The following was captured from the [volts]nut] mailing list
The experiments and the following discussion with Vishay took nearly one year, until now I am able to present a final result.
Further details about the measurements and the setup will be given later.
I now have a personal, thorough understanding of Vishays Z-foil technology and its capabilities. This can also be found between the following lines.
'The specification for the hermetically sealed, oil filled metal foil Z resistors (VHP202ZT) imply a usage as a secondary Ohm standard: Longterm stability of 2ppm/6 years and a low TCR of 0.05ppm/K between 0..60°C typically, vanishing @ 25°C, given by a parabola form of resistance over temperature: http://www.vishaypg.com/docs/63120/hzseries.pdf
I ordered 5 EA 9k9998 +/- 10ppm, to be trimmed to nominal 10000.00 Ohm +/1 ppm by low ohmic thin film resistors.
Vishay always performs measurements on VHP types, i.e. absolute resistance and TCR.
So I also ordered the measurement protocol for a small additional charge. The complete batch of resistors were measured by Vishay at 25°C to a precision of about +/- 1ppm, each one 3 times.
After delivery, I compared each resistor of the ensemble against each other by a HP3458A. The initial setup was sufficiently stable (order of 1 ppm) to determine the relative resistance deviation, and to fine tune the trimming resistors to the desired precision level.
It became apparent during these first measurements, that the setup, and the resistors were not as stable over temperature, as needed for sub ppm level accuracy, and as expected.
First, the measurement setup (i.e. environmental conditions, HP3458A and resistor assembly) was greatly improved to get a verifiable transfer accuracy of < 0.2ppm over one hour, although the HP3458A is not specified in this regard.
Each resistor was assembled into an aluminium bar, together with a precision NTC. The bar then is mounted into a small aluminium box, which carries Au plated CuTe plugs to provide 4 pole measurement for the resistor, and 2 plugs for the NTC and case grounding.
The different sources of instablities, especially thermally induced ones, could then be separated, quantified and reduced.
One resistor out of the ensemble was thermally cycled from 25°C to about 0°C, back to 25°C, and then to 60°C and back. As the resistance value at 25°C was exactly reproduced every time to within <0.2ppm, obviously no thermal hysteresis occurred.
During the measurement, the HP3458A was frequently checked by another VHP202 being at constant temperature. This demonstrated that the instrument stayed stable to within 0.2ppm transfer accuracy.
The differential TCR (dR/dT) of the DUT over this temperature range was calculated to be between -0.5 .. -2.6 ppm/K. That is an order of magnitude above the expected, “typical” TCR in the Vishay specification. Also, the upwardly directed parabolum of R vs. T could not be detected, which should be typical for the Z foil technology, according to: “Zero TCR Foil Resistor Ten Fold Improvement in Temperature Coefficient”, R. Goldstein, J. Szwarc @ Visahy.
Coming from the same batch, all resistors have the same T.C.; this was checked briefly for another DUT. This excludes the possibility to improve TCR by a serial/parallel assembly of 4 resistors.
I complained at Vishay PG about the unexpected behaviour.
Vishay had made TCR measurements at -55 and +125 before delivery and repeated this on one resistor I had resent them. They found a TCR of -0.4..-0.6ppm/K. Although they found a similar result, my complaint was completely rejected. They claimed, that this TCR is well within specification.
Although most of the datasheets about Z-foil resistors and several other publications about the Z-foil technology imply a more optimistic picture, only in Table 1 of the specification, a max. upper limit ('spread') of +/- 2.2ppm/K TCR over the complete temperature range is specified. This is a barn door, 10 times bigger than the typical values over the complete temperature range.
Additionally, Vishay defines the TCR as a “Chord Slope”, i.e. as (Rx-R25)/(Tx-T25), not as a differential one. Recalculation of my results then gave a TCR of -0.5 .. -1ppm/K, still very high.
Also, the parabolum with vanishing TCR at 25°C, (or at any other center temperature) is “typical” only, and does not have to occur on real components.
Conclusion:
The ensemble of 5 precision resistors represents a median standard resistance of 10k +/- 1ppm, if thermally stabilized to 25.00°C precisely.
The 3458A in combination with this stable resistor ensemble achieves a repeatability / transfer stability of <0.2ppm over 1 hour and an absolute stability of < 1ppm over days for the HP3458A alone.
The longterm stability of the ensemble should be less than 2ppm/6 years, which greatly improves the mediocre longterm stability of the internal Ohm reference of the HP3458A. This is to be checked during the coming years.
A goal which could not be achieved, is the usage of the VHP202Z resistors as secondary Ohm standards, as the TCR of roughly -0.5ppm/K is unexpectedly too high.'
Best regards - Frank