LFP Cell Selection based on Self-discharge Voltages

[pohjeremy]

Hi All,

Need your well-informed opinions.

The attached.xls shows 2 OCVs for 36 cells within a 47-day self discharge time lapse.

I'm trying to pick the best/most similar 32 out of 36 cells before parallel balancing. Please advice which 32.

Series assembling would then follow.

The .xls data might not be very indicative...or can it be..? That's all the testing equipment I have: A multi-meter.

Overall, I think 35 are aptly choice-picked by CALB's selection team...perhaps except Cell 27. What say you?

All opinions/advice will be cherished.

 

[dnmun]

we have shown that self discharge rates for lifepo4 cells does not correlate to actual capacity or Ri of the cells and is not an appropriate tool to use in building a battery pack.

it is best to distribute the cells according to total stored charge capacity so that each channel has equal capacity.

 

[pohjeremy]

Hmmm...I see. Thanks.

Therefore, 1 cell self-discharging about 55% quicker than the other 35 tells nothing about the overall health/state of that cell, when compared with the rest...?

Also, they were sent at 30% SOC...and they estimated 3% SOC self-discharge/mth. So they could be about 25% SOC now. What SOC %tage would be ideal for parallel balancing? Could charge or discharge them a little. Thanks.

 

[dnmun]

no, self discharge rates from full charge when removed from the charger as measured by the rate at which the cell voltage drops is not a useful tool to evaluate the total charge storage capacity of the cell. the single most important number is the capacity of the cell to store charge. they gave you the capacity numbers in that invoice so that is what you should use to pair up the cells with each other. pick higher or lower so the pair of cans together is as close to the average as is possible. you want the range of capacity on each channel to be minimized.

 

[pohjeremy]

Good stuff...

But the cells logged by the factory line 47 days ago seem very identical...
The IC told me that out of 100k+ cells, this is the most similar group he's ever sold (because I requested for that)... Now maybe I had requested wrongly and should have asked for dissimilar ones..(a significant mixture) ?

 

[Hillhater]

Personally, I think you are " sweating over minor detail" ... as far as a few mVolts on self discharge is concerned!
If you really want to understand the condition of your cells, and since you don't trust the factory data, (~40% variation in IR ?)....figure out how to measure their DCInternal resistance on discharge...it is not difficult and there are videos to show how.

 

[liveforphysics]

we have shown that self discharge rates for lifepo4 cells does not correlate to actual capacity or Ri of the cells and is not an appropriate tool to use in building a battery pack.

it is best to distribute the cells according to total stored charge capacity so that each channel has equal capacity.

It's unrelated to a cells internal resistance and capacity, but it remains the KEY indicator for which cells to scrap when making a pack.

There is no form of self-discharge that does not also result in micro-gassing over time. Cells have a finite internal volume, therefore regardless of how outstanding a cells capacity and impedance may be, if it has abnormal self-discharge and you put it in your pack you have just created a certain pack failure.

That is scary they drop at least 10mV just sitting. Personally I would scrap them all, but that may actually be par for the course for these cells, having the temp's not matched makes not-apples to apples.

 

[pohjeremy]

It's unrelated to a cells internal resistance and capacity, but it remains the KEY indicator for which cells to scrap when making a pack.

There is no form of self-discharge that does not also result in micro-gassing over time. Cells have a finite internal volume, therefore regardless of how outstanding a cells capacity and impedance may be, if it has abnormal self-discharge and you put it in your pack you have just created a certain pack failure.

That is scary they drop at least 10mV just sitting. Personally I would scrap them all, but that may actually be par for the course for these cells, having the temp's not matched makes not-apples to apples.

Thanks. Time to conduct a fair self-discharge test then....but perhaps just a 2-week test at the same temp / environment / etc..

Hopefully they discharged to 30% SOC before shipment and the initial 3.297 - 3.300V 9 weeks back wasn't @ 30%. Sounds more like 50% to me when they were determining OCV.

 

[pohjeremy]

It's unrelated to a cells internal resistance and capacity, but it remains the KEY indicator for which cells to scrap when making a pack.

There is no form of self-discharge that does not also result in micro-gassing over time. Cells have a finite internal volume, therefore regardless of how outstanding a cells capacity and impedance may be, if it has abnormal self-discharge and you put it in your pack you have just created a certain pack failure.

That is scary they drop at least 10mV just sitting. Personally I would scrap them all, but that may actually be par for the course for these cells, having the temp's not matched makes not-apples to apples.

Hi... 15-day test done. You think 35 cells are ok for a 32s selection? And wld 30-day be better?

Thanks!

Personally, I think you are " sweating over minor detail" ... as far as a few mVolts on self discharge is concerned!
If you really want to understand the condition of your cells, and since you don't trust the factory data, (~40% variation in IR ?)....figure out how to measure their DCInternal resistance on discharge...it is not difficult and there are videos to show how.

Ok...I'll figure out how when assembled.

The mV on cell 27 is suspect and CALB may send me a replacement. I love the cells...darn difficult to get them in and are supposed to be premium LFPs this era. 140wh/kg.

As for the iR variation, you mentioned it wasn't a good indicator so that data is obsolete for now.

As for the cells, I hope their real combined iR is low, even during discharge. They're supposed to be good cells.
But the Grade 2 copper (99.8 - 99.9%) bus-bars: 31(L50*W15*T3mm) + 1(L80*W15*T3mm) + 64pcs of M10*6 C1 copper (>99.9%) bolts + about 30cm of 6ga wiring...I won't know now.

Hey, thanks!.

 

[dnmun]

i will repeat. self discharge is not what you should use to select cells. you should use cells where they have the same capacity to store charge. the BMS will compensate for any self discharge and balance the pack each time that it is charged up to the balancing voltage. your cells are all documented to have very close capacity so there should be no problem getting enuff to make a pack from. it is obvious they all self discharge about the same amount from the resting voltages you have listed.

 

[pohjeremy]

...it is obvious they all self discharge about the same amount from the resting voltages you have listed.

Cell 27 does not. It self-discharges more than 5 times the rate relative to the other 35. And I trust my stringent test conditions.

i will repeat. self discharge is not what you should use to select cells...

This might not correspond to what guru-member 'liveforphysics' earlier elucidated:

It's unrelated to a cells internal resistance and capacity, but it remains the KEY indicator for which cells to scrap when making a pack.

There is no form of self-discharge that does not also result in micro-gassing over time. Cells have a finite internal volume, therefore regardless of how outstanding a cells capacity and impedance may be, if it has abnormal self-discharge and you put it in your pack you have just created a certain pack failure...

Now, if one engineer says a 15-mile suspension bridge will buckle and another says it won't, will a driver logically risk the crossing? I'm not using that cell.

Fortunately, because I treated those people right, the very ethical CALB staff is fighting for a cell 27 replacement.

...the BMS will compensate for any self discharge and balance the pack each time that it is charged up to the balancing voltage...

The cheaper BMSes did not and neither did the $1500 one. If fact, they gave the e-bike road stalls and fire in my eyes. So if you put in extra work-hours for blazing bucks but chalk up on hospital bills, why not work less yet end up with the same retirement wallet?

First, let BMSes become as likely to smoke out as an LFP cell would explode while at rest.
Otherwise, please give me the confidence that the BMS you promote is that high on reliability and that it suits this set-up.



Thank you.

 

[agniusm]

I would discard that 27th cell as well but going without BMS, well, that's something I wouldn't do. I have noticed increase of people using BMS and that is with reason. Perhaps crappie BMS is better than human BMS in the end

 

[wb9k]

You really need to do this test at 100% SOC. Here's the A123 test:

Charge CV at 3.60 Volts until current <1/20C
Remove charger. This is zero hours.
Record OCV at 24 hrs.
Record OCV again at 48 hours.

I can't give you the P/F numbers we use, but this method will give you MUCH better resolution than trying to do it at 50% where just a few mV can actually be a rather large difference in SOC. What you really want to watch is the delta V/hr during hours 24 to 48, but the raw numbers at 24 and 48 hours can be clear enough to show you what a real failure looks like.

I'm with Luke (and not dnmun) on this one--HSD cells should be culled if they are severe enough. In a small pack that is used regularly it's less of an issue. Packs with this issue coming out of storage can be a real pain to restore balance in if you lack certain tools. Leave a typical EV car with this problem sitting for a few months and you can be in for a re-balance time of days or weeks. Trust me, the OEM will send the battery back as defective. Even in the best cases where the balancers can cope with the issue, the other cells will be spending more time with charge applied at 100% SOC, so they will endure more stress as a result of the one bad cell. There are all kinds of potential ripple effects that range from subtle to debilitating. This IS a critical characteristic.

 

[pohjeremy]

You really need to do this test at 100% SOC. Here's the A123 test:

Charge CV at 3.60 Volts until current <1/20C
Remove charger. This is zero hours.
Record OCV at 24 hrs.
Record OCV again at 48 hours.

On it, pal...I'll try doing it 6-hrly...thanks!!

 

[pohjeremy]

I would discard that 27th cell as well but going without BMS, well, that's something I wouldn't do. I have noticed increase of people using BMS and that is with reason. Perhaps crappie BMS is better than human BMS in the end

Thanks!!

New Topic: How does one ensure a PRC made BMS is reliable and won't ruin the whole ($7000) set? e.g. Smoked MOSFET, Wrong ga, 2nd-hand parts, inauthentic parts...never-ending, darn.

This is the ONLY 2.5-3.65V, 32s, 140 - 150A, BMS 'reseller' whom I can access to, that is affordable (the $790rmb one) is: http://item.taobao.com/item.htm?spm=2013.1.20141001.2.OsiSSm&id=42190888184&scm=1007.10115.2077.i26262124428&pvid=de7844bf-fb6d-4226-b425-859a4defe3ea

If you see his wiring, it's 10ga for 150A....WHAT!?

The budget is 500USD and the BMS cannot be >15mm in height (it's going inside a fixed-height battery case). Any leads..? If not, what should I tell the 'reseller' to change..?

Pic 1: The leftover space for a BMS.
Pic 2 - 6: Why I'm now convinced a BMS is integral to a set-up.

Last resort: Regress to stone age; 32 individual chargers (3.51 - 3.56V) + 32 voltmeters (all on hand).

Darn some (not all) of those PRCs...acquire integrity quicker!!! It's not that I don't trust some of em'; It's just that...I don't trust some of em'!

Thank you all in advance!!!

 

[pohjeremy]

Sorry, forgot charging details.

Charger: 113VDC, 6A.

Voltmeters are +0.03xV (So 3.34 = 3.31V). Near 10mA each. The 3.34V base was after a test charge for about 30 seconds.

The actual OCV was 3.286V(multi-meter)-- all 32 cells. They had been balanced in parallel for 48hrs. Bulk charging was done at 75°F ambient.

Observation : Any random cell would rise to 3.5x - 3.6x while the other 31 remained 0.1 - 0.2V lower than that.

Each bulk charging trial lasted less than 2mins. Once any showed 3.68V, I switched the charger off .

How can random Mid-SOC cells rise that quick? My incorrect wiring method..?

:?

 

[pohjeremy]

By the way, what wire gauge would anyone recommend for an 135kg e-bike, 40A (estimated average), occasionally bursting to 140A for <30s..?

The hub motor's phase wires are 7ga (2ft total).
I was thinking 9ga for cells in series (3ft total) and 8ga (1ft total) for cells<--->controller..?

The wires are the flexible 200°C silicone jacketed RC types.

IMO, they are overrated online e.g 10ga, Max 140A for RC purposes.
8ga, Max 200A for RC purposes.

Who's gonna let them get past 100°C anyways..?

Thanks!!!

 

[dnmun]

12AWG wire will handle 30A continuous and 120A surges.

 

[pohjeremy]

12AWG wire will handle 30A continuous and 120A surges.

Thanks, dnmun.