I found out the reason that most <= 6.5 digit DMM manufacturers still use the LM399 (as opposed to the provably more stable LTZ1000) is that the LM399 exhibits remarkably low hysteresis between power cycles-- and the LTZ1000 may actually have more (without some kind of [proprietary] "pre-conditioning" process). So, for a bench meter that gets used day-to-day (and sometimes in "non-laboratory" environments), the LM399 is a better choice.
The LM399 is absolutely fabulous for short-term work, but as you probably already know, it takes over 10 years of operation to reach ~1ppm/year stability-- so for very long-term stability, the LTZ1000 is the preferred device. From what I understand, the DMM manufacturers have racks full of LM399's that they "burn in" for a very long time before they are used in a final assembly. Some manufacturers (like Fluke) burn them in much longer than others (Rigol, Keithley-- with Agilent/Keysight somewhere in the middle). That's also why very old DMM's (like the HP3457A) can be scary-stable-- their reference (and the cheap metal film resistors they used in the design) have had decades to become very stable.
The LM399 might also be used because of a pricing decision, not only on the part itself but by the required support circuitry and mechanics (dont forget the mold for a injection molded cap which is already shipped with and mounted at the 399 but strongly sugessted for serious LTZ application), while matching to what you usually want from a 6.5 digit device. I am quite happy with my HP3456A at home and the 34401As at work.
What I dislike about the LM399 it is the FLUKEing very hot temperature setpoint - of course required to use it all over the world even in crammed racks without aircon, but here in mild climate of germany I wouldnt mind if it is anything above 40°C. the fridge with ice packs is close, for those rare very hot days.
The REF194AG reference was made because parts laying around, easily good enough for 3 digit meters but frustrating for transfer from work to home - the backpack trip of 20 minutes has thrown it off for more than 1 mV in any direction. But it is very small and handy as you see.
The LTZ1000A was selected because part available on ebay and known to be a instant near-perfect device without any hassle. I tried to keep out most voodo which doesn't appear inside the datasheet but at the omniscient landfill aka the internet. It works well, it travelled from home to work to hobby room and back, to Dr. Frank and another guy. My cheap and lazy open source design is good enough as shown by Frank that it attracted other people already; selling leftover PCBs helped mitigate the part costs a bit
And in the short time since its birth, it did very well.
The LTZ reference even gets some non-meter-and-references-comparing exercise, e.g. I do internal calibration of some 34401A and (painful) 34405A at work; when one of the main meters come back from calibration, I use those fresh ones to check some lower-class meters which we need only in certain ranges; which I can do with a Dekavider. Also, during testing sensor front-ends i like to use the LTZ and the dekavider. Either to inject a voltage via a buffer, or null out some voltages.
The LM399 was more or less an afterthought as I felt that I skipped an important step. This is now a battery-powered reference; only the standard application again, a battery undervoltage monitor and again cheap and lazy.
When the babies allow it, I will compare the buried zener devices against each other sometimes, and against the workplace DMMs, and Dr. Frank had my LTZ two times for now. All my refs are allowed to travel
Home
Help
Search
Login
Register
