TSM TestStand socket time has a tail

[jonghkim]

Hello,

I would like to understand why there is a tail for socket time distribution(far from the main population).

i thought It have to be ideally consistent for every single batch(or DUT) or it should have at least Gaussian distribution. 

even though the number of tail samples is quite a few, i like to have more insight on this.

Here i am attaching the distribution that showing socket time i got from stdf file.

by the way, i used the batch model for the measuring.

Thanks!