"Q" of capacitors in RF tuned circuits, how to pick the best ones?

[cdev]

I've got a lot of unknown capacitors and when I use them in filters its often trial and error to find the ones that work best. Two capacitors even of the same type, (ceramic, for example) may be significantly different.

Is there any short guide to this or simple measurement procedure (besides actually building the circuit which is basically what I do now, plugging in ones I think may work better.

I have a lot of through hole polypropylene and mylar caps, of unknown manufacturers, most of which are well marked, although I have no idea who made them, also a fair number of both marked TH and unknown value SMD ceramic caps and some AVX DIP ceramic caps - that are of around 80s vintage, a developers kit. They are MLCCs, and are basically shaped like a 2 pin DIP would be.

I am wondering which of these caps would be the most consistent performers in RF filters (for receive) I currently have been using the AVXs because they seem to work okay and I have a developers kit so they are all marked known values and easy to use.

How do polypropylene compare to ceramic for RF? It would be great to have a quick jig I could rig up to determine some idea on likely "Q" in an HF application.

[T3sl4co1l]

Would expect the DIP ones are for bypass, X7R if you're lucky, Z5U more likely.  (The kit didn't come with a sheet saying what's in it?)

PP are excellent, it's just uncommon to see film caps in small values.  They're also leaded, so between the modest values and stray inductance, they aren't much use above, say, 10s of MHz.

PP are what powers the biggest "RF" generators on the planet, like induction heaters.  Their loss is quite acceptable.

C0G ceramic are excellent in all applications, up to whatever value you can get them in (which isn't much, around 0.1uF tops).  Downside, the cost.  (Small values, under 1nF or so, are competitive, even when you don't need the stability.)

Q is best tested by setting up a resonant circuit and measuring the loss resistance at resonance.  So, basically what you're doing now.  Don't forget to subtract inductor Q, which is usually dominant!

Type 2 ceramics can be identified by the C(V) and C(T) dependency.  They typically have a modest Q of maybe 10-50.

Tim

[cdev]

The AVX caps did come with data, they are X7R, also strangely, despite being non-polarized, there is a stripe on one side that looks like a small i, it's broken.

They are labeled. for example i A470 = 470 pf ceramic cap.

Apart from the ceramics, I have a lot of the PP TH capacitors.

Plus the tantalums I've mentioned elsewhere.

[DaJMasta]

For sorting unknown caps, the measurement procedure is at least an LCR meter to get a ballpark figure or a network analyzer/impedance analyzer to actually characterize it... not generally the cheapest of options.  Either device could give you a Q factor reading for the frequency it tests at, just generally if it's over a couple of MHz an LCR meter simply isn't going to have the frequency range.


With a single measurement frequency you can probably get an idea of the dielectric by matching up with some manufacturer example curves for the same value/size cap, but with a few frequencies, ideally nearish to where they'll be operating, you can take some points and try to fit it to an example curve, which, while relying on a lot of assumptions, should get you close in many cases... but at that point it's probably less effort/expense to just buy some caps that you have a datasheet for.


If you've got a sig gen that can produce the frequency you're interested in and an AC voltmeter with that bandwidth, you could try that way with a little math.  Since Q factor isn't really concerned with phase, as the network analyzer or LCR meter is, you can measure the sig gen through the shorted test fixture to the meter, then the same measurement with the capacitor in the fixture.  The difference should be the dissipation of the capacitor (since the inductance of the jig has been measured with the shorted measurement), and your Q factor is 1/dissipation factor.  Still need a fair bit of bandwidth on your voltmeter if you're looking for higher frequency stuff, but not looking at phase information means much less expensive equipment requirements.

[cdev]

Actually, I just realized I may be able to rig something like this up by scripting an older digital scope (Bitscope BS-50) I have that I rarely use, plus a small signal diode detector.. is a 1n34 the best choice? I think it likely is.

The BS-50 even has an AWG I think. I haven't used it those capabilities of it at all, ever because I have another logic analyzer and AWG now that I am familiar with- it has scripting capabilities and frankly, the Biscope is I think USB 1 and a bit strange - the software has never worked so well for me, its kind of slow compared to modern digi-scopes. However that may just be me, other people seem to like it. Its built well and seems like it would be useful for something like this especially.

For applications that use the extra pins you are usually supposed to use it with a pod that gives the extra pins a bit more in the way of protection - which I don't own, BUT given that this is a self contained for the most part totally passive application it would likely be ideal for this and not need extra protection. The pod also would probably not be at all difficult to make.

I wonder how high the signal generator goes. The only manual I have for the whole package is an ancient scanned PDF that looks to be around ten years old.. will have to dig into it.

It may be a diamond in the rough. They are to be commended for supporting the old HW with their new software, that makes it worth a try right there.