Bench Talk

Most of us learned in school that there are two types of capacitors: Polarized and nonpolarized. The critical difference is that you should never put polarized capacitors in backwards. As we begin working with capacitors, we learn that capacitors have a much more complicated function than just serving as two electrode plates separated by a dielectric. Depending on what materials the dielectric and the electrode layers consist of, the behavior and characteristics of the capacitor can be very different. 

The surprise twist here is that capacitor datasheets have become notorious for not giving complete information on components. For example, often manufacturers do not publish the self-resonant frequency (SRF) for their multilayer ceramic capacitors (MLCCs). It has become more common for capacitor manufacturers to include “typical” graphs for performance characteristics. Alternatively, in the case of some dielectrics, they only provide an individual chart for each parameter. While being useful information, all these approaches still require the end designer to do a bit more math.

Before looking into the case of whether capacitor datasheets lie or not (spoiler: They do not!), let’s first look at ceramic capacitor losses as an example.

Example: Ceramic Capacitor Loss Parameters

Ceramic capacitors are by far the most common type of capacitor used today, yet they’re one of the least understood. Ceramic capacitors, or MLCCs, can lose capacitance for three key reasons. The first one is obvious because all passive components suffer the fate of what’s called the tolerance band, and while it is possible for a device to be near the plus side of the band, this is rarely the case. Two other key issues—temperature and applied DC voltage—commonly cause a loss of capacitance in ceramic capacitors as well.

Temperature Loss

When you see a capacitor labeled as “C0G” or “X7R” you are seeing the dielectric’s temperature coefficient. Common coefficients are C0G, U2J, X7R, and X5R. There are more than these, but these coefficients represent the more popular types. C0G and U2J are considered “ultra-stable” Class I dielectrics. They change very little with temperature. C0G, for example, means that the capacitor changes ±30ppm across the rated temperature range. X7R and X5R fall into Class II dielectrics. Each letter has a specific meaning, as Figure 1 shows.

Figure 1: This table represents a ceramic capacitor’s Class II temperature coefficients. (Source: KEMET)

The first position is a high temperature, the second position is a low temperature, and the third position is an expected change. The X means -55°C; the "7" means +125°C; and the "5" means +85°C. Finally, the R means ±15 percent. So X7R means between -55°C and +125°C with a capacitance that may vary up to 15 percent. (Hint: It rarely goes “plus.”). Usually, the capacitors are peaked so that at about room temperature there is no loss with a bell-shaped curve towards the extreme temperatures (Figure 2). 

Figure 2: This graph represents MLCC relative permittivity change vs. temperature (Class I, II, and III). (Source: KEMET)

Applied DC Voltage

The third reason for ceramic capacitance lose has to do with a “DC bias” or “voltage coefficient.” In other words, an effective capacitance drops at the of application of a DC voltage. Paraelectric capacitor types, like C0G, do not exhibit this issue. However, all ferroelectric capacitors, like X7R and Y5V, exhibit loss at the application of a DC voltage. Several factors determine the level of loss in each capacitor type. Today, there is no commercial specification for rating this loss. Still, as a “rule of thumb,” the thinner a ceramic capacitor’s dielectric layers become, the more evident this effect becomes: Which means for lower rated voltages or lower rated temperatures, the voltage coefficient will be higher. 

A simplified explanation for why this loss occurs is related to the electric dipoles in the dielectric material. The ceramic dielectric is composed of a crystalline structure that arranges those electric dipoles into separate domains (Figure 3). Spontaneous polarization of these domains occurs when no DC voltage is present. This polarization gives the dielectric material a high relative permittivity. (By the way, relative permittivity replaced the deprecated term “dielectric constant.”)

Figure 3: This illustrates the MLCC DC voltage bias, dipole example. (Source: KEMET)

Now, applying a DC voltage to the capacitor aligns the domains, locking them in the direction of the electric field. The higher the voltage, the more locked domains. This change reduces the permittivity of the dielectric layer, which means there is less room to store energy or, in other words, less capacitance is available.

As mentioned before, some capacitor vendors will provide individual graphs for each of these losses. You are probably interested in how both temperature and applied voltage might affect an MLCC; this is why  KEMET provides a capacitor parameter simulator, called K-SIM, that combines these parameters into a single graph (Figure 4). It does not perform a Software Process Improvement and Capability dEtermination (SPICE) calculation, but it can provide a netlist for use in a PSpice^®.

Figure 4: This is K-SIM in action, showing multiple parameters in a single graph. (Source: KEMET)

K-SIM can provide plots on ESR/impedance vs. frequency; capacitance change vs. frequency, ripple current calculations, S-parameters, and netlists; and for ceramics, a capacitance change vs. applied voltage (Figure 5).

Figure 5: K-SIM offers various controls. (Source: KEMET)

Once a graph is selected, there are controls to allow you to enter an applied voltage and application temperature. There are various other controls also available, but these two are the most common. For some advanced comparisons, try clicking “Compare Across Conditions,” then you can see how a part performs, on a single graph, at -55°C, 25°C, and +125°C.

Capacitor datasheets have become notorious for omitting performance details and, instead, include a single graph for “typical” performance or individual graphs for each parameter. Performance, though, isn’t measured in a single parameter but a combination of parameters. The KEMET capacitor parameter simulator helps solves this problem by simulating combinations of variables into a single graph. The combinations include ESR/impedance vs. frequency; capacitance change vs. frequency, ripple current calculations, S-parameters, netlists; and for ceramics, capacitance change vs. applied voltage, in addition to advanced comparisons across conditions.