Two-pole compensation is well-known as a technique for squeezing the best performance from an op-amp[9] ^,[10], but it has rarely been applied to power amplifiers; the only example I know is found in Widlar[5]. An extra HF time constant is inserted in the Cdom path, giving an open-loop gain curve that initially falls at almost 12 dB/octave, but which gradually reverts to 6 dB/octave as frequency continues to increase. This reversion is arranged to happen well before the unity loop-gain line is reached, and so stability should be the same as for the conventional dominant-pole scheme, but with increased negative feedback over part of the operational frequency range. The faster gain roll-off means that the maximum amount of feedback can be maintained up to a higher frequency. There is no measurable mid-band peak in the closed-loop response.

It is right to feel nervous about any manoeuvre that increases the NFB factor; power amplifiers face varying conditions and it is difficult to be sure that a design will always be stable under all circumstances. This makes designers rather conservative about compensation, and I approached this technique with some trepidation. However, results were excellent with no obvious reduction in stability. See Figure 7.4 for the happy result of applying this technique to the Class-B amplifier seen in Figure 7.5.

The simplest way to implement two-pole compensation is shown in Figure 7.1d, with typical values. Cp1 should have the same value as it would for stable single-pole compensation, and Cp2...