Group Actions
Engagement
  • Author Author: fmilburn
  • Date Created: Date Created
  • Views 5027 views
  • Likes 12 likes
  • Comments 16 comments
Related
Recommended

Testing Current Sources for a Kelvin (4-Wire) Milliohm Meter

fmilburn

Introduction

I was inspired by a recent post from on Building Kelvin (4-Wire) Test Leads.  Shabaz explains in detail and with clarity why measurement of small resistances with the two leads on a multimeter is difficult.  This got me to thinking about how I might build my own 4-wire Kelvin instrument.  This post will describe initial tests of a simple current source that can be used with a digital multimeter to more accurately measure small resistances.

 

Method and Objectives

For accurate measurement a precise current source using two wires to connect to the device under test (DUT) must be created. Separate wires are used for voltage measurement.  I have a good bench multimeter (thank you again element14 and community for selecting my project and enabling me to buy this instrument with the prize money) which will serve as the voltage measurement instrument in this test.  I am considering building a self contained battery powered instrument and if so will replace the multimeter with either a small panel meter or a microcontroller and display.

 

Here are the design objectives:

  • Inexpensive
  • Can run off batteries
  • No more than 10 mA to DUT
  • milliohm level accuracy
  • Range 1 milliohm to 10 ohms

 

Design

To start, I chose a simple circuit using a single op-amp to create as a current source to see how far it would get me.

image

The op-am is a  dual rail to rail TLV2462a I had on hand with 500 uV typical input offset voltage.  The supply voltage can be from 2.7 to 6 V single supply so it will work well with batteries and a microcontroller.  The temperature coefficient of the input voltage is 2 uV per degree C so it should be fairly stable.  The second op-amp on the IC may come in handy later.

 

The remaining circuit components and values were chosen as follows:

  1. 3V3 power rail - this needs to be precise since along with a voltage divider it will set current.  I am using my bench power supply for this test but will use a voltage regulator in a final design.
  2. 0.3 V common mode op-amp input voltage - This was selected as it is well above the lower level for input and output voltage and is also at a convenient level for setting the current.  It provides room between the lower and upper voltage output levels for a range of resistance measurements.
  3. 10K and 1K values for the voltage divider - arbitrary 1% values I had on hand which set the voltage to the non-inverting input of the op-amp at 0.3 V.
  4. 30 ohm current setting resistor - since the non-inverting input is 0.3 V, the op-amp will do it's best to keep the inverting input the same due to op-amp action with negative feedback.  Through Ohms Law the current is 0.3 V / 30 ohms = 10 mA.

 

Construction

Here is the circuit on a breadboard:

image

Power comes in from the left out of my bench power supply and is supplied to the rails of the op-amp.  The resistors for the voltage divider can be seen lower left on the breadboard and feed the non-inverting input through the yellow wire.  I measured the voltage at 300.3 mV.  The salmon / pink colored wire is the output and goes to the DUT - a coil of orange wire.  The blue wire leads to the inverting input.  I did not have a 30 ohm 1% resistor and used 33 ohm and 330 ohm in parallel instead (measured resistance of 30.04 ohms).  Since the measured voltage across the current setting resistors was 300.5 mV and the measured resistance was 30.04 ohms the current is as close to 10 mA as I can get with my equipment.  The multimeter readings are steady.

 

Testing

I tested a range of 1% resistor values I had on hand from 30 ohms down to 1 ohm.  All were within the accuracy of the resistors and matched my multimeter well.  I followed the example of Shabaz and used wire for the final test.  I believe it to be 26 AWG telephone wire and it is approximately 1 meter in length.  I "center tapped" it so that I could measure from end point to end point or from the center to either end point.  The following photo shows the results:

image

The left photo reads 0.90 millivolts across the entire coil of wire.  Since the current is 10 mA the measured resistance is 0.090 ohms.  In the picture on the right the center tap to right end is being measured at 0.44 millivolts which is a resistance of 0.044 ohms where we expect 0.045 ohms.  I got identical results using the other end of the wire and the center tap.

 

I also measured the resistance of the wire using the multimeter.  When the leads are shorted together the meter reads 0.09 ohms.  On the center tap to end measurement the meter reads 0.20 ohms so if I back out the leads I get 0.11 ohms which is more than twice what was measured using the Kelvin method and the current source.  The spec sheet states that the meter has 0.01 ohm resolution with accuracy +/-(0.3%+40)+test leads open circuit value when the range is 400 ohms.

 

Summary

I was surprised at how well this simple current source performed.  I believe that this breadboard setup is already performing better than my multimeter on its own.  Should I decide to continue developing it I will consider doing the following:

  • Use a microcontroller with built in op-amps
  • Second stage amplification to increase the range
  • Select an enclosure with panel and design a PCB

 

Thanks for reading.  Suggestions for improving this project and a way to calibrate would be greatly appreciated.

Parents

  • You'll find that the op amps built into micro-controllers have pretty dismal characteristics.

     

    If you are happy with the floating load (ie neither end of the resistance you are measuring is connected to ground) then the things that affect the accuracy of the measurements are the reference voltage, the resistors (R1 to R3) and the offset voltage of the op amp.

     

    If you use a chopper (zero drift) type of op amp the offset will be below 20uV and the drift of offset less than 60nV/C (Microchip MCP6N16 - not the best but pretty cheap). If you don't mind calibrating out the scale error caused by the current being to high or low then you just need low temperature drift in the resistors. 10ppm/C parts are affordable, 1ppm (and better ) can be obtained but get kind of pricey.

    In your design the reference voltage 300mV will result in error of 1.2e-6/0.3*100 = 0.0004% due to a 20C temperature drift of offset of the Microchip amp (a typical in processor op amp STM32F301xx is 83x worse, giving 0.03% drift error). If you used 10ppm resistors they would add 20ppm * 20 ppm drift (for 20C temperature change) = 400ppm = 0.04%. (bog standard 100ppm resistors would take this up to 0.4%).

     

    From the above, the resistors are much more important than the op amp, so you could get the drift against temperature very low by using good (10ppm resistors ),  a zero drift op amp would be probably be overkill but using an external op amp will give you a much bigger range of processors to choose from. If you use the same reference voltage for the current source and the ADC you measure the voltage drop with then the two will track and the absolute accuracy of the reference doesn't matter. Otherwise you will need a good (10ppm) voltage reference which will be quite pricey.

     

    You could use perhaps 1V across R2 and scale R3 accordingly which would reduce the effect of op amp offset and offset drift by a factor of 3.3 - allowing you an even better choice of cheapo op amps.

     

    If I were building this as a a feature in a bit of custom test equipment I would go for precision reference and resistors if it got round having to calibrate. It would be possible to get perhaps to within 0.2% by dead reckoning but would need a 0.05% resistors, the zero drift amp, and a 0.1% reference.

     

    If you don't mind adjusting to compensate for errors in the resistors then you can put a pot at the junction of R1, R2 and the op amp. The pot will have a stinky temperature coefficient but you only need it to cover 2.5% of the range if you use 1% resistors - so 400ppm/C for the pot would be OK.

     

    You could calibrate by using an accurate DMM to measure the current, or using a precision resistor and measuring the voltage drop (or if using the combined reference for the ADC and current source, it has to be the precision resistor.)

     

    MK

  • in reply to michaelkellett

    Thanks Michael,

     

    You should be teaching this stuff (maybe you already are).  From my standpoint that was an excellent review and critique of the design - described in a manner I could easily follow.  Nice and very informative analysis of the contributors to error.  Also thanks for advice on calibration.

     

    RE: microcontrollers - I don't have any background in analog and to be honest not much with microcontrollers.  But like they say, when your only tool is a hammer every problem looks like a nail.  And besides, whenever I see a microcontroller with new capabilities something in the primitive part of my brain says "oooh… new shiny object.... must have".

     

    I am going to take your points and incorporate into a new post on the design.  One question though... I was aware of the floating load and felt it was OK when I started looking at this.  I should reconsider before getting too far however.  From a practical viewpoint, when would a floating load be unacceptable in such an instrument?

     

    Frank

Comment
  • in reply to michaelkellett

    Thanks Michael,

     

    You should be teaching this stuff (maybe you already are).  From my standpoint that was an excellent review and critique of the design - described in a manner I could easily follow.  Nice and very informative analysis of the contributors to error.  Also thanks for advice on calibration.

     

    RE: microcontrollers - I don't have any background in analog and to be honest not much with microcontrollers.  But like they say, when your only tool is a hammer every problem looks like a nail.  And besides, whenever I see a microcontroller with new capabilities something in the primitive part of my brain says "oooh… new shiny object.... must have".

     

    I am going to take your points and incorporate into a new post on the design.  One question though... I was aware of the floating load and felt it was OK when I started looking at this.  I should reconsider before getting too far however.  From a practical viewpoint, when would a floating load be unacceptable in such an instrument?

     

    Frank

Children
  • in reply to fmilburn

    I don't think there is any problem with the floating load topology for a hand held meter. Even if the meter is connected to other equipment it doesn't need to be a problem - modern DMMs have input circuits that are isolated from the control and interface connections so that approach will also work with a floating load current source.

     

    In some systems it can be preferable that the device under test and the measuring circuits share a common ground (I've recently designed a laser driver where the laser diodes have their cathodes grounded) but it makes it much more expensive (in precision resistors) to drive the current into the load. To see how it's done, Google "Howland Current Pump" - but I don't think it's necessary for normal resistance measuring.

     

    MK