Building a resistance standard.

[Traceless]

Hi everyone,

I recently received a bunch of parts, among those were also a couple of peculiar resistors, peculiar because they are pretty large, axial metal cans and so far nobody could identify the vendor. However some of those resistors have pretty tight tolerances. The most precise one is the 10kOhm, 0,001% resistor shown in the attachment. By the looks of it the resistors seem to be quite old.

Gyro suggested in the other thread to use the reistor to build a resistance standard. I think that is a good idea. The reason I'm moving the discussion here is the question:

Should I just put the resistor in an ABS box slam some Aliexpress banana jacks on and call it a day or do I want to put a bit more thought into the packaging?

Here is a picture of the resistor and the measurement results I already posted in the other thread.

[Gyro]

Personally, I would go for a diecast box and decent (not exotic) banana jacks from a local distributor rather than Ali. The alloy case will slow down the effect of short term ambient temperature changes on the resistor.

At 10k it's debatable whether it i worth using Kelvin (4 wire connection). You could always fit banana screw terminals instead so that you could feed current leads under the screws terminals and plug sense jacks into the tops.

[Mickle T.]

... and so far nobody could identify the vendor.

It is strange that no one could recognize the logo of RESISTA GmbH, a division of the famous manufacturer ROEDERSTEIN, created in Germany by Ernst Roederstein in the mid-20s. last century.
I still use these resistors (albeit in the regular, non-sealed version).

[Roehrenonkel]

Hi Traceless,
 
10k 0,001% is a good start for a reference-box. I envy you - my best one is 0,005% tolerance.
The vishay S102C come just in a limited range (50 Ohm to 100 kOhm) and are "not cheap".

For your box i'd take metal and use Kelvin-connection (see photo of my R-box).
Maybe i'll glue four PT100s onto the sides, top and bottom and connect them series-parallel
onto a Lemo-receptacle to check the inside-temperature.

Best regards

[Traceless]

@Mickle T. nice one, that took a while to solve the mystery around the Resista logo. Its also interesting to see the inside of one of those. Albeit I'm always sceptic of cracking such old parts with ceramic coating. If they used beryllium oxide you really don't want to create any fine dust/chipping.

@Gyro: Okay die cast/metal box and better jacks sounds resonable.

@Roehrenonkel: Thanks for the pictures, the box looks good.

Can anyone recommend good jacks, which ones did you use (@Roehrenonkel)? Another question in this context is would I solder the resistor to the jacks - which leads obviously to a better connection but not sure about what appling heat does to the inside of the resistor. So I might better use a pressure-type connection with washers/screws. I have no clue how sensitive those older resistors are to temperature. I would imagine ambient temperature is not a big deal because they are sealed. But applying 320°C to the leads would immediately heat up the core which probably hasn't seen temperatures over 30°C in decades.

[Roehrenonkel]

Hi again Traceless,
 
since i got me the special tools (PB1200/70 and /100)
for tightening the round nuts, i use mostly Stäubli Multi-Contact.
BUT this works just if one has access and room behind the face-plate.
Sometimes i have to use the hex nuts.

Surely one could use pure copper terminals (CNC makes it possible).

Solder the resistors, they have solder-lugs for a reason. ;-)))
You can always use a plier / heatsink between soldering-point and resistor-body.

Best regards

[Traceless]

@Roehrenonkel thanks for sharing. I also performed the inductance measurements you were curious about on the 10 kOhm 0.001% resistor, the 3.6k and 3.75k (0.02%) resistors and the 817k (0.02%) resistor. For the 817k one I didn't get any stable inductance readings. The other ones stabilized somewhere between 200Hz and 1kHz. Here the results:

[DavidKo]

For non DC standard I will use construction like in YHP/HP/Agilent 04285-61001 https://groups.io/g/HP-Agilent-Keysight-equipment/message/105764 and SMD foil resistors from Vishay/Alpha Electronics.

[bastl_r]

Hi

Should I just put the resistor in an ABS box slam some Aliexpress banana jacks on and call it a day or do I want to put a bit more thought into the packaging?

Never use low price Ali for this Resistors!
You will get many fault by EMF.
Recetly i bought any dualplugs from ali. For precise measurements they are useless.
The minimum for a referencebox are goldplatet jacks from Stäubli, Pomona and similar manufacturers.

Regards

[trobbins]

I'm not too sure about the inductance measurements.  I would have thought an impedance versus frequency sweep would have been more enlightening, to identify the phase shift with increasing frequency, and to identify a section of frequency where phase shift was about 90 deg and hence a simple lumped L value could be calculated, prior to onset of first resonance.

[guenthert]

Hi

Should I just put the resistor in an ABS box slam some Aliexpress banana jacks on and call it a day or do I want to put a bit more thought into the packaging?

Never use low price Ali for this Resistors!
You will get many fault by EMF.
Recetly i bought any dualplugs from ali. For precise measurements they are useless.
The minimum for a referencebox are goldplatet jacks from Stäubli, Pomona and similar manufacturers.

Regards

    Do as you see fit, but I like to point out that early (and some still used) resistance standards didn't bother with low thermal EMF connectors.  Many digital resistance meters offer a function where during a measurement that EMF is first determined before a current is applied, so that it can be calculated out.  Resistance bridges generally offered to invert the measurement current.

[Traceless]

Thanks to everyone for the feedback on the enclosure design and connectors.

@trobbins: I don't have the equipment (VNA) here to perform a proper sweep. Not sure if this information helps but I used my function generator and oscilloscope to measure the .02% 3.6kOhm resistor. Channel 1 (yellow) shows the original signal from the generator Channel 2 (blue) the same signal across the resistor. At about 447kHz (which is well beyond the 200kHz my LCR meter can measure) we see ~90 degree phase shift and also quite some attenuation. At about 245 kHz, the signal appears to be slightly amplified. If I'm not mistaken I'd need another (non-inductive) known value resistor of equal or higher precision in order to calculate the inductance with this function generator/oscilloscope setup.

[EC8010]

My Hameg 8118 LCR meter only goes up to 200kHz, but it's perfectly adequate for showing what a wirewound resistor looks like. As mentioned earlier, you need a sweep of Z against frequency, to which you then fit a model. Your resistor appears to be multi-layer, and from measuring Mann Components LT450/C (also multi-layer) I can tell you that a simple model of resistance and inductance won't fit well. Multi-layer wirewound resistors really are for DC only.

[trobbins]

I just use an audio soundcard and REW software to do impedance sweeps up to 96kHz.  That made it easy to assess the manganin coils in a vintage 4-decade resistor box, and I could easily then 'compensate' the inductive impedance of the Units and Tens resistor elements by adding a shunt capacitor across each step element to restrain phase shift to less than 1 deg out at 90kHz.  The 100 and 1k elements were effectively 'capacitive' so couldn't be easily compensated.  These were long thin diameter coils of uniformly wound folded-back wire, so not the same coil construction method as your reference.

As indicated, multi-section wound parts of what appears to be scatter wound sections are likely to have multiple resonances and may only have a valid simple R-L model up to only lowish frequency, but I guess that depends on what frequency you want to test up to.

[Traceless]

My Hameg 8118 LCR meter only goes up to 200kHz, but it's perfectly adequate for showing what a wirewound resistor looks like.

I'm not familiar with the Hameg 8118, if it supports a sweep mode that is awesome. Unfortunately with my LCR meter I can only switch between a fixed set of measurement frequencies (40Hz、50Hz、60Hz、80Hz、100Hz、120Hz、150Hz、200Hz、250Hz、300Hz、400Hz、500Hz、600Hz、800Hz、1kHz、1.2kHz、1.5kHz、2kHz、2.5kHz、3kHz、4kHz、5kHz、6kHz、7.5kHz、10kHz、12kHz、15kHz、20kHz、25kHz、30kHz、40kHz、50kHz、60kHz、75kHz、100kHz、120kHz、150kHz、200kHz). So basically I can only collect data points but not sweep over an entire range.

So get it "pretty good," as far as materials, and let the temperatures stabilize, and then always assume you're going to need to reverse the excitation current and average the two readings...

So far the resistor doesn't even seem to be that sensitive to ambient temperature change I assume because it is sealed in the metal shell.

That's easy to do on a bridge - my meter "zero" (the midpoint) often slowly shifts around as I flip the generator polarity back and forth. I balance for the midpoint, rather than worrying about zeroing the meter or trying to eliminate the EMF. At balance, the needle doesn't move:
-0 + TEMF == +0 + TEMF.
I think digital meters do that, too.

My meter has an auto zero function (basically you measure a short and open leads). As long as the ambient temperature stays roughly constant and I don't swap the leads zero point calibration works pretty good. Of course If I swap the leads or ambient has changed I need to repeat the process.

As mentioned earlier, you need a sweep of Z against frequency, to which you then fit a model.

I just use an audio soundcard and REW software to do impedance sweeps up to 96kHz. 
...
As indicated, multi-section wound parts of what appears to be scatter wound sections are likely to have multiple resonances and may only have a valid simple R-L model up to only lowish frequency, but I guess that depends on what frequency you want to test up to.

Because of trobbins original comment:

I would have thought an impedance versus frequency sweep would have been more enlightening

I came up with the scope/function generator setup as workaround for the missing sweep feature of my LCR meter. Basically since we know R = 3600 Ohm, and we can measure the phase angle for any given frequency with the scope we could load the data into a spread sheet and calculate Z for all datapoints. Unfortunately my scope is not up to the task. First data recording with the software does not work. Also while the scope is good enough to calculate a phase angle at an arbitrary fixed frequency, when performing a sweep there are glitches where the phase angle suddenly jitters significantly even though frequency steadily increases. I recorded a short demo showing the problem here. This might just be an averaging/on-screen display issue but without working data logging I don't know for sure.

So currently I'm limited to measuring Z/phase angle at any of the frequencies listed above with my LCR meter. However I'd be curious to know if such a measurement setup works better with higher tier scopes.

[EC8010]

The Hameg 8118 doesn't do a true sweep, it has 69 different frequencies between 20Hz and 200kHz. The BK981 has 200 frequencies between 20Hz and 300kHz and is 0.05% uncertainty rather than 0.1%. But even the 69 frequencies is good enough for most components. The advantage that these LCR meters have is that they use a lock-in amplifier. What that means is that they have a pair of multipliers looking at the current forced through the DUT, one fed with sine of the applied frequency, the other with cosine. The output of each multiplier is DC plus 2f. But because the two multipliers operate 90 degrees apart, those DCs are A + jB. So far, so good. But if you low-pass filter each DC with (say) a 1Hz low-pass filter, you have effectively synchronously demodulated your original signal with a bandwidth of 2Hz (1Hz either side of the carrier). And that means you don't get much noise, whereas your oscilloscope unavoidably has a rather large noise bandwidth, which is why it doesn't measure phase terribly well. If you know the voltage applied, and have determined the current and its phase, it becomes easy to express the impedance in whatever form seems appropriate.

If you take A + jB impedance data to a spreadsheet, you can fit it with a model, and when you do so, you effectively average across all your measurements, further reducing random errors (but not systematic errors). The long and the short of it is that an LCR meter that works by doing an Ohm's law experiment (rather than being a true bridge) can be very good indeed if you can get its data out to a spreadsheet. You can spot whether they do an Ohm's law measurement by the fact that to do so, they must make a four-wire connection and have four BNCs on the front. It seems to be an informal standard that the BNC centres are separated by 22mm.

You may be able to make your existing LCR meter do what is required...

And to answer your first question about connectors, one of these (meant for measuring thermocouple voltages) has twelve low thermal EMF terminals on the front:

https://www.ebay.co.uk/itm/256470899285?itmmeta=01HTT3ZXCK6QMXZ8WZ66Z53PDE&hash=item3bb6db5a55:g:vkIAAOSwk7VlpCM9&itmprp=enc%3AAQAJAAAAwMCh0uJx1ygUYbk4q3azM3Gsn9l1eUaox4t6I79HIU8LSwNgLndLUGfqutstYI7H57XB0ufGWkm%2B3c%2FeS%2BcgKPHdSp%2Br%2BzeLTf4x9zBYhDMhb%2FR4Hjv94ypUz5nyF%2FbU1ByNThj4wmBgH7JVgquERmc%2BKMSRVkjiBYkodyMGtmWPqO0U74YLBZA1VF6UEuj7j2HHt0aB2y6gjTXrSuCHKxqr%2B3Wlb%2BIgZlwvtdo5%2FC2Bhv%2BnZmLK0uTd5gdv%2B20VYg%3D%3D%7Ctkp%3ABk9SR7bW_8PWYw

I shall be using the box from mine for a low noise bench supply.

[mawyatt]

I came up with the scope/function generator setup as workaround for the missing sweep feature of my LCR meter. Basically since we know R = 3600 Ohm, and we can measure the phase angle for any given frequency with the scope we could load the data into a spread sheet and calculate Z for all datapoints. Unfortunately my scope is not up to the task. First data recording with the software does not work. Also while the scope is good enough to calculate a phase angle at an arbitrary fixed frequency, when performing a sweep there are glitches where the phase angle suddenly jitters significantly even though frequency steadily increases. I recorded a short demo showing the problem here. This might just be an averaging/on-screen display issue but without working data logging I don't know for sure.

So currently I'm limited to measuring Z/phase angle at any of the frequencies listed above with my LCR meter. However I'd be curious to know if such a measurement setup works better with higher tier scopes.

The new 12 bit low cost DSOs from Rigol and Siglent might help by using the built-in Bode Plot function, and very reasonable price wise. The Siglent Bode Plot implementation is quite good and useful, of course you may need a AWG unless the DSO has such built-in. Results aren't going to be as accurate as a proper LCR meter, but finding the equivalent reactive part of the DUT might prove sufficient.

Here's a link that might help.

https://www.eevblog.com/forum/testgear/admittance-measurements-with-dso-awg-with-bode-function/msg4491952/#msg4491952

Best,

[Traceless]

@EC8010: Thanks for the detailed explanaiton and the reference to the low noise connectors. So far the supported frequencies of the MCR5200 have also been more than enough for my purposes. It does have datalogging capabilities so I could even log measurements, but would either have to switch through the frequencies manually or write a program to do that for me. Either way I'd of course "only" get measurements at the predefined frequencies.

@mawyatt: The new Siglent/Rigol scopes are indeed tempting. Since both are rather new it might be worth to wait a while till all the early quirks have been fixed. Also currently I wouldn't know how to choose between the two product lines, both seem pretty capable. If I understand EC8010 correct though for this particular experiment (measuring phase of a wirewound resistor) scopes are generally not a particularly good tool. I do own a 60 MHz 2-Channel AWG, which I used to perform the sweep in the experiment demo I linked above.

[trobbins]

You could always clip a capacitor across ref resistor and repeat the sweep, and deduce if the resistor is net inductive or capacitive.  If it is net inductive then a SOT shunt capacitor may well suppress any phase shift until quite high up in frequency.

[EC8010]

Traceless: Your 5200 has (if I've counted correctly) 36 different frequencies and is based on an Ohm's law experiment. You should be able to get quite a reasonable result from it. I fitted BNCs to a die-cast box that hangs from the front of my Hameg 8118. (I added extra support for the larger box to avoid straining the BNCs.) The capacitance of the supplied four-wire leads gets converted into an effective 100nH inductance, so it's better to do the measurement close to the instrument in a screened enclosure. Small four-wire crocodile clips are available (Muller, I think). If you line the box with foam, this will prevent your DUT getting close to the screened enclosure and reduce stray capacitance. If you add folded tin-plate screens between the wires and their crocodile clips, it becomes possible to reduce stray capacitance between them to 5fF - enabling measurement of little capacitances.