The Miller Effect and Pole Splitting

Engineers frequently design systems to be dominated by a single pole. Aside from being easily analyzed (certainly an extremely attractive property in its own right), such systems also have the highly valuable attribute of being able to tolerate large amounts of negative feedback without stability problems.1 While it is impossible in practice to build a system that is truly single pole, it is not hard to approximate single pole behavior over a broad enough frequency range to be useful. Consider, for example, the Miller effect: it can increase dramatically the time constant associated with a capacitance that feeds back around an inverting gain stage. Usually, this effect is considered undesirable (because it degrades bandwidth), and we therefore often expend a great deal of design effort to avoid it (through cascoding, for example). However, the Miller effect can also be useful; it can be exploited to make a system’s open-loop transfer function approximate simple firstorder dynamics over a wide range by creating a dominant pole.