal so that the whole thing becomes positive or negative. It never crosses the zero point. So let’s say you want a 100W into an 8Ω Class A amplifier. That requires an RMS output of 28.3V, which means a peak output of 40V. So 40V DC is applied to shift the whole thing.
Which means that the system as a whole is dissipating 100W of power, when there is no signal at all! Horrible inefficiency for sure!
So Class A amplifiers tend to be relatively low in output power.
The distortion from Class B is more or less constant in level, regardless of the output level. So it’s a high level of distortion in comparison to the signal when the signal is low, and a much lower level of distortion when the signal is high.
Class A/B designs take advantage of this. Like Class A, they apply a DC bias. But only a relatively small amount. The idea is to move the signal by a few volts so that it’s working in pure Class A mode at low output levels, and Class B when the peaks of the signal exceed the level of bias. As, almost by definition, the signal is higher in that latter state, the distortion if far less problematic, and can be reduced further with modest amounts of negative feedback.
I mentioned the rail voltage earlier. Remember, an amplifier works by using the input signal to control flow of current through a transistor. The voltage creating that current is applied from the main power supply by conductors traditionally called ‘rails’, even though they’re just bits of wire or lines of conductor on a printed circuit board.
In Class A and Class A/B designs, that’s a fixed voltage. Class G mitigates their inefficiency, particularly with Class A/B, by operating with several different rail voltages available. It analyses the signal and switches between the voltages as required to ensure that the appropriate output can be created, using the lowest voltage position it has available to permit that. Lower voltage means reduced wasted current and less wasted power.
Class H takes the same idea as Class G, but rather than having a choice from several specific levels of rail voltage, the system can more or less vary this continuously.
With both Class H and Class G, the underlying principle of operation of the amplifier is pretty much the same as described above.
With Class D the underlying principle is very different. Instead of a continuously varying voltage being increased in power capacity through the system, the incoming signal is turned into pulses. The pulses may be of varying widths depending on the signal at any given point, or at varying densities (that is, the number of equal-width pulses in a given time), depending on the design. The first is called Pulse Width Modulation (PWM), the second Pulse Density Modulation (PDM).
Now it turns out that amplifying simple on and off pulses is a great deal more efficient than continuously varying signals and it’s generally reported that efficiency ranges from at least 50% for low level signals to well above 90% for high level ones.
Conceptually, turning those digital pulses back into analogue high power output at the speaker terminals is fairly straight forward: a low level filter to remove the very high frequencies that constitute the pulse. In actual implementations things tend to be a bit more complicated.
So the advantage of Class D is high efficiency. If your source is digital and can be connected to the power amplifier digitally, then a digital to analogue to digital conversion can be omitted too. PCM can be transcoded to PWM or PDM quite well.
Various wild claims are sometimes made about Class D being audibly better than the alternatives. Sometimes others are made about it being worse. A well designed and made Class D is in practice very little different, but can get by with small heat sinks and is, of course, a good choice in many places where heat dissipation might be a problem.
On the other hand, how well it handles high resolution audio depends upon design choices. Insufficiently high sampling frequencies may result in an inability to reproduce the 40kHz or more that lovers of high resolution audio demand.
Incidentally, if all that sounded very digital – as in ‘Class D’ for digital – that’s because it is. Things are complicated because at various times Class D has been used as a name for switch mode amplifiers, some of which have been rail voltage switching amps, such as Class G above. However the convention these days tends to be to use Class D in the sense described here.
All of these are broad categories and there are plenty of designs around that involve the use of two or more Classes to achieve particular design goals. But they do pretty much cover the great majority of amplifiers one is likely to come across in home entertainment.