4
\$\begingroup\$

I am using a simple buck switching converter (AP3211) to convert 5V-12V to 3.3V.

The issue I am trying to resolve is the output ringing. I see a 800mV P-P ringing on the output, with a frequency of 185MHz. I've read details that I could possibly use an RC snubber to reduce the ringing.

Some references I found:

http://www.ti.com/lit/an/slyt465/slyt465.pdf

onsemi.com/pub_link/Collateral/TND396-D.PDF

The schottky diode has a Junction Capacitance of 120pF. A few different articles have a slightly different approach to calculate the R and C values for the snubber, however I've come to the following values:

R = 7 Ohm C = 270 pF

My question is: If I don't care about power consumption (to an extent, and not enough to use a linear regulator), could I use lower R and higher C to reduce the ringing even further? If so, what are suitable values for R and C?

Schematic showing RC snubber in RED

\$\endgroup\$
  • 1
    \$\begingroup\$ Do you have a schematic? \$\endgroup\$ – Peter Smith May 11 '16 at 13:42
  • 1
    \$\begingroup\$ It is very unusual to have to use a snubber network to stabilize these switchers. Are you sure that you have the correct input (C1) and output (C2) capacitors? Are they low ESR types? Is the layout that recommended by the manufacturer? \$\endgroup\$ – Steve G May 11 '16 at 13:51
  • 1
    \$\begingroup\$ @AdamB can you attach a 'scope trace? What load do you have on the output? \$\endgroup\$ – Steve G May 11 '16 at 13:58
  • 1
    \$\begingroup\$ 185MHz? It's measurement artifact. The phenomena of such frequency in DCDC are switching rise time, boost cap charging, etc. If you really see them, probably layout is poor or you actually see radio interference. \$\endgroup\$ – Gregory Kornblum May 11 '16 at 14:01
  • 2
    \$\begingroup\$ To be sure, please, post oscilloscope screenshot. And, you have to perform the measurement with very short ground lead, not more than 1cm. To do that wrap a wire or solder around your probe and touch the output capacitor from both sides. \$\endgroup\$ – Gregory Kornblum May 11 '16 at 14:03
3
\$\begingroup\$

As far as snubbers go...

If you want to absorb as much noise power as possible then C in the snubber should be large enough that its impedance is near 0 and much smaller than R at the frequency of interest. R should be matched to the noise source impedance if you want to absorb the maximum amount of power from that source.

1/(2 * pi * 185MHz * 120pF) = 7.17 ohms.

So your 7 ohm resistor is about optimal. The only improvement you may be able to make to the snubber is to increase the 270pF capacitor. But be careful that the resonance frequency of the new capacitor is several times larger than your 185MHz noise frequency.

In general...

800mV of ripple does seem excessive in my opinion. If you have enough capacitance on the output of the regulator you shouldn't be seeing that much ripple to begin with. I would try adding more ceramic capacitors on the output of the regulator.

If that is insufficient, then a low pass filter consisting of a 10nH inductor, 10uF capacitor, and 63mOhm resistor would form a second order low pass filter with a cutoff of 3.16MHz. This should essentially eliminate all your noise at 185MHz. When designing the filter be sure that R > SQRT(4*L/C) to prevent ringing, and be sure that C has a high enough working frequency (or use multiple capacitors in parallel to get what you need).

\$\endgroup\$
0
\$\begingroup\$

Any ringing on the switched line is due to the inductor and parasitic capacitance and, the main culprit of parasitic capacitance is the MOSFET inside the AP3211 chip. It might have 100pF of source-drain capacitance and, when the device open circuits (every cycle), the inductor (4.7 uH) resonates with this capacitance. The resonant frequency is (using 100 pF and 4.7 uH): -

\$\dfrac{1}{2\pi\sqrt{LC}}\$ = 7.3 MHz i.e. nowhere near 185 MHz.

To get a resonance peak at 185 MHz means that the internal parasitic capacitance is a tiny 0.2 pF (and this is most unlikely).

Basically my answer is poring water on your observations about what you have seen or measured. I will also add that putting a snubber where you suggest is pointless because it defeats the whole point of a switching regulator.

\$\endgroup\$
  • \$\begingroup\$ The inductance that causes edge ringing on switching converters is usually the parasitic loop inductance through the switching loop and the power supply (e.g. D1, AP3211 MOSFET, C1). Usually this is in the nH range, and 6.2 nH matches up a bit better with the numbers and expectations. \$\endgroup\$ – W5VO May 11 '16 at 14:56
  • \$\begingroup\$ @W5VO not true - the coil will resonate with the mosfet capacitance when the inductor energy left can no longer produce a low enough voltage to cause the flyback diode to conduct. From this point until the mosfet switches back on it rings due to these components. \$\endgroup\$ – Andy aka May 11 '16 at 15:02
  • \$\begingroup\$ We don't know what the load conditions are, (nor do we have oscilloscope pictures) which makes it hard to say with certainty what is happening. The AP3211 datasheet makes no mention of discontinuous mode operation or low current operation in general. If you are correct, this issue would only manifest itself on light loads. Otherwise it would get worse with higher inductor current. \$\endgroup\$ – W5VO May 11 '16 at 15:19
  • \$\begingroup\$ While it could certainly be a measurement artifact or error, and I agree with your assessment that given a 4.7µH inductor you won't get anywhere near 185 MHz. I also believe that there are other resonant loops in the circuit, and given the right circumstances they can act up. \$\endgroup\$ – W5VO May 11 '16 at 15:26
  • \$\begingroup\$ @Andyaka, if putting a snubber where I suggest is pointless, then why is it discussed here: fairchildsemi.com/application-notes/AN/AN-4162.pdf \$\endgroup\$ – Adam B May 11 '16 at 15:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.