Why do these two sets of seemingly identical capacitors behave differently?

Question

I have two sets of almost identical-looking capacitors which show different behaviour:

The ones on the left (which are definitely these MLCCs) act as I would expect, but the ones on the right (which we think were a previous order of the same but cannot be sure) have odd effects in an RLC circuit as below:

^{simulate this circuit – Schematic created using CircuitLab}

At the resonant frequency, the trace of \$V_\mathrm{FG} - V_\mathrm{r}\$ shows a triangle wave:

and consequently the FFT of the \$V_\mathrm{r}\$ channel shows odd harmonics for the "bad" capacitor:

but the good capacitors give an almost pure sine as expected when swapped

They both measure the same approximate capacitance (100 nF) on a multimeter. Also, when sweeping the frequency across the resonance, the "bad" capacitors have a much faster phase change when going slightly above the resonant frequency. What imperfection or property of these capacitors could be causing this nonlinear behaviour?

Answer 1

Classic type 2 dielectric behavior. All capacitors of this class have a voltage dependency (ferroelectric saturation), meaning C decreases as |V| rises. Evidently you have a particularly sensitive example.

Tolerable voltage, for a given amount of C loss, varies widely with material type (temperature coefficient code), chip size, and construction. Basically it depends on the thickness of layers, but thickness and layer count are up to the manufacturer. So you really just need to look up the characteristic sheet for the part -- if you know the part number, and if the manufacturer provides such.

Random junkbox parts, you have very little hope of that, so, at best you can characterize them. Which is a bit of a PITA...

Most capacitors aren't too sensitive, but especially larger values, and at higher voltages, the decrease can be surprising. It's a common pitfall.

Type 1 (C0G, etc.) ceramics, and film capacitors, do not exhibit C(V) dependency, so are a good choice for resonant circuits.

---

Interesting aside about the circuit: it has hysteresis with respect to frequency. As amplitude grows, average capacitance (averaged over a cycle, roughly speaking) drops. So as you adjust frequency down, it seems to stay in resonance, and amplitude continues to increase. Eventually, amplitude peaks, then frequency drops below the passband (given the driven amplitude) and amplitude suddenly drops. Ramping frequency back up again, a much smaller resonant peak is seen. Not very useful, but a nonlinear curiosity.

Answer 2

Y5V is a nasty Class II dielectric- the worst much in present-day use, I think.

I wouldn't even use it for a bypass capacitor let alone resonate something with it (it might be acceptable for a toy that never has to work over much of temperature range). +20/-80% over voltage/temperature and a fast aging rate.

Possibly they substituted something like X5R- on one of the orders. A better dielectric will meet all the specifications of the crummy one easily. Or maybe the construction is just different (perhaps from different factories). If you want nice behavior where the capacitance value and loss characteristics matter, NP0 or a film cap would be best.

Y5V is bad enough that a friend of mine suggested the possibility of making an amplifier with it.

The voltage coefficient varies greatly with manufacturer and construction, here is a Z5U MLCC from a the same maker with similar voltage rating:

The other characteristics are similarly unpleasant:

Answer 3

Ceramic capacitors have different dielectrics and the dielectric is labelled by class, not material composition. So that means even two capacitors both labelled as X7R might actually be made of different materials.

Answer 4

Either the capacitance or dissipation factor of the capacitor is voltage dependant. And in the case of the dissipation factor, really poor. With your cap meter put a good cap (larger, say 1 uF) in series with the suspect cap like this meter neg - test cap - good cap - meter pos. Then you can use a resistor (say 100K) connected to the junction of the caps and add DC bias to the test cap and see if its capacitance changes with say 0, 1 and 2 Volts across it. You can also try measuring its leakage, but you need to know its voltage rating. Any way hook it up in series with the most sensitive current meter that you have to a power supply. Bias the cap at its working voltage, you should measure no current. However whatever is happening to create the distortion it needs to be a (changing) function of the applied voltage.

Answer 5

In a series RLC circuit operating at resonance, the applied voltage (your V1) (V_FG), should equal your Vr.

In a series RLC circuit operating at resonance, Vin equals Vr.

Since your input voltage V1 does not equal Vr (you get a triangle wave), you have something important missing from your equations.

The missing part of your equations appears to be a resistive portion of capacitance or the resistive portion of your inductor.

Since you have not mentioned that your inductor has a core other than air, it seems reasonable that the difference in your output lies with differences between the capacitors.

Therefore it seems reasonable that the difference between your capacitors is a non linear component of resistance in the capacitors dielectric.

Answer 6

They might be rated for different voltage, or they may be made from different material.