Edinburgh-based Wolfson Microelectronics has significantly cut the typical power consumption of battery headphone amplifiers using variable voltage rails, claims company CTO Peter Frith.

"In order to achieve 20mW maximum audio power, you need a 3V power rail," Frith told EW, "but listening typically the power delivered to the load is 0.1mW which is about 0.1V at the output. From 3V that is an efficiency of three per cent."

Calling the technology 'Class W', Wolfson is not powering its audio output stage from the battery.

Instead it is generating supply voltages using a two-level charge pump which allows the rails to be halved when large amplitudes are not needed.

"Halving the supply rails saves 50 per cent power," said Frith. "We take in 1.8V and produce [+/-]1.8V for 25mW out or [+/-]0.9V when the signal allows."

The firm has avoided tracking the output envelope exactly because switched capacitor generators are at their most effective doubling, inverting or halving voltages. Efficiency drops dramatically if other multiples are required.

However, it has developed a control algorithm for switching between normal and half voltage that Frith claims halves losses again with typical audio content.

Operating when coupled to a digital music codec, the algorithm gets amplitude data from the front of the decode process, allowing it time prepare the rails for the expected output voltage excursion. "The charge pump gets information milliseconds ahead of time. Noise and distortion at the output is not compromised."

The algorithm has control of the frequency and capacitance used in the voltage rail generator, said Frith, and also, for example, uses the natural decay of the capacitively-supported rails between full and half settings to follow reductions in output amplitude.

Music from rock to classic was used to develop the algorithm, according to Frith, but the firm found the filtered pink noise signal IEC-60268-5 particularly useful. "It represents music fairly well," he said.

Wolfson also investigated power saving using an inductive ΔΣ Class-D topology, but the necessary high switching frequency meant high quiescent current. "If you reduce the switching frequency at low amplitude, hiss and noise increases," said Frith, "and you need an inductor."

Having split rails in Class-W means no output coupling capacitor is required - saving a pair of 220µF capacitors, said Frith.

A DC servo circuit cuts output offset voltage to reduce losses in the load, and a clamp holds the output until the servo has operated to remove output 'pops' at switch-on.

The headphone amplifier has been implemented in the firms WM8903 codec, which also includes newly developed power saving switched capacitor DACs.

"This enables DAC to headphone power consumption of just 4.4mW and includes a capability for programmable performance vs. power profiles," said Wolfson.

Power supply rejection ratio, at 60dB, is clamed to be high enough for the analogue and digital parts of the chip to operate from a single 1.8V rail.

"1.8V is a common supply in portable systems. It is a typical voltage used to for RAM, digital I/O busses and some digital peripherals," said Frith. "The WM8903 is designed to deliver high performance audio even when it is connected to a noisy 1.8V supply that is shared with digital logic. The 1.8V in a portable system is typically generated from a switched mode power supply."

See also: the Electronics Weekly focus on Digital technology and loudspeaker design