Class-G Power Amplifiers.

Page updated:
7 Feb 2002

I have just published a Class-G power amplifier design I am rather proud of. It is very likely the lowest-distortion Class-G amplifier yet. It has appeared in Electronics World issues for December 2001, and January, February 2002. I thought I should add some more information here. The EW articles include (naturally) a complete amplifier design, and there is a PCB available to make building it simple.

The history of Class-G amplifiers goes back a long way. The earliest reference I am aware of is an article in Radio-Electronics, dated August 1976. Hitachi is credited with putting the first Class-G design into production- their 1977 Dynaharmony HMA 8300 power amplifier. However, only in the last two or three years has Class-G made a serious impact in the market. In terms of modern technological advance, this is a enormous length of time. Why so?

One reason is that the quality obtainable with Class-G was for a long time suspect. Problems with commutation-diode glitching were well publicised, and there was little enthusiasm for adding another set of glitches to the ones already produced by crossover distortion.

Secondly, the improvement in amplifier efficiency and the reduction in heat evolved were not of the order that could give a noticeable reduction in final cost. Certainly the heatsinks would be smaller- possibly half the size, but they would still be necessary and would still have to be drilled and tapped or whatever. In fact, they would probably need a good deal more machining to mount the extra power devices, of which there are now twice as many. These extra power transistors cost money, in an era when silicon was a good deal more expensive than it is now. For the kind of domestic power levels in use at the time, around a maximum of 60 Watts, the savings were pointless, especially if the technology introduced all sorts of ill-defines fears about poor audio quality.

Class-G is now out and about. So what has changed? This would be easier to answer if there had been more published on the technology; as it is I have to rely pretty much on my own research. I found that Schottky power diodes have eliminated the commutation-diode glitching. Having peeled this skin from the distortion onion, linearity was still signficantly poorer than a Blameless Class-B design. This proved to be due to Early effect in the output stage devices, stemming from the sudden changes in collector voltage they experience in Class-G operation. I had my own ideas about how to deal with this, and you can read all in Electronics World for Dec 2001 and Jan, Feb 2002. There is a fully worked-out design, and even a PCB available to ease building it.

Multichannel amplifiers.
The increased demand for multichannel amplifiers- ie 5 or 7 channels instead of a mere two- gives an incentive to minimise the heat liberated per channel. Sanyo is one manufacturer going the Class-G route for this.

Class-G has become popular here, possibly because people feel that the quality requirements for a subwoofer are less demanding. A typical product: "The largest and most powerful of Fosgate Audionic's subwoofers, the FA15.0 couples our 200 watt Class-G amplifier with a 15" down firing monster woofer. (200W RMS @ 4 Ohms)"

ADSL drivers
As Texas Instruments puts it:
"This full-rate ADSL and G.lite capable ADSL line driver, designated the THS6032, is the first in the industry to incorporate a Class G amplifier architecture rather than a Class AB amplifier architecture that most ADSL line drivers use. Unlike Class AB amplifiers which have two power supplies, the Class G architecture uses four power supplies, typically 5 Volts (V) and 15V, to obtain its low power consumption. The THS6032 draws from the low-voltage supply and can reduce power consumption to 1.3 Watts on a full-rate 25-ohm ADSL line by automatically switching to the higher voltage power supply when needed. Due to TI's advanced process technology, the THS6032 keeps total harmonic distortion low even when the driver is switching between power supplies."