We found that regardless of whether or not the companion chip starts up correctly, the VIO output rail will rise to ~.8V soon after VBAT/VIN is applied. When the chip does start up correctly, VIO will then later (hundreds of ms) rise to it's correct output voltage of 1.8V. When it doesn't start correctly, VIO will remain ~.8V.
When VIO remains in this ~.8V state, its output voltage tracks the VBAT/VIN voltage. The higher VBAT/VIN is, the higher VIO is.
If the chip does not power up properly, we discovered reducing the VBAT/VIN voltage below ~3.2V will cause the chip to turn on. At the same time of reducing VBAT/VIN, VIO will reduce until it hits a threshold of ~.7V and it's at this point the chip seems to power on. Does this behavior make sense based on the internals of the IC? Is there an internal diode that could be latching on when it shouldn’t be?
We noted that if we lower the external power supply voltage to <~3.2V and then apply power, the chip seems to power up correctly everytime. If we set the power supply voltage >~3.3V and then apply power, the chip fails to power up, unless we subsequently reduce the input voltage to <~3.2V. The ramp-up times for both the external power supply and the BFT power supply board are ~5-10 ms. We discovered that if we slowly turn up the power supply voltage by hand, the board starts up correctly as well.
Toggling the PWRON line while the power was applied had no effect on the unit.
Adrianna noted that the markings on the TPS65920 indicate they are pre-production parts and in theory should be the same as the production parts, but she will send us production components to test out and verify with.
Todd suggested there could be a pre-charge on the VIO rail and to try adding a loading resistor (100-1k ohms) across the output of VIO to see if that has any effect. He also suggested the possibility that another circuit on the board was interfering with the startup. The only other circuit that gets powered up at the same time is the VCC15 rail which is powered from a step-down converter off of the VCC33 rail and suggested to try disabling it to make sure there is no interference occurring.
This morning I added the loading resistors as Todd suggested. I added a 100 ohm resistor to two boards on the VIO output and they both started up correctly every time even with the voltage set to 3.5V. I then replaced both 100 ohm resistors with 1k ohm resistors. The boards started up more often than normal, but still failed 50% of the time at the 3.5V input. When I lowered the input voltage to 3.3V, they both seemed to start up correctly 90% of the time.
One other thing we noted was when the board is powered up correctly, the REGEN output toggles at 1.6 MHz. This behavior is unexpected and not documented. Tony noticed this before and we reported, but haven't heard back on possible causes. See Figure 5 of the attached plots.
Summary: It appears that the companion chip doesn't like being powered up to a voltage >~3.2V. We need to figure out why this is occurring and will try the following steps to see if any changes occur. Adding a 100 ohm loading resistor to VIO seems to have fixed the problem. I will do further testing to confirm this, but we also need to figure out why to make sure it is the correct solution.
Other things I will try (if the loading resistor doesn’t turn out to be the solution):
1. Replace the pre-production part on the board with the production part
2. Disable the VCC15 regulator from starting up
Todd or Ghandar:
- Can you explain why adding the loading resistor may have helped and if it could be the correct solution to this problem? (We don’t want it popping up again later).
- Can you explain the delay on the VIO rail between the rising to .8V and then 1.8V?
- Can you explain the toggling on the REGEN output?
