My basic component stock is 0805 resistors and capacitors, with capacitors being C0G up to 1nF, X7R beyond there, and larger chips (1206, 1210) in the larger values (couple uF up). Resistors more diverse than capacitors of course, and capacitors at least 1, 2.2, 4.7 series. And of course I've collected many other values over the years, for specific needs, from client projects, etc.
You don't actually -- or not necessarily, anyway -- want parts as low-loss as C0G for bypass. A little loss helps dampen out supply fluctuations and keeps PDN (power distribution network) impedance flatter; anywhere you have resonances in the PDN, you have valleys that are lower ("better"), yes, but you have complementary peaks where the impedance is higher.
It's not a big deal anyway for RF circuits, where impedances are fairly modest to begin with (50 ish ohms), or logic circuits of similar nature (single chips draw comparable switching currents). Unless of course they're not, in which case more work is necessary: an SoC or FPGA will generally need many bypasses in parallel, many supply pins, and wide pours; an RF power amp may use resonant traps (1/4 wave between subsequent blocks of bypasses; staggered values/chip sizes) to optimize PDN attenuation / isolation at the center frequency.
And so the question becomes: not whether you need some amount of capacitance, or ESR, but how much. Of both. Ceramic caps are generally quite low ESR, so you have that corner covered; higher ESR can always be obtained by adding more externally (at some expense to ESL, as the body length of C+R+traces adds up), or using [selected*] polymer, or tantalum or electrolytic, types.
*Al poly are generally quite low impedance, but they are available in a wide range, including fairly mid ESR values. Dry tantalum cover a wide range as well, averaging mid, with some low values available.
It's worth remarking on, or contemplating, just how low ESR C0Gs really are. Qs are typically 1000s, with 3k, 10k even, not being unusual. ESR is so low that it's difficult to measure, even with precision equipment. It's so low that connections you would normally think innocuous, are actually prone to resonance. Chip parts in parallel, adjacent on the PCB, resonate with the few nH between them. Even power components (10s nF say), which can contribute EMI peaks in the 10s MHz range.
As long as such resonances aren't a problem in the application, or can be dampened by connected components, that's fine. Just to be aware of it, and quickly model or test to verify it.
Tim
Home
Help
Search
Login
Register
