Ian J. Smith of Lucent Technologies (email@example.com) writes:
I recently re-read your article on operating decouplers above resonance and was glad to see such a common-sense approach, especially for those digital designers who can't yet face analogue!
One area which was omitted, is the situation where designers put several capacitors in parallel, close together, of differing values. The idea is that each capacitor will decouple a frequency range. I suspect this is more of an issue for non-multi-layer boards, but won't this potentially cause a parallel-resonance problem?
A good old rule of thumb was to use values with a ratio of 100:1, but I'd rather see engineers actually looking at impedance plots, ESRs and tracking.
How about a follow-up article in EDN (or have I missed it)?
Thanks for your interest in High-Speed Digital Design.
I discourage engineers from combining together different-valued capacitors if they share the same package format (for example, 0.1 uF and 0.001 uF, both in 1206-SMT packages). Since the lead inductances are the same, you aren't really buying anything at the high-frequency end that you wouldn't have gotten with just two of the larger-valued components in parallel. In addition, you have introduced the possibility of a resonance occurring, as you point out, between the lead inductance of the larger capacitor and the capacitance of the second, smaller-valued component. My advice is this: What's important when combining values is to make sure that we are maintain the same L/C ratio in each new section of the design: smaller C, and smaller L at the same time. That's what makes for a good combination-valued array. For example, if the bulk decoupling capacitor has a value of 10,000 uF and 10 nH, I might combine that with an array of 200 bypass capacitors, each having 0.05 uF and 2 nH. The effect of the array is the same as if I had a single capacitor with 10 uF, and 0.01 nH, values 1000 times less than the those of the bulk capacitor. This balance tends to keep the power-to-ground impedance more or less constant (within a factor of 30 or so) over a very wide frequency range.
By the way, The best method for controlling the resonances between sections of the power system is to buy cheap, low-Q bypass components in the first place.
About EDN, thanks for the good idea. Starting in the Fall of 1997, I'm writing a new column in EDN magazine, and I may get the space to go back over the bypass capacitor subject again.
Dr. Howard Johnson