## Self-resonant Frequencies of Some Ceramic Capacitors

In that version of the GPS+voice interface, I sprinkled 100 nF and 100 pF SMD caps across the input lines in the hope that they’d reduce EMI on the audio board. The board worked fine for years, but now that it’s time to build another board & box, I figured it’d be good to know a bit more about their actual response.

So I cobbled up a test fixture with a 3 dB pad from the tracking generator output and a 20 dB pad to the spectrum analyzer input (both of those are bogus, because the cap impedance varies wildly, but work with me on this):

Pulled an assortment of 100 nF ceramic caps from the stockpile:

And rubbed them against the HP8591 spectrum analyzer & tracking generator:

Their self-resonant frequencies are much lower than I expected:

The attenuators produce about 17 dB of loss with no cap in the circuit, so the disk caps are pretty much asleep at the switch from VHF on up. The small bypass cap in the top photo is OK and the SMD cap is pretty good, but they’re all well past their self-resonant frequency and acting like inductors.

The relevant equations:

• FR = 1/(2π √(LC))
• XC = 1/(2π f C)
• Q = FR / BW
• ESR = XC / Q

The drill goes a little something like this:

• Find resonant frequency FR and 3 db bandwidth BW
• Knowing FR and C, find parasitic L
• Knowing FR and BW, find Q
• Knowing XC and Q, find ESR

In round numbers, the 100 nF SMD cap has L=2 nH and ESR=60 mΩ.

Now, it turns out a 100 pF SMD cap resonates up at 300 MHz, between the VHF and UHF amateur bands:

So I think the way to do this is to pick the capacitance to put the self-resonant frequency in the VHF band, parallel another cap to put a second dip in the UHF band, and run with it. A back of the envelope calculation suggests 470 pF and 47 pF, but that obviously depends on a bunch of other imponderables and I’ll just interrogate the heap until the right ones step forward.

Just to show the test fixture isn’t a complete piece of crap, here’s a 12 pF cap resonating up around 850 MHz:

For the combination of components, sweep speeds, bandwidths, and suchlike in effect, the spectrum analyzer’s noise floor is down around -75 dBm. I think the 12 pF cap is actually better than it looks, but I didn’t fiddle around with a narrower resolution bandwidth.