The boards I sell have good volumes, but highly intermittent demand. I'm currently building a crap load (in fact a metric crap load!) and the issue of tantalum cap shorts is raising its head again.
I had started a thread on this a couple of years back, but I don't see where a clear conclusion was found. Many people said "just don't", but that's not an option. The design is fully qualified and a change to this part would require revalidation which is not acceptable to the customer.
My understanding is the issue is micro-shorts caused by the heat of assembly. At the cap factory they do whatever they need to do so good parts are shipped, but the heat of board reflow creates new micro-shorts, some not quite so micro. We see something on the order of 1 failure in 100 or 300 units when initially powering the UUT. Once it passes that, we don't see further failures. One time in some 8,000 units we had a severe short that damaged the PWB.
The literature indicates these parts will self heal if the current is limited on the initial application of power. So my thinking is to install a current limiting resistor in place of a current measuring resistor designed into the test fixture. This resistor has a jumper. The initial application of power would have a current limit set by this resistor with the jumper removed. The power can be removed, the jumper installed and power reapplied and the other tests performed. This should allow the capacitor to self heal and no more tantalum issues messing up production.
I'm not sure what value of resistor to use. The UUT uses 47 ohm, 75 ohms and other values of 1k and above. The 47 ohms or 75 ohms would appear to be in the ball park, but none of the references I've found give a clear indication of the current required for self healing or the current limit to prevent real damage. Other values could be used, but we'd need to order them.
Any suggestions about the resistor value for this application? The circuit is +12V. I assume the self healing would be virtually instantaneous, no? So no time value would be required to assure the power was on long enough?