Why does the LD39050 have a maximum capacitance value?

[MAlde.3]

We have been seeing some intermittent failures of the LD39050 in our products.

I can't see anything wrong with the design other than the output capacitance.

The datasheet specifies a maximum of 22uF for stability, but our design has approx 30uF on the board.

I have tested the stability of the LD39050 by step changing the load, as well as introducing step changes on the inputs. I've done this testing with reduced output capacitance (~10uF), as well as increased output capacitance (~50uF) and I see no stability issues. If anything, the greater capacitance was more stable.

When the part fails it loses regulation. We are running the adjustable version stepping 5V down to 3.3V. When it fails, the LD39050 feels hot and the output is 4.5V.

I am completely out of ideas so I am hoping someone will have some to share!

[MAlde.3]

To answer the obvious first questions...

The load is approx 100mA and rarely (if ever) sees any step change of more than a few mA.

I've measured the temperature under normal operation, the LD39050 doesn't get more than 2-3 degrees above ambient (which is room temperature).

Input capacitance is 10uF.

The regulator is supplied by either an on board 5V regulator, or an off-board 5V source (e.g. USB) these are all diode OR'd to the input of the regulator.

All caps are ceramic. I don't know exactly what their ESR is but since the stability looks good I didn't see any need to find out.

When I say the stability looked good, I couldn't detect any oscillation at all with a 300mA load step change.