David,

Depending on the time frame needed for these resistors you can order many
(but not all) of the parts from Farnell. I believe the Rhopoint resistors
are the recommended brand. The high value resistors could be Caddock USF
series.

Another alternative would be to contact Edwin Pettis. He manufacturers
custom resistors and does ship internationally. He is working on a
'project' for me that requires a replacement precision resistor. I am not
sure if he can make the highest values.

Your other alternative would be to order a couple of these
http://www.precisionresistor.com/Digital-Multimeter-Calibrator/

They have made me some custom sets, so I know it can be done. The 10M
resistor prices are extremely high and the turn around for the custom sets
can be long. They do not make 20M WW resistors last time I checked. You
will probably still need to buy separate resistors from Caddock.

I am guessing the Fluke 5450A is difficult to get or is too expensive where
you live.

Todd

On Thu, Aug 6, 2015 at 11:25 AM, David C. Partridge <
david.partridge@perdrix.co.uk> wrote:

On 06/08/2015 16:25, David C. Partridge wrote:

I don't necessarily need 0.001% parts.  Low TCR is important, and I will
need to know that actual values to 10ppm or better.

Presumably long term stability is more important than TCR - not many
resistors are guaranteed to drift less than 25ppm/year, even unpowered.
E.g. Welwyn 4800 metal foil 25ppm/year, Caddock USF ultra stable 200 &
300 series 20ppm/year dropping to 10ppm/year after 10 years. Short of an
SR104 or the like I doubt you can do better than the 2ppm 6 year life
stability of the hermetically sealed Vishay metal foil resistors
including the HZ series you mentioned or the VHP101, VHP203 etc.

Frankly I find it hard to differentiate between the Vishay sealed foil
resistors - they seem to have almost identical specs. For example is the
VHP203's .05ppm typical TCR (0 to 60C) better than the 10ppm max for the
VHP101 (15 to 45C)? I've no idea but given that the maximum TCRs are
usually at least 10 times worse than the typical figures, maybe not.

Is the 'VHP101 < 10 ppm (+ 15 °C to + 45 °C)' even a maximum given that
Figure 4 says typical TCR is .3ppm and has a 'typical' curve showing
approx 8.5ppm max change from 15 to 45C (and has a TCR of .55ppm from 15
to 20C)? It looks to me that the 10ppm figure isn't specifying a maximum
TCR at all which means you could get almost anything! Interestingly a
note in the HZ series datasheet says 'For maximum TCR < 1 ppm/°C, see
VHP100 and contact application engineering'

The VH102Z seems to have the best guaranteed TCR of .6ppm maximum (100
to 100k ohms) compared to 2ppm for the VHP203 and the HZ series.

Edwin Pettis states his resistors drift is typically better than 2ppm in
the first year, so pretty good but you'd still need to have them
measured every few years. If you have to get them professionally
calibrated it may be cheaper to buy the Vishay parts. Edwin could
provide the measured values as could Vishay if you bought directly from
them.

The spec for the Fluke 5450A's 10k is 6.5ppm max drift/year, but an old
one may be a good solution as hopefully the resistors will be very
stable by now. The only way you could be sure though is to have it
calibrated at least twice to see how much it actually drifts which
wouldn't be cheap.

An alternative is to buy one or more 10k VH102Z as a standard and buy or
build a 10:1 and perhaps a 100:1 Hamon divider and use the 10k reference
to calibrate 1M, 100k, 1k and 100R resistors in a bridge arrangement
using a null meter or suitable DVM. Adding a 2:1 would address the 20K,
200K etc. requirement. This approach might not be suitable for very high
or low resistances, but would allow you to have confidence in many of
your resistors relative to your 10k standard. The divider could also be
useful for voltage calibrations or checking your other calibrators.

Dr Frank describes the one he built, achieving uncertainties of 0.2 /
0.5 ppm for 10:1 / 100:1 releative to output, in reply #8 and #10 here:

[url]http://www.eevblog.com/forum/testgear/hp34401-measurement-of-linearity/?nowap[/url]

I thought he had a more detailed description somewhere but I can't find
it just now. You could probably get away with using cheaper resistors at
the cost of slightly increased uncertainty.