Dear Experts,
                     Designed a traditional current summing bandgap ( sub 1V). Verifying  with  global and local variation.

Opamp used was pmos diff-pair folded cascode. Now if I change the OP to an wide-swing folded cascode ( bot n and p diff pair) get a little better result. Means less number of bad samples compared to the p-only diffpair. May be because when you have two pair to do your job you have more chance to survive variations  ;) . However I am affraid wheteher this conclusion makes any sense because the result of montecarlo depends a lot on the seed. Can anybody hint at the extra cost of going with wide-swing folded cascode ?