Incredible 3458A Stability!

[CalibrationGuy]

I was talking to the head of Keysight's Roseville, CA standards lab and he shared some numbers with me regarding our 3458A that is there for calibration. These are the numbers he shared:

Calibration: 03/2024 - 10.0000254 10VDC Range
Calibration: 11/2025 - 10.0000319 10VDC Range

That's 0.65ppm in 20 months! I am really impressed by those numbers. Stunned, actually. The guys who designed this meter were legends. Well, needless to say, I'm never selling this meter!

TomG.

[Kleinstein]

The 10 V performance of the 3458 is essentially from the LTZ1000 reference. In the 3458 the reference runs rather hot and thus usually drifts quite a bit when running 24/7.  When not used 24/7 that can be hysteresis / humitidy effects that may take several weeks to settle. This could have effected one of the 2 calibrations. So the value of the 2 readings is limited. There is true long term drift but also short term effects depending on the use in the last few weeks before.

[CalibrationGuy]

The 10 V performance of the 3458 is essentially from the LTZ1000 reference. In the 3458 the reference runs rather hot and thus usually drifts quite a bit when running 24/7.  When not used 24/7 that can be hysteresis / humitidy effects that may take several weeks to settle. This could have effected one of the 2 calibrations. So the value of the 2 readings is limited. There is true long term drift but also short term effects depending on the use in the last few weeks before.

I've never turned the 3458a off.

TomG.

[CalibrationGuy]

Also, have been constantly comparing it to the bank of voltage references (30min after running ACAL).

TomG.

[Kleinstein]

The drift of a LTZ1000 ref circuit is a mix of drift of the actual reference and the resistors around it. Getting the very low dirft value, chances are this is from compensation between the 2 parts. The LTZ1000 chip alone is expected to have a higher drift due to the high temperature.

So while stable in the past, it may well get worse in the future as the 2 parts may well have different charackteristics.

With a good reference to compare too, repeated comparison on a weekly / monthly scedule would be more telling than 2 calibration points.

[CalibrationGuy]

I agree, that's why we have it connected to the reference 732c. I've checked it at least 2-3 a week for the last year now (except when it's at Keysight, where both units are now).

TomG.

[exe]

What is a "usual" drift a hp3458a?

[Grandchuck]

I have no idea what a usual drift 3458A behaves like.  Mine has an uncertainty of about 3.5 ppm on 10V DC according to the documentation that came with the meter, which seems about right according to my observations over a several year period.

[MiDi]

My unit from '89 running 24/7 has similar drift around -0.5ppm/a.
That is a typical drift for LTZ1000.

[CalibrationGuy]

What is a "usual" drift a hp3458a?

The one year spec for the ten volt range is plus or minus eight parts per million without the high stability option.

TomG.

[aronake]

i leave one of my 3458a that was calibrated with 1ppm uncertanity in 2016 unadjusted since then. It now show around 1.5 ppm lower at 10V than it did by the time of calibration.

Its from 1998 and used to serve as reference meter in the inhouse calibration laboratory of a test equipment manufacturer. It has pretty much been running constantly since the 2016 calibration. Cal 72 has moved around 0.01 PPM / day over the last year. It seems to be a quite unusually stable 3458a.

For the other 3458a I have, i do not have good enough data to know their long term drift, but working on that.

[miro123]

I was talking to the head of Keysight's Roseville, CA standards lab and he shared some numbers with me regarding our 3458A that is there for calibration. These are the numbers he shared:

Calibration: 03/2024 - 10.0000254 10VDC Range
Calibration: 11/2025 - 10.0000319 10VDC Range

That's 0.65ppm in 20 months! I am really impressed by those numbers. Stunned, actually. The guys who designed this meter were legends. Well, needless to say, I'm never selling this meter!

TomG.

Traceable and accredited calibration reports gives uncertainty. You should not be worry if the total  uncertainty is order of magnitude higher that the drift rate

[mendip_discovery]

Mine,
Feb 23, 44 uV, 10.0000248
Feb 24, 30 uV, 10.0000010
Jun 25, 15 uV, 9.9999953

3 calibrations 3 labs. So a little unknown for now.

[CalibrationGuy]

I was talking to the head of Keysight's Roseville, CA standards lab and he shared some numbers with me regarding our 3458A that is there for calibration. These are the numbers he shared:

Calibration: 03/2024 - 10.0000254 10VDC Range
Calibration: 11/2025 - 10.0000319 10VDC Range

That's 0.65ppm in 20 months! I am really impressed by those numbers. Stunned, actually. The guys who designed this meter were legends. Well, needless to say, I'm never selling this meter!

TomG.

Traceable and accredited calibration reports gives uncertainty. You should not be worry if the total  uncertainty is order of magnitude higher that the drift rate

The uncertainty for that measurement was quoted at twenty eight nanovolts!

TomG.

[exe]

How measurements are done for calibration? I assume there is some averaging over some time. When I do a measurement, I take an average value over, say, 2-5 minutes to average out noise and "motorboating" drift. I also monitor stdev, if it's above 600nV, I discard the measurement and do it again.

What does the metrology say about that?

[mendip_discovery]

This is the settings from my Cert for the 3458A, as done this year by KS UK.

[Kleinstein]

For testing the 3458 to very high resolution, it may make sense to no only use long integration like 500 PLC and may be some extra averaging, but also repeat the test with a new ACAL call. There is also a little noise in the internal gain test and a new ACAL cause may give slightly different gain, just from noise there. The CAL72 value has not just drift, but also some noise.

[drhex]

Something strange has happened to mine:
Entry Date                         TEMP?       CAL72                      Drift to Day1     Av since      Rel last test
                                                                                                                Day 1
1   20.11.2025      35,30   1,00084219            
2   21.11.2025      34,90   1,00084204   1   -0,149873778   0,07494   0,149873778
3   22.11.2025      35,50   1,00084219   1   0,000000000   0,00000   -0,149873800
4   23.11.2025      35,80   1,00084223   1   0,039966341   0,00999   -0,039966341
5   24.11.2025      35,30   1,00084196   1   -0,229806459   0,04596   0,269772789
6   25.11.2025      35,10   1,00084192   1   -0,269772800   0,04496   0,039966350
7   26.11.2025      35,70   1,00084170   1   -0,489587674   0,06994   0,219814933
8   27.11.2025      36,00   1,00084153   1   -0,659444622   0,08243   0,169857031
9   28.11.2025           36,30   1,00084145   1   -0,739377304   0,08215   0,079932734
10   29.11.2025      37,30   1,00084055   1   -1,638619971   0,16386   0,899243332
11   30.11.2025      37,50   1,00084018   1   -2,008308623   0,18257   0,369689258
12   01.12.2025      37,50   1,00084005   1   -2,138199230   0,17818   0,129890868
13   02.12.2025      37,30   1,00083999   1   -2,198148741   0,16909   0,059949639

Sudden increase in temp and drift, sitting on my desk in a non temp controlled environment (but that is relatively stable in the winter), shorted, 170 nV at rest - any thoughts?

[aronake]

It looks like its getting hotter in the end. Some 3458a get more drifty when hot. Not much hotter though, so maybe not too probably nut one idea. Maybe humidity changed? Let it run for a couple of weeks more, log every day and see what happens. That makes it easier to make conclusions.

[Grandchuck]

@drhex

Check the fan filter.

[drhex]

Thanks! First thing I did, this still has the old cup style filter which is free. Will keep monitoring and report back in a few weeks.

[drhex]

OK, have to eat my words, it only looked clean. Took the filter off and blew it through, results below. Seems like my TC isn't particularly good...

Day  Date                                  TEMP?       CAL?72                    Drift rel day 1     Av drift      Drift rel last
1   20.11.2025   12:00:00   35,30   1,00084219            
2   21.11.2025   12:00:00   34,90   1,00084204   1   -0,149873778   0,07494   0,149873778
3   22.11.2025   12:00:00   35,50   1,00084219   1   0,000000000   0,00000   -0,149873800
4   23.11.2025   12:00:00   35,80   1,00084223   1   0,039966341   0,00999   -0,039966341
5   24.11.2025   12:00:00   35,30   1,00084196   1   -0,229806459   0,04596   0,269772789
6   25.11.2025   12:00:00   35,10   1,00084192   1   -0,269772800   0,04496   0,039966350
7   26.11.2025   12:00:00   35,70   1,00084170   1   -0,489587674   0,06994   0,219814933
8   27.11.2025   12:00:00   36,00   1,00084153   1   -0,659444622   0,08243   0,169857031
9   28.11.2025   14:26:00   36,30   1,00084145   1   -0,739377304   0,08215   0,079932734
10   29.11.2025   16:00:00   37,30   1,00084055   1   -1,638619971   0,16386   0,899243332
11   30.11.2025   11:16:00   37,50   1,00084018   1   -2,008308623   0,18257   0,369689258
12   01.12.2025   20:00:00   37,50   1,00084005   1   -2,138199230   0,17818   0,129890868
13   02.12.2025   09:50:00   37,30   1,00083999   1   -2,198148741   0,16909   0,059949639
14   03.12.2025   20:30:00   37,80   1,00083925   1   -2,937526045   0,20982   0,739378929
15   04.12.2025   23:00:00   34.6   1,00084081   1   -1,378838756   0,09192   -1,558691868
16   05.12.2025   10:37:00   35.2   1,00084076   1   -1,428796682   0,08930   0,049957995

Also, sitting at 70 nV shorted again which went up to 200 nV earlier. Much happier.