Placement of dc-link current sense resistor

[onemilimeter]

Hi,

I wish to monitor the dc-link current to protect the power MOSFETs in a 3-phase inverter from short-circuit current or overcurrent. A current sense resistor (R1 in attached figure) of 0.015ohm is used. Besides aluminium electrolytic capacitors (C1~C4), I would like to add a 2.2uF (or 0.1uF or 1.0uF) metallised polypropylene (MKP) capacitor (C5) as shown in the figure. However, I'm not sure if I should place the MKP capacitor to node A or B (i.e. before or after the current sense resistor). Kindly comment.

Cheers.

[tesla500]

To properly protect the inverter, you should have a resistor in each of the phase legs. It's possible (depending on the load) to get an overcurrent circulating through the IGBTs without it going through R1.

In any case, the resistor shouldn't have a capacitor (C5) between it and the IGBTs. That will slow down the transient response. The resistor needs to be a physically small, low inductance type, placed after any capacitors. Another thing to watch out for is the inductance between the bank of C1-C4 and C5. If this is just wrong, C5 and this inductance can resonate at your switching frequency and generate large voltages.

Will you have any sensing of the 3 phase output currents? If so, that could be used for short circuit protection.

[onemilimeter]

Will you have any sensing of the 3 phase output currents? If so, that could be used for short circuit protection.

The inverter will be connected to a three-phase motor. Each phase current is monitored using a LEM LTA-50-P/SP1. However, the LEM current transducer is not placed between the source of lower MOSFET and negative dc-link. Hence, it can not detect short-circuit current if shoot-through happens in one inverter leg.

In any case, the resistor shouldn't have a capacitor (C5) between it and the IGBTs. That will slow down the transient response. The resistor needs to be a physically small, low inductance type, placed after any capacitors.

I will use a Welwyn OAR5 current sense resistor (0.015ohm) as R1 (http://onecall.farnell.com/welwyn/oar5-r015fi/resistor-1-0r015/dp/1200375), which has inductance of approximately 10nH. Do you think 10nH is consider low enough for my application? The rated current in my application is 10A and the short-circuit trip point will be set at 25A.

[tesla500]

I wouldn't recommended protecting against shoot through by monitoring the current. Shoot through protection should be by designing the gate drive stage so it can't shoot through, no matter what the microcontroller or whatever is upstream tries to do. The resistive sensing may still be useful for short circuit protection if the LEM sensors aren't fast enough in case of a phase-phase short, or in the case that one IGBT has already failed short and you want to avoid a bang.

That resistor should be acceptable. It has more inductance than say an SMT resistor would, but I'm assuming this is a high voltage inverter (hundreds of volts), so L*di/dt voltages won't be as big a problem compared to a low voltage inverter. For an inverter of this size, you will probably have current slew rates on the order of 0.1A/ns, which is 10nH*0.1A/ns = 1V across the inductance of the resistor, so you won't be overvolting the IGBTs due to this. Your current sense circuit must filter these spikes without tripping though.

You definitely will want to connect the gate drive returns directly to the IGBT emitters, not to the node on the capacitor side of the the resistor. Any inductance in the emitter path is very bad for gate drive (L*di/dt voltage during switching cancels out your gate drive! This could form an oscillator in the worst case).

[onemilimeter]

I wouldn't recommended protecting against shoot through by monitoring the current. Shoot through protection should be by designing the gate drive stage so it can't shoot through, no matter what the microcontroller or whatever is upstream tries to do. The resistive sensing may still be useful for short circuit protection if the LEM sensors aren't fast enough in case of a phase-phase short, or in the case that one IGBT has already failed short and you want to avoid a bang.

In my application, Si8233 isolated gate driver will be used. According to the datasheet, it includes features like overlap protection and programmable dead time. The latter is crucial because the microcontroller in my system is not able to generate dead time. Please share if you know any other gate driver (capable of supplying high peak output current) with similar embedded features.

That resistor should be acceptable. It has more inductance than say an SMT resistor would, but I'm assuming this is a high voltage inverter (hundreds of volts), so L*di/dt voltages won't be as big a problem compared to a low voltage inverter. For an inverter of this size, you will probably have current slew rates on the order of 0.1A/ns, which is 10nH*0.1A/ns = 1V across the inductance of the resistor, so you won't be overvolting the IGBTs due to this. Your current sense circuit must filter these spikes without tripping though.

Unfortunately, mine will be a low voltage inverter using power MOSFET IRFP4110 (VDSS=100V, ID=120A, http://www.irf.com/product-info/datasheets/data/irfp4110pbf.pdf). The DC-link voltage of the inverter will not be more than 72V. Since a SMT resistor has lower inductance, I will try searching for an equivalent one for current sensing. I'm considering to add a low pass filter or simply a bypassing capacitor across the current sense resistor but I wonder if this will increase the response time of the short-circuit protection scheme. Many always say that, as a rule of thumb, the short-circuit withstand capability (SCWC) of an IGBT is approximately 10us. This information helps to design an short-circuit protection scheme with response time shorter than the SCWC. I've been searching for the similar information (as a rule-of-thumb SCWC) for a power MOSFET but to no avail so far. What is the typical response time of a short-circuit protection scheme for a power MOSFET? Do you think a 5us response time is sufficient for my application?

Thank you.