394 Chapter 13 set by 1K3 and 100 feedback resistors, their values being kept low to minimise common- mode distortion. It also underlines the point that in some circumstances an 8.1 mA current source gives worse results than the 5.4 mA version. When extra common-mode distortion is introduced by the presence of a significant source resistance, this extra distortion is likely to swamp the improvement due to output biasing. In a 5532 amplifier stage with a gain of 3.2 times and a substantial source resistance, the basic output distortion with a 1 K load at 9.6 Vrms, 1 kHz out was 0.0064%. A 3K3 output biasing resistor to V+ reduced this to 0.0062%, a marginal improvement at best, and an 8.1 mA current source could only reduce it to 0.0059%. Earlier I said that the practice of output stage biasing appears to be pretty much universally misunderstood, judging by how it is discussed on the Internet. The evidence is that every application of it that my research has exposed shows a resistor (or current source) connected between the opamp output and the negative supply rail. This is very likely based on the assumption that displacing the crossover region in either direction is a good idea, coupled with a vague feeling that a resistor to the negative rail is somehow more "natural" and looks more like the familiar drawing of a load to ground. However, the assumption that the output stage is symmetrical is usually incorrect; as we have seen, it is certainly not true for the