I'd have to see your circuit to really know for sure, but it's pretty easy to figure out based on the expected voltages, impedances, etc. in the circuit.
The most revealing facet should be the most apparent: if you're connecting to this capacitor with other leaded components (or traces), surely those inductances add to the intrinsic inductance of the capacitor alone? In which case, how few components do you need before your condition (SRF > Fo) is completely blown to hell from unavoidable geometry alone? And surely the same concern must befall the commercial guys, yet they don't seem to suffer such problems..?
It always bothered me that there is such a stated property as "SRF". The inductance is essentially proportional to length dimensions, so it varies directly and inseperably with circuit layout. Perhaps you can determine this figure in a fixed, standardized jig situation, but who defines that standard, how accessible is it (open vs. free; expensive spring clamps vs. carved hunk of FR-4?), and how useful/applicable is it (does it work on 1206s but not 0603s? Do you need a different one for disc types?)?
So, fortunately, the answer is: no, it does not matter, and you don't have to worry about it. A 1uF 0805 has exactly as low impedance as a 0.01uF 0805 at high frequencies. There is no need to use smaller caps in parallel to "distribute resonances" (indeed, this can make things much worse at certain frequencies).
The general design of RF circuitry is to have signals transmitted with relative freedom along the signal paths, to have them well blocked by supply and ground nodes (where applicable), and either pass or block when signal and bias cross paths (i.e., bias tees, coupling capacitors.. those sorts of things).
Passage is achieved by matching the impedances of source, load, transmission line, filter, etc. (Often this impedance is something other than the small signal, maximum power point, because active devices often have input and output impedances very different from a more practical impedance, like the highest-undistorted-power point. Resistors often get involved to modify those dynamic impedances -- example, a grounded-base amplifier has an input impedance of approximately zero ohms, so that even with a matching transformer, some damping resistance is worthwhile.)
Blockage is achieved by using an intentional mismatch: very high or very low impedances. It doesn't matter what the angle of that impedance is, so long as the result is very different from system impedance.
Indeed, part of the power supply network need not even be low inductance at all; a 1/4 wave trap or stub can be used to make an open or short circuit at the system frequency (assuming it's an acceptable approach for the inevitable harmonics and parasitics of the system, as well). This approach is frequently used in microwave microstrip construction, where brute force bypass capacitors simply aren't possible (the body length of the capacitor is simply more transmission line length!).
That said, maintaining a sufficiently low, and dissipative (lossy, resistive, attenuating) power supply network is an important goal, since you don't want that transmitter power feeding back to earlier stages (let alone to the receiver, if they're in common), or anything like that. Which usually means maintaining nice low impedances with lots of bypass caps, and also punctuating that with chokes and whatnot (hopefully with some lossy electrolytics or anything else with ESR to dampen the LF resonances thus created).
Tim
Home
Help
Search
Login
Register
