We are using a TPS65023 PMIC chip to power an AM3505AZCN processor. On some of the boards (3 of the 4 tested), the 1.8V_VDC3 (DDR Power) rail is acting strangely. At powerup, it goes to 1.8V, as expected. However, on some boards, it will switch to about 2.5V after about 400 msec, and stay there.
If we do a hot reset, it will switch back to 1.8V and stay there.
We thought that it was the software changing the voltage, but we disconnected the I2C interface and the results were the same.
If you hold the reset switch, while powering up the unit, it stays at 1.8V. You only need to hold it for > 400 msec.
We used the same PMIC and processor on another design and did not see this problem.
One problem that you might notice is that the inductors are connected wrong. The boards were modified to correct this, but i'm wondering if something is shorted on the boards that are not working. However, i can't think what might cause this strange behavior.
Any help would be greatly appreciated!
Are you able to disconnect the rail from the processor and check to see that the rail does/does not exhibit the issue? If disconnecting corrects the issue, it could be that one of the higher voltage rails is pulling this one up through the processor. You may need to look at the processor errata for a workaround such as a 10k resistor from the 1.8V rail to ground.
Have you verified that the converter is stable by looking at the output voltage with an oscilloscope?
You mention that the inductors were initially connected wrong and corrected; this might be an issue since switch mode power supplies are sensitive to layout. Can you please provide schematic and layout and please note the fix?
It turns out that the cap on the PWRFAIL SNS was too large. A 1uF was used, for some reason, instead of the 1000pF suggested in the TI docs.
It held off the 3.3V from turning on for about 400 msec. Not sure why this would cause the problem, but i changed it to 1000 pF and the output is now 1.8V.