Dear all,
I've designed an embedded system which roughly consist of an microcontroller + External Flash, CAN transciever and the LM26001 buck converter from TI.
During testing of the first prototypes of the embedded system i have discovered that sometimes my microcontroller seems to crash and result in a dead microcontroller that i cannot program again or use in it any other form.
This event seems to happen when the embedded system is powered on and when I hold the PCB within my hands. Digging into the problem I've discovered that I'm causing some kind of interference of the LM26001 IC what causes a oscillation on the output of the regulator. When my fingers touch the PCB or components that belong to the LM26001 the output starts to oscillate. This oscillating produces a high enough voltage that kills my microcontroller.
The Buck converter is configured to produces a 3.3V output and powers the microcontroller, CAN transciever logic section and Flash memory. The input voltage is 24V, during testing a linear lab power supply is used to power the device.
The following screenshot shows the oscillating behavior of the buck converter when i'm touching the PCB/components around the LM26001 with my fingers. It can be seen that a too high voltage is generated which causes my IC's to fail.
The following image shows the schematic design of the power supply section, the design is based on a TiWebbench simulation.
The following image shows the PCB layout section of the LM26001. The components that are visible are placed on the bottom layer (L6) of the PCB. Layer 5 is a solid GND plane which covers the entire PCB.
below: PCB Layer 6
Below : Layer 5 thru 1
My question is if there is anything i could to in order to improve the immunity of the LM26001 IC in order to make it more resilient against accidental touches of fingers while operating.
My goal is to make sure the microcontroller doesn't get fried.
Whats I have already tried:
- Reducing the feedback network (R38 + (R39 // R40)) to a series impedance of max 3k2 Ohm in order to evaluate if the high input impedance of the feedback network could be the cause.
- Recalculated the feedback compensation network by hand with the actual parameters of the used components (ESR, Cout) -> C67 = 10nF, C68 = 16pF and R37 = 44k. This gave no measurable improvement.
- increased the SS capacitor from 2.2nF to 100nF. ( no improvement)
Thanks for your effort, any suggestion is more than welcome.
Arjan