Efficient amplification for real-world audio

Consumers generally cherish a warm sound from their audio systems but they are becoming less patient with the physical warmth generated by wasted electricity from their power amplifiers and active speakers.

In contrast to the relentless pace of development in digital signal processing for audio enhancement and compression, amplifier design has been much slower in evolving and has not seen the same order of power-efficiency gains. But consumers are demanding more convenience, not just from portable audio devices where battery life is a prime concern, but from home systems too.

Wireless speaker systems are becoming more common as they give home users more flexibility on placement without worrying about the mess caused by thick speakers cables. Yet these still require power cables as the efficiency of existing amplifier circuits gives a very limited battery life.

The shift to digitally stored media on portable devices, such as laptops, phones and tablet PCs, has also led to an increased demand for USB driven and wireless speakers which don't rely on an additional mains power supply. However, the USB port's 5V, 2.5W output – and wireless speakers battery constraints - places a hard limit on the power such speakers can draw, meaning amplifier efficiency has become a prime concern in this audio-system design.

Although the range of amplifier topologies available to designers has expanded in recent years they still largely fall into a few broad classes of architecture. Many of them are based on traditional linear circuit designs rather than more efficient switched topologies.

The traditional amplifier
The simplest amplifier topology, Class A, boasts the greatest theoretical linearity; but this comes at a price. Current always flows through the output devices, and most of the power drawn is wasted, indeed its maximum efficiency is just 20%.

To rectify this the Class B topology is split into two halves, with the output devices of each only conducting for one half of a cycle – one in the positive, the other in the negative region of a cycle. This improves the best-case efficiency to 50% but the switching between positive and negative states tends to introduce distortion.

A third topology, Class AB, allows both halves to conduct as the signal nears the crossover point. This reduces Class B's non-linearity with comparatively little loss in overall power efficiency.

Class AB remains one of the most commonly used audio system topologies, but recent years have seen it being supplanted by Class D amplifiers. This is particularly true in battery-powered systems as this class of amplifier can, in theory, provide much higher best-case efficiencies.

Class D topologies operate the output transistors as binary switches, delivering either the full output voltage or 0V. Switching occurs at non-audible frequencies – typically 300kHz to 2MHz – and generates a train of electrical pulses, the width of each corresponding to the input audio signal. A low-pass filter after the output stage turns the PWM signal into a smoothed, audio-range output. This switching topology enables Class D amplifiers to reach efficiencies of more than 90%.

However, this apparently high efficiency is only reached in artificial circumstances: with a fully sinusoidal input and when the volume is turned up so high it is practically at the clipping point. Under typical usage conditions Class D amplifiers see single digit efficiencies, with much of the input power still being wasted as heat and little realistic improvement over Class AB.

This wastage can be seen in Figure 1, with efficiency only approaching 90% at the very top of the power output range. Even with the volume turned up to maximum only the very strongest peaks will be created at close to 90% power efficiency, for all other outputs switching losses begin to dominate.