How to test the capacitance of ceramic capacitor under dc bias?

[onemilimeter]

I have a bunch of multilayer ceramic capacitors in 1206 package. As we know that there is no label on ceramic chip capacitor, the only way to tell its capacitance is to measure it. However, other details (e.g. voltage rating and type of dielectric) are not known. How do we measure the capacitance of a ceramic capacitor under dc bias?

Thanks.

[Conrad Hoffman]

Methods vary for different capacitance bridges and meters. You may be able to isolate the cap being tested with a couple much larger caps, then apply a bias through some large value resistors so as not to mess up the capacitance reading. Since there's no DC path with the isolating caps, the cap under test should charge up to the full value of the bias supply. Consult the manual for whatever meter you're using, as they often describe how to do this.

[Short Circuit]

You can't tell the voltage rating from applying DC bias either. MLCC's are very rugged when it comes to overvoltage,
so they might work quite well at 2 or 3 times rated voltage (for a short while).
At best you can determine the type of dielectric Y5V vs all others.
Why the effort? Except for exotics, these things are dirt cheap.

[saturation]

DC bias is often a 'contaminant' in measuring capacitance using medium priced meters like a hand held LCR meter.  But if your need is to measure the actual capacitance while in a circuit in the presence of a DC bias, it requires a good LCR meter, the $30,000 'impedance analyzer' variety.

http://www.home.agilent.com/agilent/product.jspx?nid=-33831.536879654.00&cc=US&lc=eng

http://www.eetimes.com/design/analog-design/4009883/Capacitance-measurement-Understand-and-use-the-right-technique-to-dramatically-improve-results

DMM onboard capacitance meters apply a "DC bias" and measure the time constant to say 63% of applied voltage and use a formula to estimate capacitance.

[onemilimeter]

You can't tell the voltage rating from applying DC bias either. MLCC's are very rugged when it comes to overvoltage,
so they might work quite well at 2 or 3 times rated voltage (for a short while).
At best you can determine the type of dielectric Y5V vs all others.
Why the effort? Except for exotics, these things are dirt cheap.

I learned from this forum few weeks ago that the capacitance of a MLCC will be lower if a dc bias applied across it. Some of the MLCC I have are 10uF and they are quite expensive (based on the price quoted in Farnell) compared to those with values lower than 1uF. I thought if may be possible to guess its voltage rating (approximately) by observing the variation of its capacitance against applied dc bias voltage (see attachment). Unfortunately, like you mentioned it's possible.

[amspire]

Measuring the capacitance with a DC bias is pretty easy if either the meter or voltage source is floating (like a 9V battery) or if they can share a common ground.

Just add the DC bias via a large resistor (1Meg is probably fine) to the DC voltage source. If connecting the resistor changes the reading too much, make it bigger.

Connect the capacitance meter via a large non-SMD cap, or maybe a small electrolytic.  Measure this capacitor first, and it is easy to make a graph in Excel or OpenOffice Calc to correct the measurement for the series cap.  It is basically

C_unknown = (1/C_series - 1/C_measured)^-1

This methods will show the capacitance change with voltage easily, and you can always measure directly just to get the accurate capacitance value.

I would not use this arrangement if you are going to increase the DC to the point the capacitor may fail as that will result in a nice big transient pulse going back to your capacitance meter. Similarly, always take the applied voltage back down to 0v before changing test capacitors for the same reason.

Richard.

[saturation]

A DIY setup is probably most beneficial to find out that DC bias does affect the rated capacitance, that is after one checks it with and without DC bias, but what the value actually will can be more elusive unless other measurements are taken into account. 

Just as example see:

http://www.cliftonlaboratories.com/capacitor_voltage_change.htm



http://electronicdesign.com/article/power/turn-capacitors-and-inductors-into-active-passives.aspx

[Short Circuit]

Gee, always knew Y5V wasn't very stable and never used them. But after seeing the charts at that site, one wonders why anybody bothers to manufacture these at all. What a crap !

[amspire]

Gee, always knew Y5V wasn't very stable and never used them. But after seeing the charts at that site, one wonders why anybody bothers to manufacture these at all. What a crap !

They do a really great job as a power supply decoupling cap. You get a lot of capacitance in a small package. They are not meant for any application where accuracy is required.

[IanB]

Can someone explain in terms of physics why a DC bias affects the capacitance value of certain capacitors? Is it because the electric field from the bias voltage does something to the dielectric?

[amspire]

Can someone explain in terms of physics why a DC bias affects the capacitance value of certain capacitors? Is it because the electric field from the bias voltage does something to the dielectric?

I cannot tell you the exact details, but here is a rough explanation. If I get this wrong, please correct me.

Many common insulators like glass, plastics, mica have a dielectric constant of about 2 to 5 -  ie 2 to 5 times better then a vacuum.

The best behaved ceramic capacitor - the NPO cap - has a dielectric constant of about 60. So you can get a over 20 times the capacitance then you could with a common insulator.

Y5U and Y5V can have dielectric constants over 10,000, so you get a massive increase of capacitance for a given size of capacitor. This happens because the particles in the ceramic have charges that can move slightly under an electrical field. If you apply enough electric field by increasing the DC bias, you start to saturate these particles - they have aligned their charge to the field as much as they can and they cannot do any more. At his point the dielectric constant plummets downward, and so the capacitance decreases proportionally.

In the dielectrics with a low dielectric constant, the material will fail from voltage breakdown long before this situation occurs.

Richard

[Short Circuit]

Gee, always knew Y5V wasn't very stable and never used them. But after seeing the charts at that site, one wonders why anybody bothers to manufacture these at all. What a crap !

They do a really great job as a power supply decoupling cap. You get a lot of capacitance in a small package. They are not meant for any application where accuracy is required.

It seems you can get a lot of capacitance indeed. But after derating, there is not that much left. When used close to their working voltage, the Y5U has only ~25% of the nominal capacitance left. At room temperature that is. When temperature increases, and usually it does, we start losing capacitance again. Up to 82% at 85 degrees C to be precise; so the net capacitance left is only 4.5% of the nominal value! So it you want a certain minimal decoupling value to be available over the entire operating range, say 22nF, you'd need to use 470nF caps when using Y5V.

Dielectrics coding: http://www.niccomp.com/Products/TC_Ceramics.pdf
Y5V temp graph: http://www.johansondielectrics.com/technical-notes/product-training/basics-of-ceramic-chip-capacitors.html
X7R volt&temp charts: http://www.niccomp.com/catalog/nmc2.pdf

Of course, this is not a big deal for the simple bypassing of logic ics and micros because of the easy availability of 25-50v mlcc caps, so the voltage derating does not apply that much. But when you move towards DC converters and such, voltages tend to get higher, as do capacitance requirements. For example many of my designs have some 10uF/50V/X5R/1210 in the power sections. The same thing is also available in Y5V at ~50% of the price. But after all the derating, I'd probably better off with a 1u X5R instead (at ~10% price).

Anyway, I understand better now that many companies have design constraints that prohibit parts like 5% resistors or Y5V capacitors (unless explicitly required for some reason). Better safe than sorry after all.

[Joshua]

Im a newb so there is no reason to take what I'm about to say as any sort of fact, but if you are just trying to figure out the capacitance, couldn't you just use a precision resistor, stable voltage source, and record the time constant on an oscilloscope? Wouldn't this give you the capacitance since you know the resistance and voltage?

[IanB]

Im a newb so there is no reason to take what I'm about to say as any sort of fact, but if you are just trying to figure out the capacitance, couldn't you just use a precision resistor, stable voltage source, and record the time constant on an oscilloscope? Wouldn't this give you the capacitance since you know the resistance and voltage?

I think you could. Once you know the appropriate circuit theory you can measure things like capacitors and inductors using experiments on the work bench. However, it is typically a bit time consuming and fiddly to set up the experiments, record the results and do the analysis. And then you have accuracy to worry about--how do you calibrate the test setup, eliminate experimental error and ensure reproducibility?

The advantage of LCR meters is convenience and speed in light of the above questions. But for a one-off and a learning exercise you can always improvise.