How to screw up a capacitance measurement

Consider the task of performing a capacitance test using an LCR meter or a good multimeter such as the Fluke 287 True RMS Multimeter. If the capacitor under test is an electrolytic or film-based device, the main measurement concern is often just ensuring the capacitor is not under power or doesn’t have a stored charge–electrolytics in particular can store and release large amounts of current at high voltage after an extended amount of time. So a primary measurement task is to first ensure any hazardous stored voltage is bled off to avoid severe shock. This done, you should be able to attach a multimeter or LCR meter and get a capacitance measurement.

However, the measurement task isn’t so simple for some kinds of multilayer ceramic capacitors (MLCCs). The reason lies in the way LCRs and multimeters make the measurement. Capacitance measurements with these instruments typically involve charging the device under test with a known current and then measuring the voltage. Here’s the rub: MLCC manufacturers specify device capacitance under the conditions of a specific applied dc voltage. And in fact, some MLCC devices exhibit a capacitance that varies with applied dc voltage. So if the MLCC ends up experiencing a voltage during the measurement that differs from that used by the manufacturer to specify the level of capacitance, the MLCC can look as though it is out of spec.

This effect is called voltage coefficient of capacitance or VCC. It arises in Class two and Class three MLCCs. (The Class two and three monikers denote capacitor dielectrics with a high permittivity which gives these devices high capacitance.) The capacitance of these devices drops as applied dc voltage rises. This capacitance

drops happens regardless of who manufactured the MLCC and is a function of the design and material properties. Class two MLCCs are made from BaTiO₃ ferroelectric material. As dc voltage is applied to the device, an electric field influences titanium ions so to lock them in place within the crystal lattice structure of the ferroelectric material. This action prevents the capacitor material from being influenced by applied ac voltage and thus reduces the dielectric constant of the material, resulting in a measurable drop in MLCC capacitance.

To further complicate matters, not all Class two and three MLCCs exhibit the same level of capacitance loss versus dc voltage. As pointed out by MLCC manufacturer Kemet, some MLCCs may lose 10% capacitance at rated voltage while others having the same case size can lose 70% at rated voltage. One reason: Higher voltages cause higher electric fields on each active layer, pinning those titanium ions hard in place. Also, the ferroelectric material in the capacitor can include dopants for various performance reasons that can significantly worsen VCC. Further, different MLCCs use ceramic dielectrics whose thickness can vary from 10 μm down to below 1 μm. The thinner the layer, the higher the electric field acting on it, and the more pronounced is the VCC effect.

Thus VCC can be particularly acute in super-small MLCCs that necessarily have thinner dielectric layers. A further point to note is that ac voltages also cause electric fields within the MLCC dielectric. The magnitude of the electric field on each layer is proportional to the peak ac voltage across the MLCC. So capacitance can change with the applied ac voltage as well as with applied dc voltage. Thus for making an accurate capacitance measurement, the ac voltage of the meter should be the same as that specified in the MLCC datasheet.

Finally, if capacitance sensitivity to voltage is an issue, consider going to a Class one MLCC. These devices have a dielectric based on CaZrO₃ which is a paraelectric material. These capacitors provide minimum change or drift in capacitance with temperature and voltage. The paraelectric dielectric material has a relatively low permittivity, so their capacitance values are generally in the low picofarad to microfarad range.