effects of TVS capacitance on BLDC driver

[Simon]

I have been looking into issues with a motor driver that keeps shutting down due to a "short circuit". This can be when the motor is just not moving but the controller is interacting with it.

So there are some TVS diodes from each phase to controller negative supply. This was to stop issues where the motor stopping suddenly or slowing down would dump excess voltage into the controller and blow it.

So I took the diodes off. The problem went away. I probed the motor phase to see what was happening. It looked fine and dandy without the TVS, nice square waves (well as square as can be expected) but the motor was not loaded just driving some mechanism. So we went to a setup with the TVS diodes, immediately  ringing was obvious on the rising edge of the square waves.

Then I removed an EMC choke that was on the motor phase wires. The ringing got worse. the phase wires where just hanging about so I twisted them to reduce their inductance. No the ringing was so bad that clearly there were 2 peaks that triggered the TVS diodes at nearly twice the supply voltage and the motor controller started complaining of short circuits again.

Is this all because the TVS diodes go to ground and are not just across the phases where they might not involve controller circuitry in their ringing?

[T3sl4co1l]

That will still happen between phases, yes.

The equivalent circuit is: inverter -- filter and wiring inductance -- (TVS capacitance || motor inductance and resistance).

With the TVSs to GND, you have a "wye" equivalent network, which can be transformed to an ideal (floating) "delta" network plus a separate neutral (common mode) circuit.  As far as the ringing is concerned, this doesn't matter.  (The exact voltage drop on the TVSs will matter, for instance.)

By removing the filter inductor, and reducing wiring inductance, you've lowered the impedance of the LC network relative to the motor -- which, I'm assuming the motor is dominant loss in this case.  (Core loss should be a fairly high resistance, in terms of the parallel equivalent.)  This raises the Q and resonant frequency of the network, hence the higher peaks.

TVSs shouldn't be necessary, nor do much of anything, with a proper controller.  If the controller can't handle locked-rotor or startup conditions, or regenerative braking (or handles them poorly), I would rather suggest replacing it with an actually-good one.  (Mind that regeneration needs to go somewhere; if the supply doesn't have capacity by itself (like a battery), you'll need a braking resistor and probably some means to control it.  In that case it may be preferable to simply not permit regeneration at all, and use mechanical braking.)

Tim

[Simon]

I think the answer is an electric brake rather than the half assed solution proposed by the controller manufacturer. The controller runs from 48V but can work up to 60V. I don't know if this is enough headroom to put a capacitor in the supply rail so that when the motor regens it has somewhere to go. I don't know if 12V is enough headroom although the abs max of the controller should be more than 60V if it is rated for that.

I'm assuming that as the TVS is bypassing the negative switch it is causing some imbalance in a measurement the controller is making and triggering a very sensitive short circuit protection.

[floobydust]

You'd have to see how smart the BLDC controller is, the ones that monitor bridge return current will get confused if they see TVS current at that node. The current spike is from the TVS conducting, but it thinks it's from a mosfet staying on too long ("short-circuit"), if the TVS is in the wrong spot on the circuit.

For short term braking (inertia), just the capacitor on the DC bus is not good as it quickly gets charged over 60V and capow.
I have used a big clamp on the DC bus, either active or TVS depending on how much regen energy there can be, to stop from blowing mosfets.

[Benta]

I'm having major difficulties following the argumentation here.

We're talking about a BLDC motor, whether star/wye or delta is of no consequence.

These are generally well-behaved, and like brushed DC motors work by balancing supply voltage with back EMF.

1: the reason for the TVS's eludes me, unless they're protecting the controller when using long cables. They are otherwise unnecessary.
2: why would the motor generate a higher back EMF when coasting down (controller off)? Please explain how this could charge a line capacitor on the controller. In my experience, it just means that the motor will take longer to stop (it's unbraked). Nothing critical here.
3: fast edges from the controller may create ringing from the motor coils' inductance/interwinding capacitance resonance. That's why the line inductors are there, they slow the edges.

Now if you want regenerative braking it's a different story. That's significantly more complicated and will need active work from the controller's power
stage.

[Siwastaja]

As Benta said, remove the TVS's, they are not doing what you think they are doing. If you spin a motor with, say, a 60V supply, the motor itself can't generate more than the same 60V when spinning down, no matter how much inertia it has.

Regen braking may need something to dump energy into, but that would be on the DC link (input) side of the controller (and designed by the controller designers).

If there is a possibility that external forces rotate the motor to higher speed the maximum input voltage would be capable of, then you would need again something to dump the energy into before the maximum voltage ratings of DC link caps (and switch Vds_max) is exceeded, but again, this would be on the input side of the controller, not at the phase outputs.

[Simon]

It turns out the reason for the TVS's is static.... so they have been put in the wrong place, they need to go across the source of static not across the thing that needs protecting....

[T3sl4co1l]

Heh... inverter outputs should be quite immune to ESD.  More importantly it seems, that needs to be a properly grounded circuit, or in conduit (which will help with emissions besides)?

Tim

[Simon]

Heh... inverter outputs should be quite immune to ESD.  More importantly it seems, that needs to be a properly grounded circuit, or in conduit (which will help with emissions besides)?

Tim

Yea I was just starting to think that myself. The the 3 phase transistor output bridge is also a 3 phase rectifier, it will just pump the static into the power rails. The TVS diodes should go across the power supply not the motor.

I'm going to look at increasing the common mode choke the phases go through as there is resistance to taking the diodes off as the controller maker recommended them

[Siwastaja]

Yes, over the power supply is the correct place. Another possibility is to increase the DC link capacitance to absorb whatever it needs to absorb. What comes to ESD, you surely have enough capacitance there already, no?

Be aware that an external mechanical power source rotating the motor so that the BEMF voltage exceeds the TVS voltage causes prolonged dissipation which is really hard to deal with. But often this is a non-issue, just rate all components to high enough voltages so that this would only happen at some ridiculous RPM, then document that such external extreme speed rotation is not allowed.

[Simon]

The motor won't be forced round but it can have some inertia behind it when it stops. There is an electric brake system in place for this?

[Benta]

Just braking a BLDC motor is simple. There are two main options:
1: brutal: this is done by simply shorting each of the three motor coils. Depending on mechanical load, the motor will come to an almost immediate stop. Lots of mechanical and electrical stress.
2: softer: instead of switching in shorts, you switch in brake resistors. Depending on resistor choice, stopping will be faster or shorter. This option can be implemented using the existing power stage, resistor braking being simulated with PWM. The algorithm is complicated, though, and will probably need a DSP.

Regenerative braking is more complicated, and only pays off with high-inertia loads (think electric car/truck or electric train or a really big flywheel).

[Siwastaja]

The motor won't be forced round but it can have some inertia behind it when it stops. There is an electric brake system in place for this?

I'm assuming you did not originally use "field weakening" to drive the motor.

Inertia means it's slowing down, and generating less and less back-EMF voltage. If you drove it as fast as you can (without field weakening), then the back-EMF voltage, at most, equals the supply voltage and ramps down as it slows down. No problem.

If coasting down is not an option but you need electric braking, then you indeed have two options (well, one really...)

* Full blown short circuit. Not usually possible due to exceeding safe currents, blowing up the MOSFETs and at least causing mechanical damage. This dissipates energy in all parasitic resistances; windings, wires, Rds_on... But sure, if you really short the motor, then it won't generate any high voltage...

* Regeneration; for example, in FOC terminology, just change the sign of torque-generating current (Iq), the correct commutation will follow. But this makes the motor inductance and the bridge act as a boost converter, motor BEMF being input, DC link being output, supplying the DC link to almost arbitrarily high voltages from even quite low RPMs. So you need a safety voltage comparator for the DC link, which turns off regeneration, i.e., defaults back to coasting by not driving the FETs. Or, use resistor banks that are turned on before DC link max safe voltage is exceeded, regulating it at safe max voltage. A simple TVS-only solution would be expensive because they are not rated for very high continuous power, so you would need a lot of them. The "T" in the name has a meaning, after all!

[Simon]

The TVS diodes are there for static which is silly. But as the controller manufacturer/distributor suggested it, it stays. I guess if a resistor was put in series with each TVS diode it would kill the Q factor of the current resonant circuit and stop the oscillations that are triggering the TVS diodes.

[wraper]

So there are some TVS diodes from each phase to controller negative supply. This was to stop issues where the motor stopping suddenly or slowing down would dump excess voltage into the controller and blow it.

AFAIK normally there are protective and/or parasitic diodes in MOSFETs within the driver which clamp the output to power rails. And the power rail voltage is clamped if it's possible that it may exceed a safe threshold when motor is acting as a generator.

[Simon]

So there are some TVS diodes from each phase to controller negative supply. This was to stop issues where the motor stopping suddenly or slowing down would dump excess voltage into the controller and blow it.

AFAIK normally there are protective and/or parasitic diodes in MOSFETs within the driver which clamp the output to power rails. And the power rail voltage is clamped if it's possible that it may exceed a safe threshold when motor is acting as a generator.

Yes and apparently there is a regen break too, so the diodes are not for that but static which should be conducted by the bridge back into the supply anyway which is where the TVS diodes should be.

[Siwastaja]

You can't build a proper motor driver that works properly and passes EMC without quite significant amount of capacitance (typically tens to thousands of uF) on the DC bus, part of which (tens of uF) is fairly low ESR and low ESL. This easily absorbs any ESD scenario without any extra clamp devices.

The MOSFET body diodes do acceptable job of diverting current into said DC bus capacitors. MOSFET parasitic drain-source capacitance also helps here, also the parts are likely avalanche rated (for much larger energies than ESD ever is). Power MOSFETs simply are not susceptible to ESD damage, for large devices even the gate is quite immune, and here the gate is not exposed anyway.

Given the avalanche rating, MOSFET itself is equivalent to a TVS. Compare the datasheet-specified avalanche pulse energy to the ESD pulse energy of the relevant ESD standard.

I have never heard ESD being a concern before.

[Simon]

i don't know much about ESD. No idea how to measure it or quantify it.

[Simon]

Well in the end I spoke to the motor controller manufacturer who denied that they would recommend doing such a thing and this was the advice of an ex employee. So we are taking them off. Turns out the motor rotor was already grounded anyway.