Repeatability when measuring resistor

[Conrad Hoffman]

Regarding interference, you might keep some large ferrite toroids around and try a few wraps through them to kill pickup. I know this works for my precision capacitance measurements, but haven't tried it with resistors.

[Tryer]

I would never expect any precision measuring a 300 Ohms resistor in 2W mode. There are so many sources of error. You have the cables, the connectors, the front-back switch, the relays don't know what else. Your ppm results are a miracle.

Certainly they are high by the cable resistances (at least 0.1 Ohm = 333 ppm of 300 Ohms). In case you have two more cables, you can check this by repeating the measurement in 4W mode.

Regards, Dieter

Hi Dieter and thanks for your remarks,
it is only about repeatability and not accuracy. I am looking at the noise of that instrument and the effect of the relay switching. My PPM results are also refered to repeatability. They only say that after e.g. 100 samples my standard deviation is abt 150e-6 Ohm and this is 0.5ppm of the (end)value. Where is the miracle?
My quetions were about the effect: i have some resistance in the relay contact which is not stochastic, and decays as if it were a discharging curve of an RC combination, with no irregularities and repeatable. Yes, in a 4 wire constellation it will be suppressed more or less.  I will check that and report. The problem is that i have to remove the bridges below the relays.
Greetings, Tryer

[dietert1]

There are mercury wetted relays like coto 3400 and 3500 series. For most other relays a contact resistance instability of 10 mOhm is absolutely normal and the front-back switch may develop contact resistance of 1 Ohm unless you operate it every now and then. The problem arises mainly for contacts with small currents ("dry switching").
Mechanical contacts may be good during the first years but later they become unstable. This applies to potentiometers and fuse holders as well. Even many connectors have this, e.g. IC sockets. Instead of buying new equipment you just have to move the switches and re-seat connectors.

Regards, Dieter

PS: Your observation about slowly relaxing error may be caused by out of range measurement.

[Tryer]

PS: Your observation about slowly relaxing error may be caused by out of range measurement.

T have checked that by interrupting the connection from the outside (pulling out the banana connector). It comes immediatly to the end value.
What confuses me, is the fact that it decays smoothly. And it happens only if the relay contacts participate.
regards, tryer

[dietert1]

Ok. that means your banana connector is better than the front/back switch. If you operate that switch 20 or 30 times to clean the contacts, the effect may vanish. Really, you cannot expect ppm measurements in 2W mode, except with resistors above 100 KOhm or so.
Recently i happened to measure the old mains switch of a HP 3456A DVM. Switching dry (or almost dry) it had between 2.5 and 1.6 Ohms. I assume that resistance also vanishes after power-on.

Regards, Dieter

[Tryer]

Look how it looks like. Every time the same. You must imagine the 3457 measures in a loop once per second.
Then, at the point of the jump, it had switched to rear for 200ms, switches back to front and continues measuring in the same loop.
That means i have a contact with ON-resistance  x mOhm, jumps after interruption to  6+x mOhm and then slowly back to x mOhm in such a smooth way. At least i would expect wild small jumps that happen so fast for a very small time-not to capture with an fs of 1 Hz. Why does it take so long ?
But OK, i have so many things that i cannot explain, one more or less doesn't matter.
Thanks again and regards 

PS  : And of course i have tried to clean the contacts switching the relay with a current of up to 1 A, many tens of times.
As i told the other relay has a similar behaviour but reduced by a factor of 3.

[MegaVolt]

Perhaps these are some kind of thermoelectric processes. Time is needed to equalize the temperature.
It is also possible this is a consequence of thermal heating of the measured resistor. After all, a current is supplied to it and it begins to come into equilibrium with the environment.

[Tryer]

It is also possible this is a consequence of thermal heating of the measured resistor. After all, a current is supplied to it and it begins to come into equilibrium with the environmen

If the current is interrupted by pulling the connector out and reconnecting again this effect does not happen. It comes directly to the end value. So, it cannot be the thermal heating of the resistor.

[Anders Petersson]

One year later... Now I have bought a Vishay VHP101 100 ohm resistor (Y4078100R000T9L), which should be the best in its price class with the advertised "typically 0 TCR". (I'm aware these specs can be debated.) I mounted it like the previous Vishay Z series (Y1453100R000V9L) in a shielded metal box. I have also upgraded from Keithley DMM6500 (6.5 digit) to Keithley 2010 (7.5 digit).

K2010 settings: Four-wire mode, offset compensation, 5 NPLC plus 20 samples averaging (12 seconds in total). I use scan card 2000-SCAN to alternate between the two 100 ohm resistors.

A three day run was telling. At first, there's a big dip for 1.5 hour while the K2010 warms up. The specs say the warm-up time is 2 hours so this is ok (although much slower than DMM6500).

After that, there's a clear correlation between the readings of the old Z series (yellow) and new VHP101 (green) resistors. The VHP101 starts out at roughly -6 ppm and ends at -1 ppm. The Z series starts at +2 ppm and ends at +7 ppm. I blame the K2010 for this overall +5 ppm drift. K2010 specifies the 24-hour accuracy as 15 ppm of reading + 9 ppm of the range. While this specification isn't the exact same thing as drift, I guess it's an indication of what to expect. (DMM6500 specifies 20+20ppm so almost the double).

I haven't calculated the short-term noise yet, as was the original topic of this thread. Logging the temperature could be helpful too, but so far the drift continues unabated so I don't think it's due to temperature variations over the day.

EDIT: The K2010 short-term noise is about +- 1 ppm peak-to-peak for both channels, which is roughly half the noise I originally measured with DMM6500.

[Kleinstein]

With many DMMs the accurcy of the ohms current source is limited and may show quite some 1/f noise. This is also reflected by the often relatively poor accuracy in the ohms ranges compared to the voltage. Often a singel current source circuit is used and this is more like a compromise over the large range. So the lowest and highest ohms ranges are kind of compromises.

For really accurate resistor measurments there are special resistor bridges.

[Anders Petersson]

Thanks Kleinstein, the numbers in the K2010 datasheet agree with you. Here's the accuracy in ppm for the voltage and resistance ranges:

Voltage:
100mV     37+9
1V        25+2
10V       24+4
100V      35+5
1000V     41+6

Resistance:
10ohm     60+9
100ohm    52+9
1kohm     50+2
10kohm    50+2
100kohm   70+4
1Mohm     70+4
10Mohm   400+4
100Mohm 1500+4

Resistance measurements are about double the inaccuracy and go off the chart in the high ranges.

The unexpected conclusion of this experiment is that both resistors are useful as reference for my other measurements. I'm not likely to splurge on a dedicated resistance bridge. (But with test equipment addiction looming, who knows...)

[dietert1]

Most of these DMMs use a constant current source and measure the resultant voltage to determine resistance. And they don't have different test currents for each resistance range. You can use another meter to check what is the test voltage when measuring a 100 ohm resistor with the 2010. I think you will be surprised and understand why accuracy is 5 digits and not 7 digits.

Regards, Dieter

[Anders Petersson]

I don't need to test -- the current is specified in the datasheet. Both DMM6500 and K2010 have the same: 1 mA for both range 100 ohm and 1 Kohm.
1 mA over 100 ohm is 100 mV. That suggests that the accuracy of the 100 mV range applies. The K2010 100 ohm range (52+9ppm) has an additional 15 ppm 1 year inaccuracy compared to 100 mV (37+9ppm). Perhaps this is reasonable for the additional current source to measure ohms.
DMM6500 however specifies 30+35ppm for 100 mV and 85+20ppm for 100 ohm, so it has a bigger discrepancy between the two ranges than K2010.

Just interesting to ponder.

[dietert1]

The interesting fact is that these meters measure 100 mV to 5 digits or worse, although they are called 7.5 digits. The explanation is that avoiding thermal EMF and drift at a level of some uV or below is expensive (speak: missing). Even specialized instruments like the micro-ohm meter 34420A don't do it much better although it would use 10 mA for the 100 ohm resistor.
The drift you logged was a factor 10 below measurement uncertainty and not significant.
Your arrangement with two high quality resistors was a first step towards making a bridge. I got a SR1010 from ebay with 12x 1 KOhm precision resistors and wired it as a bridge to characterize a 500 Ohm hermetic precision resistor and demonstrate its faults at a 0.01 ppm level. It's in the thread with resistor temperature coefficient measurements.

Regards, Dieter

[Anders Petersson]

The interesting fact is that these meters measure 100 mV to 5 digits or worse, although they are called 7.5 digits.

Let's see.. K2010 1-year specs for 100 mV was 37 ppm of reading + 9 ppm of range. I suppose this could be called 5.5 digits. But even 3458A offers 9 ppm of reading + 3 ppm of range (100 mV), so then 3458A would count as 6 digits. As long as we realize that DMM resolution doesn't equal accuracy I'm ok with the drift. I think short-term noise is a more serious problem. My +- 1 ppm noise limits the K2010 to 6 usable digits.

I'm also annoyed that calibration labs consider any reading within the 1-year tolerance as 'pass' without warranting adjustment. A freshly calibrated unit then has an uncertainty of the 1-year tolerance, and then it's supposed to drift for a year until calibration again.

The drift you logged was a factor 10 below measurement uncertainty and not significant.

Well it was significant for my understanding.  I now understand that drift is a fact of life even for an aged instrument, making a stable reference all the more important.

Your arrangement with two high quality resistors was a first step towards making a bridge. I got a SR1010 from ebay with 12x 1 KOhm precision resistors and wired it as a bridge to characterize a 500 Ohm hermetic precision resistor and demonstrate its faults at a 0.01 ppm level. It's in the thread with resistor temperature coefficient measurements.

It's a 45 page thread... the closest post I found was this and it doesn't go into detail: https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg2700574/?topicseen#msg2700574
By "bridge", are you referring to a wheatstone bridge? How would that work?

[dietert1]

In my understanding an uncertainty of 50 ppm means 1 part in 20 000, so this would be 4.5 digits of resolution.

I think your way to of using two precision resistors to attribute an observed drift to your instrument instead of a precision resistor is the same idea as using a bridge. In a bridge circuit you can assert balance and precision by exchange of elements, thereby removing unknowns.
Example setup: https://www.eevblog.com/forum/metrology/hp-3456a-terminal-block-and-wires/msg2759646/#msg2759646.
Results: https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg2751598/#msg2751598 and https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg2767686/#msg2767686.

Regards, Dieter

[MegaVolt]

https://www.allaboutcircuits.com/technical-articles/hardware-techniques-of-linearizing-resistive-sensor-bridges/

[Kleinstein]

The calibraion uncertainty is essentially allways quite a bit high than the resolution - it has to. Often digits in addition to the cal uncertainty also can be used, e.g. to see short time changes of the DUT.
Still the erros and noise in the resistance mode of the usual DMMs is quite a bit higher than with votlage. It starts with often measuring a small voltage like 100 mV, which is not the best range with most meters. The 7 digit meter may well be only 6 useful digits for the 100 mV range.

There is additionally the uncertainy in the current source: In most cases this includes 1 ref. resistor and 1 resistor pair to transfer the reference voltage to a higher reference point. The Keithley2002 is an exception here with a charge pump for this. There are than usually 2 OPs that can contribute an offset. Of these one is often non AZ a high impedance version to also cover the µA level current. So this OP can contribute quite some low frequency noise.

Some of the old Keithly 19x meters measure resistance with a kind of simplified bridge circuit and not a constant current source. So they may actually be relatively good with resistance, although the performance can vary (different from the current source case) over the range.

The bridge circuits for precision resistor measurements are usually more like a Kelvin bridge or similar. So one can compensate lead resistance and use 4 wire mode. An alternative configuration (especially low resistance) uses a current mirror. For the high resistors (e.g. >> 1 M) the better way is to apply a constant voltage to the resistor and than measure the current - so more like a conductance measurement.

[Anders Petersson]

In my understanding an uncertainty of 50 ppm means 1 part in 20 000, so this would be 4.5 digits of resolution.

That can be a useful measure but it's better to call it accuracy or drift-free resolution. Like Kleinstein said, the extra resolution is still good for something.

I think your way to of using two precision resistors to attribute an observed drift to your instrument instead of a precision resistor is the same idea as using a bridge. In a bridge circuit you can assert balance and precision by exchange of elements, thereby removing unknowns.

Thanks, yes I agree it's the same principle. This removes a big error term. Some remaining terms are DMM nonlinearity and noise.