Multi-layer Ceramic Capacitors

Background

Multi-layer ceramic capacitors (MLCC) have become very popular on circuit boards (earlier designs used leaded disc type ceramic capacitors). Though they have very good characteristics of stability, low-losses and volumetric efficiency, designers should be aware of their drawbacks. This application note provides some guidelines and pitfalls that may come with these capacitors.
This article does not explain the details of MLCCs or how they are classified. The electronic industry association has done a good job of creating the EIA-198 series of standards relating to ceramic capacitors. This article is meant to highlight critical elements of these capacitors that can be of concern to a circuit designer.
MLCCs are used primarily as decoupling capacitors. These are capacitors connected between power and ground, and primarily provide the switching energy for circuits. This mitigates unintended coupling of energy (at higher frequencies) through the power buses and prevents malfunction. The primary reason that these capacitors are not used in precision applications like filters or oscillators is because their absolute tolerance is not good. In order to manufacture them cheaply, most manufacturers have relaxed the tolerance band for the capacitance value of MLCC components. In general, the previous statement applies to dielectrics classified as class-II ceramics (X5R or X7R). Class-I ceramic capacitors have good tolerance and characteristics (though they are more expensive). Hence, circuit designers use class-I based capacitors (C0G or NP0) for applications where the value of the capacitor is critical.

Electrical Characteristics

The first thing to note is that a capacitor is an ideal element. Most practical capacitors are close to the ideal capacitor (as specified by electrical behavior) only over a specific range of frequencies. Ideally, capacitors are two conductive plates, with a dielectric sandwiched between them. Capacitors have series resistance - due to the contacting metal. At sufficiently high frequencies, the metal leads of MLCCs (or other capacitors) also behaves as an inductor. For the most part, MLCCs are very good capacitors. They have low ESR (equivalent series resistance) and have a high self-resonance frequency. Hence, in most decoupling applications, they behave better than leaded capacitors. However, there may be differences in electrical characteristics, depending on the size, type and manufacturer of the capacitor. For the purpose of this article, some capacitors made by TDK Corporation and some by Murata have been chosen.

Impedance Response:

Capacitors are intended to provide low-impedance at higher frequencies. The impedance response of a few MLCC capacitors is shown in figure-1. All of the capacitors have a nominal value of 100nF. The figure shows the comparison between capacitors of different sizes (0402, 0603, etc.) with the same nominal capacitance value. Though the curves seem to overlap, it is observed that capacitors of smaller sizes (0402) have higher self-resonance frequencies. So, if the application needs better decoupling at higher frequencies, one is better off choosing 0402 (or even smaller sizes - though 0201 capacitors have not been included in this study) - if all other factors are the same. The Murata 0402 X5R capacitor seems better than TDK 0402 X7R capacitor at higher frequencies (above 25 MHz), but this difference is within the tolerance range of capacitors and hence should not be relied upon - for a design that has to be mass manufactured.

How much capacitance should you put on your board? Guidelines and rules of thumb abound in this area. Some recommend one capacitor for each switching ASIC and others go by number of capacitors per square inch of the board. The difficulty is that all of the guidelines have inherent assumptions behind them. As long as your board fits within those assumptions, they will work in your situation. When your application is more critical - for instance, a mother board switching at 1-GHz clock rate or a sensor with very small signals, then it may not be enough to base your design on the rules of thumb. Many times, you will have to iterate the design of a board to come to the right optimum (lower the cost of the board versus lower design cycle time). A general guideline is to estimate the amount of switching current on your design, the frequencies at which these currents occur and then lower the impedance of the power-ground plane to a reasonable level - so that the resulting power supply disturbance is within limits. This implies that larger capacitors are required for lower switching frequencies.

It is interesting to see if using multiple values of capacitors on the supply helps in improving the decoupling. For instance, fig-2 shows the impedance of various capacitors - all in the 0402 MLCC form factor. It is interesting to see that at high frequencies, the curves overlap. Hence, if the designer is interested in decoupling at higher frequencies, then using one large value and another smaller value capacitor does not improve the situation. It is better to choose the largest capacitor available in the smallest form-factor and use that capacitor. However, note that the voltage rating of large value capacitors drops (while maintaining the same size). Hence, it is recommended to use the maximum capacitor that can provide the dielectric withstand for the given supply voltage. Inherently, this makes sense, since the parasitic inductance depends on the package size. Hence, all capacitors in a given package size have similar behavior at higher frequencies.

If you have to use a larger value of the decoupling capacitor (to meet the desired voltage rating of the application), then it makes sense to use a smaller capacitor to lower the impedance at higher frequencies. This is shown in Fig-3, where a 2.2uF capacitor in size of 1206 is used in parallel with a 100nF capacitor of 0402 size to achieve overall lower impedance. Note that the capacitor in form-factor 1206 has lower impedance at lower frequencies, but has higher impedance at higher frequencies due to larger parasitic inductance. Hence, using two capacitors in parallel makes sense here.

Voltage derating:

All of the MLCC capacitors with class-2 dielectrics have a horrific voltage coefficient. For most decoupling applications, the voltage non-linearity of the capacitors is not important - but if you are counting on the absolute value of the capacitor to guarantee stability (like required minimum loading capacitance for a low-dropout regulator), then one needs to account for this derating factor. Figure 4 shows the typical derating factor for MLCCs. For comparison, the curve for a class-I dielectric is also shown in the figure-4. It is observed that smaller form factor capacitors have a larger derating factor. The capacitance value falls by as much as 80% of its original rated value if the voltage bias is 40volts (for a 50volt rated capacitor). Note that all of these capacitors are of the same 100nF value in this graph. Hence, if the application calls for high voltage stress on the capacitor, one should apply the appropriate derating factor - or use class-I dielectric capacitor.

Mechanical Factors

In industrial, automotive and aerospace applications, a lot of mechanical vibrations or bending forces are applied on the circuit board. Most electrical engineers do not pay attention to these details - as it is relegated to the domain of mechanical and packaging engineers task to address these issues. However, the most economical solution to the problem may be in the electrical domain. Ceramics are stronger in compression but weaker in tension. When a capacitor is mounted on the board and the board is flexed, repeated vibrations can cause the capacitor to develop cracks and fail. As expected, larger (longer) capacitors experience more stress and are more prone to failure. This is especially noticeable if the form factors of the capacitor is 1210 and higher. Vibration tolerance is more difficult to measure in smaller geometries like 0402 and 0603. In general, using smaller form-factors makes a design less susceptible to mechanical failure (from vibration).

When capacitors fail, they may cause catastrophic failure - by causing a short between power and ground (especially for decoupling capacitors). This causes the whole system to fail. In order to permit graceful failure (where only the module that has the capacitor fails), it is necessary that the capacitor fails by open-circuit and not short-circuit. To improve these failure scenarios, either capacitors with soft termination (metal terminals) or open-mode capacitors can be used. Soft-termination capacitors try to isolate the stress from reaching the ceramic by using stress-isolating flexible termination. Open mode capacitors reduce the amount of overlap between parallel plates, so that if a crack occurs (and most of the cracks occur closer to the solder joints as shown in figure-5), they do not cause a short circuit. Another alternative is to use flex- boards, so that vibrational stress is not induced in the capacitors.
In power applications, MLCC capacitors may experience vibrations due to the piezo-electric property of the ceramic dielectrics. When mounted on a rigid boards, the electrical energy can get transformed to mechanical energy and then into acoustic energy - resulting in capacitors emitting hum or audible sound. Larger MLCCs are likely to exhibit lower resonance frequencies and hence emit lower frequency sound. In such cases, it may be advisable to use low-distortion or metal terminal capacitors, so that any audible noise is minimized.

Thermal Issues

As opposed to the voltage derating, MLCCs have much lower derating factor for temperature change. As long as they are used within their specifications, temperature induced changes in the capacitor value is within the tolerance specifications given by the manufacturer.
Since capacitors are reactive components with high Q-factor, they do not dissipate real-power. However, if a significant AC current flows through the capacitor (for instance in power electronics applications), then the ripple current can cause the capacitor's temperature to rise above the ambient temperature. This increase is primarily due to the ESR (equivalent series resistance of the capacitor), and is a function of the frequency. Capacitor manufacturers provide graphs indicating the rise of temperature that can be used for design (see figure that also correlates to the ESR values of the capacitor for the frequencies of interest (see fig-6). Designers should ensure that capacitors used in such applications are still within their datasheet specified temperature limits - after accounting for the increase of temperature due to the thermal heating effects.

Manufacturing Concerns

Capacitors placed near the edge of the board or where scoring takes place are likely to see more strain due to the board separation process. This strain can cause the capacitor to crack. It is recommended that capacitors be placed at a minimum distance of 30-60 mils from the edges. In general, it is also recommended to avoid side fillets when the soldering process takes place - to minimize cracking.
When drawing the land-patterns for capacitor placements, it is recommended that the pads be identical (on either side of the capacitor). This is more important for smaller form-factors, as the smaller capacitor is held in place due to the surface-tension of the solder. Uneven melting of the solder can cause the capacitor to tomb-stone, resulting in manufacturing headaches. The IPC -SM-782 recommendation for land-patterns is to be the size of the width of the capacitor - especially for the larger form-factors. This will avoid the formation of side fillets. The size of the land-pattern also determines the size of the toe-fillet (See figure-7). Improperly formed fillets can cause more strain on the capacitors. Though this is not as important for commercial applications, other applications that are subject to large temperature changes will exacerbate the strain on the capacitor due to improper fillet. The profile of the fillet is also examined to see if the electronics assembly is compliant to IPC-A610 acceptance criteria.

Conclusion

Many factors go into picking the right de-coupling capacitors for your application. For your application to be manufactured without defects and work well in the field, you should pay attention to decoupling capacitors. Otherwise, you run the risk of your circuit malfunctioning in situations that you don't expect - such as system level integration, creating noise on the power supply. If you need assistance in getting your design production ready, we can provide design consultation and help.