Kelvin measurement, 4 Wire connection, Inner Resistance

Cherry67

Cherry67

Member
Joined
May 13, 2018
Messages
518
I have been babbling around IR (Inner Resistance) for awhile, in several threads now.

I should come to the point.

As an introduction, i see discussed, regarding usage of the cells,
- Capacity,
- Self discharge
- and inner resistance.

The first two are described quite well from many users here, and im rather sure that most agree in taking these in check and account.

The third one, i assume, is well understood as a factor to be taken in account on packs with higher discharge rates, for instance but not only as a factor of incresed losses on higher currents.

But, i dont see a common understanding on checking IR, and taking it account on building packs.

That may either be the case because of measurement problems for accurate IR, or IR is considered less important.

The Basic problen on IR is, that it is very low, and that requires a 4 wire measurement. Resistance may be 20 to 50 mOhms for a cell, and that is much to low to be measured 2 wire via contacts. It is to be understood that contacts itself have resistances in the 10 to 20 milliohm range, and that adds (two times) to th measured IR, plus the cabling, plus another 2 contacts where the plug goes into the instrument. Additionally, contact resitances are not constant, but change starting fron actual pressure to many other reasons.

There is a thread here already, which discuss proper IR measurement with Kelvin connection, the link is:

https://secondlifestorage.com/t-4-Wire-Kelvin-testing?highlight=kelvin

All the basics of a 4 wire Resintance measurement are correctly shown, but with all respect little seems to be understood.

Especially the video shows all, but what i have seen here and elsewhere is some misundestanding about the REAL implication how to connect the 2 additional wires.

The point is - 4 wire measurement does NOT mean (only) to connect the 2 wires (of one pole) BEFORE the contact to one of the cell poles.

It does mean to use a separate contact for all 4 wires to the battery poles.


image_fwkfbz.jpg


This is the Kelvin Clip assembly of a professional 4 wire Bridge (Hioki). In the picture the rubber rings, which close the clip, are removed.

It is to be understood that the 2 brass pieces (of one clip) are NOT connected in their turning point, but isolated, and each has its own wire to the bridge.
When the clip is connected to a wire, it makes TWO contacts to it - one to carry the current into it, one to pick up the voltage there.
And since the voltage pickup does not carry significant current (in contrary to the other contact) its contaxt resistance hardly influences the measurement result.

Example for the well used Opus charger, it does not only suffer fof the wiring between board and contacts to the cell (especially the spring) it suffers of two SEPARATE ADDITIONAL contacts to the cell to pick up its voltage.

Example for the known B6AC charger, here the situation is quite different. This charger usually charges packs, which are soldered/welded together. The BMS connections go between the cells, soldered as well. THIS is a nearly perfect 4 wire connection. Therefor a B6AC has a good chance to supply proper readings, IF you really use a soldered pack.
If you have installed several sockets for easy replacement/remove, you again have the contacts ruining your IR results.

I hope this makes things clearer how to get proper IR Results. If this will result in using them might be another discussion.
 
Good write up. When i test for it on 18650 i have used holders where I soldered the voltage wire on the tab as close as possible to the contact point to the cell. Ideal world above clamps is whats needed.
 
daromer said:
Good write up. When i test for it on 18650 i have used holders where I soldered the voltage wire on the tab as close as possible to the contact point to the cell. Ideal world above clamps is whats needed.
Click to expand...

Sorry, but thats just the suboptimal version.

What i want to point out is that you need to use 2 separate contacts to the battery.... for a better result.
 
I guess you didnt read what I wrote then? :) Never stated it was the best sollution
As I stated ideal world need above with 2 clamps to it but using the same connection with the current wire on one end and the voltage measurement on the other will work very very close to using 2 clamps as long as the current isnt way to high. I have been using that compared to proper clamps and it is darn close. Close down to below 1% inaccuracy is what I have had. But on other hand I have not had equipment that can go closer than above anyways


Its not optimal as I also stated but it is one way to get close to it at least.
 
daromer said:
I guess you didnt read what I wrote then? :) Never stated it was the best sollution
As I stated ideal world need above with 2 clamps to it but using the same connection with the current wire on one end and the voltage measurement on the other will work very very close to using 2 clamps as long as the current isnt way to high. I have been using that compared to proper clamps and it is darn close. Close down to below 1% inaccuracy is what I have had. But on other hand I have not had equipment that can go closer than above anyways


Its not optimal as I also stated but it is one way to get close to it at least.
Click to expand...

Sorry.
In essence the difference is the contact resistance of two contacts.
 
Yeap exactly for sure is!

Your write up was very good !
 
daromer said:
Yeap exactly for sure is!

Your write up was very good !
Click to expand...

It was indeed, but I much prefer your survival of the fittest approach.
 
What i was doing was describing the method for best accuracy.
You have to decide by yourself, If You need it for your method.

I will update with measurement results soon.
 
/Update
I just learned the english word consistency. I have used relative accuracy for it, just to point it out.
/End update


I have tried to make some measurements results to point out the advantages of "real" Kelvin connection.
As well i want to try to point out how to do it practical.
At moment, when i write this, not all results and work is done, so i dont know where it ends.

For the measurements as a term or process, you have to consider the following issues:

- i am well aware of accuracy issues. Measurements are done or rechecked with an old, but calibrated 4 digit Fluke. If otherwise, it will be noted.
- i am aware of proper industrial procedures to define total inaccuracies on measurement. I will use a simple method of 5 measures, which are done separately. Separately means, every test like a separate one, like having a socket then put out, put in again, measure, put out.....
- i have chosen three single cells as samples. One is an SD, one is a low cap, one is a well used but rarely good one. SOC is held in Mid-range, Voltage between 3.7 and 3.9 Volt.
- i could have chosen an HIOKI Professional equimpment and show ist results as proof. Thats doesnt help you in evaluating how you can trust your equipment you have, and how you can evaluate that by yourself. So i did what we all can do, with what we mostly have.

Since i will talk about the accuracy of (IR) measurements under different conditions, i have to explain a few things for the understanding of the less experienced, especially with look as the conclusions which will follow.

Summarized under "accuracy" you have to understand

- relative accuracy (consistency)
- absolute accuracy
- long term accouracy
- short term accuracy
Leaving out most of all scientific details, view and talk, some explanations.
Absolute accuracy is the match between the (standard) deviation of the results and the correct/true result.
relative accuracy is the standard deviation of the results.

Explanation for the difference:
The forged weigh of the cheating seller may have a relative accuracy of a few gram, means in several tries you come to the same weight plusminus these few Grams. But, the absolute accuracy is missing , say, 100 Grams per kilogram, the forged amount of the weigh.

The other way round :
Imagine you have wire snippets of exactly two meters length. Unluckily you need meter pieces. when you now measure and snip every wire in the middle, the you get wires where the absolute part of the accuracy is perfect, but the relative accuracy is still a few millimeters, your inaccuracy in finding the middle.

In summary, when we look at the "accuracy", the quality of a measurement, we have to look at two different things:
- the repeatability of the result, do we get the same resulte every time or are there minor or bigger variations
- the absolute accuracy, which can (usually) only been determinded if we make several tries the them calculate the average of the results. ( when they vary).

And what Long term and short terms adds to this, should be quite obvious.

Ok, lets go !
You should understand that my knowledge is, that IR (of an 18650) is a value in the low to medium MilliOhm range, and that the scinteific knowledge is that you need 4 wire connection for sure whne your result is close to 0.1 Ohm or lower.
Further i assume that the Opus is "not that accurate" on IR, and an B6AC (genuine) is better. ( I have only those two).

The samples have been chosen different to get differen results, it should be possible to recognize cell A,B and C ... if the measurement in question is "accurate".
In the first run i used the B6AC, 3 cells as 3s connected with proper BMS connection to each cell.

Then i chnaged the connections, i inserted sockets for each cell.


image_dsymuw.jpg


The comparison of this test is frightening, even taking in account the sockes are cheapo chinese. All results in Milliohms.


image_hkelco.jpg


While the soldered connections giv a rock solid result, the sockes give as well an unstable result, which has a different average far off.
(look at it as relative error plus absolute error).

So far, so bad.

Now, before i wanted to show how bad the Opus is, i had at least to show somehow, that the result of the B6AC at least has "some" absolute accuracy, the relative aspect is obvioulsly good (at moment).

Having no other IR measurement, i invented one. I took my small Oscilloscope, switched it to AC, 10 mV range, clipped it to a cell. The I attached a 50 Ohm resistor to the whole chain ( so all cells have the same current at all times, even when i measure the cells one after the other). Or, Better, i did not attach the resistor permanently, but continuously tipped it to the cell. What i got what current flowing or not, and an Display which showed a voltage drop (flank) at connection, a voltagage rise (flank) at disconnecting, between the the value approaching of the AC input. By reading and guessing the flanks, i got Millivolt readings which i recalculated with Voltage and Resitor to IR values.


image_cwjyhg.jpg


The values show upt to be alittle bit different that the B6AC soldered variant, unlike the sockets. The differences may mainly come from inaccuarcies of my reading the values from the Oscilloscope screen., which is 1,5 Inch only.
(I could have, but Ii did not fake the results to get a better match. Its that what i measured, and i want to show that Interpretation of what you see is most important when you are dealing in accuracy).

Now on for the Opus!

First i tried thes cells in an unmodifed Opus, in slot one and two.

Then I modfied slot one with the ominous wire to bridge the spring, and remeasured.


image_onbmwo.jpg


The results shhow a shocking difference to the B6AC, and to the Oscilloscope as well. This is not only a difference frome measure to measure ( relative) but a difference of the average as well. A big on, say 50 or 100 %. Whatever it is, it is unusable.

And it doesnt REALLY get better with the wire.
PooFF goes the said wire improvement through the roof.....

For now, we can draw the following conclusions.

- The stable relative accuracy of The B6AC looks promising. If it is correct on absolute level may still be in question, but is supported (against the Opus) by the Osilloscope results.
- The Opus shows wwhat is common Knowledge, imho even worse.
- The ominous wire doent really improve anything.

Now, theoretically i could have finished, but i wanted to go a bit further towards Kelvin clip usage.

First i repeated the first test (B6AC soldered) on another day.


image_ukyrki.jpg


Relative accuracy good, but..... deviations in absolute value !
A is measured lower, B higher, C more or less matches. What could be expected was, that the B6AC is not that good in a total long term stability, like a professional equipment is.

Now, to investigate that further, i changer the order of the cell in the pack. Had i done A.B,C (ground to Plus) i did ACB now. An easy and nice test, of the three (used) channels of the B6AC are equal.


image_dtnhfl.jpg


The values follow the changed order of the cells ( luckily), but here we see a "jump" in the values between try 3 and 4. My wife had called me away ( luckily, for the sake of insight here), and after i came back after an hour the values had chnaged. Probably a temperature effekt, i remeber it had cooled down.
So we have proved the B6AC has quite a good relative stability, but the absolute values canot be trusted beyond a 10 something% value.

THEN, finally, i wanted to use the KELVIN Clip. Remember, all good measuremets of the B6AC until now where strictly soldered setups.

I replaced the inner position of my setup with the Kelvin clip, and used the same cells as before.


image_hqvyas.jpg


SHOCK! Whats that? First i saw all the above go havoc. Then brain switched in again.

The inner Cell is C! 186 is an acceptable value for a kelvin-ed Cell measurement.

But why are the others, the End cells of the 3s so FAR off? Its the current path of each kelvin clip, which is a wire between the two cells each. This Voltage drop on the Clipwire is added to the "outer" cell, indeed plus the contact resistance of the current side of the clip.
(Pherpas this has to be explained with a circuit diagram)
Phew ! Every thing is fine, i can measure quite acceptable values without soldering.

Now i wanted to see, how far this can go down to even lower IR. I searche d my stock (of Cell pairs), and soon found one with 41 mOhm and one with 46 mOhm.

CHECK: if we do them in parallel, what do we get ?


image_hereoc.jpg


We can expect something of 22 mOhm.


image_qhsywu.jpg


This quad in the modified Opus slot measured 44.

image_zazeme.jpg

in the unmodified 47.


Summary.
From scientific view smal resistor values should be measured with Kelvin (4 wire) technology. No contact resistance in the measurement path is allowed.

The Opus is hardly usable for IR, even mith the wire mod.

The B6AC is better, but doubt may be allowed if it is good enough re relative stability (temp).

The Kelvin 4 wire technology is mandatory.

What is left?
We need a simple cheap clean method to measure acceptable IR.
 
Ok, here's the proposal:
take an Arduino and one of the available PWM switches for Arduino.


image_fsqhbo.jpg


Build a kelvin Swiper.

Switch a 15 Ohm resistor with the Mosfet over the cells voltage, measure differential voltage ove the two kelvin sensing inputs while switching the resistor load on and off.

The rest is calculating and Display of the result.

Where are the Arduino Geeks ?
 
Ok, i have a plan for the swiper.

The Load can be switched by an NE555 with the PWM switcher (basically a power mosfet) of an Arduino (see above picture.)

And for the Voltage drop to measure - ill try to use your beloved standard Multikmeter in AC position.

If that works there is an easy meathod to measure for everybody.
 
Ok, i built a swiper. quick and dirty.

Magnificent idea!

And it looks like, that a Kelvin Swiper ( 2+2 conatcts) can be done and makes my IR Project possible....
 
Cherry67 said:
Ok, i have a plan for the swiper.

The Load can be switched by an NE555 with the PWM switcher (basically a power mosfet) of an Arduino (see above picture.)

And for the Voltage drop to measure - ill try to use your beloved standard Multimeter in AC position.

If that works there is an easy meathod to measure for everybody.
Click to expand...

I built that up. Works !



image_vmbxjb.jpg
 
And here he is:



image_vyefzt.jpg


The 4 contact swiper:


image_dpjmgh.jpg


Left, on the swiper, the Arduino MosFet, and the loadung resistor (here 12 Ohm).

The bigger board left is just a signal generator, NE 555. Adjusted to 50 Hz.

I just saw the Chinese supply Signal generators
https://www.google.com/search?q=Sig...equency-Duty-Cycle-Adjustable-Module-LCD-3-3V
for 3 Bucks.

Thats it. For the function i will supply later, you can start to guess. :D
 
Ok, for the Circuit above.

The voltmeter is connected to two of the four Kelvin Connections. Obviously, on different sides. ( who asks why ??)

Voltmeter is set to AC, 200 mV range, or such. (we will have to measure 6 mV AC)

The (Arduino) Mosfet has a resistor in serial, 20 Ohm to start (it has to be adjusted). That is parallel to the second set of Kelvin connections. Means, there is NO connection between Mosfet and Resistor to Instrument.

The Input of the Mosfet (about 5-6 Volt for ON) is coming from an Signal generator, 50 % duty cycle, 50 Hz, 0-6 (5) volts.
I do it with an NE555, the link above shows a ready chinese universal Signal Generator.

Do i really have to draw a circuit plan ?


For the Theory.
The Resistor is sitting in parallel to the Cell in Measurement, the Mosfet switches the current on and off.

By the current (about 4V/20 Ohm it is about 200 mA) the Voltage of the Cell in Question will drop by its Inner Resistance.


Let us guess the cell has 60 milliOhm. On 0.2 Amp it will drop by 0.060*0.2 equal 0.012 mV. So the AC part of the Voltage looks like a 0.012 Vpp (Peak to Peak !!!) Square Wave AC. The Meter will look for AC, 0.012/2 mV, means ABOUT 6 mV Square Wave.

But what will it display ?
In reality the Meter is designed for 50 Hz AC Sin Wave. It will show the average of the Voltage ( Usin = Upp/sqrt(2).
On Square Wave it should show Uac = Upp.

So, for the above Resistor, we get a display of 6 (in mV). Which is about the tenth of the IR, as figure, in mOhm. Its not that a surprise, because our current is close to 0.1 Amp, which will give a thenth of the Voltage drop.

6mV to display is surely a bit challenging for the Meter. You can easily rise the current by 10, to about 2 Amps, and will get a display of 60, for 60mOhm.

The above mentioned adjustment has to take in account th REAL voltage of the cell, which is perhaps 3.8 instead of 4.0 The error is 4.0/3.8 which is about 5 %. Either adjust the Resistor for your favourite testing Voltage, or correct the result by a quick calculation in mind.

I should add that you can attach a second meter in parallel to the first, switched to DC Voltage, to display the cell voltage. IR AND Voltage in one "swipe".

This is the lot. An easy, straight forward measurement with little reasons for errors, except the obvious one.
Tomorrow i will add a nice little feature to check/ calibrate the display of the meter for the square wave.
 
Nice setup.

Very well documented but for us not so astutein the electrical engineering department a schematic would be nice. :D Please.

The electro mechanical part looks easy enough.

Wolf
 
wattwenger said:
Nice setup.

Very well documented but for us not so astutein the electrical engineering department a schematic would be nice. :D Please.

The electro mechanical part looks easy enough.

Wolf
Click to expand...
Stay tuned. schematic will follow... :shy:
 
"what is worth waiting for" :D

First is the absolutely most easy circuit plan i ever have drawn by hand:


image_ayapsw.jpg


To drive this with an AC signal, you can take the signal generator i linked before, or a simple NE555:
(copy out of datasheet)


image_uaaudm.jpg


for "calibration" you replace the resistor with the following divider, and measure the voltage there.
It is obvious you have there (BattVoltage)/10 as ACpp, and can look what your AC Multimeter says to that.


image_xfckjg.jpg


Personally, i will have a try with 4 Ohm Resistor as Load. That will be about 1 Amp peak, 0.5A average, and a AC Voltage about 60 to 100 Millivolt, which is a bit easier (and more accurate) to be measured. OK, the display will be half of that, but whatever. If you insist of the proper display figures, you can choose 2 Ohm with about 2 Amp load.

I think you all can easily figure out the absolute simplicity of the circuit, and the little margin of errors it can have.
Most and single ones are The voltage of the cell itself (changes the current of a given load), and the behaviour of the Multimeter in AC.

As i have shown, you can have two meters simultaneously on the swiper, having Voltage and IR at the same time.
You can as well add a small boost-up converter, supplying the power for the signalgenerator from the cell itself. Still totally easy.

Any comments ?
Update: The Mosfet i used is IRF 520. You can use just any N type with min 2-3 Amp max load and a convenient low internal resistance.
Or: IRFB 4110
 
Thank you.
This will be one of those weekend projects in the soon to be winter keep your ass indoor time that is quickly approaching.
We are expecting 4 to 8 inches of snow tomorrow. :(
I have already moved all my batteries, chargers and such(to the dismay of my housemate) :D into the Livingroom.
I think she may draw the line when the soldering station tries to make a migration. :angel:
You are right it is a very simple circuit and thats good cause I am a simple person ;) aber ich kann noch immer Deutch sprechen Die Muttersprace kannman ja nie vergessen!

Humor me by telling me what type of mosfet and maybe a partnumber the rest looks easy enough.

Danke sehr viel

Wolfgang
 
 
You must log in or register to reply here.
Back
Top