Nice little Milliohm Meter over at Circuit Cellar

[Dr.Krieger]

That's an amazing device sir, measuring uV with "long" leads is a difficult task, congratulations!

Based on my limited Googling there is a big diminishing returns effect in the number of ADC bits vs noise, you can go from 16 to 18 to 24 but noise is still noise and there is only so much you can do when you are trying to measure uVolts with "long" test leads, from what i understand its difficult even on PCB (chips that use using "guard pins" etc)!

A great way to increase accuracy would be to increase the current when measuring really low resistances, maybe with a adjustable Constant Current Source (but still use a "known" resistor to measure the current for good measure?) so you can measure a 1Ω res with lets say 100mA (10mW res dissipation and 100mV drop across it) and switch to 1A for under 100-200mΩ so you still get a good voltage drop (tens of mVolts) and don't chase micro-volts, assuming your shunt under test can handle 100mW of dissipation, which a sub-100mΩ should be able to do.

[Dr.Krieger]

Related:
Analog Front-End Design Considerations for RTD Ratiometric Temperature Measurements

"...For example, some designers may only be able to get 12 to 13 noise-free bits from a 16-bit to 18-bit ADC. The front-end techniques introduced in this article will enable designers to achieve 16+ noise-free bits in their system designs...."

[kripton2035]

@Dr Krieger
I did try to implement some constant current source with a pnp transistor and a zener, see attached schematic
but I've got some spurious oscillations I can't remove, and it gives totally random results in very low ohms...
if you have any idea(s) ?

[Dr.Krieger]

@Dr Krieger
I did try to implement some constant current source with a pnp transistor and a zener, see attached schematic
but I've got some spurious oscillations I can't remove, and it gives totally random results in very low ohms...
if you have any idea(s) ?

Like i said i'm not an expert or anything, but the things i was theory-crafting about Constant Current are these:

Instead of your PNP + Base resistor, add just another PNP to make it a Current Mirror setup, in simulation it keeps the current pretty stable.

For more accuracy i thought of a TL431 CCS but that needs to drop 2.5V across it so you only left with less than that for your DUT (for 5Vcc). Very stable current though and perhaps it will perform better due to less thermal drift (although the Current Mirror setup should be stable too with a 0.6-1W res)...

Also when i was messing with the ADS1115 and though that perhaps the multiple gain routine was the thing that slowed the process down i thought of the same TL431 CCS but with a Arduino-controlled BJT that shorts one res so you at least have a two-range setup you can use with a fixed gain (+-2.048V range) to measure up to a KΩ or so, which is more than enough for me.

Note that in both cases the current is "pretty stable" but not 100% the same for min/max DUT value... :/
Also i didn't do any transient analysis, circuit stability is unknown.

I can't think of anything else easy, for maximum current stability you can go hardcore and make a good OpAmp + TL431 CCS, much like a mini Dummy Load:
Just an OpAmp that measures the Vdrop on a res compared to the 2.5V of a TL431 to keep a stable current. Below is an example of a setup like that for a complete Milliohm Meter using a DMM but a good OpAmp like your AD8628 and a TL431 would be pretty good.
Depending on what you want that could be a good idea or complete overkill, still i think that measuring a 2nd "known" resistor is better but ofc i could be mistaken since i haven't compared the two choices (stable current or known res) in theory or in practice XD

[SilverSolder]

@Dr Krieger
I did try to implement some constant current source with a pnp transistor and a zener, see attached schematic
but I've got some spurious oscillations I can't remove, and it gives totally random results in very low ohms...
if you have any idea(s) ?

Try a 100nF capacitor across each of the two 33K resistors for stability.

[Shock]

You probably need to start by defining what resolution and accuracy you need.

There are a couple of inexpensive designs shown on youtube. I'd also consider looking for a 4 wire kelvin (w/guard) capable LCR meter as they can go low DCR.

Myself I along with a few members here have IET/Cambridge Technologies LOM510A these are really nice micro-ohmmeters for out of circuit work. Last video shows Robrenz's one in the first part of the video. Not only does it have guarded 4 wire but it takes one shot measurements. Temp, current, oxides and shielding affects the measurement at low ohms so you need to consider probes/fittings and technique.

[Shock]

Just noticed another one as well. This has the PCB up on PCBway.

I have additionally have a bunch of milliohm capable meters so no point in me personally building any of these projects for measurement. But Kripton2035's shorty is pretty cool tool. I'll eventually get around to making his or sooner if he decides to sell a board/kit.

[Dr.Krieger]

I did a few more tests increasing the Samples Per Second like imo suggested and it works, i can take more measurements now although the "final" measuring speed doesn't double when doubling the SPS.
There is still work to be done like figuring out how fast i can without having massive Ohm's measured differences, what N/M ratio is best (i think more samples and less cycles is better than the other way around for the same overall speed), etc but i don't think it makes sense to continue with this on the breadboard:
Already i am getting around 1mV difference between readings on the 10Ω Vdrop, that is unacceptable for a  16bit ADC at the +-1.024V range. 10Ω's Vdrop is right in the middle of that range when close to max current (60+ mA), probably on purpose. OFC as the DUT value climbs the current drops (since its in series in the loop) and the current drops to less than 10mA on "high-value resistances", for example for a 330Ω DUT the Vdrop on the 10Ω is ~125mV, with a 470Ω it drops to about 90mV and its less than 50mV for a 1KΩ DUT (ofc the PGA has switched to the +-256mV range for all of those).

I think i will also try a 1-2Ω res instead of the 10Ω once i make this on a veroboard (the next step), so the same PGA is used for both resistors in the important to me sub-1Ω range (the loss of accuracy in the >10-20Ω DUT range does not bother me at all).
BTW at 330Ω you almost maxed-out on voltage (~4V on the DUT) but Arduino's pin output voltage does raise as current drops or better the voltage sags as load increases and that is how the original circuit can get away with only 30Ω pin resistors (that would burn-up the Arduino at ~200mA total or 50mA per pin at 5.00Vout). In my Arduino Nano with external 7805 supply i can get 5.00Vout from the digital pins without load but with my 8x100Ω setup i get ~4.140V pin-to-pin, thankfully both voltage and current stay exactly the same in both current directions (as far as my DMM can tell).

For the videos i removed the Vbat voltage and icon in the debug menu and added the N and M values and the SPS rate used. Also i made a video with many different setting on screen at the same time so you can better see the speed difference.

Just noticed another one as well. This has the PCB up on PCBway.

I have additionally have a bunch of milliohm capable meters so no point in me personally building any of these projects for measurement. But Kripton2035's shorty is pretty cool tool. I'll eventually get around to making his or sooner if he decides to sell a board/kit.

That is a nice, analog circuit but the issue is that you rely on the Panel Meter's accuracy... I would prefer the ADC method, so you can for example use code to calibrate any loss non-linearity error, like if the
ADC (of the OpAmp/IntAmp front end) reads 1% lower at very low voltages, almost perfectly at the middle of the range and 1% higher at the highest voltages, you can easily add a "calibration" code that adds or subtracts a little from the measured value, depending on the voltage level.

That being said, before i fround the "Milliohm Meter" circuits i was planning to make a simple 0-1A Constant Current Source using a 5V USB charger, just a TL431 and a AD8628 (a AD620/AD8221 would be better but require negative supply) with a Darlington BJT and a 0.1Ω/5W to be able to drop 1A to up to a couple of Ohms. I have a 5mA 5-digit Panel Meter that i will use to measure the current (with the help of a 200Ω trimpot parallel to the 0.1Ω) because i want to be able to see that the current stays stable under load and it don't change by the second like in my failed LM317 "1A CCS"...
I made it on Proteus and it works nicely, i was about to start soldering but then got on the Milliohm stuff. I still want to make it because it can be used for other things like measuring diode's Vdrop vs Current (and/or temperature) and i also will add a VOM1271T to it so i can easily drive MOSFET Gates (i already use this in a temporary LM317 50mA CCS setup).

[iMo]

Nice!
Btw., when you target 5 digits results you have to consider the temperature coefficients of the parts used, especially with such things like TL431, LM317, and your standard trimpots(!) and resistors.. Also the ADS1115 is not TC free..

[Dr.Krieger]

I'm not getting 5 stable/accurate digits, i know it and i'm fine with it, in the case of the Bay Panel Meter i assume i can get an accurate two decimals on the 1A range (so i know its at least that stable), the LM317 CCS even with 4V input drifted like crazy because of the heat when i dropped that 1A on a 0.1Ω (3-4W of dissipation of the LM). That is the downside of all TO-220 regulators, the pass element is in the same case with the control circuit so when the thing heats up it all goes to shit :/

Like i said i have no prior experience with ADCs so i want to see how accurately the ADS1115 will measure the 10Ω on a soldered board, i hope for a much more stable operation with a 9V battery etc...

And i know that a normal trimmer will mess up with the tempco of a nice precision XXppm resistor but i don't have a choice, i need ideally a 0.10009Ω shunt/100Ω trimmer in parallel to get the correct 0-1.000 on the Ammeter for 0-1A...


BTW sometimes the videos in this page are all messed up, like wrong video in the wrong post, is it just my Firefox or does anyone else has the same issue? First time today i notice that....

[Dr.Krieger]

I haven't put anything on the breadboard in the "USB 1A CCS" case but i played around with it on the simulation and it seems to work good, as long as i can get the right shunt/trimpot for the meter ratio...
I even got as far as to plan how to put on the breadboard. The Meter is the other way around (before the shunt/trimpot) in the PCB version so i has common ground with the USB and i can be powered by it.

I guess you could instead use the 5mA Panel Meter as a voltage meter and convert the thing into a Milliohm Meter like in the analog circuit that Shock posted, use one OpAmp to make the CCS and then another to measurethe Vdrop on the DUT's and have the meter on its output.