Voltage supervisor not supervising correctly - TPS3831L30

[Stefano]

Hello,

I am experiencing issues with TPS3831L30 (voltage supervisor), which I use on a battery powered device I designed to disable three voltage regulators when battery voltage drops below 2.63V (factory trimmed), see attachment for circuit schematic.

Full_scenario.png is oscilloscope view of battery voltage (yellow) and RESET signal coming from TPS3831L30 (blue).
Note that battery range is supposed to be 4.2V to 2.5V.

Explanation of picture:
Simulating the battery with a power supply, when the battery is fully charged at 4.2V, the RESET pin of the voltage
supervisor is HIGH as it should be, because 4.2V > 2.67V, which is the fixed threshold of the voltage supervisor.
As I discharge the battery, the RESET pin goes low when Vbatt < 2.67V, all good here.
Now the problem: as I recharge the battery, I would expect the RESET pit to ho HIGH again when Vbat > 2.67V, but infact it does so at around 3.8V.

While recharging, if I hold 2.67 < Vbat < 3.8 (glitch.png) I see some glitches where the RESET signal tries to go high, but it does so just for a moment.

For some reason the range 2.67V to 3.8V has some quirks, which I first noted when I discharged and recharged the battery, as shown above.
Later I discovered that this happens also "on the way down", meaning that, starting from a charged battery, if I discharge it to somewhere 2.67 < Vbat < 3.8 and I turn off the power supply
to simulate turning off the device, when I turn it back on the same behaviour is observed: RESET pin low with glicthes unless charged back to above 3.8V.

All of this happens only when the RESET pin has a "load", meaning it is attached to the three EN pins of the voltage regulators.
If the RESET pin is left floating, its signal is what I would expect.
I initially thought that the RESET pin cannot source enough current to drive the three EN pins, but the datasheet says 8mA max on RESET pin, and trying to drive the EN
pins with a power supply it draws less than 1mA, so clearly this is not the issue.

Any idea why this is happening?

I condensed all this info in diagrams.png.

I was troubleshooting something else before this issue, and in doing so I reflowed the area at least a dozen times at 250 °C.
Is it possible that to much heating/cooling cycles have cooked the component such that it results in this behaviour? Everything else on the board is working absolutely fine.

Thanks

[Peabody]

I haven't checked the datasheets of the three regulators, but it's possible something is going on at their EN inputs that's causing this behavior.  That's particularly likely if the /Reset output behaves correctly when unconnected.  I found this in the supervisor datasheet:

9.2.2.2 Bidirectional Reset Pins

Some devices have bidirectional reset pins that act both as an input and an output. For applications where the TPS383x devices drive a bidirectional reset pin, place a series resistor between the TPS383x output and the reset pin to protect against excessive current flow when both the TPS383x devices and the reset pin attempt to drive the reset line. [Figure 16 shows a 47K resistor.]

Just as a test, you might also run the /Reset output through some non-inverting buffer.

Another test would be to connect /Reset to Vbat, or ground, but not to the regulators, through a resistor to check its drive capability, and see if it still switches at the proper voltages.

Finally, testing with the regulators connected one at a time might tell you something.

It just seems that if it works properly under all reasonable conditions when not connected to the regulators, the problem may be with the EN pins of the regulators, not the supervisor.

[selcuk]

Is it possible to disconnect regulator enable pins from the supervisory chip? TPS63031 has a different input stage and it is not powered from VCC input. That glitches may be caused by buck / boost transition. As suggested, you may need some sort of isolation if this is the case.

[Stefano]

[Figure 16 shows a 47K resistor.]

I tried this, no change.

Just as a test, you might also run the /Reset output through some non-inverting buffer.

I tried this with a double NOT gate and then with an AND gate with both input shorted together and the result was the same in both cases.

Another test would be to connect /Reset to Vbat, or ground, but not to the regulators, through a resistor to check its drive capability, and see if it still switches at the proper voltages.

I tried to drive an LED through the /Reset pin, my PSU said I reached 10mA, didn't went any further. Drive capability seems ok.

Finally, testing with the regulators connected one at a time might tell you something.

Indeed I got something out of it: the 1.9V LDO is the one keeping the /Reset line low until 3.8V, the 5V boost converter "only" needs 3.1V, same for the 3.3V buck/boost.

It just seems that if it works properly under all reasonable conditions when not connected to the regulators, the problem may be with the EN pins of the regulators, not the supervisor.

I think the above test confirms this.

I confirmed before that it's not a current issue, and per their datasheets 2.67V is plenty to register as a logic HIGH on the EN pin, so I'm out of options right now.

[Peabody]

Well, I'm stumped.  You say /Reset switches at the correct voltages if it's not connected to anything, but misbehaves if it's connected to the inputs of an AND gate.  If those are CMOS AND inputs, I don't think there should be any difference in behavior between the two.

[Stefano]

but misbehaves if it's connected to the inputs of an AND gate

AND it is connected to the EN pins, if is disconnected it behaves properly.

The current PCB Im working with is a mess of jumper wires, next step is to properly isolate the various components by using the other copies of the PCB I got with the original order.
Will report back.

[Peabody]

but misbehaves if it's connected to the inputs of an AND gate

AND it is connected to the EN pins, if is disconnected it behaves properly.

I don't understand.  /Reset is connected only to the AND gate inputs, and the AND gate output is connected to the ENs.  What happens then?

[Stefano]

I tried the configurations in the attachment, all of them have the same behaviour, which is that the /Reset line goes low (before and after the AND gate) when connected to the EN pins, not to other loads, for example I tried a red led.
At this point I seriously think I might have overcooked the component.

I'll consider buying some new ones and test.

I'll think about it for few days before proceeding.

Thanks for the helps so far