[time-nuts] Voltage standards

[Bruce Griffiths]

Greg Burnett wrote:
> The Linear LTZ1000 is a pretty wonderful voltage reference, but they're not 
> all created equal. It would be nice to pick the "cream of the crop" from a 
> sample of 100 of them, but that would get expensive (and time consuming) in 
> a hurry.
>
> For use in their 3458A DMM, Agilent burns-in the LTZ1000 references and 
> hand-selects the ones deemed good enough. The few, "cream of the crop" are 
> set aside for use as high stability references in Option 02 3458A DMMs (spec 
> = 4ppm/year) and the references not quite that good are set aside for the 
> standard version 3458A DMMs (spec = 8ppm/year).
>
> I've found that, after a few years use, the vast majority of standard 3458As 
> easily meet the option 02's tighter 4ppm/year spec. In fact, I've seen a lot 
> of 3458As drifting less than 2ppm/year, which is really good.
>
> I'm guessing that the LTZ1000 can exhibit, perhaps, as much as 0.5 to 0.7 
> ppm hysteresis when powered down and back up again, because I've 
> occasionally observed shifts that large after power down/up and 
> restabilization.
>
> Greg
>
>
>   
Greg

Fluke largely overcome the hysteresis due to temperature cycling by a
thermal conditioning process.
The LTZ1000 temperature is varied cyclically with a decreasing amplitude
(probably requires a microprocessor to do this).
Of course, the dwell times, initial temperature excursion, and the
number of cycles are not stated in the publicly available literature.
They also operate the LTZ1000 with a junction temperature somewhat lower
than that used in the datasheet circuits.
There is an error in the datasheet, the negative regulator as drawn will
probably destroy the chip.
The heater connections should be reversed.
The latest datasheet shows the internal connection of the various
substrate and other parasitic diodes earlier datasheets do not.


Bruce