Hello golf32,
at first, please ignore these confusing and partly misleading explanations by Kleinstein, although he already gives some hints to your measurement and metrology problems.
The most relevant topics of your measurement issues are:
- resolution vs. uncertainty (accuracy)
- thermal voltages / e.m.f.
- leakage currents in cables
- 2W vs. 4W measurement for higher values
All resistors of the 5450A were measured to 1ppm resolution, or 6 digits, that is what is shown on the display.
That means the 1 Ohm is resolved to 1µOhm!
The 34401A can also resolve 1ppm, when set to 6 digits. To achieve stable results, you should use 6 Dig Slow, and statistics.
As its lowest range is 100 Ohm only, it is simply not capable to measure 1 Ohm (i.e. to 1µOhm resolution) and 10 Ohm (i.e. to 10µOhm) with sufficiently high resolution.
Therefore, you only can make a useful judgement about 'accuracy' and comparing 'resolution' for 100 Ohm to 10MOhm ranges only.
@ Kleinstein: For 1 Ohm and 10 Ohm, there exists DEFINITELY a resolution problem!
I hope, this answers your 2nd question.
For the low value resistors, 1 Ohm to 100 Ohm, you have an additional problem, which the 34401A cannot deal with directly, these are thermal voltages, on the order of µV to several tens of µV, depending especially on your test cables. These e.m.f. have to be removed by a method called 'Offset Compensation'.
Btw.: That has absolutely nothing to do with the 5450A calibrator, this effect occurs for every low Ohm resistor measurement.
The 34401A uses 1mA test current for the 100 Ohm and 1kOhm ranges. Its last 6th digit therefore resolves 100nV (equivalent to 100µOhm) for 100Ohm and 1µV (equivalent to 1mOhm) for 1kOhm.
If you measure 1 Ohm in its 100 Ohm range, a 1µV offset voltage would create a 1mOhm error, or 1000ppm of 1 Ohm.. and that order of magnitude is exactly what you see in your table!
There is a trick for the 34401A to manually apply this Offset Compensation method:
With 4W , 6dig slow, manual 100 Ohm range, first connect the Input+ cable to the Input- jack of the 34401A, instead to the 5450A, and take a NULL measurement.
This measures and eliminates the e.m.f. voltage at the resistor w/o the test current.
Then apply the cable back to the 5450A, and you will measure the resistance w/o e.m.f.
Use this method for 100, 1k, maybe also for 10k and 100k ranges. The measurements of 1 Ohm and 10Ohm will be much more precise as well, but anyhow limited by the resolution of the 34401A.
The replacement of U201 evidently changed the measurement of all three resistors, (10 Ohm, 10k and 10M) so it was very probably defective.
A working U201 has practically no impact on the calibration of its ranges up to 100k, because the test currents are about 106 times higher than its bias current, which is on the the order of tens of pA .
For 1M and 10M ranges, this might be different, as these are using 5µA and 500nA. Anyhow, when I replaced the U201 in my 34970A, also these ranges were accurate again to about 30ppm afterwards. I assume that the 1M and 10M ranges only might have to be re-calibrated in your case; the 100M range has no special calibration constant, as it's relying on the 10M range only.
The 1k range is now accurate to +8ppm (what was it before?), 10k to -75ppm (+400ppm before), 100k to +29ppm, which is a good sign.
If you offset compensate the 100 Ohm range , I 'd guess it will be accurate to a few tens of ppm, instead of 100ppm.
For the higher ranges, 1M to 100MOhm, you have to take care about leakage currents, but regarding your cables.
Again, that effect has nothing to do with the calibrator, because its own leakage currents are reduce to near zero by design, and otherwise are part of its calibration.
These cheapo cables you are using, might have PVC as an insulator, which might create errors on the order of 100 to 10000ppm, when plus and minus leads are drilled. I recommend to use shielded PTFE cables instead.
For 1M to 100M, you only need 2W method, to simplify your setup, and reduce leakage currents. Use short cables, and don't let them touch each other.
Please repeat these measurements with the proposed methods, and lets see what happens.
Frank
PS: I briefly tested the 100 Ohm range on my 34401A, vs. 1, 10 and 100 Ohm from my 5450A.
The complete setup revealed about 80 digits offset, which varies a bit from resistor to resistor.
That is about 8µV of e.m.f. present, yielding 80ppm error for the 100 Ohm resistor, 800ppm for the 10 Ohm plus 10ppm resolution error for the 34401A, and 8000ppm plus 100ppm for the 1 Ohm.
When using this manual Offset Compensation method, the readings agree to the 5450A within a few tens of ppm at most, although the 34401A was not warmed up at all.
Hint: My specific 5450A might have higher or additional e.m.f. compared to other units for some reason.
The basic idea for this problem stands as described, though, as well regarding the order of magnitude for this e.m.f. error.