3458A's Current shunts/resistance current source tempco study

[TiN]

Since I got my hands on damaged 3458A's A1 PCB decision was made to test critical resistors used on it for current/ohm functions to evaluate their TCR and eventually long-term stability performance.
I'll be updating this thread as the measurements proceed. Unless mentioned otherwise, all measurements to be taken with 3458A, @NPLC100, OCOMP ON with 5 sec delay.
One of the reasons why I want to go into this topic is to study possibility of pimping up 3458A's ohms and current performance and long term stability, for which I need to get baselines.
Let's get started.

Reference 40K resistor



This resistor is used as stable standard for ACAL calibration and perhaps as transfer point in calibration purposes. Hence stability of this resistor is crucial for 3458A's ohm/current specs.
Older A1 PCB have non-hermetic 40K resistor (more information wanted), while newer versions are equipped with custom-ordered Vishay PG #301031 VHP hermetic resistor. It's labeled 40.000K 0.1%.

TEST 1 (Ambient-Ambient)

To get initial TCR test going, I soldered resistor to 4-wire banana plug adapter and connected to DMM.
In this test both resistor and DMM are in free ambient, and temperature delta created by room AC.

RAW CSV-data.

Graph (LIVE version on click):



Temp-Resistance deviation:



Box TCR = +2.74 ?/K

DMM was run thru ACAL on HOT/COLD points. I expect a caveat though, as there is high probability that 3458A's own 40K drifts similar magnitude, thus obscuring test result.

TEST 2 (Ambient-Thermostat)

Test 2 using same DMM, but this time with resistor located in TEC thermostat, controlled by Keithley 2510 within 0.01 °C setpoint.
DUT resistor located in copper box filled with oil and Keithley 2001 used to measure RTD temperature, which located in same chamber.

[TiN]

Test methodology details placeholder

[TiN]

Temp chamber setup placeholder

[TiN]

Results summary and analysis placeholder.

References to other related publications

[Dr. Frank]

Hi Illya,

I assume this A1 board is from 1989, as the 2nd line on the resistor should be the date code?

Then the 40k resistor is the original one, i.e. a hermetical sealed type, specified T.C. of 1.3ppm/K (see CLIP).
The actual Vishay technology of that time were C- or K- foil types, and had T.C.s of +/- 2 or  +/- 1 ppm/K.


You have to run the 3458A in a stable environment (+/- 0.1°C), so you may well measure the DUTs T.C, independently.
Does your 3458A already contain the VHP101 resistor?

Anyhow, I'm very curious about your results.

Frank

[zlymex]

Nice topic, expecting more.

3458A is superb for 10V but not so good for ohm ranges because the current sources are small and they display only 7.5 digits.
The performance of current ranges is even worse because they use poor shunt resistors(except that 40k). It requires frequent ACAL which is not possible for continuous measurement.
In my opinion, at least foil resistors should be used for all current shunts. HP/AG even use their own brand(PR, red color) for 10mA range.
 

[montemcguire]

Those red 'PR' resistors look like metal films from Precision Resistive Products, not an internal HP brand. http://prpinc.com/ I've run across a few as surplus, and they seem to be conventional but well made film resistors, and they seem to be able to do high precision films and probably customs too.

[zlymex]

According to the component-level manual(CLIPS, page 6), the manufacturer of R211 is 28480, which is HP.

It could be CLIPS is out of date and HP/Agilent swap manufacturer as they often interchange those film resistors in different versions.

[RobK_NL]

Most probably, those resistors are in fact made by PRP, but selected by HP.

[Dr. Frank]

In HP BOMs, custom specific parts (but produced from other manufacturers) were mostly specified as "28480".

ICs in these cases carry a HP part number, even for ordinary OpAmps or logic, then they may be simply selected parts.

HP did never produce metal film resistors.

When they manufactured DCV calibrators (e.g. the HP 740A) in the 1960ties, they produced precision wire wound resistors on their own.. as I have a collection of hp marked resistors from that era.
It's mentioned here: http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1965-05.pdf

Frank 

[ap]

Replacing the 40k by a VHP 101 and others by better resistors will certainly help. However, improving the 3458A data sheet values (, accuracy over time/aging, drifts) formally so that it can be used in traceable (not even ISO certified) cal docs will be another story. It seems quite impossible to do that, not knowing the exact propagation and thus impact of any such design change.

[zlymex]

I do have some red HP wire-wound resistors, and I've seen one photo inside a HP standard resistor showing two red resistor with PFR*** marking, this again mislead me.

[Dr. Frank]

Replacing the 40k by a VHP 101 and others by better resistors will certainly help. However, improving the 3458A data sheet values (, accuracy over time/aging, drifts) formally so that it can be used in traceable (not even ISO certified) cal docs will be another story. It seems quite impossible to do that, not knowing the exact propagation and thus impact of any such design change.

That's an interesting question.

In fact, HP already improved the 3458A over the years, at least by assembling the VHP101 from about 1995 onwards, giving probably 0.3ppm/K and < 1ppm/yr. stabilities.
But they never updated the datasheet, and also the Transfer Accuracy for Ohm mode is still missing. (Can be found for Fluke 3458HFL)

If you list the DCI datasheet parameters and the T.C.s of the different resistors side by side, you easily see, that both were correlated.
The 40k resistor is related to "T.C. With ACAL", the other resistors relate to "T.C. Without ACAL".
Bias current and thermal offset voltages mainly determine the "ppm of range" parameter.

See table below, as-is, and possible improvement.


So any improvement in the resistors specification will directly improve the datasheet.

To get a reliable, metrological clean characterisation, one would simply apply the GUM:
http://www.nist.gov/itl/sed/gsg/uncertainty.cfm

You may determine A-type uncertainties, by measuring the STD of the different DCI ranges.
For the systematic  B-type errors one would measure or estimate the range-to-range transfer uncertainty of the 3458A, and the real T.C. by changing its inner temperature.

An external precision current source would be required.

For ACI, a comparison of the frequency behaviour of the old and the new resistors could be done, and then the A-type and B-type uncertainties, of course. That's much more sophisticated, than DCI.

Frank

[TiN]

ap
My plan currently stands as testing all A1's resistors from bad board, evaluation their impact on different functions, and then repairing board and using as test vehicle in my 3458A.
By the time it's ready, I'll have suitable 10V and 10K source, and both of my 2002's calibrated (planned in August already), so I can open up 3458 and mess with it's guts without worrying about calibration.

Dr. Frank as always on top of it.

I do agree that HP back then was remarking everything, even jelly-bean LM339 and 74HC00 logic chips. Parts which the selected/aged specifically inhouse - they assigned HP logo/branding.
I saw such "HP" resistors in HP 11177B calibration module for 3455A together with ovenized 5.1V zener reference.



I have done little article about it on my site about year ago.

And yes, my A1 PCB in 3458 is equipped with VHP101 (Board is Rev.D, manufactured in 1996, by datecodes on ICs).

[quarks]

thanks a lot for sharing
looking forward to see your results