18
\$\begingroup\$

I have designed a PCB (intended as a prototyping building block) that has a IR2113 high and low side gate driver driving two IRF3205 (55V, 8mΩ, 110A) power MOSFETs in half-bridge configuration:

Schematic PCB layout Picture of the physical setup

Upon testing the circuit with a load I found out that while the low side switches quite cleanly there is a lot of ringing at the output of the half-bridge (X1-2) every time the high side switches on. Playing around with the input waveform dead time setting and even removing the load (an inductor with a power resistor in series simulating a synchronous buck converter connected from X1-2 to X1-3) did not reduce this ringing. The measurements below were taken with no load connected (nothing at X1-2 except for the oscilloscope probe).

Ringing

Apparently the parasitic inductances and capacitances are enough to cause that, but I cannot figure out why the low side works as well as it does. To me both gate drive waveforms look clean enough, with the voltages transitioning the treshold voltage of the MOSFETs reasonably fast. No shoot trough is present when switching. What are the possible causes of the problem, and what measures can I take to reduce the symptoms?

I am aware that there are many very similar questions here and on other sites, but I found the posted answers unhelpful for my particular problem.

Edit

While there was a 2200uF electrolytic capacitor at the input (X1-1 to X1-3) to supress transients and noise, it clearly failed to supress any high frequencies. Adding a 100nF capacitor (as suggested in the answer by Andy aka) in parallel with the electrolytic one reduced the ringing at the output (X1-2 to ground) by half and ringing at the supply (X1-1 to ground) by a factor of 10.

capacitors

\$\endgroup\$
1
  • 4
    \$\begingroup\$ This is an excellent first post \$\endgroup\$ Jul 1, 2014 at 17:00

4 Answers 4

12
\$\begingroup\$

Try probing on the power supply rail. I bet you see those spikes on there. It'll be due to the lead length between your bench supply and the MOSFETs. Clearly you won't see it on the lower FET side because your scope is referenced to that rail but, if you probed back at the power supply I bet you would.

Try a 1uF or 10uF ceramic across the power rails close the MOSFETs.

\$\endgroup\$
8
  • 6
    \$\begingroup\$ +1 There are harder ways to learn this, ones that might lead to little piles of smoked MOSFETs. \$\endgroup\$ Jul 1, 2014 at 17:02
  • \$\begingroup\$ A 100nF metallized polyester cap reduced the spikes dramatically, but not competely. Are ceramic capacitors better suited as bypass caps in applications like this? I sadly have no high value ceramics in my parts box. \$\endgroup\$
    – jms
    Jul 1, 2014 at 17:16
  • 1
    \$\begingroup\$ What you might be seeing now are possibly o-scope artefacts. Try connecting the scope directly across the cap with as short an earth lead loop as possible. Induction into the loop is commonly found. You should be OK with that cap. How big are the spikes now? \$\endgroup\$
    – Andy aka
    Jul 1, 2014 at 17:59
  • \$\begingroup\$ At the frequencies you're seeing, yes, a ceramic will be better than polyester. \$\endgroup\$ Jul 1, 2014 at 18:34
  • \$\begingroup\$ @Andyaka With the probe connected directly to the 100nF cap it does not matter anymore which FET switches, the ringing at the output (X1-2) is the same, and the ripple at the supply (X1-1 to X1-3) is reduced to two volts. Any suggestions on how to further attenuate the 20MHz spikes at the output? Is the board layout to blame? \$\endgroup\$
    – jms
    Jul 1, 2014 at 19:10
3
\$\begingroup\$

Assuming that you have dealt to the supply rail bypassing like Andy said and you have slowed up the gate by increasing R1 R7 and doing something to make turn off faster than turn on . If it still rings then there are still two things to try;You can place shottky diodes 60V across DS of the fets and you can place RC snubbers across DS of each FET .

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Both of these suggestions worked really well for me. I am developing a brushless DC motor driver rated at 14 Vdc, 80A using a Texas DRV8305 driver. Here is a useful paper on snubbers: ti.com/lit/an/slpa010/slpa010.pdf Using that design technique for the snubbers, and placing a schottky rectifier across the bottom transistor resulted in a reduction of the first peak of the ringing from 28 to 16V. The snubber reduced the ringing decay time to half amplitude from 300 ns to 125 ns. The transistors are 2 x PSMN8R7-80PS in parallel. \$\endgroup\$
    – Ray Wales
    Jul 19, 2017 at 2:51
1
\$\begingroup\$

I think Andy aka got the answer on this one, but I wanted to clarify that the ringing is caused by the inductance of the wires leading to the FETS and the gate capacitance of the FETs. This creates an LC circuit which resonates at a frequency based on the inductance and capacitance in your circuit. Usually the effect is reduced by using damping resistors and by reducing the lead length as much as possible.

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

Reduce the high side resistor to 22E this will most likely fix the Problem,this is often Caused from Switching the Mosfets To HARD , i had to learn the hard way

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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