H-Bridge shoot-through

[alec_t]

With the low-side N-FET switched on, the voltage-sense circuit in the high-side 443 will treat that as a shorted load. The protection logic will then respond, but its operation is a mystery and perhaps some glitch results in brief switch-on of the high-side FET?

 

[Diver300]

With the low-side N-FET switched on, the voltage-sense circuit in the high-side 443 will treat that as a shorted load. The protection logic will then respond, but its operation is a mystery and perhaps some glitch results in brief switch-on of the high-side FET?

I don't really see why the BTS443 should be turned on at all. The other BTS443 would be turned on, but not the one that can cause shoot-through to the N-MOSFET that has just turned on.

I'm fairly sure that the BTS443 isn't supposed to be used like this. I guess that is because it doesn't have any power connection when it's off and there is no load, so it seems to partially turn on its output MOSFET as the load is applied.

I'm trying to see what sort of device I could use without having to try many different ones. I was hoping someone would point out some feature that is in the data sheet that I should avoid or look for when using a high-side driver like this.

 

[shortbus=]

I'm fairly sure that the BTS443 isn't supposed to be used like this.

What is wrong with just using a regular half bridge driver pair? I tried a few years ago to use SSR for the high side and didn't work at all. But since then I found out about using half bridge drivers and isolated DC-DC convertors to get around the bootstrap part of that type circuit. The DC-DC convertors are really pretty small now.

 

[Diver300]

What is wrong with just using a regular half bridge driver pair? I tried a few years ago to use SSR for the high side and didn't work at all. But since then I found out about using half bridge drivers and isolated DC-DC convertors to get around the bootstrap part of that type circuit. The DC-DC convertors are really pretty small now.

Have you got an example? I would be interested to look at other solutions.

One reason to use high-side drivers is that they have the current sensing output. In a reversing H-bridge, it's not easy to measure the current in both directions if PWM is used.

 

[gophert]

The datasheet is full of warnings about instabilities with PWM of inductive loads and yet you are designing for PWM of a motor. A voltage drop to ground on the input causes switching. An abrupt disconnect of an inductive load can make the whole power rail look like a switching event.

 

[eTech]

I put together an H-Bridge using BTS 443 P smart switches as the high side drivers and FDMS7682-d MOSFETs as the low side drivers. The schematic is attached. All the control and the feedback come from a microcontroller running on 3 V.

The H-bridge works and I can run the motor forwards and backward.

The problem is when I tried to run PWM. The BTS 443 P are not very fast, so the idea was to leave them on, and operate the MOSFETs from the PWM signal. When the MOSFETs turn on, the BTS 443 P that is turned off lets some current through for a few microseconds.

So if I am running forwards, the forward BTS 443 P is on all the time and the reverse BTS 443 P is off all the time. For PWM, the forward MOSFSET turn on and off at 20 kHz. The problem is that each time the forward MOSFET turns on, the reverse BTS 443 P lets current through for a short time. It is quite a lot of current, in excess of 15 A, and for several microseconds. That becomes a problem when the switching is happening at 20 kHz.

Is there is an alternative high side driver that wouldn't do this, or is there some other way of preventing shoot-through?

Is there something in the data sheet for the BTS 443 P that should have told me that this would be a problem?

Maybe disable "VBB", switch direction (turn off/on appropriate mosfets), then re-enable VBB?

 

[Diver300]

The datasheet is full of warnings about instabilities with PWM of inductive loads and yet you are designing for PWM of a motor. A voltage drop to ground on the input causes switching. An abrupt disconnect of an inductive load can make the whole power rail look like a switching event.

I saw that, but I am not using the BTS443 to do the PWM switching. It is too slow.

Also the body diodes of the N-MOSFET should stop there from being any large voltages.

 

[Diver300]

Maybe disable "VBB", switch direction (turn off/on appropriate mosfets), then re-enable VBB?

I don't have any way of turning Vbb on and off.

Also I think that what you are suggesting would take too long to work at PWM frequencies.

 

[rjenkinsgb]

Look at using IR2110 or 2113 drivers for both high & low FETs in each pair.

https://www.infineon.com/dgdl/Infineon-IR2110-DataSheet-v01_00-EN.pdf?fileId=5546d462533600a4015355c80333167e

If you drive the high side with PWM, the bootstrap circuit will take care of the supply; otherwise you need a small isolated DC-DC converter feeding between pins 5 & 6 to supply the high side driver.

ps. Note the COM pin is separate from VSS, so you can add a source resistor for current monitoring.

 

[gophert]

I saw that, but I am not using the BTS443 to do the PWM switching. It is too slow.

Also the body diodes of the N-MOSFET should stop there from being any large voltages.

Yes but the power rail of your supply sees a strong voltage bump when you turn off power to an inductive load (motor). Try using a separate power supply for the control side of your circuit (with common (-) to see if the problem persists. The datasheet says something about the input getting triggered if there is any rapid change in supply voltage.

 

[shortbus=]

Have you got an example? I would be interested to look at other solutions.

Do you mean for the DC-DC convertors? Pick one - https://www.google.com/search?client=firefox-b-1-d&q=isolated+DC-DC+convertors you only need a 12V to 12V one since that will keep on any NMos.

 

[eTech]

At what duty cycle does shoot-thru occur?
If you reduce the PWM duty cycle does shoot-thru remain?

 

[shortbus=]

There are even some half bridge drivers with dead time to avoid the shoot through. I like the Onsemi stuff my self but then I'm just a home taught(with much help from both here and AAC members)

Again chose what will suit your conditions.

half bridge gate driver with dead time - Google Search

 

[Diver300]

I think that I've found what is going on, although I still can't see anything in the data sheet that would have specifically warned me of the problem.

In the BTS443, there is a charge pump and the main output device is an N-MOSFET.

The input line only needs typically 0.7 mA, so the charge pump can't be very powerful. There has to be some capacitance to keep the gate at a voltage above the input voltage. When the input is turned off, there has to be a resistor to discharge the gate.

When the device is off, just about the whole IC is at the input voltage, and the gate of the MOSFET will have to be at that voltage as well.

With that in mind, I put together a circuit to mimic the IC, and I managed to get it to behave similarly in respect to the shoot through.

That shows the original high-side driver and the possible equivalent. For the tests I wasn't intentionally turing on the BTS443 or the discrete circuit. The low side driver was simply turned on for around 30 μs. Unsurprisingly, with just a 1 kOhm load, there is virtually no current consumed and the output waveform looked like this:-

When I connected the BTS443, the waveform looked like this:-

There is a peak voltage across the 0.3 Ohm current-sense resistor of about 4 V, so there is about 12 A of shoot-through, taking about 30 μs to die down.

With the discrete circuit in connected, the waveform looked like this:-

That's very similar to the waveform with the BTS443. I think that capacitance within the BTS443 is acting like a bootstrap capacitor and turning on the output MOSFET.

My new design will be using a VN7004. That has a permanent ground connection and a separate enable input, and it avoids the need to have a driver transistor. There's no shoot-through with the VN7004.

 

[Diver300]

I spoke to colleagues and someone suggested this application note: **broken link removed**

That talks about the problem of MOSFETs turning themselves on with high rates of change of voltage.