We are observing multiple device permanent failures after powering up below circuit using MPM3840. The damaged MPM3840 devices shows a short from OUT pin to GND.
The behavior is observed after power cycling the board a couple of times. When we use a fresh MPM3840 device, everything looks normal (VIN is at 5V, OUT is around expected 3.3V, MODE pin is around 2.5V). The enable signal comes from an external card and is at 5V. However, after power cycling the circuit couple of times, this device get a permanent damage showing 0 ohms resistance between its OUT and GND pins (under power off conditions). We have removed the faulty device and tried to provide an external supply (from a power supply unit) to the load circuit (PCIe Generator circuit based on LMK3H0102RERT) and verified that our load currents are only around 25mA and the load circuit is working as expected. We are now seeing such failures on multiple boards and our project is blocked. Hence we are requesting urgent help on this topic.
Thank you for reaching out. I would highly suggest getting in touch with local support relative to where you are as they can deliver more prompt support and possibly further failure analysis if need be. Please submit a ticket using the following link: Contact MPS - Troubleshooting and Debug
In the meantime, I can help further gauge the issue in parallel. Since you have already switched a bad IC with a functional one, I have the following questions to help determine where this issue originates:
Have you done a continuity test across the output capacitance?
If there’s still a short, have you used a heat gun to localize where the short is?
The EN voltage only needs to be some voltage higher than 1.2V and has a max voltage of 6V. How stable is this external EN supply?
Has the quality of the soldering been checked?
Is the layout made in a way where the manufacturing where shorting would be easy?
Hope to hear back from you soon. Hoping we can narrow down and get to the bottom of this.
The enable signal of the MPM3840 is generated from and AND gate. One input to the AND gate is from the on board circuity and another input from the AND gate is from an FPGA card that docks to this board through a samtec connector. There was no pull down on the board for this input. I suspect that there was a loose contact at the samtec connector when we insert the FPGA card resulting in an input floating condition of the AND gate. This could have resulted in a meta state / very fast switching at the AND gate output (EN pin of MPM3840). I am suspecting that this meta state / fast toggling of the EN pin is the reason for the damage of MPM3840. After putting a 10K pull down at the input of AND gate, the issue is no longer observed. Unfortunately, we looked at all signals (OUT, IN, SW, Input to AND gate) around MPM3840 during debug on the oscilloscope, but the EN pin was never probed. So I don’t know how the EN was looking when the devices are getting damaged.
As I told earlier, the damaged devices shows that low side FET inside is shorted (OUT to GND short). We have even verified it after removing the device from the board and directly measuring OUT to GND pin resistance = 0 Ohms on a multimeter.
While our circuit conditions were not ideal, I am not sure how a fast toggling / meta state on EN pin can damage the low side FET of the MPM3840 on startup? I looked at the datasheet and was not able to find any timing related specifications for EN pin. If there is a known restriction, its better to provide a specifications such as minimum on time for EN pin to prevent such damages.
If you want to re-create the issue at your end, you can create a condition as shown in above figure and do couple of power cycles. The chip will instantly get damaged with in few cycles.
Our issue seems to be resolved for now. I will get back if we see any further issues.
Firstly, I am happy to hear that you resolve the issue and that you found a relative cause here.
While I can’t speak on the exact on time of the EN pin, I can speak on some of the parameters of the datasheet where the logic voltages and typical EN signal current is shown:
Is it possible that there were spikes per power cycle that may have sent excessive current that affected the voltage reference, spiking the FBCOMP, PWM to the COT pulse, spiking the drive that would cause the LS FET to short to GND?
