[Part 1 begins with an overview of amplifiers, and discusses basic requirements for current and voltage output, and transient response. Part 2 discusses nonlinear distortions in amplifiers, as well as amplifier classes and modes of operation, beginning with Class A and AB.]
6.5.4 Class D
Class C amplifiers have no place in audio, but do find use in radio transmitters. Class D, on the other hand, is an emerging technology. Class D amplifiers are frequently referred to as digital amplifiers, but some designs are better described as switching amplifiers. They essentially consist of a switch-mode power supply, supplying current into the load (loudspeaker)
under the control of the audio input waveform, with the output being low-pass filtered in a similar manner to digital-to-analogue converters.
Class D amplifiers are light in weight and can be relatively cheap to build. However, when extreme high fidelity is required, things can become more difficult to achieve. They are also very energy efficient, beginning from about 75% at 5% power to beyond 95% at full power. Sonic performance improves year by year.
In self-powered loudspeakers they can find willing partners because of their small size, low cost, and low heat generation. However, they still can be prone to the emission of troublesome electromagnetic interference (EMI) because of the high switching frequency used, (and the whole concept of switching, itself). Of course, they all must meet current electromagnetic compatibility (EMC) regulations, just like the office fax machine and digital radio/alarm clock, but, as this is being written, those devices must often be switched off in order to clearly hear the BBC World Service on a small, portable radio.
Fast switching always generates harmonics into the megahertz regions, and such emissions have a great potential to interfere with nearby electronic equipment. In other words, compliance with EMC regulations is one thing, but being a good neighbour with the rest of the sensitive equipment in a recording studio is another thing. This is especially so when 'vintage' equipment is in use, designed before there was a need to even think about digital switching transients. Nevertheless, as time goes on, improvements will be made, and Class D amplifiers are steadily progressing.
The promise of an efficient, cheap, lightweight and high quality amplifier is a strong spur to commercial development. At the time of writing, some questions still exist about the sonic performance of the top two octaves of the audio frequency range in Class D circuitry. The very low level switching artefacts which many designs exhibit can prove to be problematical when using high sensitivity loudspeakers, such as mid-range and high frequency horns. Achieving low noise and distortion at both low and high power still presents many design difficulties.
In effect, there are two types of Class D amplifiers. Early designs used analogue inputs, which were compared to a triangle wave running at the switching frequency. This type of switching, when the audio signal crossing the triangle wave causes the output to switch, gives rise to a PWM (pulse width modulation) output.
Some later designs, however, accept a linear PCM (pulse code modulation) digital input, and use more sophisticated modulation techniques. As such, these designs are effectively digital until the output filtering which removes the unwanted switching artefacts. The output transistors are switched on and off typically at a rate from 100 kHz to 1 MHz, so switching noise is only to be expected. It also leads to quantisation noise, because the output switching rate is finite.
However, factors such as the need for very low-jitter clocking and the use of air-cored output filter inductors are complications which are not always easy to solve. Clocking errors and ferrite inductor cores can lead to non-linear distortion. The filtering is necessary because the direct output waveform is a high frequency square wave, which, if left unfiltered, could radiate large amounts of radio frequency (RF) interference from the loudspeaker cables, which can act as transmitter aerials.
Output inductors that behave well at these frequencies but which do not exhibit much loss at audio frequencies can be difficult to make. The required clocking accuracy of less than 100 picoseconds may also be hard to achieve.
In order to reduce the potential for clashing clocks, it can be necessary to synchronise the clock of the switch-mode power supply with the clock of the amplifier, which can also be beneficial in ensuring that the maximum current is available exactly when needed. The demands made of the power supplies are quite exacting. Any changes in power supply voltage causes a proportional change in the output signal, whereas in Class A, the current drawn is relatively constant, and in Class AB designs the audio feedback circuitry tends to compensate for the moderate fluctuations. The power supplies for Class D amplifiers may also have to deal with absorbing the power which can be reflected back from the output filter inductors.