Please suggest how to characterize a bandgap tempco curve.

[bluesmaster]

If I want to give mass product to a bandgap chip with 50 ppm tempco, shall I test each part with full temperature range? It will be very expensive. Or I can test several part on bench and get a optimized voltage ,for example, 1.23v. Then I will trim all the bandgap voltage to 1.23V. But how can i prevent the drift of corner?
Anyone have experience please suggest how can I granrantee the 50 ppm tempco.
thanks

 

[hk2004]

I have the same quesiton. The experienced engineer help on this, please.

 

[student_of_analog]

I tested about 40 chips, but I measured each one individually, and each one had to be trimmed slightly differently
to obtain 1.25V.
I guess to select a common trimming for all chips, you need to look at which trimming combination gives you the closest
to desired voltage for all chips.

 

[dick_freebird]

If you want it trimmed for tempco that's a 2-pass trim
(like room & hot), burn the code on the second pass for
tempco and output voltage centering.

With enough production experience and maybe a reading
of resistor value and diode Vf, you might be able to do a
single-temp trim using these to predict tempco and Vout.
But that's after you've covered some process space to
learn these relationships. And any class of defect that
makes the reference deviate from ideal (like an anomalous
diode characteristic on just one of the rack) will be a test
escape for tempco, if tempco is not actually tested. You
would have to gauge the odds and consequences of that.