[volt-nuts] Ye Olde HP3458A

[Frank Stellmach]

Hi John,

I also picked one up, a mixed one, ie old analogue PCB from the 90's and 
a younger processor board from 2001 (perhaps replaced due to SRAM 
battery discharge). Cost me about 3k instead of of 8k new, and its 
working perfectly - but stable environmental temperature conditions are 
required, thats the weak point of the design.

Maybe its an old design, but currently, no other DMM or calibrator can 
beat it essentially in certain parameters.

It has got the best (differential) linearity, over Keithley 2002, Fluke 
8508A, Datron 1281, Fluke 57xx and even the Primary Ratio standard 720A! 
Only the JJ array can test its linearity!


OK, the 3458A's internal references are not that stable, but it is not 
intended as a secondary volt or ohm standard.

For that you need additionally something like a 732B and a SR 104, or 
the quantum standards, respectively.
But all others of the above mentioned, newer DMMs are not much better.


For DCV, the 3458A obviously has been designed for a very broad 
temperature range of 0..55°C (military use??), which gives an internal 
temp of at least up to 80°C in a rack mount.
Placing its internal volt and ohm standards in a lower and more stable 
environment would have been better, but then, it could not have been a 
DMM-in-one-box.

So the internal LTZ1000A reference has to be running  on 90°C.
If powerered constantly, this gives at least 20 times higher drift rates 
over time compared to a Fluke 7000A, which is running on 45°C.
Other DMM are specified for metrological temp. range and have certainly 
slightly better drift rates (two times).

I have set (pimped) the LTZ to about 55°C for lower drift.

The HV divider cannot be corrected for power dissipation effects, so the 
1000V range is quite mediocre.
I have built my own 100:1 Hammond type divider (~ 752A) to get around 
1ppm for 1000V.


The ohm ranges obviously is its weakest mode.
It relies on an elder hermetically sealed Vishay resistor, with high 
time and temp. drift, and additionally the resistor is exposed to the 
strong internal temperature variations.

Today, by using a selected VHP202Z resistor, one might improve time 
drift to <1ppm/year and stability (with respect to temperature) to < 
0.2ppm over the complete "metrological" ambient temp range.

Additionally, its ohms range resistors are very sensitive to temperature 
changes.

But it is possible by using it in absolutely stable amb. temp. 
conditions (+/- 0.2K), to make 10kOhm measurements / transfers on sub 
ppm stability level.

Currently, I'm working on external 10k standard resistors; but still in 
discussion with Vishay.
Report will follow.

ACV is also umparalleled by other DMMs, due to its (patented?) 
digitizing algorithms, and this can be further improved by Swerleins 
Algorithm. So I think, only Flukes AC standard 792A is "better", if 
using standard electronics.

So my advice, get one used 3458A, it's simply a nice and ultra precise 
box, and build yourself some standards which you would need also if you 
got a newer DMM.

Btw.: A good, absolute calibration is nearly impossible to acquire for 
us amateurs. Even Fluke and agilent obviously offer 2nd grade 
calibrations only.

Frank