Tempco prediction after trimming--need guru's answer

Dear all,

I am designing trimming bandgap reference. The trimming range, or trimming bit is identified from the 6-sigma monte carlo simulation. What I mean is, the worst case of monte carlo simulation results can be trimming fallen within the spec.

but how can I guarantee the tempco meet the spec after trimming in mass production?

thanks so much.

If you want to guarantee the tempco, or simply output across temp
within tolerance, you would run a two pass probe and determine
the "right answer" trim codes for temp and scale. You will still
have some scatter at the end of the range you didn't test
(low temp probe is the biggest nuisance so probably you would
test at room and hot). You would not trim for dead-flat, as the
temperature curve will be bowed, but some best-guess
negative tempco from room to hot based on what you know
about the room-cold slope, and where you care most about
flatness (typically, more interested in room to warm for consumer
products, but for other apps might be more interested in
spec range than best over a lesser span)

I do not expect you would trim based on simulation results.