Ringing in H-Bridge inverter

[Jagajack]

Hi all,

I am designing a H-Bridge inverter for induction furnace application. My DC bus voltage is 400 V and current is 15 A.

To start with I used 12 V supply to test my H-Bridge inverter connected to load resistor of 10 Ohm.

I used IXFH26N50 from IXYS semiconductor for H-Bridge and the MOSFETS are driven by two IR2110 drivers.

But during turn-on of the MOSFETS, I could see excessive ringing at both the output terminals of the H-Bridge with respect to ground and also on the MOSFET gates.

I have used shielded twisted wires to connect driver terminals to MOSFET to reduce the wire inductances and 1000µF capacitor on the DC bus, but no appreciable reduction in ringing noise amplitude.

In many papers and application notes it is mentioned that, this ringing is due to gate to drain capacitance and can be suppressed with the use of ferrite beads near gate terminal.

But I have not tried with ferrite beads, will ferrite beads suppress this ringing?

Will changing the MOSFET be a solution? If so help me to choose a MOSFET that could have lesser switching noise and suitable for induction heating application.

Please help.

 

[barry]

Changing the MOSFET won't help. The ferrite might. Board layout also can contribute to the ringing.

 

[Jagajack]

Thanks barry.
I will try with ferrite beads.
Do Input Capacitance, Output Capacitance and Reverse Transfer Capacitance of MOSFET given in datasheet inluence ringing during switching?

 

[barry]

Thanks barry.
I will try with ferrite beads.
Do Input Capacitance, Output Capacitance and Reverse Transfer Capacitance of MOSFET given in datasheet inluence ringing during switching?

Absolutely. If everything were just resistive (no inductance or capacitance) you would never see ringing.

 

[Jagajack]

Then MOSFET with lesser Input, Output and Reverse Transfer Capacitance have lesser ringing during switching?

Will shortening the gate lead reduce ringing?

 

[barry]

Yes. You already asked that in post #3!

How long is your gate lead? A couple of mm is not going to make a difference. A resistor in the gate path might help by slowing down the rise time.

 

[dick_freebird]

You must consider that simply adding probe tip and ground
capacitances can impose enough C to greatly change the
resonant behavior.

Before you get super excited (heh) about ringing, try putting
a 1Kohm series resistor on your probe tip. And try putting the
probe and clip both to the same point on your reference plane
and see how mych ringing there is, from "nothing".

 

[schmitt trigger]

I congratulate you for testing a power circuit at reduced voltage/current levels. It allows one to better troubleshoot the circuit, without damaging an excessive amount of components and hurting oneself.

Series gate resistors is an excellent approach to reduce ringing, but it also has the drawback of of increasing the rise/fall times. Depending on your switching frequency, the increased time may cause an excessive power dissipation on the mosfet.
Like everything else in design engineering, it is a tradeoff. Start with 10 ohms. Record and compare the waveforms with no resistor and with the resistor. Adjust the resistor value.

And yes, also use ferrite beads. Fair-Rite has a large selection of beads specifically designed for suppression.

But before you do...I assume that you have a good scope with properly compensated probes!! Otherwise the scope will be lying to you.

 

[ftsolutions]

+1 with all of the above. Your rise/fall times and switching frequency play a major role, and circuit layout becomes critical when switching high pulses of current very fast. In similar cases, I've found that it can be quite effective - especially if you have several inches of traces between the gate driver and the gates themselves - to place a resistor (<47 ohm) near the output of the gate driver, and have the SMT chip inductor/ferrite bead right next to the gate lead on the MOSFET. You will likely have to try some different R values and some different ferrite impedance values to "tune" it if you really must have fastest possible switching times and to dampen the ringing. The impedance of the traces on the PCB can play a part too.

To simplify things, I'd also recommend that you make the layout with all the gate connections as similar as possible in length/routing, etc. This way when you find a good combination for one it will most likely work well for all, rather than working out different ones for each.

 

[Jagajack]

You must consider that simply adding probe tip and ground
capacitances can impose enough C to greatly change the
resonant behavior.

Before you get super excited (heh) about ringing, try putting
a 1Kohm series resistor on your probe tip. And try putting the
probe and clip both to the same point on your reference plane
and see how mych ringing there is, from "nothing".

Thanks.
I tried with 1.5 kΩ in series with both the leads of probe, the waveforms were almost smooth. Is this the correct way of probing?

- - - Updated - - -

I congratulate you for testing a power circuit at reduced voltage/current levels. It allows one to better troubleshoot the circuit, without damaging an excessive amount of components and hurting oneself.

Series gate resistors is an excellent approach to reduce ringing, but it also has the drawback of of increasing the rise/fall times. Depending on your switching frequency, the increased time may cause an excessive power dissipation on the mosfet.
Like everything else in design engineering, it is a tradeoff. Start with 10 ohms. Record and compare the waveforms with no resistor and with the resistor. Adjust the resistor value.

And yes, also use ferrite beads. Fair-Rite has a large selection of beads specifically designed for suppression.

But before you do...I assume that you have a good scope with properly compensated probes!! Otherwise the scope will be lying to you.

Thanks.
I will try with ferrite beads from Fair-Rite. Just I tried 1.5 kOhm in series with probes as suggested in post #7 and waveforms were almost smooth. Is there any way for proper compensation of probes?

 

[FvM]

Is this the correct way of probing?

What's correct? The resistor creates a 10 MHz low-pass with probe capacitance, depending on the measurement problem it might be wanted or unwanted.

 

[Jagajack]

What's correct? The resistor creates a 10 MHz low-pass with probe capacitance, depending on the measurement problem it might be wanted or unwanted.

Thanks FvM.
How could I measure correctly with probes?

 

[schmitt trigger]

All of the oscilloscopes have a calibration output to which you attach the probe you want to compensate.

All of the 10X probes have a small adjustment, which you can access with a small screwdriver.

Attach the probe to that output, and adjust until you have a perfect squarewave, with no under or overshoot. Do this with the probe by itself, remove the resistor.

 

[dick_freebird]

Thanks FvM.
How could I measure correctly with probes?

I would try again with a series of lesser series resistor
values, looking at the ringing frequency vs the probe
RC figured frequency. Then, I would keep the probe
clipped on (with some reasonable series R) and observe
the effect of attaching either a duplicate un-limited
probe, or its equivalent network from node to ground
and see what the effect of the bare probe is / was.

From this you can deduce whether you are "fixing"
the node RL(C+Cprobe) resonance or the probe's
high frequency response.

1.5K*7pF=10500pS or ~ 100MHz corner, if you are
using a 7pF probe. But there are higher-C ones out
there.

You can also make very low C, 50-ohm Zout networks
and use the 50-ohm input of the 'scope if you can
tolerate more DC current on the node - but that is
also an element of unrealism that you'd have to
determine the effects of.