DIY portable, useful LiFePO4

[booster]

Very interesting, I would say, looks like you are learning a lot. It really does get into splitting hairs when .1v can be a determining factor between good and not so good. I think you seem to be seeing more evidence that especially when you go all the way full like you would for a balance, the current may well be a necessary input for the best consistency and accuracy.


How much do the voltages on these cells change with temperature, as in how much temp change does it take to generate a change of .1v as that seems to be the kind accuracy needed in this process?



The increased sag back on degrading cells that was mentioned probably would be expected, I think, nice to know typical amounts. Did any of the articles relate the change in sag back voltage to how much total capacity the cells had lost? If there is a relatively predictable relation, that could be used as a much easier way to evaluate battery condition over time.

[markopolo]

I didn't come across any mention of how to relate capacity loss to OCV.

I did see stuff that said temperature does affect measurements. That's makes it challenging to say the least in an RV environment.

I've ordered a voltage reference device and it is supposed to come with precise calibrations written on it but you don't know what you're actually gonna get until you get it at certain price points.

I'm going to guess here that 3.5VPC OCV on my cells could be 99.5% of full. The cells tested so far could have been at 99%.

2% of the energy came from the first half hour of the 25.5 hour test chart I previously posted.

That first 1/2 hour looks like a dramatic drop in that chart but sure looks different when you see it as it is happening.


first half hour.jpg

[JohnnyFry]

I will be using this device from Amazon. It is settable and has a readout. The relay will handle 10 amps.

https://www.amazon.com/gp/product/B0...?ie=UTF8&psc=1

[markopolo]

That's a good link - the reviews have clear setup instructions.

There's a similar unit out there without the display and it uses dip switches to set the parameters so might have a bit less power consumption.

No doubt you already plan to connect it to the output side of the BMS but I though I'd mention it for anyone else who's getting into this. Any item that puts even a small load on the battery needs to be on the output side of the BMS so that the BMS always protects the battery.

With a lot of these "12V" devices, input voltage might need to be considered. 12V computer fans for example have various input voltage tolerances and you can add a resistor to lower the voltage reaching the fan to keep it within specification. Someone with more electronics knowledge than me might step in here and explain how/if exceeding input voltage affects the device for the long term.

[markopolo]

My thoughts on charging voltage and resulting OCV are changing again.

These cells would never end up resting at 3.6V or 3.5V - it's just not possible. Next on the agenda for me is figuring out exactly what voltage they stabilize and settle at and then figuring out the most effective way to get them to that point.

[markopolo]

If I was confused before then I'm perplexed now.

OCV went up ......

Average OCV after elapsed time:

9 hours: 3.510 V
24 hours: 3.442 V
36 hours: 3.470 V

The room is temperature controlled with thermostat precision being listed ± 0.15ºC / 0.27ºF.

Same meter used. Fresh batteries a week ago. Meter and cells remain undisturbed on a work bench. The cells are not handled or moved to take voltage measurements. The cells are numbered and lined up in order.

Both the decrease and increase in OCV are consistent with every cell's OCV decreasing at 24 hours and increasing at 36 hours.

All of the decreases and increases are consistent with the average decrease and increase.

One thing that stands out is that the cells with greater decrease do not consistently have the greater increase and vice versa. There is a bit of a trend on greater decreases having higher increases but on the decrease side it's more random.

If it's a tiny bit of temperature variation causing this then I can't imagine how you'd ever get accurate or consistent readings in an RV even from hour to hour.

I guess I should next test using two meters to see if only one of the meters varies from test to test.

[booster]

Originally Posted by markopolo

If I was confused before then I'm perplexed now.

OCV went up ......

Average OCV after elapsed time:

9 hours: 3.510 V
24 hours: 3.442 V
36 hours: 3.470 V

The room is temperature controlled with thermostat precision being listed ± 0.15ºC / 0.27ºF.

Same meter used. Fresh batteries a week ago. Meter and cells remain undisturbed on a work bench. The cells are not handled or moved to take voltage measurements. The cells are numbered and lined up in order.

Both the decrease and increase in OCV are consistent with every cell's OCV decreasing at 24 hours and increasing at 36 hours.

All of the decreases and increases are consistent with the average decrease and increase.

One thing that stands out is that the cells with greater decrease do not consistently have the greater increase and vice versa. There is a bit of a trend on greater decreases having higher increases but on the decrease side it's more random.

If it's a tiny bit of temperature variation causing this then I can't imagine how you'd ever get accurate or consistent readings in an RV even from hour to hour.

I guess I should next test using two meters to see if only one of the meters varies from test to test.

This is the kind of thing I was thinking about when I asked the temp variation question. The voltage variations are so small in the areas that are quite important, like end of charge and near full SOC, that measuring errors of other tiny influences could alter the readings enough to make them much less useful.


Having two meters certainly would be a good idea to see what that would show. You may want to take a good look at the detailed specs, if you can get them, for the meter you are using. The accuracy and repeatability are often quite confusingly stated with a couple of layers conditions. They can also range for a simple +/- percent of full scale, to a +/- of a fixed amount, to a more complex thing like +/- "X" percent of reading +/- "Y" significant digits. When you are looking at 1/100ths of a volt even the fact that you are probably using probes can be an issue as the readings can change with pressure, surface condition and oxidation, whether you are touching the metal probe, how good the contact is on the probe to meter cord plugs, etc. Autoranging meters can make trying to figure out what your rated accuracy at a given reading is difficult sometimes.


It will be interesting to see if the batteries do stabilize after a while, and then start a slow drop over time.



Could it possibly take that long for the inside of the battery temp to stabilize?


On edit--Here is a link to a spec sheet for 3 of the better Fluke meters that shows the how they list the accuracy numbers. The % is of the reading and the number of counts is the last displayed number variation. These a very good meters and better than most of us have.


https://dam-assets.fluke.com/s3fs-pu...GRFQumdcukzR6A


And here is one for a $45 Klein (Klein makes decent stuff)



https://data.kleintools.com/sites/al...110ART_WEB.pdf

[JohnnyFry]

Something is wrong with this thread. When I attempt to load page 5 (so I could see Boosters last word of wisdom) I get immediately redirected to what looks to be a text page with no opportunity to navigate elsewhere unless I hit the back arrow on the browser. What gives?

[booster]

Originally Posted by JohnnyFry

Something is wrong with this thread. When I attempt to load page 5 (so I could see Boosters last word of wisdom) I get immediately redirected to what looks to be a text page with no opportunity to navigate elsewhere unless I hit the back arrow on the browser. What gives?

They had mentioned some changes coming, but don't know if that would cause it. Had to do with https sites linking or such, I think.



Page 5 is working normally for me, although there certainly may not be any wisdom on it, at least from me

[markopolo]

Page 5 seems OK to me.

I think I learned a bit about DMM accuracy after Boosters post sent me searching for info.

Multi-meter accuracy really varies a lot. If it's needed for checking RV related DC voltage then pay close attention to the accuracy rating for up to 20VDC.

I'll use the budget Fluke 101 and the range suitable to test individual cells as an example.

Range: 6.000 V
Resolution: 0.001 V
Accuracy: 0.5%±3 digit

I think the math looks like this for a single cell: ( from https://www.designworldonline.com/ho...eter-accuracy/ )

(3.500)(0.5)/100 +0.003)

I put it in Google's calculator here: (hopefully I did it right)

https://www.google.com/search?q=(3.5...2F100+%2B0.003)

The result is 0.0205 which I guess would be rounded up to 0.021 if four number show on your DMM display. So my 3.5V cell could display as 3.479 to 3.521!

A DMM that has ±(0.1% + 3 counts) accuracy would be so much better.

Using Google calculator: https://www.google.com/search?q=(3.5...2F100+%2B0.003)

The result is 0.0065. If five numbers show on your DMM display then the 3.5V cell could display as 3.4935 to 3.5065. That's so much better.

I saw a Dawson Tools DDM645 on Amazon (US) for $40 so I figured I'd take a chance on it. It might be discontinued as the Dawson Tools website is gone. They might be part of Aya Instruments now - https://ayainstruments.com/ - but the DDM645 isn't there with the other Dawson units. The DDM645 datasheet can be found here: https://www.jameco.com/Jameco/Produc...DS/2208462.PDF - it shows ±(0.1% + 3 counts) up to 20VDC so it should be good enough fo RV related use.

I think the UNI-T UT61E - https://www.uni-trend.com/html/produ...ies/UT61E.html - would have been nice to get but it wouldn't have come direct from Amazon & that can be problematic for me if having to return it.

-------------------------------------

Edit to add:

48 hr OCV average: 3.461V, down 0.009V from 36 hr measurement.
2nd meter concurs, down 0.00875 (avg).


-------------------

[booster]

Originally Posted by markopolo

Page 5 seems OK to me.

I think I learned a bit about DMM accuracy after Boosters post sent me searching for info.

Multi-meter accuracy really varies a lot. If it's needed for checking RV related DC voltage then pay close attention to the accuracy rating for up to 20VDC.

I'll use the budget Fluke 101 and the range suitable to test individual cells as an example.

Range: 6.000 V
Resolution: 0.001 V
Accuracy: 0.5%±3 digit

I think the math looks like this for a single cell: ( from https://www.designworldonline.com/ho...eter-accuracy/ )

(3.500)(0.5)/100 +0.003)

I put it in Google's calculator here: (hopefully I did it right)

https://www.google.com/search?q=(3.5...2F100+%2B0.003)

The result is 0.0205 which I guess would be rounded up to 0.021 if four number show on your DMM display. So my 3.5V cell could display as 3.479 to 3.521!

A DMM that has ±(0.1% + 3 counts) accuracy would be so much better.

Using Google calculator: https://www.google.com/search?q=(3.5...2F100+%2B0.003)

The result is 0.0065. If five numbers show on your DMM display then the 3.5V cell could display as 3.4935 to 3.5065. That's so much better.

I saw a Dawson Tools DDM645 on Amazon (US) for $40 so I figured I'd take a chance on it. It might be discontinued as the Dawson Tools website is gone. They might be part of Aya Instruments now - https://ayainstruments.com/ - but the DDM645 isn't there with the other Dawson units. The DDM645 datasheet can be found here: https://www.jameco.com/Jameco/Produc...DS/2208462.PDF - it shows ±(0.1% + 3 counts) up to 20VDC so it should be good enough fo RV related use.

I think the UNI-T UT61E - https://www.uni-trend.com/html/produ...ies/UT61E.html - would have been nice to get but it wouldn't have come direct from Amazon & that can be problematic for me if having to return it.

Yeah, finding a good meter that is also affordable for a tough application like you have is certainly not easy. Of course there is also the inability to check the calibration without sending them in periodically and the fact that the non-mainstream brands and suppliers may not have real traceable calibration certificates.


This is also probably an indication of just how difficult it might be to build in a balancing cycle and try to hit that totally full without too full point on all the cells all while trying to monitor it remotely so more connections to mess up the readings. This would probably help explain some of the reduced voltages we have seen as well as the relatively generous top and bottom space put in on a lot of the full system packages. Most likely any loss of life by not quite getting to optimum conditions is much less than going past optimum losses. Sorting of like with lead acid undercharging gives the slow death and overcharging does the fast death.

[ikanode]

re: meter accuracy and stability

You can get a $20 voltage standard on eBay to check your meters for stability and accuracy.

"AD584KH 4-Channel 2.5v/7.5v/5v/10v High Precision Voltage Reference Module"
<https://www.ebay.com/itm/AD584KH-4-Channel-2-5v-7-5v-5v-10v-High-Precision-Voltage-Reference-Module-/181112567159?hash=item2a2b267577>

Scroll down and look at the photos showing the accuracy of the standard a various voltages. Each unit has those values written on it--so you can mathematically correct them.

I then write on each meter how high or low it reads. By measuring the standard's voltages multiple times over hours and days (at approximately the same temp) you get a good idea of voltage variations due to stability issues in the meter.

Then test at different temperatures (for the meter, not the standard) and see how much the temperature the meter impacts the readings.

By isolating issues with the meter, you can put varying cell readings in perspective.

[booster]

Originally Posted by ikanode

re: meter accuracy and stability

You can get a $20 voltage standard on eBay to check your meters for stability and accuracy.

"AD584KH 4-Channel 2.5v/7.5v/5v/10v High Precision Voltage Reference Module"
<https://www.ebay.com/itm/AD584KH-4-Channel-2-5v-7-5v-5v-10v-High-Precision-Voltage-Reference-Module-/181112567159?hash=item2a2b267577>

Scroll down and look at the photos showing the accuracy of the standard a various voltages. Each unit has those values written on it--so you can mathematically correct them.

I then write on each meter how high or low it reads. By measuring the standard's voltages multiple times over hours and days (at approximately the same temp) you get a good idea of voltage variations due to stability issues in the meter.

Then test at different temperatures (for the meter, not the standard) and see how much the temperature the meter impacts the readings.

By isolating issues with the meter, you can put varying cell readings in perspective.

Yes, absolutely, if you have a reference or if you are just measuring very close together readings to compare them, the actual accuracy of the meter itself gets very much less important. What does matter at that point in the repeatability of the meter you are using. Luckily, that is pretty easy to do by doing repeated readings on the same items in a short time. Normally, I would recommend not to take 10 readings right in a row on one cell, then move to the next, but instead to check all of them in 10 times serially. Going one way and then opposite order may be better, but probably not a lot. Most of the lower end meters seem to be much better at repeatability than they are at accuracy over the full range, so there is a good chance you will be able to narrow your error window a lot.

[markopolo]

Originally Posted by ikanode

re: meter accuracy and stability.............

I think that's a good idea. I ordered a voltage reference a couple of weeks ago and am just waiting for it to show up.

I like your idea of testing at different temperatures. I think the cells are getting close to a stable OCV. At 60 hours voltage had gone down a bit -0.008V. (two meters concur that there was a drop) I'm a bit surprised that they're not more stable than that yet. Once I get a more stable OCV then I could test at different temperatures assuming that some time is needed for the cell and maybe the meter to adjust to the different temperature.

I do think that getting a meter with a 0.1% accuracy or better spec should be included in a LiFePO4 budget if you don't already own one.

Re: repeatability or consistency - it's repeatability from one session to the next that concerns me about my best meter right now. That's why I ordered a new one. Within the same session, results are repeatable now. I'm leaning toward session consistency being the issue that made it appear that voltage had gone up in session/test #3. (not sure if the meter is consistently inaccurate)

[markopolo]

Rounding - this image shows the impact of rounding.

rounding.JPG
( borrowed from https://d32ogoqmya1dw8.cloudfront.ne...t_advan.v3.pdf to illustrate point )

Average voltage of 24 cells:
(charged to 3.600V & almost zero current)
Primary meter #.### display (rise or drop from previous measurement in parenthesis)
9 hrs: 3.5103V
24 hrs: 3.4419V (-0.0684)*
36 hrs: 3.4703V (+0.0284)
48 hrs: 3.4610V (-0.0093)
60 hrs: 3.4526V (-0.0084)
72 hrs: 3.4469V (-0.0058 )
84 hrs: 3.4408V (-0.0061)

*It does appear that the 24hr measurement (24 cells, individually) was inaccurate. Seems obvious to me now. I have not been able to recreate a similar error.

2ND Meter averages, #.## display
37 hrs: 3.455V
48 hrs: 3.447V
60 hrs: 3.437V
72 hrs: 3.434V
84 hrs: 3.426V

Maybe I'm recording the self discharge now.

[markopolo]

More on understanding multimeter accuracy specifications.

From: https://www.keysight.com/main/editor...158&id=1603103

The calculation shown is:
error % * voltage + counts * display range

My example (DDM645 multimeter):
error % = 0.1%
voltage = 14V
counts = 3
at range = 0.001

Measuring 14V: 0.1%*14+3*0.001 = 0.017V so 13.983V to 14.017V could be displayed. Not too bad.

If my understanding of this is correct then the tolerance of three Fluke models ( https://dam-assets.fluke.com/s3fs-pu...GRFQumdcukzR6A ) measuring 14VDC would look like this:
Model 175: 0.15%*14+2*0.01 = 0.041 so 13.96V to 14.04V
Model 177: 0.09%*14+2*0.01 = 0.033 so 13.97V to 14.03V
Model 179: 0.09%*14+2*0.01 = 0.033 so 13.97V to 14.03V

Not too bad but maybe not the best choice IMO for measuring the 14VDC range (unless new or recently calibrated).

Cross checking using the information on this site: https://www.designworldonline.com/ho...eter-accuracy/

The calculation shown is:
voltage * error % + counts in applicable display range

My example (DDM645 multimeter):
voltage = 14V
error % = 0.1%
counts = 0.003

14*0.1%+0.003 = 0.017 (no need for rounding as it displays at a greater resolution)

The answer is the same using either site's calculation.

I'll apply it to Klein Tools M400 ( https://data.kleintools.com/sites/al...110ART_WEB.pdf ):

Site 1: 1.2%*14+3*0.001 = 0.17 (rounded due to lower display resolution)
Site 2: 14*1.2%+0.003 = 0.17 (rounded due to lower display resolution)

13.83V to 14.17V could display. A 0.34V range is huge when it comes to LiFePO4 batteries.

If choosing a multimeter for measuring LiFePO4 batteries then three criteria should be considered.
1. accuracy % at the DC voltage you are measuring
2. counts at least significant digit
3. display resolution

It's important to note that the error % shown above for all meters mentioned is the range of possible error. Any of the meters might show very accurate voltage when new and would show accurate voltage if recently calibrated.


If anyone here has some spare time, it would be much appreciated if the accuracy of this post could be verified.

[booster]

I do have one question off the top. It may or may not apply.


In the formula they have all the normal stuff and then add range to it at the end where you would expect the term to be least significant digit. Range would normally mean to me the "scale" of the meter at the current setting as in zero to 999 miliamps or such. Usually if the use range at it means as I said, there would not be an entry for the actual voltage reading. Many lower accuracy meters would give the accuracy as "% of range" for instance so if you were on a 100 milliamp scale at 2% accuracy the it would be .002 milliamp for whatever reading you get so very inaccurate at low readings.


I do think the way you used it looks correct.



All this can also get messed up with some of the meters that don't do a full 0-9 on the last digit and may only do every other number or even just 0 or 5. I assume on them the count range in your formula would be .0002 or .0005 for one of them.


I think it is a lot like the "massaged" battery specs we see all the time lately, with the meter manufacturers using new or repurposed callouts or abbreviations, and different calculation methods, to make things look better than they really are. Of course some of it could be just translation issues as most now are imported from Asia. We fought this kind of stuff all the time while I was working, as the tech sales folks would constantly try to confuse the facts to get an order. Many would go silent when we would tell them we require tested spec data, and a post purchase performance guarantee for our particular application. Good vendors had no problem with any of that stuff.

[markopolo]

Thanks for checking it Booster.

Another link explaining it: https://bkprecision.desk.com/custome...specifications

I'll use their numbers to show how display resolution greatly affects the accuracy specification:

0.5%*2+10*0.0001 = +/- 0.011
0.5%*2+10*0.001 = +/- 0.02
0.5%*2+10*0.01 = +/- 0.11

Fluke uses error % * voltage + counts * display resolution here: http://support.fluke.com/calibration...00_ENG_A_W.PDF

0.003%*10+2*0.0001

You can just copy and paste that into Google search, then adjust the values to match the meter being considered for purchase.

https://www.google.com/search?&q=0.003%25*10%2B2*0.0001

0.003% = accuracy percentage of listed voltage range being measured
10 = voltage being measured
2 = counts applicable to listed voltage range
0.0001 = display resolution of the multimeter

[booster]

I much prefer the term "display resolution" over count, range, or least significant digit, as it is understandable to most people, I think, and it is what we are talking about. It also makes it obvious how to handle the units that don't us 0-9 in the last digit.

[markopolo]

It's really looking like lower voltage charging is a good idea for routine charging.

The most critical thing for longevity of these cells would seem to be that they are all used equally.

3.45VPC (13.8V) seems to be what a "good cell" OCV would be after a full charge and an overnight rest. That voltage would be pulled down if paralleled with "average cells" and any other "less than good cells".

When charging 24 cells in parallel, I noticed that the VPC varies. If the charge connection points are at either end or the middle the cells nearest to the charge connection points get to the highest voltage. That was expected. It was unexpected to me to see that when charge connection points were attached to opposite ends of the 24 parallel connected cells the highest voltage were at both ends and the middle cells were low. It seemed to me that middle cells voltage might have been forced lower. I'm not certain of the before charging voltage to know for sure.
It looked like this:

3.45V,3.45V,3.44V......3.41V3.40V,3.40V,3.41V..... .3.44V3.45V,3.45V

I see that as reason to limit charge voltage on long strings of paralleled cells so that the cells nearest to the charge connection points don't sit at higher voltage for too long.

I assume that a similar effect will happen when discharging.

I think that the distributed charge/discharge connection I plan to use for the 40 cylindrical cell pack will help greatly toward the goal of using every cell as equally as possible, both when charging and discharging.

4S10P_spread_charge _discharge_equally.jpg