I am developing an inexpensive but reasonably accurate meter for measuring resistance in the milliohm range. The three previous posts are listed in the related links at the bottom of this post. In this post a working prototype is presented that has milliohm accuracy down to one millivolt. A schematic and test data are presented along with plans for the next steps.
That heading above is supposed to be a pun :-). Much of the work since the last post is described in the comments of that post where a change was made to using the LM334 as a temperature compensated current source. The design includes several suggestions from shabaz which worked out very well. Shabaz also did a number of tests of the temperature coefficient of diodes and BJT transistors in TO92 packages which were used in the design. The design calculations for a 10 mA current source using a spreadsheet developed for the LM334 are shown below:
A MCP6N16 instrument amp with 100x gain was then added as shown in the following schematic:
For now, power is USB from a wall wart. There is an inline jumper so that a current meter can be inserted. Provision is made for Kelvin probes as shown in the middle box far left. A transistor is used to reduce power stress in the LM334 and a number of "Do Not Populate" resistor pads added to better allow trimming along with a potentiometer. The MCP6N16 is in the middle of the sheet and uses the recommended resistors for 100x gain in the datasheet. A useful aspect of the 100x gain is that with a 10 mA current source the output voltage reading can be directly converted to ohms (e.g. a 50 mV reading is equivalent to 50 milliohms). Finally, there is a User Interface box at far right which for now consists of a bench multimeter.
The schematic was moved off of a breadboard and pieced together on a prototyping PCB placed in a cardboard box for further testing:
Since I only have one LM334 and MCP6N16 they are socketed for easy removal. The resistors are mostly SMD and soldered between pads on the back. All other components are soldered. Testing was done on what appears to be 26 AWG wire which I stripped off of something. A photo of the test apparatus is given below:
The prototype in its box is just below the multimeter in center. The 4-wire Kelvin probes are seen connected to it. The meter is connected to the output of the instrument amp by the red and black probes far right and the reading in millivolts is being displayed. The procedure for the tests was to measure the length of wire on the tape and then connect it to the Kelvin probes and take a reading. The wire in the photo is approximately 60 mm in length and the reading is 5.71 millivolts which converts to 5.71 milliohms.
The results of the test are given below:
First observe the measurement of the 0.1 Ohm known resistor at the bottom of the table. The measured 99.72 milliohms is within the tolerance of the resistor and very close to what I have measured in previous 4-wire tests using a bench power supply as the current source. Now observe the table which is the measurements and calculations done on various lengths of the 26 AWG wire. The first column is the measured length of the wire in meters. The first value in the second column labeled Expected mOhms is the actual measured milliohms. All the subsequent values are derived from it by calculating its value if it had exactly the same resistance per unit length as the first measurement. The third column is the actual measured value on the bench multimeter. The remaining columns are calculated values.
The fourth column, perhaps mislabeled as Error, is the difference between columns 2 and 3. The fifth column is the calculated error assuming column 2 is correct. And the 6th and last column is the calculated milliohms per meter for each piece of wire derived from the measured length and measurement from the multimeter.
I was surprised at how good the results are. Even a 9 mm length of wire, expected to have a resistance of 0.8 milliohms, was measured to be 1.2 milliohms - only a 0.4 milliohm difference. All other differences were less than or equal to 0.1 milliohm. The final row with a length of 0 meters was obtained by clamping the probes to each other. So, there is a roughly 1 millivolt (milliohm) offset from zero.
The biggest issue at the moment is extending the range. It is currently 0 - 4 ohms and I would like to extend to at least 40 ohms and preferably 100 ohms or a bit more. This could be done by reducing the current from the current source are reducing the gain. Without a microcontroller to adjust or some additional circuitry this would mean the resistance would not be directly output in the same range as the output voltage however. It is also slow to settle but eventually quite steady.
- Add a LDO to the power section - probably 4.7V
- Have a look at behavior on the scope
- Add an on / off switch
- Add LED to indicate status of power
- Consider how to increase the range to at least 40 V
- Make a decision on the output - bench meter, panel meter, or microcontroller with screen
- Design a PCB and send for fabrication
- Get a proper enclosure