Tool/software:
Hi,
We recently had a batch of 20 serializer boards assembled at a local manufacturer. Out of this batch > 75% of boards had a 1V8 to GND short on the boards:
The boards that did not have the short, briefly locked with the TI DS90UB954 EVK and we were able to even read/write remote registers, but over time the boards gradually started heating and we saw a reduction in the net resistance between 1V8 and GND.
On the boards that had a short, we isolated that the short is happening directly under the package (EPAD) or on the side exposed copper shown below:
The Xray images of the shorted boards are attached to this thread.
Further more, we are able to reduce the shorting (see increase in resistance between EPAD/GND and 1V8)
simply by heating the QFN pins with solder braid. These boards are pulling about 120mA on the 1V8 rail with no other loads connected other than the serializer, and serializer locks with the partner.
We followed the CPL spec in the QFN soldering guidelines per this app note: www.ti.com/.../slua271c.pdf
We suspect excessive paste or stencil issues with the EMS but would like to get TI's opinion on what could be going wrong to fix this issue in our next MP build.
THanks in advance,
Dhruv
