Strange moisture test results

[chungylau]

We noticed that after a die is packaged in BGA and subjected to high humidity and temperature, 85C/85%, for 96 hours, the momcaps on the die appear to have shifted in value. They seem to be 10 to 20% higher than before. This is a 90nm low-leakage RF CMOS process. Our VCO's also seem to have reduced loop gain, perhaps due to the momcaps in the resonator being more lossy after moisture tested. Has anyone seen anything like this? Different molding compounds do not seem to help.

When we moisture-test the wafers (unsawn and ungrinded), we don't observe this shift. We grind the wafers to either 5 mil or 10 mil thickness before assembly.

 

[steadymind]

Could be due to package stress at high humidity. This is more of processs reliability issue with momcaps.do other devices exhibit strange characteristics

 

[chungylau]

Pressure on die due to molding compound absorbing moisture, leading to flexing/deformation of the die, is one of the possibilities. We have tried decapping a package and then moisture test, with the same results (shifts in parameters observed). So not sure if pressure on the die is the culprit at this point.

We have not observed any other significant changes in performance/measurements. The changes that we observed have a common denominator: they can all be caused by momcap value and loss changes.

 

[dick_freebird]

I would look for measurement errors that might -appear- as
capacitance drift but are not strictly capacitive. For example
a shunt leakage will increase the capacitance indication from
a ramped voltage or impulse charge capacitance measurement
method.

Your MOMcaps ought to be way deep in the passivation stack
and their dielectric should have negligible moisture exposure
or ability to absorb water. Especially if the top layer is nitride.

You might take some virgin units and subject them to temp
only burnin of the same duration & stimulus (or lack) to see
if there is truly a moisture effect.

 

[chungylau]

There are several symptoms that point to momcaps. The RC oscillator frequency changing, the VCO frequency and tuning range changing, and the S22 of an LNA changing.

The LNA has a momcap at the output. What we observed is just as if the momcap is increased by 20%.

High temperature life test does not show any problem, so moisture is necessary to cause this effect. One possibility is that during the molding process, the top passivation layer is cracked. That allows moisture to leak in and change the value and loss of the momcaps.