2
\$\begingroup\$

For my application, I am using 4S10P Li-ion battery (with NXP's RDDRONE-BMS772) to power the following load:

  • 8 ESCs to control thrusters
  • 1 Raspberry Pi with hat to control ROV (powered from 5V buck converter).
  • 1 high power LED

For my system, I used 4 n-channel MOSFETs TPW1R005PL controlled by LM9061M gate driver as a high-side switch that is capable of running more than 120A continuously, the switch is only in On or Off state and has no switching characteristic. The schematics is as follows:

MOSFET Switch circuit

In which, D2 is TVS diode SMBJ18A clamp at 18V, D1 (at output side) is flyback diode to prevent back emf from the ECSs.

I am able to run the load at 120-150A without any problem. However, I have encountered transient voltage upon turn on the circuit in quick succession (turn on and off and on quickly) and ended up fried multiple BEC. Transient voltage upon turns on At input side: Transient voltage at input side upon turns on

I suspect that it is caused by switching effect and inductive load on the circuit, I have tried adding extra TVS diode parallel to the flyback diode, but it cannot suppress the output voltage. What is the recommendation for this case? should I put an RCD snubber across Drain and Source of the MOSFETs to suppress this transient voltage?

Edit 1: I have tested my system with loads connected (thrusters connected to ESCs, and UBEC to power dummy resistor load) and without load connected to the output and received transient voltage for both scenarios.

Edit 2: Input side's signal attached below output signal

\$\endgroup\$
3
  • \$\begingroup\$ What is the BEC rated for? The transient shown doesn't look very threatening to me, but I have no idea if the level shown is typical of conditions when a failure occurs, or if your equipment is rated adequately to handle it. \$\endgroup\$ Sep 11, 2023 at 6:57
  • \$\begingroup\$ Hi Tim, the BEC is from 7-26V. The spike goes unpredictably, the highest I measured was 28.4V so it probably killed the BEC. But it is interesting for me to know where this transient voltage is coming from and how to effectively reduce this spike. Replacing the BEC to a higher rating one is a good idea but if this problem still lurking around, it may cause some other nuisance one day. \$\endgroup\$
    – Quang
    Sep 11, 2023 at 7:05
  • \$\begingroup\$ Some current flows from load (or whatever) back to source because there is almost 2V at the output before turning on (according the scope). \$\endgroup\$ Sep 11, 2023 at 7:22

2 Answers 2

2
\$\begingroup\$

From SMBJ18A datasheet, it has the following characteristics:

  • Reverse standoff 18V: it is guaranteed to NOT conduct at 18V

  • Breakdown 20-22V at 1mA

  • Max voltage 29V at 20A

So it looks like it's doing its job, clamping around 24V.

From your scope shot the spike lasts 20-50µs which indicates a resonance frequency in the 20-50 kHz.

I see two culprits:

  • Coupling through the FET gates. If the driver sends a lot of current into the gate this will couple through the FET's internal capacitance into source and drain, causing a spike. But the height of this spike is proportional to the ratio between MOSFET capacitance and decoupling capacitance that absorbs if, and you have enough capacitance on your board to make this a non-issue. So, probably not.

  • Ferrite bead (or other inductance) resonance with decoupling caps.

This is a common killer when you plug in the power connector: it excites a LC circuit with a step, causes a resonance, and fries something. The problem here is that I don't see any inductor or ferrite bead, but there' are probably some hidden somewhere, in your DC-DC modules for example.

Note you don't need a load on the output for this to occur: turning on the FETs will cause a current spike due to charging the output caps on your board, so it will cause a voltage down-spike on the input side, which may also excite any LC circuits present on the input side. You should post a scope shot of the input side too.

Since the spike occurs even with no load present, you can reproduce it without frying any more BECs, so you should be able to track it down.

A solution is to add damping on the input side, for example an electrolytic capacitor with enough ESR to dampen the resonance.

\$\endgroup\$
6
  • \$\begingroup\$ Note wiring is in the ballpark of 0.5uH/m, so it doesn't take much to resonate with the capacitors shown. \$\endgroup\$ Sep 11, 2023 at 15:09
  • \$\begingroup\$ I have tested the board again just now and as you suggested, the input side has a huge down spike. that seems to be matching with the spike at the output of the switch. I will upload this measurement on the post and try to go for your reccomendations. \$\endgroup\$
    – Quang
    Sep 11, 2023 at 23:36
  • \$\begingroup\$ @Quang During turning on you are switching 4uF of capacity (C7..C10) directly what draws a huge current from input. At such a huge current even small inductance of wires and caps is enough to to store energy to make a spike. E=0.5LI^2 so the energy rises current squared. To avoid slow down the turning on process adding a resistances to gates like 220ohms (maybe more) to hold mosfets in linear region a bit. \$\endgroup\$ Sep 12, 2023 at 2:48
  • \$\begingroup\$ Once the FETs are on we have "6µF" on the board, assuming X7R caps at 20V that should be 2-3µF. From the scope shot it resonates around 15kHz. which corresponds to a LC circuit with 3µF+37µH. The low resonance frequency means inductance is too high for just wiring (unless it is coiled around something ferromagnetic) so there's probably an inductor/ferrite somewhere at the power input of one of the boards. This needs 100µF electrolytic with 2-3 ohms ESR on the input for damping, so either a general purpose 100µF aluminium cap which should have that ESR, or a low ESR cap with resistor in series. \$\endgroup\$
    – bobflux
    Sep 12, 2023 at 7:09
  • 1
    \$\begingroup\$ Great news! However, ESR of electrolytic caps strongly depends on temperature, so you may find it provides damping at room temperature, but not in very cold or hot temperatures. If you want a wider temperature range, it is better to use a low-ESR cap with a resistor in series. This cap's ESR also depends on temperature, but if it is small compared to the resistor, its variations won't matter as the resistor will remain constant. \$\endgroup\$
    – bobflux
    Sep 13, 2023 at 10:21
1
\$\begingroup\$

It would be useful to see the layout and length of the connecting wires, as well as the details of the load. A you say, I would guess this is an inductive issue, especially give that you say the worst problem is a quick on-off-on.

Try reducing the loop area of the wires, twisting feed and return connections. You could also consider increasing the output capacitance.

Also worth checking is you get the spike if you use purely resistive dummy loads. This might help track down the exact cause.

\$\endgroup\$

Your Answer

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

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