SMPS output voltage ring

[wakoko79]

Hi,

This is my first smps design. I used a LM5088 for this (buck, non-sync). test setup:
Vin=48V
Vout=19V
Iout=5A (5.5A max, computed)

My problem is the voltage ringing at the output (see attatchment).

I tried to use a short probe so that I would know if the probe is responsible but it is still there.
I have a RC snubber across the schottky diode (2.2nF, 10k).

Is it possible that This is caused by feedback compensation? (but i think it is not cause by that, aren't feedback compensation there to regulate dynamic load changes?)


Another thing for lm5088 users. I tried to overcurrent(I use an e-load), the voltage and current output folded back as it should. I lowered the current load and it went back the usual operation. BUT when I tried to draw more current (10A), the voltage and current folded back again, but the controller was destroyed(it only outputs 4V after). What happened there? I'm puzzled since I don't know what can destroy it.

Please help. Thanks!

 

[crutschow]

Post your circuit and the layout.

That looks like ringing on the ground due to ground currents. Try connecting the oscilloscope ground clip to the circuit ground and then touch the oscilloscope probe to the ground. I suspect you will see similar ringing.

How is the circuit laid out? If you didn't have a good ground plane and a careful layout of the circuit with short connections for the high current paths, then you will typically get high ground noise.

What part failed in the circuit?

 

[wakoko79]

here are the traces.
the bottom has the ground plane (rightmost)
The output is on the center right
MOS source pin center left
The Vin is on top left.

The board is only made up from presentisized pcb, I'm waiting for the manufactured pcb, I don't want to destroy more controllers.

 

[crutschow]

Your layout post is not much help since it doesn't show all the components.

I don't understand the connections for the two diodes. Are they connected in parallel?

Why is the RC snubber not connected to ground?

Why such a large C snubber value? Typically that capacitor is much smaller. See the end of this for info on snubber design. Was yours designed with that criteria?

 

[wakoko79]

Your layout post is not much help since it doesn't show all the components.

I don't understand the connections for the two diodes. Are they connected in parallel?

Why is the RC snubber not connected to ground?

Why such a large C snubber value? Typically that capacitor is much smaller. See the end of this for info on snubber design. Was yours designed with that criteria?

The diode is STPS30170C, d2pak package. 2 diodes in parallel in one package. According to the datasheet, it has ~100pF capacitance per diode considering my Vin. I originally have 1nF to get the criterion but since the ringing is so big, I increased the cap to increase damping.
They are not connected to the ground since the snubber ckt is trying to damp the diode (actualy from lm5088 sample ckt in the datasheet).

Here is the complete layout, sorry about that. Red is top, blue bottom. The square thing is the inductor.

What trouble me most is why the controller getting destroyed. I'm just hoping that my handmade pcb fab is the reason so that there will be minimal changes in the design.

 

[crutschow]

I see no obvious problem with your layout.

Why do you have the current limit at 10A if you only need 5A?

Where are you measuring this ringing? Have you looked at the MOSFET output/inductor input?

I would revisit the snubber values you used, using the reference I posted as a guide.

It can be difficult to determine why something died, particularly under overload stress conditions. You might try slowly increasing the converter load while observing all the pins on the controller for any abnormal looking voltages.

 

[wakoko79]

I see no obvious problem with your layout.

Why do you have the current limit at 10A if you only need 5A?

Where are you measuring this ringing? Have you looked at the MOSFET output/inductor input?

I would revisit the snubber values you used, using the reference I posted as a guide.

It can be difficult to determine why something died, particularly under overload stress conditions. You might try slowly increasing the converter load while observing all the pins on the controller for any abnormal looking voltages.

I meant I stressed the ckt to 10A to test the current limit function.
It will be really difficult to determine what went wrong. The IC always dies.

Is there any reason for the ringing at the output? The ringing are observed in turn on and turn off of the FET. Is the snubber the only suspect for this behavior?
I looked at the waveforms at the gate. They are also ringing. BTW, I'm using logic level FET.

 

[crutschow]

I understand you stressed the circuit and you found that you exceeded the limit. You haven't answered my question as to why the current limit is set so high.

The ringing is due to the stray inductance and capacitance in the circuit. That's why I suggested you revisit the values for the snubber circuit.

Did you do the test I suggested, where you connect the oscilloscope ground clip to circuit ground and then observe the voltage on the ground with the probe?

If you don't follow my suggestions or answer my questions, then I am spinning my wheels here.

 

[wakoko79]

The current limit is 5.5A +10% (or so for margin, this is ABS MAX CURRENT DRAW), I don't know if that is high for you, but that is in the spec. I can't do anything about it. The 10A is a test parameter. I simply set the current draw of the e-load to 10A.

The snubber values are made (originally to R=4.7ohms, C=1nF) so that they follow that conditions (actually, I read that appnote you linked even before you posted it, among many other appnotes and datasheets). The values seen on the pictures I provided are actually updated to the last test I made (because of the reasons I previously stated). And yes, I tested the original values initially.

Did you do the test I suggested, where you connect the oscilloscope ground clip to circuit ground and then observe the voltage on the ground with the probe?

Sorry, I haven't done this yet since my board was destroyed and I haven't fabricated another (hopefully tomorrow I already have another one).
Does circuit ground mean power ground? Well, the circuit only have one ground. But what does "observe the voltage on the ground with the probe" mean? Where do I probe it?

The waveform above is measured at the output. I don't have any filters (like pi filters, since I don't understand them yet) at the output, so that waveform is measured at the inductor output WRT ckt ground.
I also remember measuring the MOSFET output (but I forgot to take a picture), There is also ringing there (switching from negative to Vin as it should + ringing).

I'd like to test more but my boards were delayed so much because of the Chinese new year. Just ranting.

 

[crutschow]

.............................
Does circuit ground mean power ground? Well, the circuit only have one ground. But what does "observe the voltage on the ground with the probe" mean? Where do I probe it?
....................

You probe the ground trace while the ground clip is connected to wherever you had it during the measurements. I think you are seeing ground loop noise and that will check its value.

 

[ronsimpson]

Part of the ringing is because of the ESR and ESL of the large output filter caps. Try adding a ceramic 10uf or 1uF cap.

Also I often add a small LC on the output of the power supply. The L can be very small. You are only trying to get rid of noise that is too high in frequency to be killed by large capacitors with too much internal L and R.

 

[ronv]

That chip drives the FET fast as well. You might try adding 5 to 10 ohms in series to the FET gate.