Resistance measurement

I'm trying to get a feel for the best option for measuring resistors. I'm looking to measure from 10 ohm to 105 ohm and hoping to be at 0.5% across the range. That means I'd want to measure +/-0.05 ohm on the 10 ohm end and +/- 0.525 ohm. One issue I think should be mentioned is that the circuit for measurement will be located about 1 meter from where the resistor is attached, and will have a relay, a cord and a couple sets of terminal blocks for connection. I have been looking into various methods, but am unable to get a feel for the resulting accuracy that can be expected from different methods. A couple of points that I'd like to gather some information on if anyone can provide it.

1. Is 4 wire necessary in this case? It seems the resistance isn't that low, and most examples I've seen are for milliohm measurements. Plus, the componenents between the electronics and the actual resistor to be measured may cause some issues (or maybe not). I'd always have wires / terminals and a relay between the op amp / adc inputs and the resistor. I use 4 wire measurement for very low values on a bench meter, but never for higher ones simply because of a higher accuracy requirement.

2. I will build calibration into the system for environmental changes. Suggestions as to the method to be used would be great (PGA, digital trim pots, etc.) Known resistors can be used as a reference.

3. I'm planning on using a microchip microcontroller, and some are available with 10 and 12 bit adc. If I have one with a 12 bit, and drop the last bit, i've got a 105ohm / 2^11 = 0.0512 V. It would seem that I'm right on the edge, but maybe by ignoring the last bit, I've given myself some room. Any thoughts on this line of thinking would be great.

If you have any suggestions or ideas about the best way to go about reaching this goal, they'd be most appreciated. Thanks for your time.

Yes, 4, 5 or 6 wire ohms is necessary, You can easily expect a few ohms of contact resistance. Traditionally, you force a current and measure a voltage across the resistor. Your voltage measuring points define the resistance.

Look here https://www.google.com/url?sa=t&rct...8jO6z_LkJeQ-fNLSHn0R0xw&bvm=bv.52164340,d.dmg

Here https://www.google.com/url?sa=t&rct...wcyfG2o68-Ef0P9HgoWHOYA&bvm=bv.52164340,d.dmg is a link to the 6 wire method.

The 5 and 6 wire methods depend primarily on ratios if I remember correctly.

This https://bit.ly/15o3lNi will get you the above and other links.

Yes, 4, 5 or 6 wire ohms is necessary, You can easily expect a few ohms of contact resistance.

Click to expand...

Would it not be possible to simply calibrate the contact resistance out? I will never be able to eliminate either the contact resistance or the terminal block resistance.

Traditionally, you force a current and measure a voltage across the resistor. Your voltage measuring points define the resistance.

Click to expand...

Of course. That's the way I plan to do it.

I understand using 4 wire to eliminate the resistance of the cable / leads. However, it seems possible to eliminate that through calibration. Place a known resistor in, measure it, then add or remove an offset in software to account for the lead resistance (and the terminals, solder joints, contact resistance, etc.) Then, store the offset and use it when measuring future units.

The current source accuracy is another source of error that I haven't put much thought into. Maybe a sense resistor in series would allow for a second adc channel to measure the current and adjust the final value accordingly?

I appreciate your thoughts and am interested in what you have to say about my calibration thinking. I'm going to draw up a schematic of what I think everything pre-measurement point would look like, including all possible resistances in the system. Maybe this will change my calibration ideas completely.

Sure, you can calibrate the lead losses out, but it gets to be a real pain. I've used Fluke meters with zero controls on them.

I also did 4-terminal measurements as my line of work as well as measurements into the pA region.

Forcing a current, even a mA when your voltmeter has an input Z of say 10 Meg ohms is negligible for low resistances.

It's also easier to get an accurate resistor ratio than a resistor. All you have to do is laser trim one of them. So a mated setwithin 0.01% or better is easy, but an absolute 0.01 resistor is harder.

There is another thread somewhere on ETO that discusses a low ohmmeter.

Also, remember that you can buy 4-terminal resistors.

KeepItSimpleStupid said:

Sure, you can calibrate the lead losses out, but it gets to be a real pain.

Click to expand...

Place known resistor into terminal blocks.
2. Take measurement.
3. Actual - Measured = Offset
4. All New Measurements = Measured + Offset
Did I miss something there?

I've used Fluke meters with zero controls on them.

Click to expand...

How do I use this information? I will only be using a 4-wire meter for validating the calibration resistor.

I also did 4-terminal measurements as my line of work as well as measurements into the pA region.

Click to expand...

Ok.

Forcing a current, even a mA when your voltmeter has an input Z of say 10 Meg ohms is negligible for low resistances.

Click to expand...

I think I follow. At 10 ohms, a voltage variation of 0.01V would result in 1mA. and the 1mA would result in a measured error of 0.01V, and you add this to the above (0.0512), and I'm out of spec. So I need to go to a 16bit converter. I think that is the point.

It's also easier to get an accurate resistor ratio than a resistor. All you have to do is laser trim one of them. So a mated setwithin 0.01% or better is easy, but an absolute 0.01 resistor is harder.

Click to expand...

You're talking about the current sensing resistor? Do you expect I'll need 0.01% for this appication? My variation would be most @ 105 ohm. This would correspond to near full scale on the adc. Say I'm using 5V. 0.1% variation in the resistor would cause an additional 0.1% error in the resistance (0.01ohm @ 10 ohms, and 0.105 ohm at 105 ohms).

There is another thread somewhere on ETO that discusses a low ohmmeter.

Click to expand...

I can't seem to search properly (get little to no results) since the update / upgrade. It's painful and I hope they fix whatever it is soon.

Also, remember that you can buy 4-terminal resistors.

Click to expand...

Yes, I am aware of them. Does this mean that you do indeed agree with the secondary measurement of the current source in order to add an additional offset based on variations in the source? Would that then remove the issue above resulting in the additional 0.1% error?

I don't know what your measuring, but an easy way to do 4-terminal measurements is with a Kelvin clip lead. That maes probing easy.

e.g. These: https://www.google.com/imgres?imgur...40yUtq6JcrE4AP1yYDYCg&ved=0CEcQ9QEwAA&dur=770

I invested a lot of time and effort into developing magnetic adjustable probes using modified standard kevin wafer probe bodies. One application was high temperature in a vibrating environment.

----

You can force whatever current your system can bear to be effective. If you want to use 1 AMP, then use one AMP. I don't have all of the constraints such as max V, max I etc. when troubleshooting one of the systems, I used two probes and zeroed the meter and then took a reading. That only compensates for lead length and not contact resistance. If you need a couple of current sources, then use a couple of current sources.

---

As I understand it, the alternate muti-wire techniques only depend on ratios of resistors used in making the measurement, but a traditional "sense resistor". , therefore the total accuracy of the current source used is basically irrelevent. That's the point. You do still need to measure to the required resolution.

Here is a better description of the 6-wire technique. **broken link removed** In one of Keithley's early service manuals the measurement is ratiometric. I just can't find that manual.

----

Kelvin clips look nice, and i may actually get a set for my meter. However, this needs to be a fixed mount type connection point. I think i'll try a terminal block with two connection pins to the pcb. i can run current to one, and the voltage measurement to the other.

jnnewton said:

3. I'm planning on using a microchip microcontroller, and some are available with 10 and 12 bit adc. If I have one with a 12 bit, and drop the last bit, i've got a 105ohm / 2^11 = 0.0512 V. It would seem that I'm right on the edge, but maybe by ignoring the last bit, I've given myself some room. Any thoughts on this line of thinking would be great.

Click to expand...

That is the accuracy of the ADC.. you'll need a very accurate current source in order to trust the measurements. The noise floor of ADC is roughly about 7.2 desibels per bit.. so 10 bit measurement has ~70 db noise floor. Close to 80 db in a perfect world.

jnnewton said:

3. I'm planning on using a microchip microcontroller, and some are available with 10 and 12 bit adc. If I have one with a 12 bit, and drop the last bit, i've got a 105ohm / 2^11 = 0.0512 V. It would seem that I'm right on the edge, but maybe by ignoring the last bit, I've given myself some room. Any thoughts on this line of thinking would be great.

Click to expand...

According to data sheets, depending on the processor, 12bit ADC has an accuracy around +- 20 digits. However, with good calibration and massive oversampling you can do much better than this - down to 0.5-1 digit.