Design Article

IMG1

Buck converter: Key is selecting the right capacitor

Josh Mandelcorn, Group Technical Staff. Texas Instruments

6/22/2009 9:12 AM EDT

Traditional power-supply design practices place much emphasis on the selection and placement of the output capacitor to meet tight ripple and noise requirements. Customers are willing to spend money for high-performance parts, but it's the often ignored input capacitor that by far is more critical for a successful buck design. Its high-frequency characteristics and placement will determine the design's success or failure. There's actually more latitude in selecting and placing the output capacitor. Even to meet output noise requirements, selecting and placing the input capacitor can be more critical.

The stresses associated with the input capacitors are greater than those associated with the output capacitors in two areas. The input capacitor will see a higher rate of change in current, making both placement and selection critical to limiting voltage stresses on the main switches, and limiting noise propagated through the system. Also, its higher root mean squared (RMS) current stress and potential component heating makes its selection more critical to overall reliability.

The current's rapid rate of change
The first area of stress is the rapid rate of change of current, or dI/dT, which shows up as a voltage across any internal or stray inductance. This can put excessive voltage stress on switches or clamp diodes operating off the input capacitor and radiate high-frequency noise into the system.

The input capacitor sees a square wave of current going from zero when the high side buck switch is off, to about full-load current when the switch is on. The current rise times of modern MOSFETs, and in turn in the bypass capacitor, are on the order of 5 ns. It's this rapid rate of current change (dI/dT), multiplied by the total stray inductance (L) that creates voltage spikes on the buck switches. On the other hand, the output cap sees a current waveform smoothed by the output choke and is limited to the peak-to-peak current in the choke. Generally, output choke ripple currents are limited by design to 40% or less of the full-load current.

For a buck operating at 500 kHz and at 10% duty cycle, this means a rise in current of 40% of load current in 200 ns. That's 100% in 5 ns vs. 40% in 200 ns represents a 100 times greater rate of current change and, in turn, voltage across a given inductance. For designs with higher duty cycles or lower ripple currents in output choke, this ratio can be much greater than 100.

1  2  3  4 

print

email

rss

Bookmark and Share

Joinpost comment



Comments


Sator

7/1/2009 3:49 AM EDT

You ought to consider the polymer Al type capacitor as well for this application. They'll handle lead-free reflow, and have much better life-temperature derating than wet types. ESR and ESL for the cylindrical type are slightly worse than for ceramics, but there are no piezo buzzing and voltage dependecy is negligible.
Tantalums are inherently risky, and the mining is blood-stained, menacing both humans and great apes.

Sign in to Reply


howardc123

7/1/2009 11:50 AM EDT

I think that we should use leyden jars since they are more environmental friendly.

Sign in to Reply


Treth

7/1/2009 2:14 PM EDT

Nice article! My only complaint is that Figure 1 was unreadable. Could you possibly repost it, enlarged?

Sign in to Reply


Josh56

7/10/2009 11:19 AM EDT

Thanks for reading my article.
Lowest ESR and ESL is most critical. I just reviewed one user's "Trouble Report" complaining about 160MHz noise being generated by four moderate power buck converters. Layout is also critical as mentioned in the article. And sometimes switches have to be slowed down to meet noise requirements.

Sign in to Reply


Please sign in to post comment

Navigate to related information

Product Parts Search

Enter part number or keyword
PartsSearch

FeedbackForm