Shoot-through in half-bridge using LTSpice

[Rusttree]

I'm modeling a half-bridge in LTSpice using Linear's LT1158. The two N-channel MOSFETs I picked in the simulation are reasonably close to the components I selected for the actual project. Oddly, I'm seeing shoot-through during the gate transitions. Supposedly, this Linear part has anti-shoot-through built-in. I've attached the LTSpice .asc file if anyone has a moment to take a look at it.

Am I missing something obvious? Is this just a simulation artifact?

Thanks!

 

[ronv]

The shoot thru is only a few ns and a few amps.

 

[crutschow]

It likely is typical of what you will see in a real circuit. But it's so short that it shouldn't cause any problems.

 

[Rusttree]

Agreed it's very short. However, it causes a problem in the full version of the circuit. I'm using two LT1158's to form a full bridge for a DC motor controller. The extra little current that gets pulled into the motor, along with the shoot-through current, causes the over-current protection to kick in and shutdown the device.

Based on the voltages present on the gate drivers, it doesn't seem like the shoot-through should be happening at all. For example, when the high side turns on, the voltage on the low side gate driver is about 1V. Yet an 8A spike of shoot-through current is shown. Wouldn't the Rds of the bottom FET at a Vgs of 1V be way too high to allow 8A of current?

 

[ronv]

It takes the FETs a while to turn off even after the gate voltage is gone. I guess I'm suprised the current sense circuit is so fast. If it is an op amp often a small cap across the inputs will take care of it. Sometimes stray capacitance in the motor winding will cause the same problem with high switching speeds.

 

[MrAl]

Hi,

You could be seeing a few different things here...

1. The diode reverse recovery takes a pulse of current to 'reset' the diode in the opposite transistor. This would show up as a spike.
2. The model of the MOSFET allows it to turn on at a lower voltage than the real life device would.
3. There isnt enough dead time built into the chip, or else the model doesnt show the dead time as well as the real life chip.
4. Input capacitance of the MOSFET is larger than the driver chip is designed for.

There's a chance that the real life circuit wont show this or that it would be less. For example, even a small inductance in the MOSFET power leads could limit the di/dt which would limit the shoot through no matter what the reason is that caused it.

 

[ericgibbs]

hi,
The gate capacitance of the MOSFET you have chosen is much higher than the rated driving capability of the LM1158.

If you plot the LTSpice the MOSFET gate waveforms you can see due to the high gate capacitance the rise/fall times are degraded and overlap.

As suggested by Mr Al, choose a more suitable MOSFET.

E.

 

[MrAl]

Hi,

Yeah Eric a different MOSFET might work better.

Also, add some symmetrical resistive load to the output. That's say 50 ohms from output to ground and 50 ohms from output to +Vcc. That could make a big difference.

 

[ronsimpson]

You don't have shoot through. I only looked on the falling edge. The top MOSFET turns off, the output voltage does not fall. After a while the bottom MOSFET turns on. Now there is a 9A spike that is caused by the large capacitors inside the MOSFETs. The current is caused by charging/discharging 10000pF of MOSFET capacitance.

To prove that: I changed the resistance of the 24V source from 1 ohm to 0.1 ohm. Connected a 12 ohm resistor from output to a 12V source (0.1 ohm). Now you can see when the top MOSFET turns off. The voltage drops slowly from 24V to 12V at a rate that is related to 12 ohms and the load capacitance. Then after the 'cross over delay' from the IC is done you can see the bottom MOSFET turn on and drag the output down from 12V to 0V. This fall time is much faster and there is a higher current charging the capacitance.

This is all normal and there is no shoot through. There is a delay in the IC that should not allow shoot through.
Amy time you have a MOSFET Drain moving 24V in 20 to 50nS there will be large current flow to charge the capacitors. Cds, Csg, Cgd

 

[Rusttree]

Thank you all for explaining this behavior. I definitely see now that it's the inherent capacitors in the FETs that are causing the current spikes, not shoot-through. The actual MOSFET and ended up choosing had a much lower gate charge, so I think this problem should be pretty well contained in the real circuit. But at least I understand it now!