Spikes in output voltage in Buck Converter

[sabu31]

Hi All,

I am testing a buck converter rated at 120V input and 48V output with 750W output power. The switching frequency is 100kHz. I am observing that the output voltage is having spikes during the switching . The LC filter of buck converter is 100uH and 100uF. The waveform is yellow one shown in figure . Its captured in AC Coupling.

How can I eliminate spikes. Do I need to add another LC filter apart from the existing one. Thanks.

 

[KlausST]

Hi,

The LC filter of buck converter is 100uH and 100uF.

"100uF" does sound like an electrolytic capacitor. These are not suitable for fast spikes. They are for low frequencies only.

--> use a ceramics capacitor or a foil capacitor (value 100nF ... 1uF) in parallel to the electrolytics capacitor. Connect the scope directly at the ceramics/foil capacitor.


Klaus

 

[Easy peasy]

your 100uF should be very low ESR/ESL - but, putting that to one side - it is just as likely your scope is picking up common mode impulses - try shorting the probe and connecting to pos out - if the spikes are still there - you have CM pickup ...!

 

[asdf44]

"it is just as likely your scope is picking up common mode impulses"

Yes what Easy said. I just started typing exactly these words.

 

[dick_freebird]

Additionally the probe "loop area" (between tip+shaft,
and ground clip) is an inductive pickup for all the
inductor leakage field, and your choice of ground-
point embeds more or less I*R, L*dI/dt voltage
noise from circulating ground "plane" currents to what
the 'scope channel sees.

Output filter wants low ESL/ESR, but the inductor also
contributes spiking through its interwinding capacitance.
If you think the switching noise is unexpected, try adding
a shunt C to your inductor model that corresponds to the
advertised SRF. This is what your output filter must knock
down. If you have 100V input and you want 100mV spike
amplitude, that's 1000:1 attenuation and if the winding
has 10pF shunt C, you'd need >= 10nF of really good HF cap
returned as tightly as possible to prime ground (along with
the fat baseband cap to soak up the 100kHz designed
ripple current).