I've no real background in electronics, but i'm jumping in the wrong end of the pool and trying to make a milliohm meter.

I've been using LTSpice to model the ideas and in this process, trying to understand what's going on. Below is where i'm at and the system is working (according to LTSpice):



However, clearly battery V1 is taking all of the test load and will eventuially discharge at a higher rate.

I've been able to get the system to almost work with a single battery, but once the DUT starts getting down around 70mOhm the opamp starts bottoming out and hitting its output offset voltage limit and the output sits up at around 70mV.



I've looked at a negative rail supply and have modelled an LTC660 to give the op amp a negative supply, but LTSpice really shows a noisy output that's mostly accurate.

Could I ask for some guidance on how I might either get a two-battery system to work or get the op amp -ve supply just low enough to get me to 0V output?

 

Okay, so I think i've worked it out. I've used an unbalanced voltage divider and a follower (is that the right word?) to feed both the inamp and its reference pin. I found that if the divider was balanced, the low end of measurements (at 1mOhm for the DUT), I was getting inaccurate gain. I think this +ve offset, this is a good thing because i'm only working in the 0-2V range.

According to the simulations, i'm getting accuracies ranging from 0.65% at the 1m Ohm measurement end down to 0.015% at the 2 Ohm measurements. Interested in anyone's thoughts on how to improve this.

Current draw is around 1.5-2mA when not in use, so i'll put a switch on it. LCD screen also to be integrated.

The resistor network I've used R1-R7 is just me thinking about how i can get the parts I need to actually build this. Any ideas on how i better achieve an accurate outcome here without costing the earth?

 

Multiple resistors in parallel (R1-R7) aren't needed for simulation purposes.
Where are you going to buy a 75.1245575 Ohm resistor?
If input offset is an issue, perhaps trimming components to cancel that could be added.
Please post your asc file to help forum members to help you.

 

Thanks @Alec_t,

Only new here and getting used to the way things run. My background is architectural (eg drawing buildings) and we often resolve things, and not forget them by actually drawing them out. That's part of the reason why I drew so many resistors.

Appreciate that a 75.1245575 resistor is going to be difficult to find - it was really a prompt to make me think about how I would trim the resistances.

Attached is the *.asc. Hope that is of help to anyone who would like to provide some input here.

 

I've worked on the accuracy part of this some more. Whether it's required in the real world, who knows?

I've added the suggested output voltage offset circuit from the Lt1789 datasheet. This is needed at the lower end of the milliohm testing range (1m ohm) where the input voltage to the Inamp is only 100uV and the voltage offsets become relatively larged compared to this input. Not sure how well that 10M/100ohm resistor divider will work in the real world though.



I'm also not getting stable outputs in some ranges. Anyone offfer any advice there?

 

Some more progress. I think it's almost where it needs to be. Includes a power supply now as I have the option to power a voltmeter from this circuit. Includes trim adjustment for the 100mA constant current supply and trim to counteract the offset voltage for setting <10mR resistor range - has negligible impact on the 10mR+ ranges.

Problem now is how much will it all cost? Probably too much.

 

Alternatively, you can use a laser trimmed dual Op Amp with an ultra-low input offset of 0.12uV (LMP2012MM $5) with one as a Howland 1mA current source, the other to amplify x1000 from 1 mV to 1V using a 1V LDO MIC5365-1 for the current source Vref+. Then the output is 1mV/mohm using a single Li battery to a DMM.

 

How accurate does the meter need to be? and what resolution is demanded? And how much current is allowed to be driven through whatever resistance you are measuring? One more question is how do you possibly hope to get accurate readings if you do not apply the source voltage to the resistance through separate connections???

Can the tester live with a 12 volt battery?? (eight "D" sized cells??)

20 years ago I designed a resistance measuring system with a full scale reading of 1.999 ohms. It was for an automotive production line tester that had to be correct every time. It replaced a system that had a very high gain amplifier and needed re-calibration almost weekly. The new design was checked daily for a month and never drifted. And it continued to never drift. The only variations were when the product test connector became dirty. Part of the solution was using a KELVIN connection arrangement. Milli-ohm resolution demands that.

 

How accurate does the meter need to be? and what resolution is demanded? And how much current is allowed to be driven through whatever resistance you are measuring? One more question is how do you possibly hope to get accurate readings if you do not apply the source voltage to the resistance through separate connections???

I started out wanting to keep this simple. But the enjoyment of learning about the circuit and what was achievable with my limited knowledge has led me to here - trying to get very high accuracy. So, it's not really just about making a cheap circuit that's good enough for Australia.

I'm happy with the 100mA current.

My diagram is schematic only, but you'll see that a 4-wire test connection is shown. I propose to use this.

Can the tester live with a 12 volt battery?? (eight "D" sized cells??)

No - way too big. I wanted to balance size, capacity and didn't need such high rail voltages, so I've settled on 2x 18650's.

20 years ago I designed a resistance measuring system with a full scale reading of 1.999 ohms. It was for an automotive production line tester that had to be correct every time. It replaced a system that had a very high gain amplifier and needed re-calibration almost weekly. The new design was checked daily for a month and never drifted. And it continued to never drift. The only variations were when the product test connector became dirty. Part of the solution was using a KELVIN connection arrangement. Milli-ohm resolution demands that.

Kelvin connection will be used here. Quality ones would be important, noting I have some cheap ones right now that have differing cable resistances (measured using a cheap multimeter) that vary from 0.2 ohms to 0.4 ohms. This would impact the measuring accuracy.

 

OK, the low range ohm meter will have a well regulated current source to force 100.00 mA through the resistance, and also an accurate DVM with one millivolt resolution to read the resulting voltage developed. Depending on the maximum resistance needing to be measured that might be either a 3 1/2 digit meter or a 4 1/2 digit meter. Those meters are smaller and cheaper today than they were 20 years ago. The current source could be an accurate three terminal one wired as a constant current regulator.