Fluke SL935 Resistance Standard project (rare prototype teardown included)

[texaspyro]

The only time you might find an SR-104 going for less than about $2,000 is if it has been messed up by an idiot

Or was listed on Ebay by somebody that could not type.   I got my SR-104 for around $200.  The auction listing title was incorrect and nobody else bid on it.  doG loves dyslexics...

I asked IET about calibration.  I think that they originally quoted around $1100 and I did not respond.  After a few weeks they came back with an unsolicited $800 offer.

I have the SR-104 mounted in an insulated box along with a Thunderbolt GPSDO and a 10V reference.  I use the temperature control feature in Lady Heather to control the temperature in the box at one of the two temps where the SR-104 is at "exactly" 10K using the Thunderbolt as a heater and a PWM fan to control the temperature.  Seems to work rather well.  Temperature swing is less than 10 millidegrees and 24 hour RMS temperature is in the low tens of micro-degee range of the setpoint.  Of course, those are not absolute accuaracy, just a relative indication of stability.

If you have the cal data (ALPHA, BETA, R_DEV. TEMP_DEV) for your SR-104 (on the chart in the lid)  this program will calculate the resistance over a range of temperatures.  You need to change the #defines at the start of the code to match your device's cal constants and re-compile the program.

[ManateeMafia]

I send mine to Process Instruments. Their pricing is almost half of IET. I just sent mine today with a few other standards and they have been great to deal with on the phone and email. IET quoted me a similar price as yours a few years ago.

The pricing is fair and they can do tempco's if needed. They also work with lab built standards if you want it calibrated.

[TiN]

Aren't they beautiful together? SL935 is out for proper calibration now  .
Will be very interesting to see standard's alpha/beta values.

[TiN]

Okay, Fluke SL935 standard is now back from calibration at Process Instruments Inc. 

It didn't go 100% smooth, as unit arrived from CAL with LOW BAT light on and no IN CAL light. Perhaps calibration tec did not hook it back to AC power after the measurements.

It was shipped to calibration lab fully charged, with battery life minimum 120 hours. That was enough time for SL935 to receive calibration in "hot" state, without any temperature interruptions,
even if it was never plugged into mains. Shipping back from PI was 2-day service, which took just 52 hours from pickup to delivery.

Click for full PDF-report data, on both resistors.


My loved resistor data is available for amusement now, with amazing tempco. Thanks to integrated +35C oven.

I mean, really, just look at these alpha and beta coefficients. 
Worst error is +0.07ppm on 10 KOhm with temperature from +23.00 to +28.9C. 

Overall I think it was worthy calibration investment (bit over $1K total for cal+tempco measurement+shipping), as now I have solid resistance reference to take care of. Will be interesting to monitor it's long-term stability, as I do plan to send it for same cal next years as well.

As bonus, we can perform some math acrobatics, to attempt our own meters accuracy, using previously captured data.
Here are my own measurements which I did at home, prior to sending SL935 for calibration + ManateeMafia's meter:

HP 3458A, meter 2 over 10KOhm, direct 4-wire, OCOMP ON, DELAY 3, NPLC100 = 9999.971 Ohm, +26C ambient, 18 May 2017. Error from PI value : -0.37 ppm 
HP 3458A, meter 1 over 10KOhm, direct 4-wire, OCOMP ON, DELAY 3, NPLC100 = 9999.985 Ohm, +25C ambient, 20 May 2017. Error from PI value : +1.03 ppm   
HP 3458A, meter 2 over 10KOhm, direct 4-wire, OCOMP ON, DELAY 3, NPLC100 = 9999.965 Ohm, +24.5C ambient, 25 Sept - 28 Sept 2017. Error from PI value : -0.97 ppm  

Meter 1 calibrated by using Vishay VHP103 resistor at 10Kohm as transfer standard, which are in turn measured against 90-hour fresh cal'd ESI SR104 (by PI too) in August 2016. Meter is 24/7 powered.
Meter 2 calibrated directly vs same ESI SR104 back in January 2017, shipped to me in Feb and in 24/7 powered use since then.

HP 3458A, ManateeMafia's meter over 10KOhm, direct 4 wire, OCOMP ON, DELAY 3, NPLC100 = 9999.966, +26C ambient, 3 Nov - 7 Nov 2017. Error from PI value : -0.87 ppm 

Based on this brief data, with adding uncertainties and error factors, I'd have pretty good confidence stating that such transfer path can provide <8 ppm absolute resistance accuracy at 10 KOhm. Noise + short-temp variations are usually under 3 ppm.

Now similar data for 1 Ohm.

HP 3458A, meter 1 over 1 Ohm, direct 4-wire, OCOMP ON, DELAY 3, NPLC100 = 1.00006111 Ohm, +22C ambient, 5 May - 25 May 2017. Error from PI value : +2.1 ppm , noise about 30 ppm. 
HP 3458A, meter 2  + TE 9823 calibrator sourcing +100mA and -100mA over 1 Ohm, direct 4-wire, NPLC 100 = 1.00006657 Ohm, +23.5C ambient, 13 October 2017. Error from PI value : +7.5 ppm
HP 3458A, meter 2  + TE 9823 calibrator sourcing +20mA and -20mA over 1 Ohm, direct 4-wire, NPLC100 = 1.00006574 Ohm, +21C ambient, 21 Sept - 22 Sept 2017. Error from PI value : +6.7 ppm , noise around 2 ppm

HP 3458A, ManateeMafia's meter over 1 Ohm, direct 4 wire, OCOMP ON, DELAY 0, NPLC100 = 1.00005831, +24C ambient, 31 Sept - 3 Nov 2017. Error from PI value : -0.7 ppm    Noise is around 30ppm.

Looks like despite the large noise impact from the 3458A (because 1 ohm is just 10% of the minimum range), calibrated 3458A still able to resolve better than 10ppm accuracy for 1 Ohm resistors, if proper care is taken. Bodging external calibrator to source +/-current into 1 ohm resistor rendered worse absolute accuracy, so this remain to be investigated further.

I'd also like to thank ManateeMafia for great help with measurement, testing and support  .

[Vgkid]

That is impressive , thanks for posting that.

[babysitter]

It didn't go 100% smooth, as unit arrived from CAL with LOW BAT light on and no IN CAL light. Perhaps calibration tec did not hook it back to AC power after the measurements.

I think thats enough reason to ask for a partial refund. A full refund if you told them about the Battery.

[Macbeth]

The type of resistors inside are not likely to show any temperature induced mechanical hysteresis, and so I think this adds up to a big "nothing burger".  The calibrated value should be good after it heats back up to calibration temperature-- then let it sit for a day.

I disagree. The amount of work TiN put into this making sure the battery and charging circuitry would be perfect, ensuring the logistics, and I can only imagine a sick feeling to the pit of his stomach when after all that work and cost he finds the device returned with a LOW BAT and OUT OF CAL indication.

Yes, perhaps you can surmise that the SL935 contains resistors that "shouldn't" drift once they get back up to temp, so no harm done - but the cal technician is not the one to make that judgement at all and TiN went to best efforts to make sure it would remain powered all along - just like a 732B would be. Perhaps he didn't specify that the battery should be charged while in the cal lab, but surely this is standard procedure on any Fluke "732B like" device and plainly obvious?

It brings into question how you can trust this firm at all if the technician is so slovenly. How do you measure such uncertainties? 

I do think some compensation is in order - perhaps a free recal next year, with maybe only shipping to be paid for? 

[HighVoltage]

Those are some very impressive calibration data
Thanks for sharing the PDF

[TiN]

I'd agree with DiligentMinds.com that it should be fine, and this theory also backed by the actual measurement.
SL935 now is connected to MM's 3458A which just completed fresh calibration at Keysight last week.

After few ACAL's deviation is <1ppm already at 10K.

Macbeth
I did sent article covering desired calibration procedure to PI prior to shipping unit. I don't expect their resistor calibration tech calibrate prototype ovenized standards very often, so I'm not filing any complains or refunds at this time. Many labs may not even accepted obscure non-production unit for calibration in the first place (since they cannot be sure of performance or stability of unknown device).

IN CAL light is just driven by simple comparator, that will trigger if oven temperature go too much outside of +35C NTC reading. It doesn't really mean that it's out of cal.
Keeping constant temperature is more important for DC zener reference, and bit less so for hermetic film resistors.

So, "she'll be alright". Next step would be comparing it to freshly calibrated SR104 and SRL1  .

How do you measure such uncertainties?

Is indeed a bigger (and expensive) question there, as I have no any forecast for ability to measure even at this noise resolution, leave alone uncertainty.

[ManateeMafia]

So, "she'll be alright". Next step would be comparing it to freshly calibrated SR104 and SRL1  .

Sounds like the 242D is going to be put to use.

[dl1640]

VERY amazing re-work.

I wonder what kind of cabling PI used in this calibration
not sure if they did as you requested of using clean copper spade lug

next time you better send in your cables for combined calibration with the SL935

[Andreas]

Hmm,

what I do not understand at the 10000 Ohms resistor how they get that resolution:
8,9 nano Ohms! per deg C as alpha on a 10K resistor sounds incredible to me.

but in the formula they use the value obviously as 8,9 ppb / K.   

with best regards

Andreas

[CalMachine]

Hmm,

what I do not understand at the 10000 Ohms resistor how they get that resolution:
8,9 nano Ohms! per deg C as alpha on a 10K resistor sounds incredible to me.

but in the formula they use the value obviously as 8,9 ppb / K.   

with best regards

Andreas

That's got to be a typo.   I would think that the certificate not including a 10 kOhm reference, is also an error.

[Dr. Frank]

I'd agree with DiligentMinds.com that it should be fine, and this theory also backed by the actual measurement.
SL935 now is connected to MM's 3458A which just completed fresh calibration at Keysight last week.

Don't you  have before / after travel comparison data, to check whether your SL935 has the same value, or may have been altered by the loss of power?

Your VHP203 should be stable enough to identify such small deviations

Frank

[babysitter]

Do you know the conditions - e.g. temperature extremes, mechanical shocks - during transport? Also, it proves some level of "sloppy work" if they were aware but neglected it.

[ManateeMafia]

The box was shipped 2nd day. The packing was in the original box and showed no signs of external damage.

The temperature on the day of delivery was ~27C.

A mistake was made on my part because of the holiday delivery workload and my expected work schedule. I anticipated being away from home that day and depended on them to deliver it to my home where someone would sign for it. Since it was under battery power, I didn't want to hold it for pickup and not make it to FedEx that day. They ended up delivering it in the evening (8pm) when I typically get it around noon.

I hold all my standards and 3458A for pickup except the 732B. I usually hand deliver it to Tektronix and then drive there a couple of hours on the day it is picked up. This will definitely be an issue shipping the 732B to Fluke since they will probably ship it without asking for special delivery instructions. This is the first time I have had to receive a reference under battery power where it was unplugged for more than a few hours.

[ManateeMafia]

DiligentMinds,

I would think that there would be some mention about this potential issue in the Fluke MFC manuals. Power cycling a 57x0A calibrator should show signs of shifting since there are so many heated hybrids inside these units.

[Andreas]

Does anyone have any objective evidence that a hermetically sealed (or even a non-hermetic) thin-film (aka: metal-film) resistor exhibits any mechanical hysteresis due to small temperature excursions? 

Hello,

My measurements of (metal foil) VHP202 resistors show 1-2 ppm hysteresis over a 30 deg C temperature span:

https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg1143103/#msg1143103

Hysteresis gets much better when the temperature excursion is reduced to a 14 deg C span:
(with my setup I cannot measure it)

https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg1170353/#msg1170353

I have also some results of molded metal film (UPF50) which were not too bad (mostly below 1ppm over 30 deg C) regarding hysteresis on slow temperature gradients (0.12 K / minute).
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg462301/#msg462301

But if you let rest the resistors over night most of the hysteresis is usually gone the next morning.
Of course there is also some permanent (ageing) drift.

with best regards

Andreas

[TiN]

The magical resistor box today arrived back to lab, finally.

Second calibration report included in selfie:



Now it's time to determine error on my 3458A/2002's and transfer ohms to the SR104 and my other resistance standards.

Previous cal data for reference: