Fluke 5440B Problems

[Dr. Frank]

@Dr. Frank: Its a little late today, i will let both instruments run over night and make those measurements tomorrow. Can you specify the needed datapoints? Otherwise i will just measure
+-10V,15V,22V,100V,220V,300V,400V,500V,600V,700V,800V,900V,1000V. How much time is needed for the 3458A to thermally stabilize at the high voltages? I know the 3458A will be a few ppm off at the high voltages.
.....

GAIN SHIFT
     +10:  -285.21uV
     +20:  -347.95uV
     +Hi:  -370.59uV
     -Hi:  +463.96uV
     -20:  +441.90uV
     -10:  +378.15uV

END OF INTERNAL CALIBRATION

Hi,

The 3458A is not compensated for self-heating, therefore at 1kV, you will have an additional deviation from linear.
This is specified 12ppm *(Ue/1kV)^2 (note 6 on page 82 of the calibration manual).

In my case, it's about -6ppm at 1kV. This parameter is varying over different instruments.
The time constant for the 3458A is about 1 minute, until it stabilizes.

The 5440B also needs some time to stabilize at 1kV, also about 1minute, but the oven compensates the self- heating completely.

The deviation and the time constant were determined identically by my 5442A, and a precision Hamon divider.

At 300V, this self-heating is negligible, so I propose for these three ranges:

10V, 20V, 250V, 300V, 1000V
At first, you might test the linearity, i.e. 0... 11V in 0.1V steps, to calculate the non-linearity factor of the D/A, between 300V and 1kV.

I have not fully understood, how the autocal process of the 5440B works, regarding these listed parameters, or what the real limits are.. but I assume, that your gain shift factors may be too high, and that your 5440 needs a proper external ratio calibration.

 This should be visible by the above mentioned ratio checks.

Frank

[Echo88]

Didnt have time to write a fitting script because my LTZ1000A turned up and eagerly wanted to be mounted in the KX-Boards. So i just took measurements by frontpanel and didnt already calculate linearity and so on.
During the whole measurements i also found out:
-i indeed should clean the relays, since on one point the relays did clatter without ending (one gentle stroke on the case ended that, the relay apparently found contact and the Internal Calibration continued without errors) 
-up to 800V the calibrator works okay, at 900V it says Output Limit Fault, Output Undervoltage 

Measurement Results:

3458A and 5440B/AF, thermally stabilized the whole day, ACAL and Internal Calibration just before the measurement, 100NPLC, after 100V i did wait >= 1min after every calibrator-value-change to thermally stabilize

3458A 10V Range

0.1V...1V...11V ->

0.1000009V
0.2000015V
0.3000018V
0.4000023V
0.5000032V
0.6000031V
0.7000031V
0.8000030V
0.9000027V
1.0000029V
2.0000032V
3.0000030V
4.0000029V
5.0000020V
6.0000018V
7.0000009V
8.0000003V
8.9999991V
9.9999981V
10.9999975V

100V Range

10V,20V,50V,100V ->
9.999983V
19.999960V
49.999980V
99.999930V


1000V Range

100.00000V
249.99982V
300.00103V
400.00127V
500.00177V
600.00231V
700.00300V
800.00397V
900V -> Output Limit Fault, Output Undervoltage

Now i somehow want a 752A and a good aged 732A/B 

Edit: I suppose the Guildline 7520 is just an automated Fluke 752A + LTZ1000A + Nullmeter (but optical  )? -> http://www.guildline.com/new-products/released/7520-automated-precision-voltage-divider

[e61_phil]

I was interested in the 10V range. Therefore, I throwed the measurements in Excel.

[Dr. Frank]

Now i somehow want a 752A and a good aged 732A/B 

Your 5440B already serves as a (Super-) 732A/B, as it is well aged, and it contains an ovenized, double SZA263 reference. So no wonder, that your 3458A shows exactly 10V on its 10V range. Your instrument has a very good linearity of 0.25ppm INL, as calculated by e61_phil.

The ratio auto-calibration seems also to be ok, although your 3458A should have been nulled in its 100V and 1kV ranges, before you have taken the measurements. There's an offset of about -15µV in its 100V range, and about -740µV in its 1kV range, if you calculate the regression line of the different ranges.

These offset errors of the 3458A are directly visible, when you compare the 10V measurements in its 10V and 100V range.
That's already a -1.5ppm error between these readings, consisting of this offset error, and the autocal ratio error of the 3458A..

If you calculate the ratio error of 20V / 10V and 100V / 10V, measured in the 100V range of the 3458A, these are about 0.75 and 1.36ppm (with compensated offset), so the 22V and the 275 ranges are probably precisely auto-calibrated in the 5440B.
The 1100V range is a bit off at about 3.4ppm, if you calculate the 300V / 100V ratio error.
But these measurements in the 1kV range of the 3458A are all not well consistent, 800V already show -7,9 ppm deviation from a linear relation.
That may be due to the heating effect inside the 3458As divider, or may be the HV generation is not working properly at that voltage already, so you'd first have to repair this.

I also recommend to properly zero your 3458A, when you calibrate the 2.2V and 220mV ranges of the 5440B.

Frank

[Echo88]

I already thought about using the SZA-reference inside the 5440B as a good aged reference, but wasnt sure if using the reference-voltage via an extra cable would have an impact on the calibrator. Also: The 5440B draws considerable more current than a 732A/B (have to measure what it really draws and the resulting electricity-costs) and cannot be shipped to be calibrated to serve as "The Volt" in my home.
I didnt know about the 100V/1000V-offset in the 3458A, good to know for the next measurement. I did buy the 3458A mainly because of its unbeaten linearity and accuracy in the 10V-range, which interests me most. You are right about the nulling of the 3458A for those low divider-voltage, i wanted to check the drift/low-frequency-noise of my 3458A in those small voltage regions for a long time, to compare it to my Keithley 181 Nanovoltmeter. 

In the coming time i will clean the relays in the 5440B and maybe will make separate Pomona jacks and connections to the 5440B-SZA-reference. This could then maybe be used as my main-reference, which has a different drift characteristic (SZA drift up, the LTZ1000/A down...or just the opposite) compared to the 7x LTZ1000A that i have.

[Dr. Frank]

I already thought about using the SZA-reference inside the 5440B as a good aged reference, but wasnt sure if using the reference-voltage via an extra cable would have an impact on the calibrator.

That's not good an idea, because that would be a path for disturbances.
Also, it's not necessary at all, because the 5440B outputs a precise 10V already, due to its very linear D/A, and its precise gain auto-calibration.

Also: The 5440B draws considerable more current than a 732A/B (have to measure what it really draws and the resulting electricity-costs) and cannot be shipped to be calibrated to serve as "The Volt" in my home.

Well, for volt-transfer you could as well use your new LTZ1000.

I didnt know about the 100V/1000V-offset in the 3458A, good to know for the next measurement. I did buy the 3458A mainly because of its unbeaten linearity and accuracy in the 10V-range, which interests me most.

The best feature of the 3458A is its very precise autocal process, i.e. the 10:1 ratio transfer, which of course is a consequence of its ultra linear ADC.
This transfer is typically precise to about 0.33ppm per range, see hpj 4/1989.

But that may vary. It's mostly spot on, but sometimes it's off by up to about 1ppm.
That's simply the natural variance of this technique, including gain and offset errors.
 
I check that comparing the measured 10V in the 10V and the 100V range, 100V in the 100V and 1kV range, and analogously for the 2V and 220mV ranges.
Either there's an offset, that can be compensated, or it's necessary to repeat the ACAL procedure.

That way, you can afterwards check the 5440B ratio much more precisely by the 3458A.

You are right about the nulling of the 3458A for those low divider-voltage, i wanted to check the drift/low-frequency-noise of my 3458A in those small voltage regions for a long time, to compare it to my Keithley 181 Nanovoltmeter. 

The 220mV is not so precise at 3mV.. an additional precise Hamon 100:1 divider is better suited for that purpose.

In the coming time i will clean the relays in the 5440B and maybe will make separate Pomona jacks and connections to the 5440B-SZA-reference. This could then maybe be used as my main-reference, which has a different drift characteristic (SZA drift up, the LTZ1000/A down...or just the opposite) compared to the 7x LTZ1000A that i have.

Again, I would simply use the 5440B as-is, that's sufficient.

You may monitor the 5440B vs. the 3458A, and the LTZ1000 vs. the 3458A.
Then you have 3 references, and maybe you can decide, which one drifts...
The more LTZ1000 references you have, the better, but they usually have a negative timely drift.
The 5440B may drift upwards, but as it's off for most of the time, drift will probably be very small, same goes for the 3458A.


Frank