milliohm measurements

[badcockr]

OK,

Following hot on the tail of the PSU questions .....

I want to make repeatable and reliable measurements of milliohm values (motors) and don't want to buy a commercial unit. Now to my thinking this is a relatively straight-forward measurement - just measure the current and voltage drop directly applied across the terminals.

So if I was to use a LM317 in constant current mode (1A) then I should get a drop of 1 mV / mOhm across the terminals. Amplify this with an instrumentation amp with gain of 100 then I would have an output of 100 mV / mOhm which I could then read with any meter. This should be a usefull adaptation of the four wire technique.

This looks so simple I feel I must be missing something - what is it?

Rod

[AndyOne]

Why not just use a any cheap old 3.5 digit DVM on the 200mV range, this will give you 0.1 mOhm resolution for 1A.
Make sure the 4 wire connection has the volt sensing connections connected inside the current connections.

Andy.

[Cerberus]

measuring milli ohms is always going to be troublesome, your multi-meter leads alone are 12 ohms. Your resistane to the soldering junction is going to be high, in other words, I believe your contact resistance will probabaly be a major source of error... i.e. your model only works if your amp is connected immediately to the motor terminals with zero contact resistance.

[AndyOne]

Cerberus,

A four wire connection eliminates any measurable effect of contact resistance. The current is fed into the resistor under test by one pair and it is measured using a meter connected by another pair.

Using a constant current supply ensures that any contact resistance has no effect on the current in the circuit.

A seperate pair of connections to the a high resistance volt meter makes sure that the contact resistance at this point is neglegable compared to that of the meter.

The system is called a Kelvin connection and is well documented.

Andy.

[mussotte]

Depending on the motor, 10A current can provide better resolution. Just pay attention not to feed this current for a long time...

DC resistance is also a function of winding temperature. it is often measured on a warm motor.

hope this helps.

[KreAture]

It's a matter of being able to supply the current reliably while measuring too.
Too short pulse would cause issues with inductance.

This is something I'd like too. I think I'll make a purpose-built meter :p
I have just the instrument-box for this project and a LCD that would be perfect.

[Cerberus]

Thanks Andy, I was actually referring to the original solution proposed, but suggesting the use of mult-meter probes was probabaly misleading on my part. Yes, your solution works well for this purpose.

[badcockr]

Thanks Andy, I was actually referring to the original solution proposed, but suggesting the use of mult-meter probes was probabaly misleading on my part. Yes, your solution works well for this purpose.

As Andy says:
A four wire connection eliminates any measurable effect of contact resistance. The current is fed into the resistor under test by one pair and it is measured using a meter connected by another pair.
The solution proposed *is* a four wire kelvin connection.

I take it from the other comments that this should work OK - time to prototype.

This is something I'd like too. I think I'll make a purpose-built meter. I have just the instrument-box for this project and a LCD that would be perfect.
.. So Kreature, if you want to collaborate and code using a PIC microcontroller we can work together and I'll get some nice low-resistance PCB's made at Olimex - what do you think?

Rod

[Mr.RC-CAM]

If your measurements will be under 1 ohm, and 1milliohm resolution is your target, then the basic elements of the uMetros project could help out, including the PIC's code.
http://www.rcgroups.com/forums/showthread.php?t=359214


It could be used as-is if a low impedance voltage source was connected with the DUT winding. Or, change the 2-state current sink circuit into a 2-state current source for a more integrated solution.

RC-CAM

[badcockr]

Kreature,

What about using a National Semi LP3866 3A adjustable regulator (free samples online) configured as a 1A constant current source and using the shutdown pin for enabling / disabling via a microcontroller?

Should enable plenty of measurements without undue heating.

Rod

[badcockr]

If your measurements will be under 1 ohm, and 1milliohm resolution is your target, then the basic elements of the uMetros project could help out, including the PIC's code.
http://www.rcgroups.com/forums/showthread.php?t=359214


It could be used as-is if a low impedance voltage source was connected with the DUT winding. Or, change the 2-state current sink circuit into a 2-state current source for a more integrated solution.

RC-CAM

Thanks,

I'll have a look,

Rod

[badcockr]

Okay,

Based on Kreature's input here is my thinking:

A 16F628 connected to a standard Hitachi display (14 pin connector) handles the measurement and display.
A push switch connected to portb.5 activates the measurement routine and enables the 1A supply.
The amplified measurement is fed into the AD on porta.5
Internal OSC used.


Maybe?

Rod

[KreAture]

badcockr
It's not just a matter of 3A reg. It's not even about the power you have to deliver, it's what you don't have to deliver.

A 100 mOhm motor at 1 amp current will only have a 100mV voltage over it. This means if you use a 5v x-amps supply, that your regulator would be dissapating 4.9 watts. This is actually way high for such a reg without a BIG heatsink.

So, short pulses is the key. But not so short as to cause issues with inductance.
Also, I think I'll use a avr as I got a C-compiler for it I like very much.

Oh, btw, low resistance isn't necessary on the traces. The current going back from the motor poles to the ADC is minimal so the resistance in the wires will not influence the results much. The resistance one is measuring isn't all that small anyway.
I'll see how accurate I can get it very soon.

[Leftyretro]

A Kelvin setup is the way to go. I might question the choice of a 1 amp constant current source. There will always be a heating effect that will effect the device under test (DUT). So one might want to have as small a constant current source as can be tolerated. However motors, solenoids, transfomer winding frequently work at high current ratings and one might need as actual resistance reading at the normal WORKING current that the part is going to be used at, which might be well above 1 amp. So I guess it's all about the purpose you need the accurate resistance reading in the first place. If it's for matching of components or identical windings then a smaller current would be fine.

Lefty

[KreAture]

If using the right AVR one can enable differential mesurement mode and a gain of 200x.
This should help mesasuring without havng to run so much current.

Since a constant-current supply won't be all that stable anyway, I think using a extra shunt is smart. The shunt should be in the 1-4 Ohm range so that it will allow a good voltage-drop for the regulator acting as current-supply. This will reduce the heating of the reg and move it to a power-resistor instead.

Further more, by reading voltage over shunt one can look at current and wait for it to become stable. This will make sure the measurement is taken after current in inductor is allowed to settle, but not waste any energy waiting too long.

As i said before, the current going to the ADC will be in the uA range and compared to the 1-amp test-current it will not present a big enough error to be bothered with. After all, the ADC has a 0.5 LSB conversion error anyways in addition to a absolute error and a linearity error. With the 200x gain setup the current can be reduced to somehting like 50 mA for a range of 0-500 mOhm if system is running on 5v reference.

I got a few tricks up my sleeve. Will be trying to churn out something tomorrow :)

[badcockr]

Since a constant-current supply won't be all that stable anyway, I think using a extra shunt is smart. The shunt should be in the 1-4 Ohm range so that it will allow a good voltage-drop for the regulator acting as current-supply. This will reduce the heating of the reg and move it to a power-resistor instead.

Further more, by reading voltage over shunt one can look at current and wait for it to become stable. This will make sure the measurement is taken after current in inductor is allowed to settle, but not waste any energy waiting too long.
<SNIP>
With the 200x gain setup the current can be reduced to somehting like 50 mA for a range of 0-500 mOhm if system is running on 5v reference.
OK - that sounds logical and more elegant and only requires two AD inputs for the current and voltage measurement. This would also avoid having to 'switch' the Vreg on and off for measurements - I like your thinking!

Rod

[Acetronics]

Hi, everybody

I suggest you to have a look to an AD 630 ( Analog Devices ) application AN 306 about µohms measurements, also published by ELEKTOR magazine in the july/August 1999 issue, at page 24.

this measure is not as simple as a " U = R x I " formula application.

good reading

Alain

Alain

[KreAture]

Actually, it is as long as we don't need micro-ohms...
I've done reading before of both mV and mA.

We'll know tomorrow...

[Acetronics]

:rolleyes:
Hi, Kreature

Th'at's not you I will learn that if you need precise measurements, resolution must be good ...
Here, you just measure milliohms ...but real milliohms !!!

not +/- 25 or 50 % !!!

here , output is 15.7 mOhms per volt ... as reading .001v is quite good ... so you read by 15.7 µOhms steps ...

...

Alain