This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TPS3820-33 not resetting MCU. Maybe disabled?

Other Parts Discussed in Thread: TPS3820

Hi,


in a current project I've used a TPS3820-3.3 watchdog which is connected to the micro controller RESET line. To test the functionality of the watchdog circuitry I've disconnected the WDI input during operation. The RESET stayed high.

In this forum I read that the watchdog is not enabled if WDI is left floating or toggled within the first 120 - 200 ms. This leads me to the assumption that the watchdog is not enabled after start up.

Checking the waveforms:

  • WDI is toggled within 13ms for the first time after RESET transitions to high for the first time.
  • WDI pulse width is 16.6 µs and toggled every 10 ms

This values seem to be well within the specified limits.
Circuit and waveforms added for reference.

Any ideas?

Regards, Martin

Not triggering:

Start up:

Pulse Width

  • Hi Martin,

    Am I correct in that your problem is the watchdog should be timing out and forcing a reset but it is not asserting reset?

    And does the MCU have an internal pull-up on PORST/? If not is R21 populated when using the watchdog?

    Regards,
    David
  • Hi David,

    yes the problem is that after I disconnect the jumper P6 and WDI is no longer toggled, it should issue a reset. You can see in the first waveform the pulses stop after the second div. and there is no reset. The timeout of the TPS3820 is max. 300ms.

    R21 is not fitted, but the MCU does have an internall pull up. Also I had the watchdog issuing a reset when the supply rail dropped below the threshold voltage, so it is capable of pulling the line low.

    Could it be that the trigger pulse on WDI is too short? Even if the datasheet suggests a minimum of 100 ns.

    Regards,
    Martin
  • So for everyone,
    I found the solution to my problem:

    To test if the watchdog is working correctly i pulled jumper P6 (see 1st post) and interrupted the WDI line. This leaves WDI-pin floating and that disables the watchdog timer. The ability to disable the watchdog by leaving WDI unconnected is a function of the device, but that you can disable it during operation is a "feature" I didn't expect.
    Let me reiterate: If WDI is floating, the watchdog is disabled regardless if it is floating before start up or during operation.

    With WDI floating, I connected a pull-down resistor to WDI then the controller was reset as expected.

    So, what if the pin toggling WDI assumes tri-state due to some error state? I agree this is highly unlikely, but in that case the watchdog does not reset the controller.

    In summary i would urge everyone who uses this part to connect a pull-down resistor to the WDI line.
  • Hi Martin,

    I have verified your solution and found the root cause. The internal logic of the device generates its own WDI pulse to allow WDI to float and not assert reset. This means that if your WDI generator goes high impedance after its first RESET\  the internal WDI generator will take over and keep RESET/ from asserting. We will be updating the datasheet to clarify on this behavior - I found values above 10kOhm to ground will cause issues. For the time being I would advise a 4.7kohm or smaller resistor to ground, but ideally a 1kOhm to give plenty of margin over temperature and outliers.

    The below waveform shows WDI being pulsed and a 10k resistor to ground, and its effects on RESET\. You can clearly make out the internally generated pulses in this capture (~600mV peak) that have been pulled down by the external resistor.


    The next is a 47k to ground:

    470k to ground:

    10Meg to Ground (Disabled Function Generator):

    Regards,

    David

  • Scope shots didn't post:

    10k to ground:

    The next is a 47k to ground:

    470k to ground:

    10Meg to Ground (Disabled Function Generator):

  • David,


    thank you for recognizing the issue.

    If you are updating the datasheet, maybe you can also clarify the part which describes that the non-A version's WDI must not be toggled if /RESET is asserted. There are a lot of misconceptions floating around this issue on the forum.

    Martin

  • Hi Martin,

    I will see on what I can do to clarify the wording on that section as well.

    Regards,
    David