You seem to be making a chunking error -- or not reading and understanding my post correctly. Or I didn't write it well. Communication is always two ways, after all. I think it's unlikely I was unclear, but perhaps I shall review it, and add more specificity.
I agree that via and traces add inductance more than a small MLCC capacitor's ESL. I think you are referring the article in the Signal Integrity Journal about usefulness of frequency graphs. The manufacturers did some measurements and shared the graphs with us to save our time while picking the parts. I don't agree that we should ignore the frequency graphs and assume more generic knowledge as a rule of thumb or a starting point.
I'm not referring to anything, actually; these are things I know independently. They may be found (to varying degrees of correctness) in various publications, perhaps SIJ among them, though, at a glance, the above link doesn't really illustrate the frequency shift specifically that I mentioned.
Or, these things can be derived from first principles, applying EM theory and a bit of network theory; or by experiment, setting up test jigs with exactly these varying conditions and measuring them. (For what it's worth, I arrived at my state of knowledge through a mix of all of the above. Not so much direct experiment, in this frequency range exactly, but inferring from other cases which I have encountered directly, and also knowing, by theory, and others' experiments, that there does not exist a cutoff, or fundamental shift; everything is still just as applicable at smaller dimensional scales and proportionally higher frequencies, my inference is justified.)
For one, if I meant "
disregard the impedance plot", I would've said so. I very specifically said:
The notch frequency is not meaningful.
I didn't say to
disregard the notch in general: its depth, or width, nor the asymptotes flanking it. You seem to have misconstrued this.
I suspect you've made a category/chunking error: I said "ignore specific aspect of thing", but you've parsed it as -- you've stuffed it into a bin labeled "ignore thing". What I said, is more nuanced than what you appear to have internalized.
Moreover, I said it in the context of:
a few mm of trace and via inductance (or even the pads themselves, depending on where exactly the test jig is open/short/thru calibrated to) will shift that resonance down by one or two octaves easily.
I demonstrate a network of knowledge: the impedance curve has the specific shape that it does, because it fairly reasonably resembles a lumped-equivalent circuit, C + ESR + ESL; by mentioning inductance, and knowing AC steady-state circuit analysis, we know inductors in series add; and knowing the general scale of things (1mm trace length ~= 0.5nH ballpark*). From this, we can readily conclude that, yes, frequency will shift down, going as sqrt(L_total), and -- if the added inductance is lossless -- the valley resistance (ESR) remains constant. (Whereas if Q remains constant, ESR goes up as sqrt(L) as well. Or maybe Q goes down, who knows.)
*I could've detailed the reasons justifying this figure, for example; but thought it unnecessary, leaving it simply as a point of reference, something understood "by those skilled in the art". (For those curious, it's mu_0, times a geometric factor determined by the cross-section of the trace (as a transmission line) geometry. The factor is usually less than 1, so rather than 1.257nH/mm, it's more like 0.5 for average microstrip.)
Or perhaps I'm taking you too literally -- perhaps you wrote "ignore" in the context of "ignore [aspect]". I have little choice, in technical matters, but to take words literally, at face value -- and assume you wrote what you meant.
The alternative anyway seems worse: if you think even ignoring a minor and readily mutated aspect of a curve should be avoided, that implies you take
altogether too much confidence in the curves, or, perhaps even that they depend on circumstances at all -- in which case, my earlier point remains correct, you've just declined to accept it. Even more chillingly, this would suggest you assume, at least less "generic knowledge" (if not necessarily none), particularly as a rule of thumb or starting point (but, what is a "rule of thumb" anyway, but a piece of "generic knowledge"?) -- which sounds like you must be hardly capable if at all in electronics, a field that assumes vast swaths of generic background knowledge to work at much depth within! So, this seems anywhere from "worse" to "contradictory", and I wouldn't tend to assume this interpretation.
Tim
Home
Help
Search
Login
Register
