Hello Ben, sure, this has been discussed intensively, also on: http://www.eevblog.com/forum/projects/ultra-precision-reference-ltz1000/ 85°C is NOT commonly used neither for LTZ1000A, nor for LTZ1000. Only in the HP3458A, the LTZ1000A is mistreated, running on about 95°C. Usually, if the LTZ is run w/o further heat sources in the surrounding applications, you may run the LTZ1000 : 45°C, 12k / 1k (gives about -0.8ppm/year stability) LTZ1000A: 55°C, 12k5 / 1k LTZ1000x: 65°C, 13k / 1k (see datasheet) 50°C will be somewhere in between. LTZ1000 and LTZ1000A both use the very same die (chip), but a different die attach epoxy. The A version therefore is thermally more isolated from the TO99 case (400°C/K) than the non-A version (80°C/K). Therefore, the A version heats up 10°C higher than the non-A, due to the self heating effect caused by the reference circuit itself, and therefore must run 10°C higher; 45°C or even 40°C is not feasible for the A version. I always recommend the LTZ1000, non-A version, on 45°C for best stability, and it's also much cheaper. Frank