Random Failures in Latest Build

I made an error on the post, we use both supplies on our board but the chip we are having troubles with this run is the ADM7150 chip.

1.a Schematic was unchanged around all 6 supplies from the last run of boards. 

1.b The capacitor used on the BYP pin changed from TCSOPL1A226M8R to T529P226M010AAE200.  Both caps are 22uF tantalum caps, 200 mOhm ESR, and 0805 footprint.  The only difference between the 6 on the board is 2 supplies have their enable pin controlled while the other 4 have enable tied to VIN.  Failures have been observed for both cases.

1.c Gerbers were checked for one of the supplies, that had experienced a failure, and no difference could be seen between previous working runs and the latest board with failures.

2. I have reviewed the Board Design section and all recommendations are implemented.  I have handed off to the manufacturer for input.

3. Will follow up with manufacturer and update post after.

4. See image below.  All 6 supplies are fed from a 5V DC-DC regulator.  Some are working and some are not working from the same rail.  It has been verified to be correct at 4.97V.

Update:

Regarding question 2

Yes, the boards have been assembled in compliance to this standard. If the maximum applied temperature is of interest—it was 238°C. According to Table 4-2 (J-STD-020C Proposed Std Jan04.pdf), packages with a thickness of less than 1.6 mm (which is the case for the ADM7150) can withstand a peak temperature of 260°C for 60–150 seconds (see Table 5-2).

Regarding question 3

Yes chips were X-rayed and sufficient solder was observed

We ordered an Eval kit for the ADM7150 to check supplies that had failed and found the following. Chips that failed require a much larger voltage drop to work than the datasheet specs.  Nominally we have a 5V input to the ADM7150 in our design and are observing anywhere from 0 V to 1.5V on devices that are not working.  On the evaluation kit the bad devices are able to regulate 3.3 V with a load of 80 mA but at a load of 680 mA it requires 6 V.  The datasheet specifies in the worst case scenario the minimum dropout required is 1 V.  What we are observing is that a drop out of > 2V is required to regulate properly on bad devices.

This chip is currently unavailable everywhere... Was there a known bad batch?

1. See table of results from evaluation board, the part used on the evaulation board was an ADM7150ACPZ-3.3.  Iout was varied using a resistive load.  These results were very similar for 4 separate chips that had "failed" on our board. 

NOTE: We are in the process of putting chips from the reel, that have not been assembled on our board, onto the evaluation board to confirm if the failures are observed off the reel or only after they have been assembled on our board.

2. On our board we see 5 V at Vin with all parts populated and 5V at Vin with no parts populated.

3. Snippet of one of the schematic for one of the ADM7150ACPZ-3.3 on our board. This portion of the schematic has been unchanged in 2 years and we have seen this supply fail on more than 1 board but not on all boards this run, of 4 total boards in the first run.

C809, C923, C732: Taiyo Yuden, LMK316AB7106MLHT

C808: Kemet, T529P226M010AAE200 (note this is a 22 uF cap instead of 33 uF as shown in the schematic)

C877: Taiyo Yuden, LMK105BJ105KVHF

Additional testing on our board, good LDO vs. bad LDO.

Good LDO

Note: Ven is 3.3 V here and Vin (5V) is not shown. Vin is the same 5V power rail for both the good and bad LDO.

Channel 1 -> Ven (yellow)

Channel 2 -> Vout (blue)

Channel 3 -> Vref (red) 

Channel 4 -> Vreg (green)

Bad LDO

Channel 1 -> Vin / Ven (yellow)

Channel 2 -> Vout (blue)

Channel 3 -> Vref (red) 

Channel 4 -> Vreg (green)