TPS65084: Buck4 voltage drop

Hello,

I am not seeing anything in the image that would indicate something wrong with BUCK4 specifically. Can you utilize this FAQ to continue debug on trying to confirm if a specific rail is the issue: https://e2e.ti.com/support/power-management/f/power-management-forum/992402/faq-tps650864-regulator-is-shutting-down-or-measuring-x-lower-than-expected---what-is-the-cause 

Hi Kevin:

We check registers that shows COLDOFF, UVLO, OCP, BUCK4_PWRFLT.

According these flags that we check the VSYS, enable, Buck1/3/4/5 out voltage and BUCK4 out voltage/current.

We can see that before BUCK4 abnormal all power rails are workable, when BUCK4 drop to around 0.6V that other BUCK1/3/5 will start drop voltage too, and after 1ms BUCK3/4/5 will restart.

Please refer the figure as below.

Then we check LX4, FB4, VOUT and I OUT of BUCK4.

We can see that FB and VOUT has no any abnormal waveform but IOUT and LX4 has some abnormal.

Why I out has negative current? and why PMIC stop switching?

Please refer the figure as below.

Please let us know your recommendations.

Thanks a lot.

Hi Statham,

Thank you for the detailed debug.

UVLO is always set on first power up (the PMIC is coming from a UVLO event when you first apply power) so no surprise there.

COLDOFF is triggered if PMICEN is low at any point

OCP indicates a power fault and BUCK4_PWRFLT is pretty self explanatory.

1. Looking at the switching waveform, definitely something strange at the negative current point - could you zoom in further to see the switching frequency and max/min voltages?
2. How often are you seeing this failure and is there any prerequisites like temperature?
3. Are you able to write the BUCK4_MODE = 1b to see if that helps at all?

Hi Kevin:

Thanks your recommendations please refer our replying as below.

1. I think which is noise please refer figure for detail.

2. It's random, so far we cannot figure out the timing.

3. If let BUCK4 in PWM mode, we keep testing it for 20 hours that we have not find out this error condition, but our device needs low power consumption in idle mode, we need to keep using PFM mode.

Please provide recommendations for us.

Thanks a lot.     

Hi Kevin:

Update the waveform for you reference.

Thanks a lot.

Hi Statham,

Thank you for the updated detailed scope shots.

This looks consistent with another issue we dealt with a few years back on the TPS65094, which is the same part but for Apollo Lake instead of Braswell (https://e2e.ti.com/support/power-management/f/power-management-forum/757046/tps65094-pmic-resets-after-glitch-on-buck5-output?tisearch=e2e-sitesearch&keymatch=TPS65094%25252525252520switching# and https://e2e.ti.com/support/power-management/f/power-management-forum/894078/tps65094-low-temperature-issue/3314225?tisearch=e2e-sitesearch&keymatch=TPS65094%25252520switching#3314225 both seem to be similar)

These cases were linked to a system level noise injection into the PMIC while in PFM mode, but both were OK switching to Force PWM to solve the issue. We were not able to replicate this on the EVM with the units returned to us even with extensive testing, so it seems to be a board layout or use case specific issue. Some of the items we recommended be looked at but were not due to time constraints:

• VREF + AGND connection clean (not connected to thermal pad or routed close to buck input/output caps)
• PVIN3/4/5 input and output capacitors GND isolated from BUCK2 (expected to be the largest noise contributor based on current)
• LX3/4/5 not routed in parallel with each other or other bucks

In one of the cases, they also mentioned that they could switch in and out of Force PWM when needed - in particular I believe they were only seeing the issue at cold temp, so they utilized Force PWM when temperature was <0 C (expected to rarely happen in normal use case).

If you could tie the failure to an event, that might help with debugging this particular board. Ways to do this:

• Linking it with a regulator:
  • Eliminate noise from a specific regulator by "over-driving" it with an external power supply
    • For example, if BUCK2 is running at 1V, supply it with 1.02V and confirm it is no longer switching
    • If shutdown no longer occurs, then we can focus on that bucks layout
    • Risk: if a shutdown occurs, the external supply won't shutdown by default - I would recommend trying to force it to shutdown with a trigger
• Linking with temperature - does it happen more at cold?