I'm using ADA4950-1YCPZ amplifiers as the last stage in a preamp before driving an AD9650. The ADA4950 is powered by +/-5V supplies. The unity gain inputs (pins 2 and 3) are driven by ADA4897s powered by the same +/-5V supplies. Pins 1 and 4 are floating. The outputs of the ADA4950 go through 50 ohm series resistors, with a 150pF differential capacitor and two 22pF capacitors to ground at the inputs of the AD9650. The Vocm input of the ADA4950 is tied to the VCM output of the AD9650 (0.9V), and is bypassed with a 0.1uF capacitor. The PD pin is connected to either +5 or -5 when enabled or disabled. The ADA4950 exposed pad is tied to ground.
Four preamps are used in a module, and I've been seeing seemingly random failures of the ADA4950s. I've seen a single channel/ADA4950 fail on 6 different modules out of 130 over the last 2 years. The failure mode is that the ADA4950 outputs are pinned to +3.7V on pin 10 and -3.2V on pin 11, regardless of the op-amp inputs. As an example, a good channel was measured at 1.98V on pin 3, 0.52V on pin 2, 0.17V on pin 10, and 1.63V on pin 11. A bad channel was measured at 2.36V at pin 3, 0.91V at pin 2, 3.73V at pin 10, and -3.18V at pin 11. Once a channel fails, it stays dead.
The resistance measured from pin 10 to 3 or 2 to 11 is still 1.0k on bad channels, so I don't think the internal resistor networks have been damaged by overcurrent. I've tried to cause a failure by inputting signals of +/-5V to the chain (for a 20Vp-p differential swing) without luck. I've also tried heating the circuit to +75C with a heat gun while operating with 20Vp-p inputs, and still haven't caused a failure.
Has anyone else experienced similar issues with this part, or have some idea of a possible root cause for the failures? I had a bad experience with the ADA4096 where I built a number of modules that failed before the datasheet was revised to say that the EPAD should be tied to -VS instead of GND; perhaps something similar is happening here? Since the inputs are protected by both ESD diodes and 500 ohm series resistors, my best guess is that there's some borderline dangerous combination of PD and Vocm during power up or power down.
Thanks in advance for any help!