Design Article
Comment
BrianDonegan
Or use C0G/NP0 MLCCs.
Dave_at_Mot
The piezo effect goes both ways, and while the audible noise problem is a ...
Power Tip 49: Avoid these common multi-layer ceramic capacitor pitfalls
Robert Kollman, Texas Instruments
6/15/2012 11:47 PM EDT
(Editor's note: Power Tips is an ongoing series; to see a linked list of entries, click here.)
Please join us next month when we will continue to discuss capacitor selection in switched-mode power supplies. The following is part 1:Multi-layer ceramic (MLC) capacitors have become extremely popular in power electronics due to their small size, low equivalent series resistance (ESR), low cost, high reliability, and high ripple current capacity. Commonly, they are used in lieu of electrolytic capacitor to enhance system performance.
MLC capacitors have the advantage of a high relative permittivity material (2000-3000) compared to electrolytics with relative permittivity of 10 with the aluminum-oxide insulation of an electrolytic capacitor. The difference is important because capacitance is directly related to permittivity. On the positive side for the electrolytics, the aluminum-oxide thickness that sets the plate separation is much less than the ceramics that can result in higher capacitance density.
The permittivity of the ceramic capacitor is not stable over temperature and DC bias and needs to be understood in the design process. High permittivity ceramics are classified as Class 2. Figure 1 shows how they are grouped with a three-digit description, such as Z5U, X5R and X7R. For instance, a Z5U capacitor has a temperature rating of +10 to +85o C with a variation of +22/–56%. Even the more stable dielectrics have a sizable capacitance variation with temperature.
Figure 1: Class 2 dielectrics are grouped with a three-digit classification. Watch that tolerance!
Things get much worse when you examine the capacitance dependence on applied bias. Figure 2 presents the bias dependence of a 22 uF, 6.3 volt, X5S capacitor that you typically would use as the output capacitor in a 3.3 volt point-of-load (POL) regulator. The capacitance drops 25 percent at 3.3 volts, resulting in increased output ripple and a significant impact on control loop bandwidth. If you tried to use this capacitor at 5 volts out, between temperature and bias, the capacitance could drop as much as 60 percent and might result in an unstable power supply due to a 2:1 increase in loop bandwidth. This is a point that ceramic capacitor vendors gloss over.

Click on image to enlarge.
Figure 2: Watch out for the decrease in capacitance with applied bias.
The second potential pitfall with ceramic capacitors is that they have a relatively small amount of capacitance and a low ESR. This can create problems in both the frequency and time domain. If they are used as input filter capacitors on a power supply, they can easily resonant with the input interconnect inductance and create an oscillator as we discussed in Power Tips 3 and 4. To see if you have a potential problem, estimate the parasitic interconnect inductance as 15 nH per inch and compare the filter output impedance to the power supply input resistance per the articles. A second potential problem exists in the time domain and can be experienced in systems like power-over-Ethernet (POE).
In these systems, the power source is connected to a load through a large interconnect inductance. The load is switched on through a switch and may be bypassed with ceramic capacitors. The bypass capacitors and interconnect inductance can create a high Q resonant circuit. Closing the switch at the load can create an over-voltage condition as the load voltage can ring to twice the source voltage. This can lead to unexpected circuit failures. For instance, in the POE, the voltage ratings on the load components can migrate as much as double the source voltage ratings.
A third potential pitfall lies in the fact that the ceramic capacitors are piezoelectric. That is, when the voltage changes on the capacitor, its physical size changes, which can result in an audible noise. Examples of problem applications are when the capacitors are used as output filter capacitors where a large load transient current is present or in “green” power supplies that go into burst modes at light-load conditions. Workarounds to this issue include:
- changing to a lower permittivity ceramic material like COG
- using a different dielectric such as film
- using leaded versus surface mount technology (SMT) components which are tightly coupled to the printed wiring board (PWB)
- using a smaller footprint device to reduce stress coupled into the board
- using a thicker part to reduce applied voltage stress and physical distortion
- restrict package size to 1210.
- keep capacitors away from high flexure areas such as corners
- orient capacitors in the short direction of the board
- keep board mounting points away from corners and edges
- be mindful of potential board flexing during all assembly steps
For more information about this and other power solutions, visit: www.ti.com/power.
Robert Kollman is a Senior Applications Manager and Distinguished Member of Technical Staff at Texas Instruments. Kollman earned a BSEE from Texas A&M University, and a MSEE from Southern Methodist University.
Navigate to related information


BooStar
7/12/2012 4:13 AM EDT
Link does not work
Sign in to Reply
a10112
7/12/2012 6:03 AM EDT
shame it doesn't work, hope that's fixed soon. I was hoping to read something interesting for once.
Sign in to Reply
Dennis.Bartch
7/12/2012 8:12 AM EDT
Link is broken!
404 - File or directory not found.
The resource you are looking for might have been removed, had its name changed, or is temporarily unavailable.
Sign in to Reply
jpmcwil
7/12/2012 9:41 AM EDT
Modify the URL link due to special character "#" in filename. "#" becomes "%23" the same way a " " becomes a "%20" .
Then the link will work.
http://www.eetimes.com/ContentEETimes/Documents/Power%20Tip%2349%20July%202012_FINAL.pdf
Sign in to Reply
nicolas.mokhoff
7/13/2012 11:12 AM EDT
This has been corrected. Enjoy!
Sign in to Reply
Dave_at_Mot
7/19/2012 2:51 PM EDT
The piezo effect goes both ways, and while the audible noise problem is a concern, the reverse effect of shock and vibration causing an electrical response can be far nastier to deal with. Any low level analog circuit such as audio or hi-res AtoD measurement needs to be very careful to avoid MLCC's in the signal path.
Dave
Sign in to Reply
BrianDonegan
9/5/2012 3:58 PM EDT
Or use C0G/NP0 MLCCs.
Sign in to Reply