I notice that the units are burned in for 48 hours. Is there any indication that it is unnecessary to have a longer burn-in period, say 1 week to 1 month?
Background: my limited experience, with a different reference, is that noticable changes occurred over more than 168 hours.
In the first 12hrs I see the output vary on some units (but not all) but then start to settle.
I found that the power cycles, i.e. temperature cycling has more of an initial effect on settling the initial output. On the very first prototype (that Dave saw in his mailbag) the reference used there was reasonable.....until you started the power/temperature cycles......it never powered up twice with the same output. The LM399AH is completely different.
There is also a full battery charge cycle from flat to full on every unit which helps.
But also it's down to the fact I am building & selling these in my spare time and 48hrs fit my production line. Also, given that the user can recalibrate themselves I have never been overly concerned with anything more than 48hrs.
There is a spec for the unit I try to adhere to, anything else is a bonus........but I do read LM399AH's can take a long time to settle (months!) going by what others say..........i'd love to know what other manufacturers do in terms of burn-in times of their LM399AH based units.
I use low tempco resistors around the LM399AH and final output, and making use of some nice LTC2050HV & LTC1250 chopper op-amps.
The DAC itself is the 18-bit DAC9881 from TI........I haven't found anything better.
PS. One thing I have learned........my 3458a is the dogs b******s, whilst my Keysight 34461a doesn't have a great tempco.....I can almost tell the temperature in my workshop using the 34461a and the uV variances.
Ian.