DIY Milliohmmeter - how to solve grounding issue?

[Avelino Sampaio]

Hi

I'm completing this project and would like a tip on how I can solve a grounding issue. Note in the image that the current of 50mA, passing through the DUT, should completely cover the AEBF path. However, this current is divided by the AEDF path, adding the resistance of the reading cable to the final value. What would be the best technique used to resolve this issue? This question seems to be fundamental so that I can read a very low resistance value (1mR).

[Doctorandus_P]

For the measurement side, do not use a grounded connection, but use a differential amplifier.

[Avelino Sampaio]

Para o lado da medição, não use uma conexão aterrada, mas use um amplificador diferencial.

I can't change the project this way. These are the best opamps I have available. There has to be another output without changing the design too much.

[Gyro]

That seems a very complicated implementation for what you are trying to achieve (it would be a different matter if you were doing AC measurement and synchronous detection). Try the attached app note (substituting one of your own opamp and reference). As long as 1A maximum measurement current isn't an issue you will achieve 1mV/mR, so at least one decimal place on a typical DMM.

Completely breaking any internal links between force and sense leads is essential when Kelvin (4 wire) sensing.

[Doctorandus_P]

I really like the simplicity of the circuit described in Circuit Cellar of September 2016, issue 314.

How it works:
the microcontroller has 2x 4 pins parallel with a 30 Ohm resistor on each of its pins, and this creates a bidirectional test current.
This test current goes through a 10 Ohm reference resistor and your DUT (Device Under Test).

Then the ADS1115 is used to measure both the voltage over the reference resistor and over the DUT, then swaps the direction of the current though the DUT with the microcontroller, repeats the measurement in the other direction. It also repeats this whole process a bunch of times and does some averaging.

It also uses 4 wire kelvin sensing. (Two "drive" wires, and two "sense" wires), as shown in the schematic.

Go buy that back issue if you like it. Unlike Elektor, circuit cellar projects always come with all source code.. In this case it's a quite simple "arduino" contraption, but it is something to get you started and to improve upon.

With the 1uF and 1k resistors there is some protection for the ADS1115, but I would add a bunch of extra diodes to short any voltages over 3Volt or so over the test leads.

Also, because there are no extra opamps in this circuit, they can also not generate any errors, and it makes the circuit cheaper.

I quite liked CC. They did not have many (interesting) projects, but they had the source code available for all of them and a subscription was relatively inexpensive. Then they had some kind of collaboration with elektor, the subscription doubled in price and I canceled (or better, did not renew) my subscription.

[Avelino Sampaio]

Isso parece uma implementação muito complicada para o que você está tentando alcançar (seria uma questão diferente se você estivesse fazendo medição de CA e detecção síncrona). Experimente a nota do aplicativo em anexo (substituindo uma de sua própria opamp e referência). Contanto que a corrente de medição máxima de 1A não seja um problema, você atingirá 1mV/mR, portanto, pelo menos uma casa decimal em um DMM típico.

Quebrar completamente quaisquer ligações internas entre os cabos de força e de detecção é essencial quando se faz a detecção de Kelvin (4 fios).

Really 1mR in this project can be quite challenging, as it will only generate 50µv at the input of the OPA277. But I think the OPA277 might surprise me. The main use of this project of mine will be to measure the RDson of the mosfets. I have Mosfet (IRFB3306) which reaches 3mR. I don't need great accuracy in this case, just that it reads within a certain acceptable range. For example, in this case I could be measuring between 4 and 2mR.

[Avelino Sampaio]

Hi

After several failed attempts, I managed to come up with an arrangement that worked. I will take some pictures of the results obtained. And the rehearsal is taking place on a protobord.

[Doctorandus_P]

The ADS1115 I mentioned earlier can measure down to 8uV (rms noise) and a resolution of about 4uV.

For measuring MOSfet RdsOn, I would use another approach. Just build a simple current source from an LM317 and a single resistor, and calibrate it to 1A. Then you can simply measure milli ohms with a small power supply, the 1A current source and a DMM in the mV range.

Only calibrating an "around 1.25 Ohm" resistor that can handle 1A is a bit fiddly. I would start with something like a 1.2Ohm 10W resistor and then add some other resistors in series / parallel in a few iterations. 10W is oversized (it only dissipates around 1.25W, but this lowers it's temperature and temperature effects. Using a big bunch of 1/2W metal film resistors is also a good option.

[Kleinstein]

The circuit still does not look good. There is still no real 4 wire sensing. There is no real need for a precision OP-amp at the current source side and the extra chanrge pump is odd and only producing extra noise.
Just for MOSFET testing and thus no high demand on the accuracy of the scale factor the LM317 typ current source should be good enough. With 1 A no amplification is needed.
If amplification is wanted, just consider the low side sense as the ground for the amplifier and output and if really needed plan for current compensation (but this needs a negative supply so not that attractive).

[Avelino Sampaio]

There is still no real 4-wire detection.

I'm considering it acceptable, given what I had initially. I don't have a differential here to implement.

There is no real need for a precision OP amplifier on the current source side.

Initially I started testing with the MCP6002, with simple font, but it had a very strange behavior for my configuration. I also tested with the MCP6022 (smaller offset) and had the same bad result. Despite being Rail-to-Rail, they all had an expressive output voltage (headroom), adding to my AZ431. Using the OP189 the voltage added to the AZ431 accurately reflects the voltage of the DUT. I am very satisfied with the stability of the Current Source, using the OPA189.

For MOSFET testing only and therefore without high demand on scale factor accuracy, LM317 type current source should be good enough. At 1 A no amplification is needed.

The project is not just for testing mosfet.

[Doctorandus_P]

If you don't do some kind of differential measurement, then it's almost useless to use a 4 wire measurement in the first place.

[Avelino Sampaio]

If you don't do some kind of differential measurement, then it's almost useless to use a 4 wire measurement in the first place.

Thank you so much for actively participating in my case.

I need to understand why it would be Useless. Assuming each cable has a resistance of 50mR and I'm measuring a DUT of 1R. Ideally I would have 1v at the Opamp output, which would give me approximately 47.6µA (1v/21k) going through the feedback resistor. The voltage drop across the cable (Vc) would be approximately 2.38µv. The input voltage would be Vin = Vr(DUT) + Vc(Cable). I would have an addition of 47.6µV to my output, leaving Vout=1.000047v.

Please correct me if I'm making a big mistake in my analysis.

[Doctorandus_P]

Start by not making assumptions on your cable wiring.

If you measure a 1 Ohm resistor, and you have 50mOhm cable resistance, then you already have a 5% error.

Or:
If you put 50mA though a cable with 50mOhm resistance then you have a 2.5mV over your cable.

Your last picture is not complete. In earlier pictures you drew a star (Ring?) GND where both the "force" and the "sense" wires are connected on both sides, which means those wires are just put in parallel, and that would then be a (static) error of 25mOhm (half of 50mOhm).

In you last picture you have drawn a triangle with a "G-20" in it.
It's a completely bogus picture, but for simplicity sake I shall assume it's your opamp and it's configured for a gain of 20.

But what signal does it amplify?
There is no way of telling without showing a complete schematic, and this last picture is too incomplete for that.

The proper way for accurate low-ohm measurement is:
1. Generate a test current though the "force" wires. (This makes cable resistance irellevant).
2. Do a differential measurement over the sense wires. Current though the sense wires is very low, voltage drop is negligible.


Also, on your "force" side. You've got an OPA899 as a buffer, then you stack a TL431 and a LM317 on top of that but the pins of the LM317 are not labeled. You have two (presumably 1W) resistors on (probably) it's output. I think you want to create a current source there, but you can't create a current source that goes into your test resistor and at the same time control the voltage with the AZ431. (TL431).
On the measurement side you have an OPA277 that may be configured to amplify 20x (too many resistors to count), And then after it you have some 7-segment display (with an ADC?), but that also measures the voltage drop caused by the 50mA though the wire on the "force" side.

The overall design is so complex and convoluted that I don't even want to go into further details.

Just curious, have you even taken a half look at the (Deceptively) simple schematic with the AMS1117 I posted?
The schematic looks simple, but it's clever and they've done everything that I can think of right (Except maybe adding some extra protection on the wires).

[Avelino Sampaio]

Start by not making assumptions on your cable wiring.

Yes, I can make a guess, since you couldn't convince me that it's useless?

If you put 50mA though a cable with 50mOhm resistance then you have a 2.5mV over your cable.

Yes, 2.5mv on the Drive cable and what is the value for the Sense cable?

There is no way of telling without showing a complete schematic, and this last picture is too incomplete for that.

Is the complete and updated schema in Reply #6?

[Kleinstein]

The upgrades shematics is still a bit confusing, but not that bad as it looks at first.
The 10 Ohms resistor in the link the sense low is essentially eliminating that link, as long as the cable resistance is way small than those 10 Ohms.  I consider the 10 Ohms more like a links for the case that no DUT is connected. A diode or 2 diodes back to back may be even better here. A considertably larger resistor would make it more obvious, that the 10 Ohms stand for a relatively large resistor were the exact value does not really matter.
So the upgraded version kind of did use the low side sense as the ground for the output circuit. The more logical path for the ground of the display would be the sense line, at least if there is no much current flowing (shared power ground).

Quite a few of the resistors (the 2x100 ohms and 1 K trimmer)  are just there to adjust the gain and are a bit confusing to the reader, but otherwise OK.

The current source is still a bit strange and not really good: the LM317 is a first crude current source and the OP deverts the the  current that is higher than set by the power resistors and the 431 ref. to the reference. This is a rather complicated circuit for relatively poor performance and the need to trim the current from the LM317 quite well.

The real test of the 4 wire ohms sense is working OK is not 1 Ohms DUT and 50 mOhms for the wire, but more the other way around. In this example the last circuit would have some 2.3 mV at the 10 Ohms resistor and as additional output voltage or about an 50 mOhms (or 100%) error to the result.  The suppression of the wire restance is only about as high as the amplifier gain in the upgraded circuit. That is better than the inital circuit with about a 2 fold suppression from using both wires, but still not good.

[Avelino Sampaio]

The current source is still a bit strange and not really good: the LM317 is a first crude current source and the OP deverts the the  current that is higher than set by the power resistors and the 431 ref. to the reference. This is a rather complicated circuit for relatively poor performance and the need to trim the current from the LM317 quite well.

Although strange, in bench tests it has shown excellent stability. So I see no reason to make changes.

[Avelino Sampaio]

Some tests:

1 - IRLZ44 (Datasheet RDS=22mR)
2 - IRFB3306 (Datasheet RDS= 3.3mR)
3-  Precision resistor of 1.744R
4 - 1R resistor

[Doctorandus_P]

At the moment you just have a simple resistor to open the FET's and you connect it to "+B".
Rdson of MOSfets change with Ugs and adding a potentiometer may be interesting (Or a simple DAC?)

Most MOSfets open somewhere between 3V and 5V, but Rdson typically halves when the gate voltage is changed from 6V to 12V (and still changes but slower above that).

Rdson values are also very temperature dependent.


You can test linearity of your meter by first testing a bunch of resistors (for example cheap 1% metal film resistors) and then measuring series and parallel combinations of those same resistors.

[Avelino Sampaio]

At the moment you just have a simple resistor to open the FET's and you connect it to "+B".
Rdson of MOSfets change with Ugs and adding a potentiometer may be interesting (Or a simple DAC?)

My main intention with testing mosfets is to find fake or damaged parts.

Yesterday I tested some P-channel mosfets and found a fake unit.

1 - IRF6218 - false unit on the right
2 - IRF6218 (datasheet RDS=120mR ty / 150mR max) - OK
3 - IRF6218 (rds three times higher than normal) - Fake
4-  IRF5210 (datasheet RDS=60mR) - OK

[Avelino Sampaio]

You can test linearity of your meter by first testing a bunch of resistors (for example cheap 1% metal film resistors) and then measuring series and parallel combinations of those same resistors.

I've done several tests in this direction and the linearity is good. All I have to do is keep the cables short and good gauge. But I'll take your advice and add a differentiator. I'm going to buy the INA143 la from LCSC (the only place I can make purchases).

[Kleinstein]

One can already get quite far with making the amplifier a difference amplifier. The difference in the circuit is not that large:  just a slight divider for the input signal toward ground.  One can check the adjustment by doing a test would a simulated 0 ohms and extra resistance in the low side ground connection.  It is not as good as most INAs, but should still quite an improvement over the current circuit.

An alternative (not very critical) would be having an additional difference stage (could be gain 1) between the amplifier and display.

[Avelino Sampaio]

Hi

Waiting for the arrival of INA143 and some resistors (50mR, 30mR, 10mR and 2mR), for a new round of tests.

[Avelino Sampaio]

Hi

I made two small changes to the design. 1) replaces OPA189 on the output with OP07. 2) I replaced the four 47R/0.25w resistors with a 10w wire resistor. The 1/4w ones were taking a while to stabilize the current. I didn't have the 1w ones, so I put a 10w one.

1- 50mA current source with very good stability.
2- Precision resistor 1R744
3- 1R resistor
4- 50mR resistor
5- 30mR resistor

[Avelino Sampaio]

1- 10mR resistor
2- 2mR resistor

[Avelino Sampaio]

Now testing some Mosfets.

1 - IRF6218 ( datasheet RDS ty = 120mR)
2 - IRF5210 (datasheet RDS = 60mR)
3 - IRF1404 (datasheet RDS = 3.5mR)
4 - IRFB3306 (datasheet RDS = 3.3mR)