Trouble finding 100uF ceramic capacitors, do they exist?

[gamalot]

Try heating them with a soldering iron then waiting for them to cool down and measuring again.

[schmitt trigger]

These high capacity ceramics have terrible DC bias characteristics. The plot below, taken from the data sheet, says it all.

Sure but that has no relevant effect when OP was testing at 0.5V AC.

You are absolutely correct.
However, it is still remains useful information.

[Siwastaja]

These high capacity ceramics have terrible DC bias characteristics. The plot below, taken from the data sheet, says it all.

Sure but that has no relevant effect when OP was testing at 0.5V AC.

"DC bias" is a misleading name - C(V) characteristics is a better name. A capacitor does not understand "DC" or "AC", it just sees a voltage, and the effective capacitance drops. This is evident if you use an oscilloscope to plot the voltage of an AC current source into the capacitor: voltage rise rate is not a constant, but accelerates when the voltage increases. And this is how component testers operate.

The point is, if the part is already rated to lose 80% of its capacitance at 5V (note: I'm saying just "5V", not "5V DC bias", which is a confusing way to say it), then surely it loses some capacitance already at 0.5V. How much, look at the manufacturer graphs (if available) very carefully, but it could be something very real and measurable, like 5-10%.

So OP would see combination of at least three factors:

• Lack of high-temperature curing (soldering)
• "DC bias" effect, even at 0.5V "AC"
• Tolerance

Remember that tolerance also allows manufacturer (who has better quality control than tolerance demands) to manufacture systematically under the nominal spec parts. For example, if component tolerance is given as +/- 20% and manufacturer is sure their process is accurate to +/-10%, they can manufacture parts which fall between -20% and 0% and save some costs while delivering a product that's completely in-spec.

[schmitt trigger]

I also agree with you, capacitance change as a function of the peak applied voltage would be better description.
The reason I mentioned DC Bias is because these capacitor’s most common “gotcha” is as a filter on a SMPS output.
Common rookie mistake, design a required capacitance value, then failing to take into account the precipitous drop in the value under actual operating conditions.

[TimFox]

"DC bias" effect is a reasonable term for a capacitor (such as a rectifier filter capacitor) where the DC value of the voltage across it is high, but the circuit requires a value of capacitance (or capacitive reactance) for small (ripple) voltage about that DC value (small AC component of voltage compared with the DC value).
A good RLC meter with a DC bias capability measures that effect at a specified DC voltage.

[thm_w]

Try heating them with a soldering iron then waiting for them to cool down and measuring again.

Then tell that to your customer in 12 months when the product fails 

These high capacity ceramics have terrible DC bias characteristics. The plot below, taken from the data sheet, says it all.

Sure but that has no relevant effect when OP was testing at 0.5V AC.

"DC bias" is a misleading name - C(V) characteristics is a better name. A capacitor does not understand "DC" or "AC", it just sees a voltage, and the effective capacitance drops. This is evident if you use an oscilloscope to plot the voltage of an AC current source into the capacitor: voltage rise rate is not a constant, but accelerates when the voltage increases. And this is how component testers operate.

The point is, if the part is already rated to lose 80% of its capacitance at 5V (note: I'm saying just "5V", not "5V DC bias", which is a confusing way to say it), then surely it loses some capacitance already at 0.5V. How much, look at the manufacturer graphs (if available) very carefully, but it could be something very real and measurable, like 5-10%.

Look a little closer at the graph that schmitt posted, and see what AC voltage was used.

[mtwieg]

"DC bias" is a misleading name - C(V) characteristics is a better name. A capacitor does not understand "DC" or "AC", it just sees a voltage, and the effective capacitance drops. This is evident if you use an oscilloscope to plot the voltage of an AC current source into the capacitor: voltage rise rate is not a constant, but accelerates when the voltage increases. And this is how component testers operate.

I'm afraid the reality of MLCCs is a lot more complicated than just DC bias effects, and the DC bias curve is likely a red herring in this case. For high density MLCCs with class 2 dielectrics, measured capacitance will depend on applied AC voltage, but in the opposite manner implied by the DC bias curve. That is, measured capacitance will increase with increasing AC test voltage. I've seen examples where this effect can change measurements by up to 50%. Some sources suggest that this is due to hysteresis effects in very thin dielectric layers.

Here's a couple links from manufacturers mentioning this effect. Good manufacturers will always mention the AC voltage they use for testing MLCCs (often 0.5V), and this is why:
https://product.tdk.com/en/contact/faq/capacitors-0007.html
https://www.murata.com/en-us/products/capacitor/ceramiccapacitor/library/solution/insufficient#Hik03

As these links point out, the actual applied AC voltage might be much lower than what your meter displays, especially when the capacitance is very large. If the meter doesn't have some sort of auto level correction, it's worthwhile to verify the actual applied voltage with a DMM/scope.

Obviously if one were to continue to increase applied AC voltage the measured capacitance would start decreasing as the DC bias effect takes over. It's unclear how the AC and DC effects interact though. I suspect that applying a DC voltage will reduce the AC "hysteresis" behavior, but have never really investigated this.

[TimFox]

The implication of all of these valid comments about voltage dependence:
Do not use Class II capacitors where you need a linear capacitor that obeys Q = CV:  e.g., audio signals, timing circuits, analog filters, L-C resonant circuits.
They still are useful for storing charge, so long as you use their actual non-linear characteristics when designing:  e.g., DC filter capacitors.
Historically, ceramic capacitors were disrespected in audio due to the non-linear Class II dielectrics used in disc ceramic capacitors to obtain useful capacitance values.
With modern MLCC construction, C0G Class I capacitors are available up to 100 nF for signal-critical applications.

[T3sl4co1l]

Heh, they're much too lossy for most LC circuits anyway, but you can get some very strange (if not very useful; but not to say useless, applications might exist) behavior, like a Bode plot that has hysteresis!  Or using the C(V) curve as a variable filter (if a crude one, given the losses, poor tolerance, and temperature sensitivity), or the nonlinearity for harmonic generation or other esoterica (parametric amplifier).

Tim

[mawyatt]

Heh, they're much too lossy for most LC circuits anyway, but you can get some very strange (if not very useful; but not to say useless, applications might exist) behavior, like a Bode plot that has hysteresis!  Or using the C(V) curve as a variable filter (if a crude one, given the losses, poor tolerance, and temperature sensitivity), or the nonlinearity for harmonic generation or other esoterica (parametric amplifier).

Tim

Sorta discussed here about Ceramic Capacitor Behavior.

https://www.eevblog.com/forum/projects/ceramic-capacitor-behavior/

Best,