DC IR VS AC 1kH IR measurements

Wolf

Well-known member
Joined
Sep 25, 2018
Messages
1,282
Discussion about the benefits of DC IR testing versus AC 1kH IR ( Impedance) testing.
2 wire vs 4 wire kelvin etc.

Primer of discussion copied from another thread.

gauss163 Wrote:

But iirc theOpus is doing alonger timeDC IR test (vs. 1ms AC). So it is also measuring somenon-ohmic parts of the IR, i.e. additional components of the cell'sIR that don't behave linearly as in Ohm's Law and take time longer than 1mstoramp up to steady-state values(e.g. charge transfer and diffusion processes). This is why at the end of (dis)charge wesee both an instantaneous voltage change (from the ohmic part of the IR- whichbehaves like a resistor), followed bya much slower exponentially decaying voltage changethat asymptotically approaches the steady-state resting voltage (from the non-ohmic IR, which takes a long time (hours) to ramp down as various elecrochemical processes converge to equilibrium).

In many cases it is thenon-ohmic part of the IR (vs. capacity degradation)that ends uplimiting the lifetime of the cell because it can degrade at 10 times the rate of the ohmic part, and this greatly drags down the voltage under load (esp. for high-current loads). Whether or nor this is a limiting factordepends on what type of loads you puton the battery. If your device only usesshort pulses (e.g. a jump starter or vape) then the non-ohmic IRplays little role since the discharges aren't long enough for it to kick in.OTOH if you are doing longer time discharges without big spikes then the non-ohmic IR plays a big role. Some loads are a mixture of both and require more complex analysis (e.g. to see if a constant load plus occasional spikedrags the voltage low enough for long enough time to trigger undervoltage protection).
 

Bubba

Active member
Joined
May 9, 2018
Messages
264
Here is an interesting read about lithium ion battery impedance. They have a resistive and inductive component. If I recall there isn't a capacities component.

https://www.google.com/url?sa=t&rct...11/1/220/pdf&usg=AOvVaw3qQjLA6GnDO3Ea1f8R86be


Z is the impedance. The m? reading we get is actually Impedance so you can use the same formula as above. .... Just most people understand that they have been checking resistance so I used in the previous formula R1.R2 etc.
Please check out this article --> https://www.allaboutcircuits.com/te...t/chpt-3/parallel-resistor-inductor-circuits/

I'll be honest it's been 25yrs since I studied battery theory and used superposition theorem but that's what were getting into.
 

Wolf

Well-known member
Joined
Sep 25, 2018
Messages
1,282
Bubba said:
Here is an interesting read about lithium ion battery impedance. They have a resistive and inductive component. If I recall there isn't a capacities component....................it's been 25yrs since I studied battery theory and used superposition theorem but that's what were getting into.
In the vernacular of Spock "Fascinating"


gauss163,

First off please be aware that I find your devotion to your DC IR measuring cause very refreshing. I myself have absolutely no quarrel with the idea that there may be a better way to determine the longevity and or quality of a USED 18650 Li-Ion cell. For that matter of fact any Li chemistry or package.
I am also not opposed to spending 10, 20 or 30 seconds to do a test if in the long run it provides me with accurate, definitive and consistent results by which I can make an educated choice whether I can use this cell or not. At this time I have not found such a tester commercially available, reasonably priced,andis accurate and consistentwhich also does not make me jump trough hoops,hang from chandeliers, andemploymath skills the likes of a Marslanding ( oh wait one of those didn't work out to well)to getproper results.

At this time the 1 kHz 4 wire kelvin method, which is incorporated into the YR1030, and RC3563 style testers,has been the most stable and consistent standard that I have gone by and the results show it. There are more than a couple of members here who have incorporated this method and found it to be fruitful and hasgiven them good results.
As they say the proof is in the pudding.

As you know if you have looked at my Harvested Cell Analysis sheet I measured the "DC IR" of the testers that allowed a relatively quick result and recorded it. And yes I did spin the cell and made sure the contacts were as clean as possible.
The problem was that the results where not consistent and consistency is important when establishing a rule to go by. Not lets try it 3 or 4 times and take an average. It's like a man with 3 or 4 watches he never reallyknows what the exact time is. I for one have a bad habit of not accepting close enough. For me I have to have a trust in a piece of equipment that it will give me accurate results always.
Hence I have 3 Fluke bench-top DMs. 8810A, 8842A and 8808A. If all 3 agree to a measurement I know its right.
The consistency of the 1 kHz 4 wire method is unmatched as I can take a cell measurementput it aside for a month and measure it again and it will be within 0.05m? of the original reading. Providing the cell is of reasonable quality to begin with.

Cells with a low or 0% SOC will have a higher IR and as the cell approaches 100%SOC the IR will drop about3% to 10% and after a 30 or so day rest just a little more.
See sheet.

image_rvskcx.jpg


Now these are real life results on USED cells. I'm not about to brake out my Algebra, Calculus, or other means of math books to determine if a cell is good or not. I'm too old for that I will leave that to the young and eager ones.
I have read a bunch of studies on this subject till my eyes have glazed over and I have to get back to reality and say to myself KISS. (Keep it simple stupid)
We are mostly laymen here not collage professors looking for a sponsorship from some agency on how to test LI-ion batteries with outcomes skewedto somespecial interest group or manufacturer.

Now all that being said I am intrigued as to how to go about measuring this "DC IR". I understand that it is a simple ohms law calculation measuring the voltage drop on a batterywith a known value resistor as a load.The issue that I have is that the cell to be tested needs to have some SOC to produce a V drop to begin with.
Where do we start? 25% 50% 75% 100%? Do we need to measure IR at each step of the way to determine a cell being good?

So my next adventure will be to figure out a "simple" micro controller project to emulate a 4 wire DC IR tester that is easy to use and has consistent results.
I have found some sketches and schematics and will peruse this in my spare time.

Wolf
 

completelycharged

Active member
Joined
Mar 7, 2018
Messages
1,008
The IR at 1kHz is more inline with the DI/DT of a fault current. i.e. the effective impedance offered up on an initial short circuit compared to the steady state DC short current which is then chemically constrained at a lower reaction rate.

Is the 1kHz suare wave or sinusoidal ? For kHz loading simple PWM should be ok and more representative of inverter PWM load switching a battery pack might see at 20-30kHz.

I do wonder if the IR at different frequencies and loading would actually give the real chemical signature (and state) of the cell
Volts no load
Load current
Frequency of loading (PWM)
Effective IR

One test would be to load a cell and then sweep the frequency up from say 500Hz to 30kHz and see if there are any specific deviations. Issue here is you then need to look at any stray inducance and capacitance in the setup. Run a dummy cell (power supply) for one test to get a baseline...

The IR does change through the charge state and the makeup of the reactions are different across the voltage range. So, one theory is that if only some of the chemicals in the cell (electrolyte mix) have degraded the cell may be still be fine in one voltage range but not the other.

Some really interesting test results to look forward to !!
 

Redpacket

Well-known member
Joined
Feb 28, 2018
Messages
1,239
I'm curious with real world in mind, eg cell in service, lightly loaded, then another appliance switched on.
Thinking of the no load & then loaded cell, ion/chemistry startup time (if that's a thing?)
Maybe fast switch on of a load on the scope might help show this? Might play with this here & see if I can find anything!
Also, our inverters tend to draw a 100Hz or 120Hz "half-sine" current depending if we're 50 or 60Hz.
Testing IR at these frequencies might be helpful?
Inverter switching frequency(eg 25kHz) would be another one to look at although any half decent inverter input caps should stop much of this.
Another way might be DC step load testing, eg you load the cell with a constant current then step it up/down but not to zero.
 

Bubba

Active member
Joined
May 9, 2018
Messages
264
Redpacket said:
I'm curious with real world in mind, eg cell in service, lightly loaded, then another appliance switched on.
Thinking of the no load & then loaded cell, ion/chemistry startup time (if that's a thing?)
Maybe fast switch on of a load on the scope might help show this? Might play with this here & see if I can find anything!
Also, our inverters tend to draw a 100Hz or 120Hz "half-sine" current depending if we're 50 or 60Hz.
Testing IR at these frequencies might be helpful?
Inverter switching frequency(eg 25kHz) would be another one to look at although any half decent inverter input caps should stop much of this.
Another way might be DC step load testing, eg you load the cell with a constant current then step it up/down but not to zero.

Harmonics ... can get complicated very fast if your not careful.
 

completelycharged

Active member
Joined
Mar 7, 2018
Messages
1,008
The batteries will see some of the HF switching, with any LF based inverter the caps are still just a parallel store to the battery so the DI/DT will be shared between caps and battery. The IR of the capacitors then has a consideration. HF types see a lower average draw because the capacitors / buffer are charged through the wave.

The way I think of a battery is just a massive sheet and at rest the whole of the sheet has a reaction potential for high current delivery, then when a current flows some of the area is depleted and takes a relatively large time to replenish the surface. So, the iniitial very brief current delivery capability of a cell when shorted out will be higher, but brief.

So when is the testing to start "Spock, Do something !" :p
 

Wolf

Well-known member
Joined
Sep 25, 2018
Messages
1,282
completelycharged said:
So when is the testing to start "Spock, Do something !" :p

Waite where did I put my Science officers uniform? Its got to be here somewhere. Ah there it is and look it still fits.............

I was thinking............. I already have the hardware and sketch for testing cells in parallel so why not modify it a bit.

Build a board with 1 proper 4 wire cell holder a 1? 1% resistor (possibly more to toggle a heavy load) severalIFR520 MOSes to control the load(s) in a variety of forms. An INA260 Current sensor and an ADS1115 for 4 wire sense AD conversion. A V divider to make it palatable for a3.3V ESP32 writing data to influx (may have node red as an intermediary for instant data)once a second. Hmm Fascinating................
And I swore I wasn't going to pull out my math books darn yougauss163

Stay tuned

Wolf
PS I do have a day job but no night life :p
 

completelycharged

Active member
Joined
Mar 7, 2018
Messages
1,008
Have a look at a servo driver module (I2C) to use for the PWM output (to switch FET) as this will give a lot more reliable and stable output wile eliminating the need for separate timing loops. Loads of code already exists for controlling them as well.

Could always end up using the PWM to vary the attached load at a later date for other types of testing (measuring voltage drop profile at different loading at different charge state)

The whole setup may end up yelling "canna take it any more Captain !"
 

Wolf

Well-known member
Joined
Sep 25, 2018
Messages
1,282
completelycharged said:
.............. servo driver module (I2C) to use for the PWM output (to switch FET) ...........................
The whole setup may end up yelling "canna take it any more Captain !"

LOL ordered.
Wolf
 

completelycharged

Active member
Joined
Mar 7, 2018
Messages
1,008
On the next generation of tester..... "Make it so...." :D
 

gauss163

Member
Joined
Jun 28, 2020
Messages
284
Since a picture is worth a thousand words, below is a graphical illustration of the points I emphasized in my remarks quoted by wolf in the first post. The graph yields a vividvisualdepiction of the large difference between DC IR vs AC IR.

An AC 1kHz IR meter measures resistance at frequency 10^3 Hz (= 1 ms pulses). In the graph AC IR is the yvalue on the cycle curves above the rightmost x-axis value (= 10^3). The graph shows this 1kHz AC IR grows very little in 100 cycles - from about 64to 70 m?. OTOH, the leftmost frequency .01 Hz uses 100s pulses, which are long enough for the IR from charge transder,diffusion etc to ramp up, which pushes the IR up much higher (above 150 m?). Further, these values increase much more, from about 154to 242m?after 100 cycles. These correspond to the values measured by DC IR meters (usually for shorter pulses of a few secs, which still yield large values, about 150 to 218 m? via the graph)

By only measuring the AC IR (= purely ohmic component) we are completely missing these other large components of IR. Only by incorporating them do we obtain an accurate idea of how the voltage drops under load during longer time DC discharges (hence the true capacity under load). The AC 1kHz IR yields this information only for very short timed loads (< 1 ms), e.g. for a car jump start pack or vaping. It is essentialto assess how these (larger) values evolve too in order to accurately assess battery health. These DC IR values will also help improve other decisions, e.g. they yieldmore accurate matching of cells when building packs - which leads tolonger pack lifetime.


image_zotwww.jpg


The graph is excerpted from p. 17 of this book:Barsukov and Qian, Battery Power Management for Portable Devices, 2013 (at libgen or scribd or amazon). The authors are leading experts at Texas Instruments, responsible for the design of TI's impedance tracking fuel gauge algorithm - which is very widely used (e.g. in most laptop batteries). It's a good place to learn about internal resistance /impedance andrelated matters because this information is crucial to information reported by fuel gauges (which includes health) so is discussed at length there. It was written to be accessible to those without any specialized knowledge of battery electrochemistry (it is much more accessiblethan the brief blurbs prepended to many research papers).
 

completelycharged

Active member
Joined
Mar 7, 2018
Messages
1,008
gauss163 : Nice post, the link does not work for me.

I would be interested in seeing how the curve actually appears at a significant load level at 100Hz to 1kHz. The results do show the chemical boundary quite nice though, around 8 seconds to stabilize.
Questions would be :
1. Was the testing 50% duty cycle of the PWM ? Raises some questions for how 10-1000Hz looks...
2. What is the correlated deviation around 9Hz and just below 1kHz ? Does not seem like resonance (missing harmonics ?).
3. Are the test series at equal spacing or a random sample, i.e. every 10 cycles ? If so why the inconsistent increments (different electrolyte degredation ?)..

Regarding question 1, argumentative, if you can decrease the effective IR by increasing the frequency then why not replace a pure DC connection with a switched 2kHz PWM 50/50 connection ? i.e. hook up all your kit with a 50% duty 2kHz switch ? Maybe the chart is not showing an equal energy per second across the scale ?


Regarding question 2, maybe these correlated deviations (9Hz and 1kHz ?) are the real fingerprints of cells and the state that is revelaed at >10sec load durations... easier way to test ?
 

Wolf

Well-known member
Joined
Sep 25, 2018
Messages
1,282
gauss163 said:
Since a picture is worth a thousand words, below is a graphical illustration of the points I emphasized in my remarks quoted by wolf in the first post. The graph yields a vividvisualdepiction of the large difference between DC IR vs AC IR.

So in a nutshell, correct me if I am wrong, I/we need to build a tester that is 4 wire and has the capability of loading a battery for 100s (extreme) to get the "chemical reaction" to take place then and only then will we get an accurate DC IR.

Sounds good to me. Got any hardware suggestions and plans?
I am working on building such a device but if you have a head start by all means please do share.

I do have a question though.
Playing the devil's advocate................
So I measure the cellsIR with a 1kHz meter and I get an acceptable reading according to my cheat sheet. Plop the battery into an OPUS at a C/D/C rate of 1A and it comes out at 90% SOH. Haven't I basically checked this battery for a high DC IR as I have exercisedthis battery for ?7to 9hoursfor the complete C/D/C cycle andit did not overheat and gave good mAh results.
The same can be said true of a high 1kHz IR reading as I have many examples of those with low mAh results and potential hot running ?45 C.

The other thing is the time it would take to get these DC IR readings for a substantial powerwall say 2800 cells for a 14s200p wall.
It is hard enough to convince some members that IR is actually an important measurement let alone add another coginto the gear.

I will have fun with this and more than likelylearn some more things. I like that.

One thing I do know though.I never trust "leading experts at Texas Instruments" (Or any other EXPERTS)especially the ones that areresponsible for the design of TI's impedance tracking fuel gauge algorithm - which is very widely used (e.g. in most laptop batteries) As I have had plenty of laptop battery fuel gauge inaccuracies cause my work to disappear.----Poof-----
These experts always remind me of the microsoft file transfer utility that tell me how long its going to take only to constantly giving me a moving target.

Wolf
 

completelycharged

Active member
Joined
Mar 7, 2018
Messages
1,008
"I will have fun with this and more than likely learn some more things. I like that." +1 on that and keen to learn from the testing results as well.
 

Bubba

Active member
Joined
May 9, 2018
Messages
264


Bubba said:
gauss163 said:
image_zotwww.jpg


The graph is excerpted from p. 17 of Barsukov and Qian, Battery Power Management for Portable Devices, 2013. The authors are leading experts at Texas Instruments, responsible for the design of TI's impedance tracking fuel gauge algorithm - which is very widely used (e.g. in most laptop batteries). It's a good place to learn about internal resistance /impedance andrelated matters because this information is crucial to information reported by fuel gauges (which includes health) so is discussed at length there. It was written to be accessible to those without any specialized knowledge of battery electrochemistry (it is much more accessiblethan the brief blurbs prepended to many research papers).


https://www.scribd.com/document/369105562/battery-power-management-for-portable-devices

I don't see this chart as useful if you look each cycle is C/10 and the IR change after 100 cycles rises to 240m? ?
100 cycles is nothing! I hope I could get several thousand cycles at that low C rate.
The chart i think is incomplete since they don't list the what the symbols are and what is going on at 1kHz and 10Hz?

My conclusion this graph should have been based on 1000+ cycles and should list % of usable capacity as the trends.
The graph doesn't indicate % loss.

Please someone correct me.
 

gauss163

Member
Joined
Jun 28, 2020
Messages
284
Bubba said:
https://www.scribd.com/document/369105562/battery-power-management-for-portable-devices

I don't see this chart as useful if you look each cycle is C/10 and the IR change after 100 cycles rises to 240m? ?
100 cycles is nothing! I hope I could get several thousand cycles at that low C rate.
The chart i think is incomplete since they don't list the what the symbols are and what is going on at 1kHz and 10Hz?

My conclusion this graph should have been based on 1000+ cycles and should list % of usable capacity as the trends.
The graph doesn't indicate % loss.

Such graphs (Bode plots) are not intended to show capacity loss. Rather their purpose is to show how IR (or impedance) depends on load frequency, in order to give a comprehensive view of how a battery responds to diverse signals. There is no need to do further cycles since the points I made above are already quite clear after 100 cycles.

The same shape graphs appear for all Li-ion batteries, though the absolute numbers will depend on chemistry etc. Such graphs are basic tools that are widely used in battery electrochemistry.

For further background on this and related matters see said book, section 1.3.1 p. 12 where they discuss the closely related Nyquist plot of the impedance spectrum, and see the following section where they explain how IR (or impedance) affects the usable capacity.
 

Bubba

Active member
Joined
May 9, 2018
Messages
264
gauss163 said:
Bubba said:
https://www.scribd.com/document/369105562/battery-power-management-for-portable-devices

I don't see this chart as useful if you look each cycle is C/10 and the IR change after 100 cycles rises to 240m? ?
100 cycles is nothing! I hope I could get several thousand cycles at that low C rate.
The chart i think is incomplete since they don't list the what the symbols are and what is going on at 1kHz and 10Hz?

My conclusion this graph should have been based on 1000+ cycles and should list % of usable capacity as the trends.
The graph doesn't indicate % loss.

Such graphs (Bode plots) are not intended to show capacity loss. Rather their purpose is to show how IR (or impedance) depends on load frequency, in order to give a comprehensive view of how a battery responds to diverse signals. There is no need to do further cycles since the points I made above are already quite clear after 100 cycles.

The same shape graphs appear for all Li-ion batteries, though the absolute numbers will depend on chemistry etc. Such graphs are basic tools that are widely used in battery electrochemistry.

For further background on this and related matters see said book, section 1.3.1 p. 12 where they discuss the closely related Nyquist plot of the impedance spectrum, and see the following section where they explain how IR (or impedance) affects the usable capacity.
So to draw a conclusion from that cart... how the trend seems to be going you will have resistances in several Thousand m? (or in several full ohms).
I would be curious to find out what variability there is in the readings with such wide range as a result of such few cycles.
If someone could find that information it would be useful regarding the 100s test. vs 1kHz I haven't been able to in a quick search.
 

gauss163

Member
Joined
Jun 28, 2020
Messages
284
Bubba said:
So to draw a conclusion from that cart... how the trend seems to be going you will have resistances in several Thousand m? (or in several full ohms).

There are many studies that show how IR evolves over the entire life cycle,but that is not the purpose of the above. Rather, it is simply to contrast the large IR differences atlow-frequency (DC IR) vs high-frequency(such as AC 1kHz IR).

Of course cells will no longer be usable for most practical purposes long before their IR has a chance to grow to monster values (so no one studies them all the way to such extreme overkill).


Bubba said:
I would be curious to find out what variability there is in the readings with such wide range as a result of such few cycles.
If someone could find that information it would be useful regarding the 100s test. vs 1kHz I haven't been able to in a quick search.

You can find some further examples of DC IR vs. AC IR in my initial post.
 

Wolf

Well-known member
Joined
Sep 25, 2018
Messages
1,282
Excerpt from1.3.3 Usable Capacity
Therefore, simple modeling assuming fixed R will not give an accurate estimate of usable capacity. Also battery manufacturers often report battery impedance at 1 kHz. This value cannot be used as an estimate for internal resistance at DC conditions, because low frequency impedance (that corresponds to DC conditions) is much higher than that at 1 kHz. Although DC resistance istypically two to three times the 1-kHz resistance for charged Li-ion cells, the ratio is unpredictable for different states of charge, aged cells, or at low temperatures, so it is better to always measure DC resistance directly. Impedance of 1 kHz is useful for detecting catastrophically failed cells during production because of the fast measurement ability afforded and widely available instrumentation.

I have to slightly disagree with these statements and here is why.
First of, if I measure a cell at 1 kHz I am not trying to ascertain the internal resistance at DC conditions. I am merely running a quick check on the feasibility that this cell is good enough to charge and test.
I am sure the ratio is unpredictable between AC IR and DC IR but I can tell you that AC IR can pretty much predict the SOH of a battery.
These are the same manufacturer cells I have been testing. I'm at 1468 cells recorded so far.
You can see that thesecells havea high IR cutoff of around 42m? measured with a 1 kHz tester. They are marginal but defiantly degraded.

image_rogtru.jpg

The lowest IR cells all are good.

image_sprloh.jpg


Have you even looked at the pivot chart in my harvested cell analysts sheet where you can select a cell manufacturer and mode number and see the IR to capacity correlation? It has over 6000 cells in it from various manufactures and the trend is as clear as crystal.

image_rmsqae.jpg


Again I am not opposed to a DC IR test and am planning of building such a tester.
By the way since you are such an advocate of this method please provide some insight on how you do your testing. Have you built a DC IR tester?Or at least show us how you do it.
It would be nice to collaborate on this project,
Also any sheets that you have to show what cells you have tested, your results as far as DC IR measurements, and its correlation to capacity resultswould be appreciative.

Thanks

Wolf
 
Top