Boost converter output voltage spikes

[jayce3390]

Hi,

I have designed a boost converter using SMD components. I expected a flat output voltage response but some spikes occured (see attached picture) when the device is switching.

1. Do you know why?
2. Does it exist a solution to clamp?


Output : pink trace
Gate transistor waveform : green trace
Drain transistor waveform (diode anode) : blue trace

 

[FvM]

Seems like diode reverse recovery is mainly causing the output spikes. In my opinion, it doesn't look severe. I guess, there's no output filter except for the output capacitor. The rather large triangular output voltage ripple suggest to increase it's value or connect multiple items in parallel, the latter involves less series inductance and can already reduce output spikes. An additional LC filter would further minimize it.

 

[jayce3390]

Yes, indeed, I spent a lot of time to find diode with low reverse recovery time, but 5-10ns is the best you can get from manufacturers.
You are rght, there is no real post output filter but I put smaller capacitors in parallel to extend the frequency range; but spikes are still remaining.

You said an LC filter at the output, but how to choose values ? where do the 1/sqrt(L*C) cut off frequency should be chosen?

 

[pauloynski]

You are using a quite high frequency. Just some hints:
1- Are you sure the spikes are not being induced in the scope cables?
2- Changing the position of the scope cables makes any difference?
3- This can be also generated by a poor layout.

 

[dick_freebird]

The asymmetry of the output spike (H-L edge large, L-H edge
small) says to me that you can discount any of the plain
passives (incl coil winding capacitance, etc.) as a primary
cause.

The chatter about ground on the switch node I think is the
source and the key. It could be that the ground bounce
is responsible for the -appearance- of spiking simply by spoofing
the 'scope probe or the differential input of the 'scope.
Try varying the ground pickup point.

There is a lot of output "activity" and maybe a snubber
would help you out.

 

[jayce3390]

To pauloynski : Yes the frequency is high that's why I m using 1.5GHz bandwidth probe providing very low capacitance to avoid loading however with normal probe you don't see any spike due to small bandwidth of the probe.

about ground, I tied scope ground and boost converter ground. How can I chose the proper ground point?

 

[FvM]

Generally, acceptable level of output interferences is strongly application dependant. Your measurement doesn't necessarily
tell about the actual interfering signal at the supplied circuit. Conducted and radiated EMC is another problem, but it has to be
determined with respective measurement methods, a particular waveform can't tell.

A discussion about suitable ground points is meaningless without knowing the exact circuit layout in my opinion. The same with a
detail discussion of effectivity of filter capacitors.

 

[pauloynski]

Even the best probes can be fooled. I suggest your limiting the bandwidht to a reasonable 200MHz or less. Measurement of conventional converters is often limited to 20MHz. With respect to the best ground point I suggest your choosing the filter cap. ground point but you have to make some tests because it depends on the topology and layout. As posted above, this measurement doesn´t tell you the whole story. Sometimes spikes are carried out by the power lines, so some kind of isolation is necessary (usually a small 1:1 toroidal transformer is used to isolate the probes, for AC signals only). Also inspect your pcb. Try to increase the copper cross section by soldering a wire in parallel where the current density is high. And do a lot of tests to see if any makes a difference.

 

[Antonyb]

Hey Jayce,
If you have chossen a good recovery diode and a filter then this output would not be seen.
Even after doing so if you are getting such response then I would agree what paul has said. Use a lower bandwidth setting and use proper ground for the probe. keep the ground tight to the boost module ground. By doing so you might not get any switching spikes. Just check the layout of the same as it seems to be placement related issues.

 

[FvM]

If you're expecting several 100 MHz interference bandwidth (possible with fast diodes. e.g. schottky), why not measuring
with full oscilloscope bandwidth?

You have to use your FET proble with a short ground tip to get an almost valid measurement. You can measure e.g.
directly across a SMD filter cap and you should also determine the amount of ground bounce still effective at your
probe my measuring ground against ground. If you can assume, that you're measuring actual interferences, and
they are still too high directly at the filter cap, you obviously should reduce it's series inductance.

If you're already using small ceramic caps, the only practical way is to use a n-fold parallel circuit. Or accept the interference
at the output node and and add a low pass filter. But if your converter is injecting the interfering currents in the (common)ground
plane due to an unsuitable layout, no filter can help. A local ground plane for the converter is the best method to avoid this issue.
I think, it's mandatory for a 10 MHz converter.

 

[pauloynski]

Withou knowing if the measurement results can be trusted any other discussion has no sense.

 

[jayce3390]

Hi all

Fvm,
I can't use the full oscilloscope bandwidth due to 50 ohms loading, I need high impedance probing config. (I m using 1156A 100kΩ,0.8pF 1.5GHz BW active probes). I m already using small SMD components.

As you said, I measured a signal through the ground (ground to ground probing), short duration spikes with 1 Volts magnitude in the ground found when the device is switching!

thank you again for the help