Group Actions
Engagement
  • Author Author: fmilburn
  • Date Created: Date Created
  • Views 7045 views
  • Likes 9 likes
  • Comments 51 comments
Related
Recommended

Even More on Current Sources and a Kelvin (4-Wire) Milliohm Meter

fmilburn

Introduction

This is the third post on the development of an inexpensive but reasonably accurate meter for measuring resistance in the milliohm range. In the first post a simple current source was described that created a 10 mA current  across a resistor that allowed the voltage drop to be measured using a multimeter and the resistance calculated.  A number of helpful suggestions were received and I ordered additional components based on that feedback.  In the second post a block diagram for the instrument was introduced and initial measurements were made with a microcontroller using the built-in ADC.  Some, but not all of the ordered parts have been received now and this post will update progress as I don't want John's popcorn to get stale.

 

A Change to the Design Objectives

I originally specified that the current to the DUT would not be greater than 10 mA.  Testing to date has indicated that meeting the desired accuracy will be difficult without amplification of the voltage difference across the DUT which adds some complexity and cost.  Accordingly, the specification is being changed to 100 mA across the DUT.

 

Component Status

First, I have to admit to making a mistake in the orders.  The MCP6N16 instrument amp comes in three versions with different minimum gain.  I wanted the version with minimum gain of 1 and ordered the version that has a minimum gain of 100.  Doh!  Always read the datasheet carefully.  For now I am substituting the MAX9619.  I also ordered a precision LDO voltage source from the TI store which has not been shipped yet.  Usually they are pretty quick. The volt meter I plan to use is still in shipment from China.

 

100 mA Current Source

This is the revised circuit, the only real changes being the addition of a MOSFET to handle the increased current and a new precision 0.1% 10 ohm resistor to set the current.  I am using an inexpensive ANENG multimeter to measure voltage but it does agree well with my bench meter.

image

And here are the results:

image

The tests are being performed the same way as previously using a coil of wire that has been center tapped.  The measured resistance of the full length of wire is 0.092 ohms as seen on top while the measured resistance of half the length is 0.046 ohms - exactly half.

 

Next Steps

The inexpensive voltmeter needs at least 4.5 V to operate so I will probably use either 4 x 1.5 V AAA batteries or USB power and a precision voltage source to set the current.  If I decide to use a microcontroller instead of a voltmeter then a 3V3 LDO will be used to power that.  The parts for Kelvin probes are on order.  Progress depends on the postal service now...

 

Past Posts on this Topic

More on Current Sources and a Kelvin (4-Wire) Milliohm Meter

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

  • in reply to jw0752

    Hi John,

    The other thing I was thinking is maybe use my hot air rework station as it can be set down to 100 C.  If I pursue this further I should probably figure out a way to do it in still air like Shabaz.  Maybe baffles and mixers to still and  cool the air or something like that.

  • in reply to fmilburn

    Hi Frank,

     

    Interesting stuff! I didn't try a hairdryer, but I noticed that even just lifting the bowl causes large jumps, and I concluded (concluded is a strong word, it was really just a guess) that different parts are heating/cooling at different speeds, since the TO92 are thick plastic, and the resistors are tiny) and so in the setup I had, I found it hard to make many conclusions in free air. But I guess since you are sustaining the hairdryer heat since you have been able to clamp it into position, this was good enough for you to get a reasonably steady measurement of 0.017 to 0.01mA/degC. Even with that level, if your ambient temperature changes by 10 degrees C in your lab, you're still in the ballpark to measure resistance with just +-0.5% error. But I reckon it will improve more once it is sealed from drafts, for me that played a huge difference, and of course any tweaking of resistance ratio - I may have got lucky that the selected values worked well enough with the breadboard contact resistances. Also, are the resistors carbon film or something? Because those can be 300 or 500ppm/degC, so that too could be having an impact.

    In your PCB layout, the PN-junction and the LM334 need to be close too, since they have to be at the same temperature, I'm wondering if unexpected drafts are flowing differently around the two devices when they are not enclosed (or maybe I'm overthinking this!).

  • in reply to fmilburn

    Hi Frank,

     

    I don't know if this is of any interest but a while back I was doing some temperature testing and I got tired of fighting with getting the temps where I wanted them and then moving them slowly in one direction or another. I took and old hot air gun and hooked its fan up so that it would run at a constant speed. I then used a small variac and powered the heater with it. I was able to vary the temperature up and down from room temperature to 340 F and stop at any point and remain more or less stable. I did not build a thermocouple into it but it could be an option. Instead I just calibrated the variac against empirical temps. I have to use an external thermocouple if I want to know exactly where I am at. It isn't perfect but it was a big improvement on trying to get readings as the temperature changes under a conventional warming or cooling.

     

    John

  • in reply to fmilburn

    Hi - wanted to make sure you saw this.  I also put a summary at the bottom of the comments.

     

    I am somewhat embarrassed to put my temperature test alongside yours but here it is.  I mounted a hair dryer on my PCB vise and pointed it towards the setup from a safe distance and on the low heat setting.  Another meter was set up to read temperature with an open air sensor alongside the components.  Temperature was observed to bounce back and forth 1 degree C during the test.   The test was done twice with similar readings both times.  Here is the setup.

    image

    It behaved in a strange manner.  The current at T= 0 is 10.008 mA.  Temperature in open air at the location was immediately varying back and forth 37-38 C and stayed there.  Current started dropping very quickly and then started to recover after maybe 15 seconds.  At approximately T = 30 seconds it was 9.983 mA having recovered somewhat.  After a few minutes it slowly came back up to 9.996 mA where it stayed for some 10 minutes until I stopped the test after 15 minutes.  This is a mess of open air stuff but perhaps still useful.  The different components make the behavior nonlinear and they actually seem to compensate for each other after a while.  So it varied from about .017 mA per degree C to 0.01 mA per degree C.  Not as good as your results - maybe the resistors or just the crazy act of blowing hot air on it.  Still, not too bad.  Notice that the voltage reading across the DUT varied only between 10.02 to 10.04 mV for these three readings.  Funny stuff.

  • This is a summary of experiments with the LM334 Adjustable Current Source as the comments are getting pretty long and possibly difficult to follow.  See the running dialog above for greater detail.  Shabaz has provided links to interesting information on an improved variation of the LM344 circuit and the advantages of using BJT transistors as a PN junction for temperature compensation.  He also ran a very interesting series of experiments to better determine the temperature coefficient of various common transistors as well as the diode recommended in the LM344 datasheet.  The forward voltage was also measured.  My experiments to date are consistent with cruder than his although and I have not run detailed experiments on temperature influence.

     

    Here is a circuit for a milliohm meter, modified using the information above, and using the LM334 to set current.

    image

    The forward voltage drop across the 2N3904 transistor and the values of R1 and R2 in series with R3 are used to set the current through the load.  The tempco of the 2N3904 must also be known in order to properly offset temperature effects from the LM334.  The trimming potentiometer R3 allows error in the measured current to be corrected.  Spreadsheets were developed to make the calculations which are described in the LM334 datasheet easier as shown below:

    image

    My current experimental setup on a breadboard looks like this:

    image

    During the test the following were relatively constant:

    • Input Voltage from Lab Power Supply: 5 V
    • Ambient Temperature: 19 deg C
    • Vf = Vd = 0.7183 V
    • R1 = 17.66 Ohms
    • R2 = 123.54 Ohms (includes trim pot, R3)

     

    The small meter at left is measuring the voltage in millivolts across the DUT which is a 1 ohm 1% resistor.  The larger bench meter behind is measuring current in milliamps.  Since the current is very nearly 10 mA and the voltage drop is very nearly 10 mV the setup is performing properly for a 1 ohm resistor.  The current was remarkably steady over the desired input voltage range of 3 to 6 Volts:

    image

    The test setup was then left to run for one hour exposed to ambient with no observed change in the current.  Tests by Shabaz in the comments above better illustrate the temperature sensitivity.