Balancing 300 18650 Batteries :/

[SackofFluor]

ahh, so much information

well then i will first sit down, calculate the average, median and std.
Discard cells that are too low (shall i keep "too" high cells) and then do proper "mixing".

Yes I will be using a Daly smart bms and thought to stop charging at 4V and discharging at around 3.3/3.4 V , I thought this would give me relatively good safety margins.

Regarding the architecture I think i will have to see what kind of setup can be build well and stimm remains somehow scalable.

I was thinking about spot welding thin nickle strips to the cells and attaching those to a copper bus bar using my WIG welder.
The nickle strips i was thinking about 5mm, 0.1mm, they should be able to survive around 2-4Amps and work therefore as a fuse.

But because i wasn't sure, i also ordered fuse wire

 

[eMark]

ahh, so much information

well then i will first sit down, calculate the average, median and std.
Discard cells that are too low (shall i keep "too" high cells) and then do proper "mixing".

Ideally, you would want to keep/use the cells with the highest storage capacity (1700 -1800mAh) for your initial 7s DIY learning powerwall build experiment.

FWIW i'd view your DIY mixed variety of 300 cells (possibly only 210-245 usable - 7s30p-35p) as a preliminary learning experience ... not your final DIY reliable powerwall build (e.g. 420-490 cells - 7s60p-70p). Chances are you may decide to build your final powerwall from just one or two brands with same cell specifications ... https://www.batteryclearinghouse.com/products/200-cells-200-lg-mh1-18605-3200mah-10a-cells-in-5-aluminum-cased-packs ... or from two or three brands with similar specification. The advantage of using salvaged cells closer to 3200mAh (when new) is say a 7s80p powerwall taking up considerably less space and build time than a 7s140p using salvaged 1600 to 1799mAh cells (2200mAh when new?) as you previously posted.

You can see why the cells are $1 each when the storage capacity has dropped from 2150-2200mAh to 1600-1750mAh (ICR18650-22P and ICR18650-22F) ... not a whole lot of remaining cycle life, but what can you expect from $1 salvaged cells. Thus your reason for a 7s140p build of 980 salvaged cells averaging only 1600-1800mAh storage capacity. Actually less than that if charged to only 4.0V and discharged to 3.3/3.4V (thus your rationale for a "7s 100p or maybe even 140p" build) ...

Yes I will be using a Daly smart bms and thought to stop charging at 4V and discharging at around 3.3/3.4 V , I thought this would give me relatively good safety margins.

Salvaged cells at $1/ea means you'll invariably discard a fair amount whether beginning with 300 or 600 cells. Anyway for all your effort you want a DIY powerwall that's reliable knowing you've done your due diligence having first practiced building a 7s30p-35p powerwall (e.g. 210-245 usable cells out of 300) :thumb:

Another reason for building your powerwall with one or more brand of salvaged cells that were originally 3200mAh-10a than four or more salvaged cell brands of varying specification (1600-1799mAh, 4.4-10amp) is a smaller build (less time, materials, frustration) as a valuable learning experience.

 

[john61ct]

The lowest cell mAh capacity dictates that of the whole pack

by definition, if doing HVC / LVC from cell level monitoring, as you should.

Which means the other 299 cells can be as large Ah capacity as you like, just the extra won't ever get used.

Which is another (slight) advantage of grouping like cells into sub packs

and then paralleling those to create your "powerbank".

When you see a given sub pack has deteriorated, can pull it out, rest stays in production

giving you time to then atomize that sub pack and weed out / replace the weak links

or discard (recycle) them all and get to work on replacing the sub pack

Again, way too much work for poor reliability IMO best to start with new.

 

[eMark]

If your budget is that limited i'd still try to find salvaged cells with 3000-3500mAh original capacity when only charging to 4.0V (e.g. LG HG2 or Sanyo GA ... than using 2200mAh (original capacity) salvaged cells that have been depleted to only 1600-1750mAh storage capacity.

How many hours do you figure is the most that may ever be required from your auxillary powerbank? If a 7s30p-40p (3000-3500mAh new cells) is a sufficient build (210-280 cells) it should be worth the extra expense to buy new cells costing no more than $6 from IMR (quantity of 200) than mess around with $1 salvaged cells that may still end up being a frustrating powerbank disappointment. So i agree with U-know-who :thumb: ... https://voltaplex.com/lithium-ion-battery/18650-battery?s_capacity_max=3000-3500

It all depends on how much you value being able to finish a good-sized container or two of home brew ... CHEERS

 

[fatty]

The lowest cell mAh capacity dictates that of the whole pack
Which means the other 299 cells can be as large Ah capacity as you like, just the extra won't ever get used.

He's not building a 300s pack, so this is incorrect. The lowest parallel string capacity dictates the capacity of the resulting pack.

As I've shown before, increasing parallel cell count decreases the impact of individual cell capacity variance.
With powerwall-scale numbers of parallel cells, even without culling cells, string variance becomes negligible.
Per the model attached, assuming 10% cell capacity variance, 40 such cells in parallel reduce string capacity variance to 1%.

 

[eMark]

The lowest parallel string capacity dictates the capacity of the resulting pack.

U-know-who detests the use of parallel "string" instead of parallel "group" even in a powerbank when the parallel cells are in a row as opposed to a parallel "group" (non-string) in a triangular ebike battery configuration.

I mention this because he was/is adamant that "string" is inappropriate even when referring to a rectangular powerwall/powerbank of parallel rows. Or my experimental 30Q 10S3P rectangular pack with ten rows of 3P "strings". He comes across as always being right even when wrong as you and others may point out (allbeit infrequently) as he's very knowledgeable and usually right.

As I've shown before, increasing parallel cell count decreases the impact of individual cell capacity variance.

Per the model attached, assuming 10% cell capacity variance, 40 such cells in parallel reduce string capacity variance to 1%.

A cell variance from 1600mAh to 1760mAh (variance of 160mAh or 10%). Is that how you'd figure? So would seven 40p "strings" possibly with a variance from 32000mAh to 35200mAh if 20 cells in the string were 1600mAh capacity and the other 20 were 1760mAh capacity? Would you stagger the 40 cells ... 1600mAh, 1760mAh, 1600mAh, 1760mAh, 1600mAh, etc, etc, etc (20-1600mAh cells and 20-1760mAh cells) if those were the only two cell variances? A total of 67200mAh capacity for each of the seven 40p strings.

Would a 7s40p powerbank of 5A rated cells be just as efficient as 10A rated cells to handle a 2A draw down load? Would 15A rated cells (or even 10A cells be overkill in such a 7s40p powerbank supplying at most a 2A power drain? Would it be of any concern if 20 of the cells were rated at only 4.4 amps MCD and the other 20 rated at 10 amps MCD?

 

[fatty]

If your budget is that limited i'd still try to find salvaged cells with 3000-3500mAh original capacity when only charging to 4.0V (e.g. LG HG2 or Sanyo GA ... than using 2200mAh (original capacity) salvaged cells that have been depleted to only 1600-1750mAh storage capacity.

How many hours do you figure is the most that may ever be required from your auxillary powerbank? If a 7s30p-40p (3000-3500mAh new cells) is a sufficient build (210-280 cells) it should be worth the extra expense to buy new cells costing no more than $6 from IMR (quantity of 200) than mess around with $1 salvaged cells that may still end up being a frustrating powerbank disappointment. So i agree with U-know-who :thumb: ... https://voltaplex.com/lithium-ion-battery/18650-battery?s_capacity_max=3000-3500

This is like telling somebody who just got a hand-me-down Toyota Camry that he has to buy a new Tesla to commute.

Capacity-tested salvage cells are ideal for a powerwall, so this is off-topic.

 

[fatty]

U-know-who detests the use of parallel "string" instead of parallel "group" even in a powerbank when the parallel cells are in a row as opposed to a parallel "group" (non-string) in a triangular ebike battery configuration.
I mention this because he was/is adamant that "string" is inappropriate even when referring to a rectangular powerwall/powerbank of parallel rows. Or my experimental 30Q 10S3P rectangular pack with ten rows of 3P "strings".

Interesting, I didn't know he held that distinction. I can't remember why I use string instead of group, but they're understood to be synonymous.

He comes across as always being right even when wrong as you and others may point out (allbeit infrequently) as he's very knowledgeable and usually right.

I agree with his safety-conscious approach (especially protected charging), but our post history shows we have disagreed on technicalities (degree of undercharging, balance charging).

A cell variance from 1600mAh to 1760mAh (variance of 160mAh or 10%). Is that how you'd figure? So would seven 40p "strings" possibly with a variance from 32000mAh to 35200mAh if 20 cells in the string were 1600mAh capacity and the other 20 were 1760mAh capacity? Would you stagger the 40 cells ... 1600mAh, 1760mAh, 1600mAh, 1760mAh, 1600mAh, etc, etc, etc (20-1600mAh cells and 20-1760mAh cells) if those were the only two cell variances? A total of 67200mAh capacity for each of the seven 40p strings.

No, I just used 2000mAh and 10% arbitrarily to illustrate the statistical phenomena. More representative numbers can be inserted into the spreadsheet, but the results will be the same.

In groups of few cell count, it is beneficial to bin by capacity to match group capacity. With 10% variance, 1p has a maximum variance of 20% (+10% to -10%); 2p reduces this to ~15%, 3p to ~10%, etc.
However, there is no need to bin or stagger the cells in such large parallel groups. The groups will statistically trend towards the mean (even out).

Play with the capacity dispersion spreadsheet to see for yourself.

Would a 7s40p powerbank of 5A rated cells be just as efficient as 10A rated cells to handle a 2A draw down load? Would 15A rated cells (or even 10A cells be overkill in such a 7s40p powerbank supplying at most a 2A power drain? Would it be of any concern if 20 of the cells were rated at only 4.4 amps MCD and the other 20 rated at 10 amps MCD?

No, this was incorrect when you posted it earlier -- none of this is of concern. Cells in parallel electrically become one -- their characteristics are aggregated/summed.
The concern is if cells of different characteristics are concentrated between rather than distributed among different parallel groups, as this can lead to imbalance.

 

[goatman]

+1
300 cells, you can build same cells in one pack
build multiple packs of each battery brand and then parallel them together
mixing and matching in one big pack would be a nightmare

This is incorrect. Building serial packs and then paralleling them introduces balance problems that can only be resolved by paralleling the packs at the parallel string level, at which point you could have just built a parallel-first pack to begin with.

youre going to get in trouble with cells micro-cycling

 

[fatty]

youre going to get in trouble with cells micro-cycling

No, "micro-cycling" has long been shown not to alter capacity fade.

 

[SackofFluor]

Well first of all, THANK you guys so much.

I think I might have actually set a couple of batteries accidentally on fire if I wouldn't have asked.

By now, because of safety and "skill" limitations I will technically really just use as the powerwall as a learning-tool, when it comes to the skills required, especially when it comes to handling and working with BMSes, and properly understanding what is happening to my cells and why

I think long term I will go with completely new LTO cells because of safety reasons, but that is at least another 2-3 years down the line and the cells and the little solarpanels i have will be great tools to learn and try out stuff...

 

[fatty]

I think long term I will go with completely new LTO cells because of safety reasons, but that is at least another 2-3 years down the line and the cells and the little solarpanels i have will be great tools to learn and try out stuff...

LTO is only attractive where rapid charge is required -- not applicable to solar storage.

Check out LiFePO4 (LFP) prismatics for safe unattended indoor storage.

 

[john61ct]

A string means cells in series, a group is cells in parallel.

And yes, a group functionally acts as one cell from a design / layout POV

and that was exactly what I meant by the Ah capacity of the bank being determined by the weakest link.

Yes some people go to the trouble of getting the group total Ah capacities of scrap cells to match up

but I think not **at all** worth the time spent, since a mixed bag will not be wearing / failing evenly anyway. By buying quality new cells from the same production run, of course they will be fairly well matched to start with.

The ideal (besides buying new) is cell capacity equals the target bank capacity, so if you want a 180Ah bank, buy cells that are 180Ah, so no grouping required at all, 18S means a total of 18 cells to worry about.

For that size House bank (very small for that use case), futzing around with the small cylindricals is IMO ludicrous.

 

[john61ct]

LTO is only attractive where rapid charge is required -- not applicable to solar storage.

That is just insane.

Yes LFP is likely just as safe wrt thermal runaway, but

they only last a tenth the cycle lifespan

and can be rendered instant scrap by letting them go too near dead flat

and also by charging at below zero temps.

The 10C charge rate comes in handy for lots of use cases too, but it is by far not the most important advantage.

Since LTO is much earlier in its adoption cycle, they are not counterfeited, at least haven't heard of that yet.

And because LTO cells last so long in normal use, and are harder to murder, when they do come up secondhand, they are more likely to be in good shape and still have many thousands of cycles' lifespan remaining.

 

[fatty]

A string means cells in series, a group is cells in parallel.

I've seen string used more commonly on the eBike subforums, but I agree that group is more correct.

The ideal (besides buying new) is cell capacity equals the target bank capacity, so if you want a 180Ah bank, buy cells that are 180Ah, so no grouping required at all, 18S means a total of 18 cells to worry about.

For that size House bank (very small for that use case), futzing around with the small cylindricals is IMO ludicrous.

Agreed

 

[eMark]

This is my last post in this [salvaged cell] thread and then it's up to OP if he wants to ask more specifics (what-Ifs) for building a preliminary experimental DiY powerbank (e.g. 7s10p) from used cells with questionable cycle life remaining. How ludicrous is it when as someone previously said "thousands" of powerbanks (using salvaged 18650 cells) have been built and still being built when new cells cost more than ever. A 2A power drain isn't ludicrous from a 7s40p powerbank of usable salvaged cells OR is it?

In groups of few cell count, it is beneficial to bin by capacity to match group capacity. With 10% variance, 1p has a maximum variance of 20% (+10% to -10%); 2p reduces this to ~15%, 3p to ~10%, etc.
However, there is no need to bin or stagger the cells in such large parallel groups. The groups will statistically trend towards the mean (even out).

What about staggering both series as well as parallel cells (see diagram) if he was going to build an experimental 7s10p with 35-1600mAh capacity cells and 35-1760mAh capacity cells (both salvaged 18650 cells) with original capacity around 2100mah when charged from 3.0V to 4.2V. He mentioned charging from 3.35V ro 4.0V (23.45V to 28V) so probably less than 1600mAh available capacity and 1760mAh available capacity (3.35V to 4.00V. He could then calculate the effective power time available with a 7s40p powerbank of similar used/salvaged cells with same cell storage capacity.

Cells in parallel electrically become one -- their characteristics are aggregated/summed.

The concern is if cells of different characteristics are concentrated between rather than distributed among different parallel groups, as this can lead to imbalance.

He would have to do a capacity check on representative cells if only charging to 4.0V with the intent to prolong powerbank cycle life longevity. He could test the difference in 2A drain time between charging to only 4.0V versus charging to 4.1V with these [depleted] used cells to see how many parallel cells would be needed (e.g. 7s140p) for his worst case beer brewing auxiliary backup powerbank scenario ...

1600 - 1760 -1600 - 1760 -1600 - 1760 - 1600 - 1760 - 1600 - 1760
1760 - 1600 -1760 - 1600 - 1760 - 1600 - 1760 - 1600 - 1760 - 1600
1600 - 1760 -1600 - 1760 -1600 - 1760 - 1600 - 1760 - 1600 - 1760
1760 - 1600 -1760 - 1600 - 1760 - 1600 - 1760 - 1600 - 1760 - 1600
1600 - 1760 -1600 - 1760 -1600 - 1760 - 1600 - 1760 - 1600 - 1760
1760 - 1600 -1760 - 1600 - 1760 - 1600 - 1760 - 1600 - 1760 - 1600
1600 - 1760 -1600 - 1760 -1600 - 1760 - 1600 - 1760 - 1600 - 1760

Assuming he can find 70 (35 & 35) usable ICR18650-22P & ICR18650-22F cels how would you advise him to arrange/build an experimental 7s10p powerbnk from these salvaged cells if a relatively equal mixture of 1600mAh & 1760mAh remaining capacity between these two brands? It would be a great learning experiment to see how long this powerbank provides power with a 2 amp drain. Then he should be able to calculate the effective time with a 7s40p powerbank using either salvaged or new 3000mah to 3500mAh 18650 cells versus 1600mah to 1750mah salvaged 2200mAh cells.

 

[fatty]

A 2A power drain isn't ludicrous from a 7s40p powerbank of usable salvaged cells OR is it?

2A per group is obviously not ludicrous. If you mean 2A per cell, be careful because a heterogeneous group of cells will share the total load, so you can't simply divide the total load by the number of cells in the group to find the exact load per cell. You can estimate average load per cell, in which case you can roughly estimate maximum discharge current as a percentage of current when new. So if the capacity test shows 80% of new capacity (80% state of health), you could likewise estimate 80% of the maximum discharge current.

What about staggering both series as well as parallel cells (see diagram) if he was going to build an experimental 7s10p with 35-1600mAh capacity cells and 35-1760mAh capacity cells (both salvaged 18650 cells) with original capacity around 2100mah when charged from 3.0V to 4.2V.

I don't think you are following the fundamentals here. Cells in a parallel group become one electrically, so the order in which they're connected doesn't matter -- physically "staggering" doesn't matter.

He mentioned charging from 3.35V ro 4.0V (23.45V to 28V)

Frankly, I think undercharging a salvage pack makes even less sense than undercharging a new pack. Since the history is unknown, you can't estimate cycle life vs capacity benefit. It's salvaged anyway -- who cares if it lasts 10% more cycles?

He would have to do a capacity check on representative cells if only charging to 4.0V with the intent to prolong powerbank cycle life longevity. He could test the difference in 2A drain time between charging to only 4.0V versus charging to 4.1V with these [depleted] used cells to see how many parallel cells would be needed (e.g. 7s140p) for his worst case beer brewing auxiliary backup powerbank scenario ...

True to get most accurate data, but again unnecessary on a salvage pack.

Assuming he can find 70 (35 & 35) usable ICR18650-22P & ICR18650-22F cels how would you advise him to arrange/build an experimental 7s10p powerbnk from these salvaged cells if a relatively equal mixture of 1600mAh & 1760mAh remaining capacity between these two brands?

As above, the physical arrangement doesn't matter to the parallel electrical group.

 

[fatty]

That is just insane.

That's where LTO is primarily (if not exclusively) used. As I showed in the LTO thread, it's not competitive anywhere else.

they only last a tenth the cycle lifespan

LFP lasts 3000-5000 cycles at 100% DoD, which is not seen in solar powerwall applications. The batteries used in this application are invariably upgraded before they are actually scrap.

The 10C charge rate comes in handy for lots of use cases too, but it is by far not the most important advantage.

Since LTO is much earlier in its adoption cycle, they are not counterfeited, at least haven't heard of that yet.

And because LTO cells last so long in normal use, and are harder to murder, when they do come up secondhand, they are more likely to be in good shape and still have many thousands of cycles' lifespan remaining.

Advantages, certainly -- but obviously not enough to offset the costs, or else they would see wider application.

Like LFP, LTO's energy density disadvantage is minimized in a static installation. But that alone doesn't make it the best choice for a powerwall.

 

[john61ct]

Cost is just one factor, irrelevant to many, certainly low priority.

So sure, specify tgat LTO is more expensive per kWh

but that hardly means "unattractive" to everyone.

And fairly regularly, batches of LTO cells do become available at very low prices on the secondary market

Even if already used for a few thousand cycles, that's barely broken in