This installment's title is borrowed from one of my favorite segments on David Letterman. The curtain rises and some strange but intriguing circus-like act plays out for just a few seconds, then the curtain suddenly comes down again. Dave and Paul have to decide if the act was of any significance or simply a put-on. Well, I have one of those in the area of embedded passives and I can't quite decide. It has to do with decoupling and the ongoing debate over what really makes for an effective decoupling dielectric layer between power and ground. I wrote a couple of articles for CircuiTree on the subject back in June and July 2003, so I won't go over much of the technical points again, except to mention the two duties of a decoupling capacitor: provide enough charge to run the chip for one clock cycle and be able to provide that charge fast enough to keep the voltage about constant. The first requirement simply says that enough nF's must be present through some combination of dielectric constant, film thickness and area, and the second says that parasitic inductance and, to a lesser extent resistance, must be kept low.
展开▼