INA301: Hardware-Only High-Side Current Protection Problem

AJ Pyne

Part Number: INA301

Hello,

Here is an 24VDC H-bridge driver circuit for a ~1A relay coil. It has Enable and Direction control bits, and an INA301 Gain=50 high-side current monitoring.

Enable bit is active-low; upon enable: drain of Q41 and gate of Q21 go high, U39 Reset goes high and switches to "latching" mode. Upon fault, alert pin of U39 goes low, gate of Q21 is pulled low. Upon removal of enable, drain of Q41 goes low, resetting fault on U39.

The problem: Circuit built on breadboard with half-bridge works correctly. Built PCA 1 has strange behavior, with the following voltages seen below after the removal of enable. The Reset and Alert lines are jumpered on PCA 1, so I'm able to break these connections and measure. Built PCA 2 has strange behavior as well, but that behavior spontaneously corrects itself some time (~1 second, with sharp transition) after the removal of the enable. PCA 2 does not have jumpers that I can disconnect.

I've replaced the INA301 with a new chip on PCA 1, put the old chip on a breakout board and put it in the breadboard circuit, where it behaves correctly, and the new chip behaves on PCA 1 poorly. Q21 and Q41 have also both been replaced, with the same behavior. All the resistors measure correctly, and the circuits ring continuity within nodes. I'm looking for a direction to go with the troubleshooting, or if I've missed some normal behavior of the INA301. I've reviewed the gerber designs of the PCAs, they look correct, and have not seen any evidence of fabrication errors.

Thanks

AJ

over 5 years ago

0 AJ Pyne over 5 years ago

Prodigy 60 points

Additional info: I've scoped all the nodes. They are all steady DC voltages, with sharp transitions. I've also tried two meters, one handheld, one benchtop. Doesn't look like this is a case of the circuit fooling the measurement equipment

0 Peter Iliya over 5 years ago in reply to AJ Pyne

TI__Mastermind 31815 points

Hello AJ,

Thank you reaching out on the forum with this useful information. I hope to solve this issue with you.

Looking these over I am curious if the removal of R86, Q13, R11, Q20 in breadboard test circuit schematic has anything to do this. Maybe there is not enough current being sourced through Q21, Q41 for RESET and ALERT pins to turn on? The current through R201 is only ~10uA and once you turn on Q19 and Q20 this could be diverting even more current for RESET and ALERT.

If you could provide any layout information and differences between breadboard and PCBs then that will be helpful for me and colleagues.

Sincerely,
Peter Iliya
Current Sensing Applications

0 AJ Pyne over 5 years ago in reply to Peter Iliya

Prodigy 60 points

Peter,

The issue is with the H-bridge circuit failing to shut off (due to voltage at the gate of Q21 when there should not be), not turn on.

When ENABLE is pulled low, The drain of Q41 is at 3.3V and the gate of Q21 is at 3V, just as expected. The current sensing and fault detection works correctly.

When ENABLE returns to high-impedance and R200 pulls the gate of Q41 back to 3.3V, I'm still seeing the voltages in the attached images, instead of the drain of Q41 and gate of Q21 being pulled to ground through R201 and R212.

Q19 and Q20 source current through Q11, not through Q21 and Q41. The 3.3V supply does not sag during activation of the H-bridge.

The physical layout of the breadboard is of course totally different from the PCB. All the mosfets, diodes, and INA301 are SMD, so those are soldered to DIP adapters. The resistors are all 1/4W axial parts.

Thanks

0 Peter Iliya over 5 years ago in reply to AJ Pyne

TI__Mastermind 31815 points

Hey AJ,

Thanks for the explanation.
1. So does the breadboard setup never fail?
2. At what point do you decide to disable the H-bridge? Is it when the ALERT pin of INA301 goes LOW or is it based upon some other factor in your system?
3. Do you know if disabling the H-bridge during or not during an over-current event (ALERT pin pulled low) makes any difference in system operation?

As a general statement we do not recommend a pull-down resistor for the ALERT pin, a pull-down resistor for RESET, nor connecting RESET and ALERT pins with a resistor. I am not certain these things are causing circuit failure, but it seems likely.

As to why to your breadboard circuit works over your PCA, I would need more information into how you are generating these circuit failures.

For example, there could be a difference in circuit operation if you disable H-bridge (HTS_ENABLE floating) while INA301 is showing an over-current (OC) event (ALERT pin pulled LOW). If this were to happen, then:

a. RESET pulled-down through effective 10kΩ resistance (transparent mode).
b. RESET is pulled LOW for > 100ns
c. OC conditions still exists and thus ALERT pin still holds LOW (so RESET still pulled down with 10kΩ resistor through ALERT pin.)
d. Once OC event goes away, both ALERT and RESET both weakly pulled down with 100kΩ.

If H-bridge disabled with no OC event present (ALERT is High-Z and RESET/ALERT pulled up to 3.3V).
a. RESET and ALERT both become weakly pulled-down through 100kΩ

What worries me is if there is a race-condition here between ALERT and RESET or there are ground currents circulating through ALERT and RESET pins. We specify leakage currents of 1µA typical for RESET and 1µA max for ALERT when VOH is 3.3V, so ground currents may not be the issue.

A race-condition could be likely since your bread board may include more trace capacitance and slow things down enough for circuit to be stable.

One possible step to improve circuit is to add a pull-up resistor directly at the ALERT pin and then insert a buffer in between ALERT and HTSR_DRIVER FAULT. This could help us narrow down the issue.

Sincerely,
Peter Iliya
Current Sensing Applications

0 AJ Pyne over 5 years ago in reply to Peter Iliya

Prodigy 60 points

Peter,

1. Breadboard setup has never demonstrated a problem

2. H-bridge is enabled and disabled dependent on system requirements

3. No problem is observed after disabling the h-bridge after an over-current event.

As a note, the datasheet does not say anything about problematic pull-down resistors or ground loops. It seems I'm otherwise uneducated about these types of chip-level problems.

The problem stops occurring when R201 and R212 are changed from 10K/100K to 1K/10K.

There are 'strange' nodal voltages on the breadboard when I increase R201 and R212, but it took ~1Meg for R212 before they were observable.

It appears a race condition/ground current is the issue here, solved (for now) by lowering the impedance of the pulldown resistors.

The datasheet recommends a 10K pullup resistor for ALERT, however does not state this as a limit, and there is no maximum set for the current sinking of the ALERT pin.

Do you recommend that both ALERT and RESET pins be buffered before interacting with Q41/Q21/R201/R212?

Thanks again,

0 Peter Iliya over 5 years ago in reply to AJ Pyne

TI__Mastermind 31815 points

Hey AJ,

Thanks for answering my questions.

The ALERT pin is a standard open-drain output so the sunk current cannot be too high so that it would burn out the output nFET. Anyway, we do have some current limit information for ALERT pin. Please refer to Figure 33 which shows the ALERT swing to ground versus the sunk current. It shows the IOL current getting up to 5mA. Note that in this Figure/test, the pull-up resistor was probably replaced with a current source, thus the pull-up resistor was effectively being varied to change the current into ALERT during an over-current condition.

Placing a buffer on the RESET pin is less important than on the ALERT pin since RESET is just a digital input.

It makes sense that reducing the absolute values of R201 and R212 is fixing the problem of voltages at these nodes when you are trying to disable the H-bridge. Doing this means you pulling these nodes more strongly to ground with less resistance (when H-bridge disabled), but you won't violate ALERT current (during OC condition and H-Bridge enabled) since the current sunk will be 3.3V/1k = 3.3mA. You will still have some leakage through R212 since ALERT can't swing all the way to ground. From Figure 33 it would sing to 75mV at 3.3mA.

I still would ensure what the state diagram of this circuit's operation would be given all possible scenarios of disabling/enabling bridge during over-current conditions or not. I say this since we do not provide any suggestions in datasheet where the ALERT and RESET pins are tied together.

Hope this helped. Please post back with any updates/questions about your system.

Sincerely,
Peter Iliya
Current Sensing Applications

0 Peter Iliya over 5 years ago in reply to AJ Pyne

TI__Mastermind 31815 points

Hey AJ,

We have not heard back from you. Were you able to remedy your issue?

Sincerely,
Peter Iliya
Current Sensing Applications

0 AJ Pyne over 5 years ago in reply to Peter Iliya

Prodigy 60 points

Hey Peter,

I ordered the next set of prototypes with the lower resistance values and no buffer, as that appeared to work well on all the reworked prototypes I have now. I will try to remember to update this thread once those come in and I have a more 'final' verdict.

Thanks for the info.

0 Peter Iliya over 5 years ago in reply to AJ Pyne

TI__Mastermind 31815 points

Hey AJ,

Have you reached a more final verdict?

Sincerely,
Peter Iliya

0 AJ Pyne over 5 years ago in reply to Peter Iliya

Prodigy 60 points

Hi Peter,

Sorry, but not yet. This circuit is part of a large board, so new prototypes that also include this change won't be available for many more weeks. I did try a 5K/50K set of resistors on the board circuits, no issue at that level either, so I'm pretty confident in the 1K/10K values I chose.

0 Ian Williams over 5 years ago in reply to AJ Pyne

TI__Mastermind 27715 points

Hi AJ,

I'm happy to hear that your circuit is now working well. I will close this thread for now, but if you need more assistance you can reply to reopen or simply start a new thread.

Best regards,

Ian Williams
Applications Manager
Current & Magnetic Sensing

Amplifiers

Amplifiers forum

INA301: Hardware-Only High-Side Current Protection Problem