Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
DC IR VS AC 1kH IR measurements
#31
This has been an quite interesting exchange, what may becoming clearer in some regards is that the Wolf tester will hopefully be able to return IR measurements form an arry of frequencies and it is the array of frequencies that will matter.

The testing so far has gravitated to a single IR measurement (or two if the DC value is counted) and Wolfs current process of an IR filter before actually performing a cycle test is the right path. The next tester might (hopefully) be able to fingerprint enough cells to actually derive a close proxy to the actual health of the cell in seconds !

I still wonder if it is possible that the fingerprint of a cell that has been held at high state of charge (medipacks) for the majority of it's life will show up a very distinct fingerprint compared to the same cell that has been cycled from a lower state of charge. Unique pattern of degredation, which shows up in the IR fingerprint.

gauss163, as you will notice there is a mix of reactions to your input and it is this type of input which derives changes and thought and progress (not necessarily always in the direction originally envisaged). For me very much welcome different perspective, however it may apply.

Wolf, the IR to Capacity correlation chart, change it to a scatter plot as it will make it a lot, lot more clearer in many ways. IR one axis, capacity other axis. The correlation will show along with outliers and rough variation pattern (part caused by different internal chemistries). If you have enough cells of the exact same make wihin the scatter plot if this one set was a different colour. What you are also starting to get an idea off is that the testing you are performing and carrying out is starting to get closer to the leading edge, hence the questions as to your "obvious" why has not somebody else done it already....

Complexity rises exponentially for additional gains in "simplicity" until you get E=MC2.
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
Reply
#32
(07-08-2020, 12:58 PM)Bubba Wrote: I too found that most of the book is too vague  https://www.scribd.com/document/36910556...le-devices
I don't think the author did their own tests.  There is too much information missing to support the conclusions they arrive at.  What did the waveforms look like?  What equipment was used?
Is this based on 1 cell or the results of many?
What  What occurred with the samples at 1kHz... they jumped.

My honest feeling is that if this information was accurate you would see more papers and data about using lower freq. to determine battery health by now.  Published in 2013 the book is 7yrs old and there have been advances.  On the surface the book looks good, but it brushes over things and doesn't seem to provide proper reference to the backing data.  It's a book to make $ lacking engineering support and focus.
I did find the BIT-BMS which uses low freq. but the low freq is only used for SOC% and 1kHz is still used to determine if cells are good or not.

As the introduction hints, that book is targeted at "engineers and do-it-yourself enthusiasts". Yes, of course, it would be better if it included more literature links but, alas, this is the norm for such popularizations of science. It's rare that leading experts even have the time or motivation to compose popularizations. We should be grateful when they do so. If you know of better introductions then by all means post them. In my experience they are few and far between.

Anyone who is seriously interested in learning more will have a good enough foundation from the book that they should be able to search the web to delve further.
Reply
#33
Found the sheet, starting to play....

IR vs Capacity scatter (only showing the bulk of points in the 80-100 range)


CGR18650CG - interesting falling off the cliff



This makes me wonder, the left axis should ideally be the "unknown" remaining cycle life (Wh throughput), because the capacity of the cell does not fail linearly and for some rather fast at the end of life. hmmmm.... The point here is even thought a pack may have high mAh cells (if IR is ignored) in some will fail far far quicker than others and the big question is can these cells be identified with the next gen Wolf tester ?
Redpacket likes this post
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
Reply
#34
(07-08-2020, 03:07 PM)gauss163 Wrote: The DC IR should reveal more, but there could also be other factors that come into play. Are all the cells from the same source? Are all cells used or are some used and some new-old-stock?

All cells where recovered from various laptop, medical, tool and whatever had 18650 batteries in it. No NOS.
I tried to stick with brand name stuff as the "replacement" batteries generally had generic chinesium cells in them.
So I would say 95% of the batteries recorded had certainly seen a multitude of cycles.
The laptop packs that had the sweet smell of soda or the district aroma of coffee usually produced some good cells as the laptop died and the battery was discarded.
Bonus for me...... Big Grin

Quote:gauss163 
I need the link to the spreadsheet with the DC IR values, and also the link to the posted graph with IR vs remaining capacity. 
Everything is here you should be able to download it and at least get open office or some other free office suite to visualize the data.
https://drive.google.com/file/d/1NujY1eO...sp=sharing
Quote:gauss163
I don't build powerwalls so I have no further data on that. But I have extensive knowledge of Li-ion batteries from prior (professional) work, so I may be able to help in various ways.

 OK that makes sense and any guidance is appreciated.

You may be able to use the online Excel viewer, editor. might have to put up with some adds but may help you.
Wolf
For Info Google Drive

Great ABB shunt breaker seller River City Industrial
Great prices and superb service.
 [Image: em2566%20(1)__19172.1521541365.jpg?c=2]

Not your average Wolf       
            Cool
Reply
#35
(07-08-2020, 03:10 PM)completelycharged Wrote: .................. What you are also starting to get an idea off is that the testing you are performing and carrying out is starting to get closer to the leading edge, hence the questions as to your "obvious" why has not somebody else done it already....

Glad you found the sheet. Excellent!

Why Has nobody done this already?
Well for one thing it's time consuming. You are looking at data accumulated over a 10 month period. Almost every day entering data and testing.
Then another 3 month or so rechecking all the cells.
 
Maybe someone has done this before we just don't know it.
Although in this day and age of the internet and information sharing you would think we would know about it.

Quote:completelycharged
.............and the big question is can these cells be identified with the next gen Wolf tester ? ............


I may have to put that project up on github and maybe some skilled sketch writers can assist as my code writing skills are still in its infancy. Although quickly growing to adolescence. Smile
Wolf
completelycharged and Redpacket like this post
For Info Google Drive

Great ABB shunt breaker seller River City Industrial
Great prices and superb service.
 [Image: em2566%20(1)__19172.1521541365.jpg?c=2]

Not your average Wolf       
            Cool
Reply
#36
And so it begins.....................

What to do first?
Yes find the 4 wire Probes for the fluke. ✔
Find 1Ω 100W resistors. ✔
Test Fluke 4 wire ✔

Test 1Ω resistors for best and closest. ✔

Solder wires to Resistor and recheck Ω ✔

Start the layout plan. ✔

Now waiting on the PCA9685 should be arriving today.✔

Post to board.✔
Ordered 4Ω 100W resistors for a 1.05A draw.✔
Wolf
completelycharged likes this post
For Info Google Drive

Great ABB shunt breaker seller River City Industrial
Great prices and superb service.
 [Image: em2566%20(1)__19172.1521541365.jpg?c=2]

Not your average Wolf       
            Cool
Reply
#37
I see Spock giving the sign of approval to the project...

The PCA9685 is good for 24Hz - 1526Hz with 12 bit PWM accuracy so minimum pulse width of 160nS (1526Hz at 1bit PWM on state) so some interesting range, but nothing above 1.5kHz for now which may turn out to be ok as it avoids in the near term the wonders of harmonics and reactive power from places and wires you did not expect.

The ADC will ideally need to be syncronised to the pulse position to get a consistent reading if sampling at pulse resolution or with a longer sample window to average out the variations. I'm thinking here of say when a high load (say 4A from 4x4Ohm in parallel) at 1kHz is attached the voltage will drop from initial steady state quite quick and the shape of the drop as the charge state decreases will be interesting. 4 seconds (4000 samples ideally, 50+ ok) to get a pulse by pulse response, although this will need a different coding approach (buffer to memory then save to db). Will help out on the code..

I'm wondering if there is some additional good info a cell gives out relating to the energy out vs volt drop vs time in those initial seconds for high loading.

Maybe have the load arrangement software switchable ? approx 1A to approx 4A ? Arrange 4 and 1 Ohm resistors in parallel and switch in/out the 1 Ohm.
If you can't quantify how much they cost, it's a deal, I'll buy 5 of them for 3 lumps of rocking horse ......
Reply
#38
(07-10-2020, 02:31 AM)completelycharged Wrote: The PCA9685 is good for 24Hz - 1526Hz with 12 bit PWM accuracy so minimum pulse width of 160nS (1526Hz at 1bit PWM on state) so some interesting range, but nothing above 1.5kHz for now which may turn out to be ok as it avoids in the near term the wonders of harmonics and reactive power from places and wires you did not expect.
I know it has been a while since my electronics theory, can you get a PWM signal without harmonics.  The sharper the on/off the higher the frequency the components to make it?

Too bad the PCA9685 doesn't allow each PWM to have it's own independent frequency.  Does anyone know of one that does?
Reply
#39
(07-10-2020, 02:31 AM)completelycharged Wrote: [...] I'm wondering if there is some additional good info a cell gives out relating to the energy out vs volt drop vs time in those initial seconds for high loading [...]

We obtain additional info by correlating the DC voltage drop with the (real part) of the impedance spectrum, which allows us to effectively decompose the DC resistance into its conceptual components: ohmic, charger transfer, diffusion, etc.  This is explained well in the Barsukov book I linked above - see esp. pp. 9-11 where you'll find a guided tour by an expert of the Nyquist-plot view of the impedance spectrum.  

I will post some links and tips soon that will hopefully be useful to wolf or anyone else that has interest in these matters (alas, the last couple of days were unexpectedly hectic so my spare time was limited).
Reply
#40
OK while you guys work out the theory here is the prototype for the hardware.
All attached devices are functional and recognized by the ESP32 love that i2c bus.
The only problem was the INA260 and the PCA9685 board had an address conflict...... . well a little dab of solder took care of that. Big Grin
Still got to hook up the V divider for the ADS1115 and get the DS18B20s  online but that's easy.
Now to start combining the code for the hardware to make some magic happen...............

Wolf
Redpacket, Bubba, completelycharged like this post
For Info Google Drive

Great ABB shunt breaker seller River City Industrial
Great prices and superb service.
 [Image: em2566%20(1)__19172.1521541365.jpg?c=2]

Not your average Wolf       
            Cool
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)