Teardown & Rebuild of a MuscleGrid LiFePO4 Server Rack Battery

rsaeon · January 4, 2024, 2:49pm

Taking a detour from:

https://techenclave.com/threads/working-with-inverter-ups-lead-acid-batteries.217418/

Led me to exploring this option. Most 48V server rack style batteries available abroad are in a 16S configuration and having learned my inverter charges at 14.2V per 12V battery, for a total of 56.8V, divided by 16 resulting in a per cell voltage of 3.55V which is pretty good for a LiFePO4. Bumping it up to 14.6V gives us 3.65V which is pretty much the recommended charging voltage for this chemistry. Float at 13.9V works out to be 3.475V which is a little over the 3.4V recommendation:

The best option of course would be to build my own battery with a kit from MDS Enterprises who pride themselves as selling only the highest grade, best of the batch LiFePO4 cells:

https://techenclave.com/threads/where-can-i-buy-lifepo4-grade-a-prismatic-cells-in-india.215698/#post-2495221

That wasn’t an option for me partly because I don’t have that kind of funding and actually well that’s the entire reason. MuscleGrid’s products are available on Amazon with no-cost emi and seeing how I just got out of a EMI hole it made sense to jump right back in.

Considering how expensive these things are, this one sold by MuscleGrid can’t be anything but a very low grade, bottom-of-the-batch rebranded/refurbished/old-stock product. Which is perfectly fine, [your first ten thousand batteries are your worst anyway](’

https://www.goodreads.com/quotes/266224-your-first-10-000-photographs-are-your-worst’

). But seriously, I knew it wasn’t going to be the best but that it’ll probably be good enough — and way better than any lead acid FLA/AGM option.

So obviously it had to arrive damaged, dented:

But first, the specifications:

In case it’s not clear, this is supposed to be a 16S unit (but it’s not, its 15S as photos will later show).

One bent hand grip. Nothing a hammer can’t fix.

Two bent hand grips. I wasn’t planning on keeping this enclosure anyway, I just want the cells.

125A DC circuit breaker, that’s good to see and reuse in my own pack.

Removing 15 screws reveals the insides:

Tidy wiring, which was I guess is the bare minimum standard you can expect.

BMS is an older unit from Shinwa:

https://www.shinwa-net.com/products-bms/

Unfortunately while it may support a 16S configuration, it was built for a 15S one:

The connector doesn’t include the B16 pins. Maybe it could be replaced or rewired or reprogrammed for a 16S configuration, I’ll find that out later when I reuse this for a powerwall project.

The battery pack itself:

That’s a lot of wires! There’s EIGHT thermal sensors in there, probably to make sure nothing overheats — either because this 15S configuration was meant for Li-Ion chemistry originally, or they expect 48V inverters to overcharge this 15S pack, at 14.4V that’s a per cell voltage of 3.85V which is well out of spec for LiFePO4.

Yeah, that’s a 15S pack.

This information is mostly useless, nothing comes up in battery databases or web searches.

But this sticker reveals it’s a Narada Power FE105C cell. The C indicates it’s a lower-grade cell, more about that here:

And it also indicates it’s a 105Ah battery, not 120Ah as claimed by MuscleGrid.

[HR][/HR]

To recap, these are facts about the MuscleGrid 48V 120Ah LiFePO4 Server Rack Battery:

It arrived damaged, despite being packed in a wooden crate
It’s 15S not 16S
It’s 105Ah not 120Ah
It uses ‘C’ grade/batch cells from Narada Power
It’ll probably last ten years or more despite all of this (benefits of LiFePO4 chemistry)

[HR][/HR]

Todo:

Purchase a single cell to rebuild this as a 16S pack
Build or buy a 16S BMS

I expect this to take about a month.

octal777 · January 7, 2024, 5:54am

Thanks for your post on this topic, I have been trying to go down this path as well. I have a few questions for you

Are you shifting to LiFePO4 primarily due to the longevity or do plan to add more cycles by probably adding a solar charge controller?
What is powered by these inverters are they typical home loads?
What BMS are you considering ? The JK BMS are recommended quite a lot due to their active balancing capability. I found the Off Grid Garage on Youtube quite helpful -

For me the major hurdles are -

I have a 12V Inverter meant for Lead Acid upgrading this makes no sense might as well move to a 48V system from Ashapower or similar
Want to have the ability to add solar capability when I shift houses down the line

Best of luck with your project. Do keep us posted on your progress. I have been following the thread on Lead Acid as well interesting stuff

ze_cook · January 7, 2024, 8:45am

rsaeon:

Taking a detour from:

Working with Inverter/UPS Lead Acid Batteries - PC Builds, Modding & Overclocking - TechEnclave

Led me to exploring this option. Most 48V server rack style batteries available abroad are in a 16S configuration and having learned my inverter charges at 14.2V per 12V battery, for a total of 56.8V, divided by 16 resulting in a per cell voltage of 3.55V which is pretty good for a LiFePO4. Bumping it up to 14.6V gives us 3.65V which is pretty much the recommended charging voltage for this chemistry. Float at 13.9V works out to be 3.475V which is a little over the 3.4V recommendation:

Recommended Charge Profile for DIY LiFePO4 Batteries *Sticky Post* | DIY Solar Power Forum

The best option of course would be to build my own battery with a kit from MDS Enterprises who pride themselves as selling only the highest grade, best of the batch LiFePO4 cells:

Where can I buy LifePo4 grade A prismatic cells in India? - Home Tech & Appliances Buying Help - TechEnclave

That wasn’t an option for me partly because I don’t have that kind of funding and actually well that’s the entire reason. MuscleGrid’s products are available on Amazon with no-cost emi and seeing how I just got out of a EMI hole it made sense to jump right back in.

Considering how expensive these things are, this one sold by MuscleGrid can’t be anything but a very low grade, bottom-of-the-batch rebranded/refurbished/old-stock product. Which is perfectly fine, [your first ten thousand batteries are your worst anyway](’

https://www.goodreads.com/quotes/266224-your-first-10-000-photographs-are-your-worst’

). But seriously, I knew it wasn’t going to be the best but that it’ll probably be good enough — and way better than any lead acid FLA/AGM option.

So obviously it had to arrive damaged, dented:

[ATTACH alt=“photo_2024-01-04 19.20.55.jpeg”]187318[/ATTACH]

But first, the specifications:

[ATTACH alt=“photo_2024-01-04 19.20.56.jpeg”]187319[/ATTACH]

In case it’s not clear, this is supposed to be a 16S unit (but it’s not, its 15S as photos will later show).

[ATTACH alt=“photo_2024-01-04 19.20.57.jpeg”]187320[/ATTACH]

One bent hand grip. Nothing a hammer can’t fix.

[ATTACH alt=“photo_2024-01-04 19.20.58.jpeg”]187321[/ATTACH]

Two bent hand grips. I wasn’t planning on keeping this enclosure anyway, I just want the cells.

[ATTACH alt=“photo_2024-01-04 19.20.59.jpeg”]187322[/ATTACH]

125A DC circuit breaker, that’s good to see and reuse in my own pack.

Removing 15 screws reveals the insides:

[ATTACH alt=“photo_2024-01-04 19.21.00.jpeg”]187323[/ATTACH]

Tidy wiring, which was I guess is the bare minimum standard you can expect.

[ATTACH alt=“photo_2024-01-04 19.21.01.jpeg”]187324[/ATTACH]

BMS is an older unit from Shinwa:

https://www.shinwa-net.com/products-bms/

Unfortunately while it may support a 16S configuration, it was built for a 15S one:

[ATTACH alt=“photo_2024-01-04 19.21.02.jpeg”]187327[/ATTACH]

The connector doesn’t include the B16 pins. Maybe it could be replaced or rewired or reprogrammed for a 16S configuration, I’ll find that out later when I reuse this for a powerwall project.

The battery pack itself:

[ATTACH alt=“photo_2024-01-04 19.21.03.jpeg”]187329[/ATTACH]

That’s a lot of wires! There’s EIGHT thermal sensors in there, probably to make sure nothing overheats — either because this 15S configuration was meant for Li-Ion chemistry originally, or they expect 48V inverters to overcharge this 15S pack, at 14.4V that’s a per cell voltage of 3.85V which is well out of spec for LiFePO4.

[ATTACH alt=“photo_2024-01-04 19.21.03 (1).jpeg”]187335[/ATTACH]

Yeah, that’s a 15S pack.

[ATTACH alt=“photo_2024-01-04 19.21.05.jpeg”]187336[/ATTACH]

This information is mostly useless, nothing comes up in battery databases or web searches.

[ATTACH alt=“photo_2024-01-04 19.21.06.jpeg”]187337[/ATTACH]

But this sticker reveals it’s a Narada Power FE105C cell. The C indicates it’s a lower-grade cell, more about that here:

Narada Power

And it also indicates it’s a 105Ah battery, not 120Ah as claimed by MuscleGrid.

[HR][/HR]

To recap, these are facts about the MuscleGrid 48V 120Ah LiFePO4 Server Rack Battery:

It arrived damaged, despite being packed in a wooden crate

It’s 15S not 16S

It’s 105Ah not 120Ah

It uses ‘C’ grade/batch cells from Narada Power

It’ll probably last ten years or more despite all of this (benefits of LiFePO4 chemistry)

[HR][/HR]

Todo:

Purchase a single cell to rebuild this as a 16S pack

Build or buy a 16S BMS

I expect this to take about a month.

I’ve been planning to get a lifepo4 battery pack for a while now. Problem with cheap batteries are mismatched cells, one cell might’ve degraded more, will drop below threshold faster and the BMS will cut the entire battery off. Even if the individual cells are rated at 105, you might not get 105 with the entire battery. Only way to know for sure is a capacity test with cell voltage logging.

I want to get a capacity tester from banggood, but they’re too expensive (8k ish) for my one off case. Please keep updating the thread.

rsaeon · January 7, 2024, 2:57pm

This battery will be powering a few ‘servers’ that I run at home and need to keep online. Occasionally, it will be used for a water pump during extended power cuts.

Both.

The decision was mostly constrained by what was available with Amazon’s No Cost EMI. It was either Luminous Red Charge FLAs or this MuscleGrid battery. Before purchasing it, I suspected this wasn’t really a 120Ah battery and that it couldn’t possibly be using new or high quality cells but went ahead anyway because of the other benefits of LiFePO4. I intend to build my own battery packs in the future so taking this one apart was a learning experience.

The load is a continuous 1kW, so I would eventually like to shift to an offgrid solar installation at some point in the future with used panels, they’re often sold at less than half the cost of new ones.

I’ve been watching that channel on and off over the months leading to this purchase, along with Will Prowse, DavidPoz, Lithium Solar and Jehu Garcia. We’re often behind the technology curve in India but with LiFePO4, it is nice to learn from the mistakes/pitfalls that the early adopters in other countries have went through.

Right now, I’m leaning towards making my own top-balancing BMS, a simple ESP8266 running Tasmota that will discharge each cell through a power resistor for 30 seconds at time when voltages go over 3.6V. It’s the basic concept of Stuart Pittaway’s original diyBMS but without the isolated I2C bus. I want to use mass produced modules instead of custom-made PCBs. Tasmota’s integration with MQTT would be very useful in creating NodeRed flows that I’ll use to trigger shutdowns when the batteries get low.

I’ll be using one ESP module for every four cells. They’ll log data through MQTT/NodeRed into InfluxDB and from there I can generate nice graphs with Grafana. It’s not as self contained as the diyBMS, but it’s a system I’ll have complete control over. I administer Tasmota devices with Ansible so it’s easy to make changes and I have automated daily backups of each device’s configuration setup with NodeRed. My backup strategy for config files is to keep last thirty days backups and then one for every month prior. The unneeded backup files are deleted with a daily cron job:

[ICODE]1 0 * * * /usr/bin/find /backups -type f -mtime +30 ! -name ‘01-’ -exec /bin/rm {} ;[/ICODE]

The backup files are named [ICODE]YYYYMMDD-indentifier[/ICODE] so the [ICODE]01-[/ICODE] part keeps only the backup made on the 1st of each month before and only files older than 30 days are deleted.

For over-current and short-circuit protection, I’ll use that DC rated MCB. I might forego under-voltage protection and rely on software controlled shutdowns to reduce the load. I might use a DC-rated contactor if I can find one or maybe a power mosfet as a battery disconnect. I need to do more research on this.

I have a battery equalizer that I purchased from Banggood, I think I’ll be using that for active balancing (four groups of four cells each):

Ashapower really appears to be the highest ‘Make In India’ standard for this market, with UTL being a very good price-conscious alternative. There are interviews on youtube with the engineers from both companies and it’s hard not to be impressed.

Yes, I suspect this is an unbalanced battery nowhere near the rated 105Ah capacity and that is the reason for the lower pricing compared to other brands. Older reviews mentioning getting only 50Ah from the battery and I really hope it’s not that bad. MuscleGrid keeps removing the older product pages whenever someone posts a negative review, which is part of the reason why this thread is here and not on Amazon.

For logging, again, Tasmota on a ESP8266 with a INA219 module costs less than Rs 500 and is plenty accurate, so all I need is a low-voltage high-amperage DC load to discharge the batteries.

[HR][/HR]

I’ve updated my Todo:

Build a high amp battery charger for 3.65V
Purchase a battery tester and do a charge/discharge test for each cell
Purchase a single cell to rebuild this as a 16S pack
Build or buy a 16S BMS

octal777 · January 8, 2024, 5:27am

Wow that’s a very hands-on approach, one query - would the efficiency go down compared to an off the shelf BMS which in my understanding discharges one cell into another or do they also use a load to bleed off charge?

The cells from the pack at-least look physically good no bulges seen in the photos

rsaeon · January 11, 2024, 1:58am

I haven’t seen any teardowns of those BMSs to see if they do active balancing. An active balancer would definitely be more efficient than burning off the extra power through a resistor, but that’s added complexity in a DIY project. I would be using those specialised BMSs if they were easily available in India like how they are in the west, but since they aren’t, I decided to make my own.

My self-imposed restrictions with building anything is that if any part should go bad, then it should be as straight forward as swapping it out with a spare that’s easily procurable and affordable. I learned this the hard way when I imported all the parts of my first new desktop PC in 1999 and had to wait months to fix anything.

As for actual efficiency, it shouldn’t be more than a few percent either way. As flawed and undesirable this MuscleGrid battery is, it is still a product that’s made and sold by the OEM (Narada Power) so they have some expectations of longevity despite using an older BMS and non-ideal cells.

Currently waiting on Sharvi Electronics to ship out the 3.3V 40A power supply I purchased, it’s adjustable so I’ll be using this to charge up each cell to 3.65V before I start the capacity tests.

rsaeon · January 17, 2024, 2:59am

As it turns out, it wasn’t that bad! That was a nice surprise. As shipped/received, lowest measured voltage was 3.28V and highest was 3.31V for a delta of 30mV. It’s not great but it’s way better than what I was dreading. Ideal would have been under 10mV.

Update time! I put together a rudimentary high-amp charger to top off these cells to 3.65V:

I’ve always wanted to use these instrument cases for something and now I finally get a chance! They’re from:

Also here is a PZEM-015 DC Battery Capacity tester, a panel mount IEC socket+fuse+switch from Sharvi Electronics, 2 meters of 12AWG silicone wire from Quartz Components cut into thirds, and PG11 cable glands from Sunrom Electronics.

First was to crimp the wires into a 10mm sq uninsulated terminal lug, these were only available in 8mm hole sizes at Quartz Components:

I’m using 24AWG silicone wire for voltage sense of the PZEM module.

Gratuitous bokeh:

Front panel layout:

I wanted everything to be accessible in the front, including the power socket and switch. It’s not conventional but way more usable, in my opinion. Switches are outward facing for the same reason: usability.

After some quality time with a rotary tool, I got the first revision of the panel ready (already found ways to improve it — but that’ll be in another update):

That’s a mockup of the exhaust/venting in the background. Funny, I wanted something angular and futuristic and ended up with something organic instead. The design evolved to rounded edges with no sharp corners so as to facilitate easy dust cleaning.

Then a week a later this showed up — 40A at 3.3V! It’s adjustable so I’ll have it set at 3.65V:

oops:

Some hours of sanding later, it’s a snug fit.

A couple of evenings of were spent wiring everything up:

Eventually the PZEM module will be powered from the same power socket to make this all self contained.

I’m also considering adding a ESP8266+INA219+Relay to cut off power automatically at 3.66V volts.

With the power supply adjusted and powered on, everything looks good:

And here it is in use, charging up these cells, one by one:

It’s taking about 45mins to add about 11.8Ah per cell to bring it up to 3.65V, that would mean these cells were shipped at 89% State of Charge, if they really are 105Ah.

Next up is to charge up all the other cells and do a capacity test of each cell down to 2.5V.

Psycho_McCrazy · January 17, 2024, 3:14am

Would that be on a CC DC Load?

And while you’re at it, might also run DCIR tests on a 1s and 10s pulse…

rsaeon · January 17, 2024, 5:25am

Yes, I’m thinking of a modest 25A load, which at 48V is ~1000W AC load at >80% efficiency using this:

https://robu.in/product/black-2-4-inch-150w-dc-usb-tester-electronic-load-battery-tester-discharge-charge-power-meter-supply-detector-meter/

The PZEM module shows internal resistance to be 2 and 3 milliohm for the batteries are sitting at 3.28V and 3.31V, that load tester will have more accurate results.

Psycho_McCrazy · January 17, 2024, 6:08am

If 1000W (After AC inversion) is the target load for this battery bank, the equivalent 25A may be a good rating.

Alternatively, the standard C3 capacity may be tested which works out to roughly 35A for the 105Ah nominal cells.
This would also shorten your test duration per cell by more than an hour!!!

rsaeon · January 18, 2024, 12:31pm

Spoiler alert — it was really bad. I topped up each cell individually and they needed anywhere between 11.5Ah to 29.3Ah to reach 3.65V until current draw fell below 1A. That’s a very unbalanced battery pack!

Cell	Vpc as received	+Ah for 3.650Vpc
1	3.328	26.8
2	3.331	13.9
3	3.328	17.7
4	3.330	11.8
5	3.331	11.8
6	3.330	13.5
7	3.328	29.3
8	3.331	13.0
9	3.331	11.5
10	3.331	12.7
11	3.330	14.4
12	3.331	14.2
13	3.331	13.6
14	3.331	13.4
15	3.331	13.6

Some months ago, I had messaged the seller on Whatsapp asking about the battery configuration, if it was 15S or 16S. I never got a clear reply but the seller did share some photos of the cells, and they were labeled with voltages, not capacity.

So It would appear they’re grouping cells by voltage instead of tested capacity — to say that is grossly incompetent is an understatement.

How do people get funding for a business like this when they don’t even know what they’re doing?

Who gives them the capital to ignorantly assemble and sell high cost items like these? Gah.

Next up is to top balance groups of cells to 3.65V until there’s less than 50mA current draw, before starting the capacity tests:

So far, for the first group of five cells, that’s 5x 105Ah = 525Ah, it only needed a 3.43Ah. This means my charging strategy for each individual cell has worked as expected.

That would make things a lot easier, but it looks like this battery capacity tester can only discharge at a max of 20A, so I’ve started the tests with a 19.90A constant current load, anything higher would trigger the OCP.

Psycho_McCrazy · January 18, 2024, 12:46pm

The SOC-OCV curve of an LFP chemistry battery cell is very flat in the middle SOC ranges. That is why such variance was probably encountered. This is also why the balancing done by BMS is at higher SOCs, where the slope is steeper and differences actually perceptible!

Tough luck about the current limit on the DC Load, ~20A means more than 5 hours per cell if they match their advertised capacity.

rsaeon · January 19, 2024, 2:57am

Looks like the only way to get a larger DC load at these voltages is to get a second one and connect in parallel. Maybe later, when I’m building more of these.

I feel like I could write a thousand words eviscerating this company’s ineptitude — they have the golden opportunity to make a name for themselves in the market as a price conscious lithium battery manufacturer but they’re drowning in their own ignorance.

I finished one of the capacity tests, the first cell. To recap, it was shipped at 74.5% SoC while the highest SoC of any cell was was 89%. It’s now been tested at 104.689Ah or 99.7% SoH with a 19.99A CC load.

There’s just no excuse — these are decent cells. They’re not A+ but they’re not terrible in anyway. But the people assembling this pack have no idea what they’re doing and they’re putting together mismatched cells, not top balancing in any way and relying on an outdated BMS to fix things in the long run. It’ll take months if not years to rely on a BMS to average out a pack that is this unbalanced.

Updated table with SoC calculations:

Cell	Vpc as received	+Ah for 3.650Vpc	SoC as received
1	3.328	26.8	74.5%
2	3.331	13.9	86.8%
3	3.328	17.7	83.1%
4	3.330	11.8	88.8%
5	3.331	11.8	88.8%
6	3.330	13.5	87.1%
7	3.328	29.3	72.1%
8	3.331	13.0	87.6%
9	3.331	11.5	89%
10	3.331	12.7	87.9%
11	3.330	14.4	86.3%
12	3.331	14.2	86.5%
13	3.331	13.6	87%
14	3.331	13.4	87.2%
15	3.331	13.6	87%

I’ve two questions for anyone who can answer them:

SoH, state of health, does this mean how much of the original capacity is left? Like, are my calculations correct? If a 105Ah cell tested at 104.689Ah, then is it correct to say it is at (104.689/105) x 100 = 99.7% State of Health?
SoC, state of charge, would this be calculated at rated capacity of 105Ah or tested capacity of 104.689Ah?

rsaeon · January 20, 2024, 4:05am

Capacity testing is completed for 5 of the 15 cells:

Cell	-Ah to 2.5Vpc	SoH
1	104.689	99.7%
2	104.821	99.8%
3	103.741	98.8%
4	100.382	95.6%
5	102.466	97.6%

I tried charging them up with that 40A charger I built but the over current protection kept tripping — I needed to lower the voltage to 3.45V to get it to charge.

Looks like I’ll need some kind of adjustment knob on the front panel that’ll adjust voltage between 3V and 3.65V. Maybe even some kind of automated CC controller with… well it gets complicated now. Knob would be easier.

If I could do it over again, I’ll probably buy a proper CC/CV power supply.

rockyo27 · January 20, 2024, 6:15pm

@@rsaeon Okh. So i am looking for a used 2kva ups for my system which can also take external batteries. Which one should i look at and what price i am looking at. Would smf batteries be good with those or tubular batteries or lifepo4 ones.
What wud be the price point for batteries in case of used one’s.

rsaeon · January 21, 2024, 1:25pm

@@rockyo27 Replied in the other thread but a good price is around 10k to 15k, this is for a 2000VA inverter and two batteries for a 24V setup. 5k to 7k is the price range for secondhand 100Ah/150Ah batteries — try and buy batteries that are as new as possible, I wouldn’t buy anything that has a billed date over one and a half year. There shouldn’t be much difference in price for secondhand SMF or FLA batteries. The best place to buy is from a dealer, they’ll be able to test properly and they have their reputation to look after. For a single battery setup, OLX is a good option. Most people just want to dispose of their inverter batteries. For OLX, if the battery is under two years old, don’t pay more than half of the billed amount.

The testing continues with another 5 cells completed:

Cell	-Ah to 2.5Vpc	SoH
1	104.689	99.7%
2	104.821	99.8%
3	103.741	98.8%
4	100.382	95.6%
5	102.466	97.6%
6	100.143	95.4%
7	104.442	99.5%
8	102.481	97.6%
9	104.049	99.1%
10	102.305	97.4%

So far, I’m pleased with the results. Each test is taking over 5 hours, and there’s life happening in between, so it’s been a slow few days.

rockyo27 · January 21, 2024, 7:32pm

Have asked a few on olx, for lifepo4 battery one of the guy selling an okaya lifepo4 battery used in ev for 25k 48v30ah, another one selling kinetic lithium battery of 60v28ah asking 15k it maybe Liion or other, don’t know.
As i don’t know how old are these what should be the used price.
Another one is 48v160ah okaya lifepo4 one with asking of 48k.
Another is some greenbolt 12v 50ah lifepo4 with asking of 8k for new. Seems a local assembler

rsaeon · January 21, 2024, 7:57pm

The 48V 160Ah battery at 48k is a very good price, the 30Ah is priced almost the same as new. The brand doesn’t matter since they’re probably all Aliexpress/Alibaba cells/packs that have been repackaged/rebranded. You will need a 48V inverter for them though.

I didn’t realize these would be available second hand. Can you ask that seller to share photos of the 160Ah battery?

rockyo27 · January 22, 2024, 5:52am

this is how it looks. Mostly these are ev batteries

rsaeon · January 23, 2024, 6:33am

That’s definitely a very good price for a prismatic pack. I wouldn’t want to travel to Delhi for it though.

The capacity testing is complete! Here are the results:

Cell	Vpc as received	+Ah for 3.650Vpc	SoC as received	-Ah to 2.5Vpc	SoH
1	3.328	26.8	74.5%	104.689	99.7%
2	3.331	13.9	86.8%	104.821	99.8%
3	3.328	17.7	83.1%	103.741	98.8%
4	3.330	11.8	88.8%	100.382	95.6%
5	3.331	11.8	88.8%	102.466	97.6%
6	3.330	13.5	87.1%	100.143	95.4%
7	3.328	29.3	72.1%	104.442	99.5%
8	3.331	13.0	87.6%	102.481	97.6%
9	3.331	11.5	89%	104.049	99.1%
10	3.331	12.7	87.9%	102.305	97.4%
11	3.330	14.4	86.3%	100.902	96.1%
12	3.331	14.2	86.5%	105.249	100.2%
13	3.331	13.6	87%	98.522	93.8%
14	3.331	13.4	87.2%	101.093	96.3%
15	3.331	13.6	87%	103.237	98.3%

Most consider a capacity over 90% as being pretty decent for a used LIFePO4 pack — so it’s nice to see that the lowest capacity of any cell in this pack was 93.8%.

The highest capacity of any cell was 100.2%, this places that cell as one of the highest grade ones, likely a new cell. I’ll need to take apart the pack to see if it’s labelled differently.

This pack should have an actual capacity of 15 cells x 3.2 volts x 98.522 amp hours = 4729Wh. This brings the cost per kWh to Rs 14,424 or about $173. Not too bad!

For comparison, a new 150Ah FLA battery is around 10k, that’s 12v x 150ah / 2 = 900Wh or Rs 11,111 per kWh. And you’ll likely need to replace it at least once before the LiFePO4 pack starts to degrade.

4.7kWH is plenty to power my 1kW load for >4 hours, we’ve recently had a change in government and there’s talk of daily >2 hour power cuts after a decade of barely any outages. The change will be jarring but hopefully not insurmountable.

Now I’ll have to decide to either make this a 16S pack for a 48V inverter or a 24S pack for a 72V online UPS — which would also have the benefit of a larger battery and longer runtime.

Or maybe even a hybrid solution, with a UPS connected to the inverter, since the UPS would take forever to charge a 100Ah pack — most 72V UPS only charge at 8A, while an inverter should be able to charge at 20A.