[volt-nuts] Some thoughts about 3458A stability

[Dr. Frank Stellmach]

Well, the voltage performance of the 3458A is best-in-class concerning 
linearity (better than Fluke 720A) and short term stability (transfer 
accuracy).

Its longterm volt stability is mediocre, compared to its long 8 1/2 
digit range and I really wondered, why.

In the service note, 3458A-18A, it's even mentioned, that virtually all 
references initially show a drift behaviour which is too high for the 
specification of 8ppm/yr. Only by aging, and later by regular usage of 
the instrument, the  drift will decrease to lower values. If not used, 
the reference may fall back into its pre-aged state!


Its internal reference, the LTZ1000, is capable to deliver a stability 
of 1ppm/year, stated by the Linear Technology datasheet. Also, standards 
like the 732B (2nd generation), the 7001 and the Datron 4910, also using 
this same reference, provide much better drift rates.

What's the reason for this obvious disagreement?

I had a quick look at the reference board of the 3458A, and I think I 
have found the reason.
The temperature of the LTZ1000 seems to be set to about 95°C, as the 
temperature sensing divider consists of a 15k and a 1k resistor.

The LT datasheet states a typ. drift rate of 1ppm/yr at 65°C, which is 
30°C less.
Typically, drift rates of such a devices detoriate by a factor  of 2 
every 10K temperature rise.
This would fit well to the observed high drift rate of the 3458A.


The Fluke 732A/B references are stabilized at 45 and 48°C, and the 7001 
maybe set to about 40°C only.

Fluke offered a "pimped" version of the 3458A, naming it HP3458A-HFL, 
with a drift rate of 3ppm/yr. only.
I assume, they just have set the stabilization temperature to 50-60°C, 
either by using their own design, or by just adding a ~70k resistor in 
parallel to the 15k.


The disadvantage of a lower reference temperature is a longer warm up 
time for stabilization, as such buried zener references show a 
pronounced hysteresis or creep behaviour during temperature-cycling.

This can be overcome by a permanent heating ( as with the 732s), or a 
"degaussing" technique to remove the hysteresis, as has been used in the 
Fluke 7001.


Frank Stellmach