Why is current going through the H-bridge in Hi-Z mode?

[robzy]

So I'm using a DRV8872 in a bit of an atypical application, but I'm still pretty sure it should work.

I've attached a schematic where I'm using a DRV8872 to switch a load. I'm turning off the load by putting the DRV8872 into Hi-Z mode by putting IN1=IN2=0volt.

In the schematic I've shown a short to 24V, but in reality its a low impedence load.

The problem is that I seem to be getting a current through my load when the DRV8872 is in Hi-Z mode, and I can't work out why. The MOSFET's should be turned off and the body diodes are reverse biased.

When I measure the current into OUT1 with a multimeter shorting it to 24V I'm seeing a current of about 30mA. Why is that?

[drvtech]

The waveforms on page 1 of the datasheet explain it.
With both inputs low it shows OUT1 at an intermediate level.

[jwet]

Do you have bypass and bulk caps in place at VM? The part has a charge pump internally- likely wants a good stiff rail at VM to develop Vm + 10v for N channel gate drive.

[SiliconWizard]

Are you "shorting' OUT1 to the 24V supply you use for VM with your multimeter?

[robzy]

The waveforms on page 1 of the datasheet explain it.
With both inputs low it shows OUT1 at an intermediate level.

What does "OUT1 at an intermediate level" mean exactly? Shouldn't all the MOSFETs be turned off? Where is the 30mA of current going exactly?

Do you have bypass and bulk caps in place at VM? The part has a charge pump internally- likely wants a good stiff rail at VM to develop Vm + 10v for N channel gate drive.

Even if I had insufficient bulk capacitance, why would that effect Hi-Z state? Hi-Z states requires the MOSFETs to be turned off, and you only need a solid GND to achieve that.

Are you "shorting' OUT1 to the 24V supply you use for VM with your multimeter?

Interesting question. I am indeed. And I'm seeing 30mA flow.

[edit]: And I'm doing this because even when I connected OUT1 to VM with a small-value resistor I was getting current flow. I was not expecting current flow under either of these scenarios.

[SiliconWizard]

Are you "shorting' OUT1 to the 24V supply you use for VM with your multimeter?

Interesting question. I am indeed. And I'm seeing 30mA flow.

[edit]: And I'm doing this because even when I connected OUT1 to VM with a small-value resistor I was getting current flow. I was not expecting current flow under either of these scenarios.

Do you have an oscilloscope? What I'm wondering is whether it's oscillating when you connect OUT1 to VM with a very low impedance like this. Could be worth looking at.

[SCSKITS]

This part is not designed to have 24V (or a load connected to the supply) connected to the Out1 pin. You may be back feeding some internal circuitry in the part. Connect the part per the data sheet with the motor or load connected between Out1 and Out2. Then measure the current through the motor (or load). Getting 0mA current when both outputs are OFF requires requires the load connected from OUT1 to OUT2.

If you just need to switch a load to ground, a simple logic level FET would suffice.

[jwet]

I'll be curious to see what you learn, please post solution.

[robzy]

Do you have an oscilloscope? What I'm wondering is whether it's oscillating when you connect OUT1 to VM with a very low impedance like this. Could be worth looking at.

That's a really interesting possibility. Seems unlikely. But I can't rule it out. I would've thought a low impedence connection would be fine, but in my case I probably have some inductance, so maybe something freaky is going on.

I do have a scope, an annoying USB one, but it ought to do the trick.

This part is not designed to have 24V (or a load connected to the supply) connected to the Out1 pin. You may be back feeding some internal circuitry in the part.

It's an H-Bridge. The datasheet shows it as being an H-Bridge, along with two FETs on the output. When not in Hi-Z it will have VM connected to the Out1 pin through a low impedence.

And even if this particular H-Bridge implementation doesn't like VM being connected to Out1 when in Hi-Z mode I'd like to understand why. Sounds like a good learning experience.

If you just need to switch a load to ground, a simple logic level FET would suffice.

I don't just need to switch a load to ground. I need a H-Bridge.

[tooki]

The waveforms on page 1 of the datasheet explain it.
With both inputs low it shows OUT1 at an intermediate level.

What does "OUT1 at an intermediate level" mean exactly?

See annotated oscilloscope image from page 1 of the datasheet, attached.

[robzy]

See annotated oscilloscope image from page 1 of the datasheet, attached.

Yes I saw that, but what does it mean? What is causing that? Is it relevant to the situation in the OP?

In Hi-Z mode won't the DRV8872 just turn off the MOSFETs and rely on the body diodes? Are the body diodes creating that "intermediate" level? Is it the motor's EMF as it winds down? And why would either of those have an impact on the circuit in the OP?

[robzy]

ChatGPT has an interesting theory:

"The 30 mA likely represents the conduction through the body diode of the high-side MOSFET, which has a small forward voltage drop when VM is applied to OUT1."

That kinda-sorta makes sense. I've been assuming that VM is connected directly to the drain of the high-side MOSFET, but if there's some voltage being dropped before it hits the high-side MOSFET then some current will get through the body diode.

Although... the datasheet does show VM being connected directly to the highside MOSFET...

[tooki]

I don’t understand why anyone would ask technical questions of a hallucinatory software system. (Am I the only one who gets annoyed when people post “I asked ChatGPT and here’s what it said…” posts? What a waste of time…)

[robzy]

I don’t understand why anyone would ask technical questions of a hallucinatory software system. (Am I the only one who gets annoyed when people post “I asked ChatGPT and here’s what it said…” posts? What a waste of time…)

Lol, I never thought I'd see an ad hominem argument applied to an LLM.

It came up with ~6 possibilities. All of them filtered through my meat brain. 5 of them weren't helpful, 1 of them actually had some legs.

Describing LLMs as "hallucinatory software" doesn't do them justice. They are massive mathematical models that link words together (and therefore concepts).

As you're well aware, they most certainly have limitations that are easily to overlook because of the way they mimic humans.


As an aside - I don't understand why anyone would use a rubber duck for debugging. It's just a piece of molded polyvinyl chloride!

[mikerj]

Lol, I never thought I'd see an ad hominem argument applied to an LLM.

It's not an ad hominem, it's a basic fact that current "AI" based on LLMs are unreliable and should not be used as authoritative sources.

[jwet]

Back from AI Space-

I think it might be related to the charge pump.  The charge pump doesn't disable until in 1 and in 2 have been low for 1 mS.  The spec sheet says that quienscent is 10 mA at 12v which sounds a bit  like 30 mA at 28v if it scales.  I don't think the charge pump is shutting off.  Are measuring the current in the out 1 leg or from the VM pin.  Have you measured ground current.

Possible tests

If you put a low impedance load between the outs, does it tristate?
Can you connect a variable supply with an ammeter up to out 1 ref'd to Vm and seen when it starts to draw current.