[volt-nuts] "*WAY* too expensive for even Keysight to redesign"

frank.stellmach at freenet.de frank.stellmach at freenet.de
Fri May 8 08:38:15 EDT 2015

Joe Geller once collected 3458A serial numbers, and concluded, that this instrument maybe sold to about 50k units in these 25 years. It's not clear, if it's still sold by high numbers, but the total turnover might have been around 400M$, or 16M$/yr.
In the end, that should be enough to finance a redesign, or a model facelift.
This is urgently necessary, even if Keysight would not see a totally new or increasing market for 8.5 digits DMMs, but only wants to still offer this instrument 'as-is'.
Many components of the 3458A are already obsolete, or endangered by PTNs, not to speak about all these through-hole components. I've already seen pictures about a piggy-back solutions for several ICs, and maybe they have to use that already for the new production, especially the two fast comparators EL2010, U142 & U181, used with the A/D.
The 68HC000 is also obsolete in the DIL package, and the SMD package is 'not for new design' already.
Therefore, a complete redesign, including the software architecture, is more reasonable.
Keysight would need the budget for that, but they are meanwhile also lacking the brains, which have mostly left the company (Wayne C. Goeke, the inventor of the A/D, joined Keithley, and Ronald L. Swerlein, the God-father of the ACV processing, well he's retired, for some personal reasons, obviously).
Then, another big problem would arise, that is the verification/validation of the traceability of the 2 source / autocalibration feature.
It would be not so easy to again achieve the acceptance of the 'metrological community', if any of the crucial parts of this instrument would be touched.
I assume, this direct acceptance in 1989 was only due to the close cooperation with the NBS then, when they validated the ~0.02ppm linearity by means of the new JJ array.

I have read a lot about the history of the very similar FLUKE 5700/5720 artefact calibration.
In contrast, it took FLUKE several years, before their instruments experienced the same reputation.
Well, the 3458A was designed for metrological use in 2nd instance only, due to the 55°C ambient operating temperature, and these many compromises they had to make concerning stability.. especially the LTZ1000A reference could have been optimized greatly (8x) with 20°C lower ambient requirements, and a bit more cleverness.
Regarding this aspect, please compare the stability specs to other real metrological instruments, like the FLUKE 732A/B, the 7001, and the 1281 / 8508A 8.5digits DMM.
So, the 3458A was  mainly intended for military conditions, but also for harsh industrial application, e.g. end of line testing at the manufacturing line, where laboratory conditions can not be maintained.
I also think, that the mediocre / cheap (copy-and-paste) design of these new 6.5 .. 7.5 digits DMMs still leaves a big field of other applications for precise 8.5 digits DMMs, as it always has been..
I used this instrument already in 1990, at university, for high SNR, low distortion digitizing @ 16bit/100kHz or 18bit/50kHz, down to -100dB / 0.001%, single shot.
The 3458A may still be benchmark in this category, probably also compared to modern delta sigma A/Ds, but for sure compared to the recent, new 6.5 and 7.5 DMMs.
For my experiments, I also had the necessity to design and to adjust several precision current sources, DCI < 0.01%, ACI < 0.05%.
That's not yet a true 'metrological' application.. But if you study the specifications of these new DMMs, even the 7.5digits 344470A will still not manage that level of uncertainty, if you take the 90 days spec, or their T.C.s.
Generally, their crucial parameters do not fit their resolution.
All of them have an A/D (multislope IV), wich are linear to 1..3 ppm only.
A 7.5 digit instrument would instead require 0.1ppm linearity, otherwise the resolution is useless. For that reason also, the featured autocal function does not work like in the 3458A, not a quarter as good!
Same goes for the mid and long term stability and the T.C.s of the references and the ranges.. these are 2 or even 3 orders of magnitude beyond the claimed resolution.. that simply does not match.
The 3458A instead gives much better reliability, and comfortableness to the user, especially by it's unique autocal feature, which relevance can't be emphasized enough. 
Also, only the 0.02ppm linearity legitimates the 8.5digit resolution, and allows ultra precise ratio measurements.

A revised version should have lower ambient temperature specification, if used in metrology, more stable Volt and Ohm references, and better resistor networks, like in the 1281/8508A.
(It's a shame, that the 3458A does not even have Ohm transfer uncertainty specification, and only 12ppm uncertainty for 1KV DC.)
Then it would be prepared also for the upcoming uncertainty improvements on the electrical units in 2018, by the planned new definition of the SI. The uncertainty will jump from about 0.2ppm down to 0.001ppm for the mise en pratique, and to zero uncertainty for the definition.

I would not buy a revised instrument (3458B ?), as I hope that my existing 3458A will survive another 15 years.
Maybe, I furthermore upgrade some key components like the 40k standard resistor, the DCI shunts, and part of the Ohm circuitry, instead..


Alle Postfächer an einem Ort. Jetzt wechseln und E-Mail-Adresse mitnehmen! Rundum glücklich mit freenetMail

More information about the volt-nuts mailing list