0
\$\begingroup\$

I am switching a 1 Ohm load with an N-channel MOSFET, which is driven by a TC4420 driver. The load is three Arcol HS50 330 mOhm resistors in series, the power supply is a 3S LiPo battery. The switching frequency is 1kHz with 10% duty.

I am trying to understand what is happening in the circuit when the FET is switching off. I started with this:

1-basic_circuit

This is the waveform in this case. Channel 2 (blue) is just the signal generator output, Channel 1 (yellow) is Vds, channel 4 (green) is the battery voltage, measured on the main battery wires (if I measure it through the balancer connector, the shape of the waveform is still the same, with smaller amplitude):

1-basic_scope

First I thought the huge peak around 55V is the avalanche breakdown of the FET, due to the load and the wiring having some inductance, followed by some ringing for the same reason plus the FET output capacitance. But then why is there that big rise on the battery voltage? This implies to me that current is flowing into the battery, which I really do not understand.

Then I tried adding a flyback diode:

2-with-diode_circuit

Which resulted in a bigger battery voltage rise, but not much otherwise:

2-with-diode_scope

To eliminate the ringing, I added a snubber circuit (without the previous flywheel diode):

3-with-snubber-no-diode_circuit

This was quite effective regarding the ringing:

3-with-snubber-no-diode_scope

So I tried with both the snubber and the diode:

4-with-snubber-and-diode_circuit

With the same additional results as before (bigger battery voltage rise, and nothing else):

4-with-snubber-and-diode_scope


So my question is basically what is happening here? More specifically:

  1. What causes the voltage rise on the battery, when I do not expect the current flowing back into the battery?
  2. Why does the avalance happen regardless the presence of the flywheel diode? (Maybe this is some other phenomenon, not avalanche?)
\$\endgroup\$
7
  • \$\begingroup\$ Bad oscilloscope probing methods. \$\endgroup\$
    – Andy aka
    Apr 21 '20 at 9:49
  • \$\begingroup\$ How can it cause a 55V peak if it is not there, or the seemingly risen battery voltage? Can I verify it somehow? \$\endgroup\$
    – vjaz
    Apr 21 '20 at 9:57
  • 2
    \$\begingroup\$ Or too much loop inductance in your battery supply or that plus a combination of bad-probing and multiple earths. Show a photo of your set-up and probing method. \$\endgroup\$
    – Andy aka
    Apr 21 '20 at 10:04
  • 2
    \$\begingroup\$ Welcome to EE.SE! Please show layout and probe setup. \$\endgroup\$
    – winny
    Apr 21 '20 at 10:13
  • \$\begingroup\$ I could not add more images to the question, so here the are: imgur.com/a/5ZNiiRu I know this is not the best setup, if the issue would just be some oscillation, I wouldn't ask anything. \$\endgroup\$
    – vjaz
    Apr 21 '20 at 10:36
2
\$\begingroup\$

You have a massive loop inductance in your measurement set-up: -

enter image description here

I've tried to indicate that with a green dotted line showing the loop. I estimate that loop inductance to be about 1 uH. If I simulate a 1 ohm load in series with 1 uH I get this: -

enter image description here

  • Vin (red) is the gate drive voltage (fed via a 10 ohm resistor but could be made much worse by having long leads as indicative of the set-up picture).
  • Vout (blue) is the drain voltage
  • i(R1) is the current through the 1 ohm load

I'm using an IRFZ44N MOSFET by the way - the same as shown in the OP's circuit diagrams.

But then why is there that big rise on the battery voltage?

Because of bad probing techniques and an ill-defined 0 volts node (also with series inductances). Just look at the grey left-most probe tip connected to the red lead from the battery - that probe earth point is back close to the source of the MOSFET but, there's a hundred mm (or more) ground wiring back to the battery negative terminal so, it's impossible to conclude that there is any movement on the battery voltage at all. Everything need to be ten times closer to make any sense of o-scope readings and even then, there'll be ringing and inductive artefacts present (but smaller).

And, I haven't modeled the inductance of those power resistors - you can get low inductance versions of those resistors but you appear to be using the standard parts and the data sheet doesn't give any indication of the inductance for those parts unfortunately.

\$\endgroup\$
6
  • \$\begingroup\$ Thank you! I get the battery-related part now. As for the loop, I tried moving the black wire on the bottom of my picture around, and when I approached the resistors (so the "other side" of the loop), there was a small but noticeable decrease in the duration of that 55V pulse, so I guess that proves this loop is the reason indeed. Has my attempt with the flyback diode falied also because of the long wiring? Was I right in thinking the 55V pulse region is because avalanche breakdown is happening? \$\endgroup\$
    – vjaz
    Apr 21 '20 at 12:39
  • \$\begingroup\$ Yes, the 55 volt limit is probably the MOSFET breaking down. My sim didn't show that. Did you use an IRFZ44 BTW? If you can twist forward and return wires together you will get a significant improvement but there is also the loop formed by the scope tip and its earth wire. Current flow in the circuit can induce error voltages into it that measurement loop. For instance, try probing at the earth croc-clip point and see how much induced voltage you get. \$\endgroup\$
    – Andy aka
    Apr 21 '20 at 12:48
  • \$\begingroup\$ Yes, I use an IRFZ44N. And yes, if I probe the clip, the same shape appears as the Vgs waveform, with about 1V peak... \$\endgroup\$
    – vjaz
    Apr 21 '20 at 12:58
  • \$\begingroup\$ OK, a couple of lessons learned! \$\endgroup\$
    – Andy aka
    Apr 21 '20 at 13:01
  • \$\begingroup\$ Absolutely! Could you please answer my other question from my previous comment, too? (Why did the flyback not do anything)? \$\endgroup\$
    – vjaz
    Apr 21 '20 at 13:06
1
\$\begingroup\$

Do these high dI/dT experiments ONLY over a ground plane.

The stored energy in the wiring:

(1) prevents diagnosis

(2) allows oscillation and self destruction (seen that)

(3) may cause huge delays in switching (seem up to 1us delay, because of 1 meter lead in gnd wiring)

(4) gets expensive for replacing FETS and GATEDRIVERS (Had a gatedriver explode as the silicon vaporized during its task of absorbing the many joules stored in a large-capacitor lab power supply; that die had 3mm by 4mm of output FETs on the silicon, and either the Nchannel or the Pchannel entered bipolar snapback AND now_its_gone, to a several-cubic-mm plasma ball)

but

------does generate a fine learning experience about INDUCTANCE.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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