Can TVS protect a power MOSFET from damage caused by overvoltage?

I have two 3-phase inverters. One is from Semikron (with IGBTs, rated 600V, denoted as Semikron-Inverter in this post). Another one is a home-made prototype.


The schematic of the home-made prototype inverter is shown in Figure 1, while its physical layout is shown in Figure 2. The specifications of the power switches used in the home-made prototype inverter (HMP-Inverter) are as the following:

IRFP4310ZPbF (power MOSFET)
VDSS: 100V
RDSon: 6mΩ
ID: 130A
VGS: ±20V
http://www.irf.com/product-info/datasheets/data/irfp4310zpbf.pdf

Each HMP-Inverter leg is driven by a IR2110 (manufactured by International Rectifier).
http://www.irf.com/product-info/datasheets/data/ir2110.pdf


More than 10pcs of IRFP4310ZPbF were killed so far and all of them were from leg-C (see Figures 1 and 2). I wish to find out the reasons that kill the IRFP4310ZPbF and why the damage was always at leg-C. When checked with multimeter (using the 'continuity' feature), the G, D, S pins of damaged IRFP4310ZPbF were found shorted. I do not know whether the failures were caused by overvoltage or overcurrent.


Then, the Semikron-Inverter was used to drive the same motor (with the same DSP controller and control algorithm). Please note that the DC-link voltage of the inverter is supplied by a Xantrex power supply (100V/50A), which has a digital display showing the supplied voltage and current. Sometimes the voltage shown on the Xantrex power supply (PSU) display fluctuated (decreased and increased). Occasionally, the fluctuation triggered the overvoltage protection (OVP) of the Xantrex PSU (the OVP of the PSU is set at 115V). It looks to me that the back-emf of the motor, due to some unknown reasons, was "added" to the DC-link voltage and when the effective DC-link voltage was higher than 115V the OVP of the PSU was triggered.


So, my guess is the damaged IRFP4310ZPbFs in HMP-Inverter were probably caused by overvoltage. When HMP-Inverter was used with the Xantrex PSU, the OVP of the PSU was never triggered. Since IRFP4310ZPbF is rated at 100V, the overvoltage (slightly above 100V) would have killed the power MOSFET first before it reached the 115V OVP limit of the Xantrex PSU.


My another guess on why the damages were always at leg-C is that, as shown in Figure 2, the power MOSFETs of leg-C were placed closer to the DC-link. Do you think this reason make sense?


By the way, if a TVS or transorb (rated below 100V) is added to the inverter as shown in Figure 3, can the power MOSFETs be protected from overvoltage?


Thanks.


Figure 1: Inverter schematic.



Figure 2: Physical layout of the inverter (viewed from the top). Gate drive circuit is not shown.



Figure 3: Inverter with a TVS or transorb.




Pim: Generally speaking, power MOSFETs can withstand over-current better than one can believe, but over-voltage is an instant killer...
https://www.edaboard.com/threads/224099/#post955299

spiba: I fully agree with you that 90 % of the mosfets are damaged due to Overvoltage rather than overcurrent...
https://www.edaboard.com/threads/224099/#post955617

richie: MOSFETs have very little tolerance to overvoltage. Damage to devices may result even if the voltage rating is exceeded for as little as a few nanoseconds. MOSFET devices should be rated conservatively for the anticipated voltage levels, and careful attention should be paid to suppressing any voltage spikes or ringing.
http://www.richieburnett.co.uk/mosfail.html

FvM: A serious problem arises however from the fact, that the bulk diodes are forward biased, sourcing high reverse current to the battery...
https://www.edaboard.com/threads/224099/#post956575

FvM: You should analyze more thoroughly what happens in the circuit. Did you undertstand the bulk diode conduction problem? It can't be handled by a inverter shut-down, you have to disconnect the battery, respectively allow the transformer center tap to rise to about 40V.
https://www.edaboard.com/threads/224099/#post956885

If the motor is actually sourcing enough power to raise the bus voltage to 115 V, then it would be surely sufficient to blow the FETs. Could happen if you switch the control circuit from full speed forward to recuperation brake without connecting a braking resistor. Without recuperation, I would rather expect Vgs overvoltage to damage the FETs, e.g. due to a bad gate driver circuit layout.

Another question is if you provide sufficient supply voltage margin. I would expect 60, maximum 80 V regular bus voltage with 100 V FETs. Raising the bus voltage from 80 to 100 V implies an energy of 36 Ws stored into the 20000 uF capacitors. That's quite a lot and rarely handled by a small TVS diode.

FvM said:

If the motor is actually sourcing enough power to raise the bus voltage to 115 V, then it would be surely sufficient to blow the FETs. Could happen if you switch the control circuit from full speed forward to recuperation brake without connecting a braking resistor. Without recuperation, I would rather expect Vgs overvoltage to damage the FETs, e.g. due to a bad gate driver circuit layout.

Click to expand...

There is no recuperation (regenerative) braking in the inverter. Would you please explain why without recuperation braking, the expected cause that kills the FETs is Vgs overvoltage and not drain-to-source (Vdss) overvoltage?

FvM said:

Another question is if you provide sufficient supply voltage margin. I would expect 60, maximum 80 V regular bus voltage with 100 V FETs. Raising the bus voltage from 80 to 100 V implies an energy of 36 Ws stored into the 20000 uF capacitors. That's quite a lot and rarely handled by a small TVS diode.

Click to expand...

In my application, the maximum DC-link voltage supplied to the HMP-Inverter is 70V. Let's the DC-link voltage is to be clamped at 80V. Do you think a LittelFuse 30KPA (https://www.littelfuse.com/data/en/Data_Sheets/Littelfuse_TVS-Diodes_30KPA.pdf), which is rated for 30kW, is enough to protect the FETs from overvoltage?

Thanks.

without regenerative braking, I don't see how the bus voltage should be raised more than a few volts. In this case, overvoltages can be expected rater low energy surges, that should be absorbed by the FETs itself through avalanche breakdown with damage. TVS aren't bad as additional protection.

I would translate the 30000 W TVS absorption capability specification very roughly to 30 Ws energy handling. That's more than a small TVS. It can be still difficult to select a voltage that guarantees TVS clamping before FET breakdown.

There is no recuperation (regenerative) braking in the inverter.

Click to expand...

There is regenerative action in the inverter. It is automatic.

If you try to decelerate your motor quickly (means faster than natural deceleration of the motor), then stored mechanical energy in the rotor will go to DC bus automatically.

MC3PHAC is a ready made controller IC only for designing three phase VFDs. You can see its datasheet, page 14, topic "regeneration control" for details.

MC3PHAC https://pdf1.alldatasheet.com/datasheet-pdf/view/184904/FREESCALE/MC3PHAC.html

If you try to decelerate your motor quickly (means faster than natural deceleration of the motor), then stored mechanical energy in the rotor will go to DC bus automatically.

Click to expand...

Yes, of course. I presumed, powersys knows about.

FvM said:

without regenerative braking, I don't see how the bus voltage should be raised more than a few volts. In this case, overvoltages can be expected rater low energy surges, that should be absorbed by the FETs itself through avalanche breakdown with damage. TVS aren't bad as additional protection.

Click to expand...

Some tests were carried out yesterday... The Xantrex power supply output voltage was set as 70Vdc, while its OVP limit was lowered to 80V. The motor was operated at 2000rpm, then the speed command was reduced to zero instantly. When the motor was slowing down, the voltage reading on the Xantrex power supply display panel increased and sometimes it triggered the OVP of the power supply. Thus, it looks to me that the motor, while slowing down, did "generate" energy back to the inverter. However, the HMP-Inverter does not have a mechanisme/circuit (e.g. braking resistor) to recover the energy.

FvM said:

I would translate the 30000 W TVS absorption capability specification very roughly to 30 Ws energy handling. That's more than a small TVS. It can be still difficult to select a voltage that guarantees TVS clamping before FET breakdown.

Click to expand...

The 30kW TVS may be able to absorb the transient energy if the DC-link capacitance is reduced to half (e.g. 10000uF). I agree with you that it still difficult to find a TVS with suitable "Reverse Stand Off Voltage (Vr)" and "Max. Clamping Voltage (Vc)" for my application. According to this article (**broken link removed**

"...Through proper selection and configuration, an effective transient suppressor combination can be achieved for almost any protection need..."

If several TVSs are paralleled , do you think the "effective" clamping voltage (not the "Max. Clamping Voltage") can be lowered for the same amount of peak pulse current (or surge current)?

Thanks

- - - Updated - - -

FvM said:

Yes, of course. I presumed, powersys knows about.

Click to expand...

True... I noted such phenomenon in some tests carried recently...

- - - Updated - - -

ahsan_eda said:

There is regenerative action in the inverter. It is automatic.

If you try to decelerate your motor quickly (means faster than natural deceleration of the motor), then stored mechanical energy in the rotor will go to DC bus automatically.

Click to expand...

Thanks. I think this exactly what happened to the system described above.

So it turns out that you do use regenerative braking. To decide if the TVS diode can manage the energy, you can estimate the energy stored in motor's and possibly connected system's moment of inertia.

It doesn't help to reduce the bus capacitance, because the energy is supplied by the motor, not the capacitors. In contrast, increasing the capacitors can possibly reduce the overvoltage. If it's true, that the bus voltage has been raised from 70 to 115 V OVP trip in your previous test, the motor energy is at least 83 Ws. So you would in fact need several paralleled TVS diodes, and they must be allowed to cool down before next braking action.

FvM said:

So it turns out that you do use regenerative braking. To decide if the TVS diode can manage the energy, you can estimate the energy stored in motor's and possibly connected system's moment of inertia.

It doesn't help to reduce the bus capacitance, because the energy is supplied by the motor, not the capacitors. In contrast, increasing the capacitors can possibly reduce the overvoltage. If it's true, that the bus voltage has been raised from 70 to 115 V OVP trip in your previous test, the motor energy is at least 83 Ws. So you would in fact need several paralleled TVS diodes, and they must be allowed to cool down before next braking action.

Click to expand...

Great explanation! Thank you. It seems that TVS is not an easy solution for this problem. Do you have other better option?

Cheers.

Break resistance (switched by a FET) is the best option. It can be automatically operated by a voltage comparator.

FvM said:

Break resistance (switched by a FET) is the best option. It can be automatically operated by a voltage comparator.

Click to expand...

Thank you. I will implement the brake_resistance + FET + voltage_comparator.