My recycled 18650 DIY "powerwall" project.
For background see:
My current setup (as at 28 May 2018):
YouTube videos of the process: https://youtu.be/UTiqkOaNElY
A new 3D printer
Battery holder design improvements
I've improved the design of my 3D printed battery holders. It will now accept 7mm wide nickel strapping OR twisted copper mains wire.
The standard model will take 12 pairs of 18650s in two rows of six. And the longer versions will do 18 pairs, 24 pairs and 28 pairs.
I've also left off the skirt for the new designs (because my new printer has a heated bed).
All the 3D printer (STL format) files are available free on Thingiverse: https://www.thingiverse.com/thing:1729730
I've created a Teespring Store for the 18650 "merch" Nerdville and I designed:
I'm going to use 7 of these low voltage meters to indicate the voltage of each of the 7 groups in my 2kWh "small blocks" pack.
I'll test them to see how accurate they are.
Here's the quoted specs:
Here's the DC-DC converter I used: 10-35V to 5V 20A DC-DC converter
And here's a better one: 8-40V to 5V 40A DC-DC converter
Download the STL file from Thingiverse: http://www.thingiverse.com/thing:1892392
This is one side of a DIN rail mount for a 105x60mm BMS board. To make the other side mirror the SLT in your favorite slicer, then print.
4 Nov 2016 cascading small blocks
Small blocks extra connector clip-on
3d printer files
The flat "T" section is a custom raft to help it print nicely in PLA on a non-heated printer bed.
DIN rail mount bracket for 1x2 18650 clips:
DIN rail mount bracket for 4x5 18650 clips:
Here's a google spreadsheet for sorting a 7S12P pack: https://docs.google.com/spreadsheets/d/1DzHPjOyKAWf_BDvVjKvI2jfDCsEAXT9AFQznFGO5Ml4
I'm looking for volunteer developers who could help create a generalised app to generate sorted packs from a input list of cell capacities.
If you're interested please email me.
My primary discharger is now an OPUS BT-3100: http://lygte-info.dk/review/Review%20Charger%20Opus%20BT-C3100%20V2.1%20UK.html
Here's the original design on Tinkercad (without the skirt): https://www.tinkercad.com/things/3MQdVcH7bOt-din-rail-mount-for-18650-small-block-v2
Here's the STL file for the small blocks battery holder I used: http://www.thingiverse.com/thing:1729730
And here's the Tinkercad source: https://www.tinkercad.com/things/lft19zWDf7f-18650-battery-holder-2x6-pairs-v6b-no-skirt
Feel free to copy and modify.
I picked up an old server rack cabinet for NZ$32 today. My plan is to clean it up and use it to house all the batteries.
And here's some 18650 amusement: https://www.youtube.com/watch?v=Bn5zhqz4LsE
I'm experimenting with not using a BMS on the 1kWh Powershelf. I'm will be monitoring the cell voltage levels manually using a cheap cell level alarm. If I see the cells drifting apart more than 0.15 V then I'll slap on a BMS.
The 1-8S Lipo/Li-ion/Fe Battery Voltage 2IN1 Tester Low Voltage Buzzer Alarm I'm using to monitor the cells costs US$1.13 each on Aliexpress.
Download PDF copy
In many recycled laptop batteries the cells are arranged as three pairs of cells, which are connected in series (3S2P).
My contention is that keeping the pairs together saves a lot of processing time without much downside.
Here are the results from my testing to see if the capacity of single cells is better or worse than keeping them in pairs.
This test suffers from very small sample size. That's because I don't want to break up my good pairs. A better test would be if one of the powerwall DIYers who does do single cells would number their next batch of cells they process then we'd get a lot more data.
Is there anyone up for that task? Let me know if you are!
Graph showing the voltages of the 7 groups in my 2kWh pack.
I ran a 6 hour test with the 2kWh pack powering my Sunraiden 1000W pure-sine wave inverter driving a 22W CLF lamp (about 1.5Amps drain), then a 177W incandecent lamp (about 7Amps drain).
As you can see group 1 (red) and 2 (blue) of my groups started sagging faster than the rest so those are the ones that need upgrading.
The graph was generated by LogView from data coming out of my iCharger in Monitor mode.
Here's the latest version of my 18650 cell pairs battery holder: http://www.thingiverse.com/thing:1534159
Some good stuff in this video: https://www.youtube.com/watch?v=pxP0Cu00sZs&list=PLia3mEYsCVebndk_2F2bacuKkoYOKSWAt
Battery pack grid template to download:
And now I'm wondering about a smoke detector too: http://www.aliexpress.com/item/Freeshipping-Smoke-Sensor-Module-use-for-Detection-smoke-Detector-W-Relay-Output/1956798697.html
I get my recycled laptop batteries from my local e-waste recyler IT Rercycla in Lower Hutt (there's also a branch in Auckland).
They are currently selling them for NZ$1 per laptop battery and they have plenty.
If you have an e-waste recycle facility operating near you - go and talk to them.
I used 7 tri-colour auto transitioning LED's wired in series across the battery, plus the diffuser and polarizing filters from a old laptop display to produce the cool & silly light effect.
Made the panel by laser cutting 6km MET 'wood'. I chose MET because I suspected I might want/need to tweak the design so I didn't wean make it from something too permanent.
I probably should have chosen cardboard.
The design file was done in Freehand 9 (because I have it and it still output clean EPS files - which the laser cutter can use.
The actual cutting was done by Graley Plastics & Laser Cutting in Petone (located on my way home) and it cost $63.25 including GST. It took 3 days for them to do it.
This was my very first laser cut project and it worked out perfect for the task.
I've since tweaked the design a tiny bit. Next time I plan to laminate two layers of corrugated cardboard from a used bike box and get that laser cut.
I like the idea of housing the recycled batteries in a recycled front panel. But I wonder if it would be significantly safer in a metal box - that I can't easily quantify.
Solar lithium charge controller and BMS:
MJLorton "What is a Battery Management System / BMS?: https://www.youtube.com/watch?v=MZyY1dpka7c
These YouTube channels cover 18650's or related electronics:
Switched to using Jaycar 0.25mm Enamel Copper Wire after running out of the 5 Amp mains fuse wire I had been using (it was too expensive).
This should be suitable for approx. 10 Amps http://www.powerstream.com/wire-fusing-currents.htm
I've just sorted the best 168 pairs of cells into 7 groups of roughly equally capacity.
The median pair was 3402 mAh.
3402 mAh divided by 2 = 1701 mAh for a median cell
1701 mAh * 48 cells (in parallel) per group = 81,648 mAh or 81.648 Ah
81.648 Ah * 25.9 V (nominal 3.6 V * 7 groups in series) = 2.114 kWh
See also: Rinoa Super-Genius, How To Organize 18650 Cells for your Battery Pack: https://www.youtube.com/watch?v=NVJNol7jq0M
I've decided to solder together some of my best single cells to make good pairs. I'll add those to the pack so I can bump out the lowest capacity pairs - thus raising the overall capacity of the pack. Hopefully I can solder the cells without damaging them.
I really want to get the nominal capacity over 2 kWh.
Before you embark on a large laptop battery project itís worth working out how long itís going to take to complete.
For example - if you want to do a battery pack for a Powerwall or an electric car you will need as many as one thousand cells or more.
I will go through my numbers step by step. Some of these will be specific to my pack design and some will be more generic. Either way youíll be able to see how to copy my process if you want to.
Here's my Google Spreadsheet so you can copy and reuse it.
Your typical eighteen-six-fifty laptop battery is going to have 6 cells inside. I have seen as few as 4 cells and as many as 12.
In order to simplify my process I try to only buy 6 cell batteries. That also means Iím very likely to get 3 pairs of cells, and Iím only interested in pairs of cells for my projects.
The Powerwall Iím working on at the moment is a 2 kilo-Watt-hour pack which will be made up of 336 cells, or 168 pairs. To get that many good cells I need to open 75 laptop batteries, each containing 6 cells which actually results in 450 cells. That is 25 percent more than I need. After measuring the internal resistance and capacity I can discard the weakest 25 percent, which I give back to the recycler.
So, the first major task is to open each battery, cut off the BMS and wires, measure the voltage, write that on the cell, and then put it in a sorted pile.
Iíve measured the time this took me for 25 laptop batteries, and the average time was 5 minutes 24 seconds. The quickest I could manage was 3 minutes 36 seconds and the longest was 9 minutes and 25 seconds, and that was because the plastic casing was brittle and had to be broken apart chunk by chunk. Another battery was glued so strongly that each cell to ages to pry out. And of course they didnít stay in pairs so are less useful to me. If I was more ruthless, Iíd discard any battery that took too long, or I didnít think I could get pairs out of.
That average of 5 minutes 24 second times 75 batteries adds up to 6 hours and 43 minutes of solid battery opening. Or one easy day. So thatís not too bad.
If I wanted to build an 8 kilo-Watt-hour pack with 1 thousand 3 hundred and 44 cells that would require 3 and a half days to open 299 laptop batteries.
The second major task is charging, and measuring the internal resistance, of all the cells. In my situation I have an eye-Charger ten-ten which can balance charge 10 cells in series at a time. And, given that Iím using pairs of cells, that is 20 cells at a time.
My i-Charger take on average 11 hours and 9 minutes to complete a balance charge of 10 pairs at 1 Amp. It can take as long as 15 hours or as little as 7 hours. I can often do 2 loads per day.
I do not recommend trying to charge these random cells fast. You do not know what condition they are in, so it is better to charge slowly this first time.
It takes about 5 minutes of handling time per load, which is not too much hassle. If I average one point five loads per day itíll take me 15 days to charge up all 225 pairs of cells. Typically I start a load in the morning and reload in the evening.
The third major task is discharging and measuring the Amp-hour capacity of all the cells. This is the largest time cost. This is where youíll need to use multiple dischargers to get through as many batteries in a reasonable time.
My i-Max B6 clones take on average 3 hours 47 minutes to discharge a pair of cells down to 3 volts at 1.5 Amps. The longest time was an amazing 6 hours and 40 minutes. Weak cells will discharge much faster.
Again, I do not recommend discharging recycled cells harder that 1.5 Aaamps per pair. You donít yet know how much you can trust them.
I typically do the discharging at work because they need attention intermittently all day.
I currently have 4 i-Maxís running at a time to speed this process along. But that means I have to do a reload at random times throughout the day. Which is a hassle.
If I had 8 i-Maxes it would double the number of random times I would have to check for a finished run. Even more hassle. This task-process could be improved.
My current process will take about 19 days on this task, to discharge all 225 cell pairs. An 8 kilo-Watt-hour pack would take 75 days to complete this task.
The fourth major task is 3D printing cell-pair holders. As you can not yet buy battery holders for pairs of cells from e-Bay or Ali-Express you will have to 3D print them.
My 2 kilo-Watt-hour pack requires 42 of my 2-by-4 pair battery holders. That involves about 1 and a half minutes of handling time per print and 57 minutes for my little Printrbot Junior to print. That adds up to roughly 42 hours of printing, but only 1 hour total handling time.
The first four major tasks can be done in parallel to some extent.
The fifth, sixth and seventh major tasks are; soldering, testing and packaging for which I estimate about 2 days all up. More if you factor in package design time.
So to complete my 2 kilo-Watt-hour pack Iím expecting about 21 8-hour days worth of time on my part.
An 8 kilo-Watt-hour Laptop Powerwall, built with the equipment I currently have, would take about 77 days, or over 16 40-hour weeks.
It might be cheaper to buy a real Tesla Powerwall. But not nearly as fun.
If you'd like to 3D print my 18650-pairs battery holder here is the STL file: http://www.thingiverse.com/thing:1324840
It requires 3.44 meters of 1.75mm dia. PLA to print one of my battery holders, which holds 16 cells (8 pairs).
A 1kg roll of 1.75mm diameter PLA has about 300 meters and costs around US$30-60 depending on the source. So for a US$50 roll, that 3.4 meters would be US$0.57 per holder - but you need two - top and bottom. My 2kWh pack needed 42 in total. So for the whole pack it was about US$24 of PLA.
Pretty cheap... as long as you don't factor in time, and the cost of the 3D printer. My little Printrbot Jr was US$400 plus shipping and taxes.
How long does it take me to open a laptop battery, separate it into pairs and measure the voltages? On average: 5 minutes 27 seconds.
That includes opening the laptop battery case, clipping off all the unwanted bits, measuring the voltage and then writing the voltage on the cell(s).
If I'm going to be doing a lot of this - it's a good idea to budget the time each task takes. So today I timed myself opening 26 laptop batteries over the last two days:
It ranges from 3 minutes 54 seconds, up to 9 minutes 3 seconds. Mostly the difference is due to how easy the plastic case comes apart. Some batteries are really stubborn and come apart in small pieces. Others just take a chomp at one corner with the wires cutters, then running the screw driver along the back seam, after which the bottom pops off with ease.
My 2 kWh pack requires 336 cells, which will come from about 75 laptop batteries, will take about 6 hours and 45 minutes to open, separate and measure the voltages.
My main rationale for keeping the cells in pairs is to reduce the overall time it take to build a pack. I don't have any comparative data to say if that is true yet.
A second benefit is that most (but not all) laptop batteries with 18650 cells are wired as three pairs of cells which are then connected in series. This meas that if a single cell in a pair fades it forces it's twin to take more charge. One cell is undercharged and it's twin is over charged - leaving the pair in bad shape. My voltage, ESR (resistance) and capacity measurements should find those bad pairs and they'll be left out of the pack.
If I separate all the cells I'm liable to find the weak cells, but also be over-confident in the stronger of the twins. Keeping the cells in their pairs avoids that problem.
It also provides an easy tab for me to solder the cells together - so I avoid having to directly solder on to the cells - thus avoiding heat stressing the cell.
Parts list for a multi-18650 balance battery holder:
And some other parts:
I'm busy printing battery cell holders.
As that's happening I'm trying to decide whether to build two 1 kWh packs (and connect those together) or one 2 kWh pack. Either option would include it's own 40 Amp breaker/switch (should probably swap that for a 20 Amp breaker) and cheap 20 Amp balancing BMS with thermal cutoff, and 45 Amp Anderson connectors.
My current design is either 1 kWh 24P7S (24 in parallel, then 7 of those in series) or 2kWh 48P7S.
The long term plan is to build more packs and work my way up to 4 kWh or more.
Pros and cons:
I'm leaning towards building the 2 kWh pack because it'll be faster and has more potential for driving larger loads in future.
Note: I'm assuming the 1C rating for these random batteries is 1 Amp each. My Enviromower pack has a peaking draw of 94 Amps! or 7.8 Amps per cell (which is probably dangerous and bad for the batteries) - so my ball-park 1C rating of 1 Amp per cell above is very conservative.
I'll be using the same 0.2 * 7 mm nickel plate/strap that I used on the go-kart battery pack.
It has a cross-section of 1.4 mm2 which provides a current capacity of around 20 Amps (source).
I've got the solar charge controller working on my 1kW 24V 24P7S go-kart pack - which I'm currently testing as a small power wall.
I've chosen to set it to charge up to 28V (4.0V per cell group). This is about 75-80% full charge - which should mean the pack will last a lot more charge cycles. But it also means the capacity is limited to about 70-80% it's theoretical full charge. (The "full charge" is based on the capacity readings I measured on each cell pair.)
I've set the low voltage cut-off at 23V (3.3V per cell group).
My current solar power setup uses 4 lead acid deep cycle batteries.
I'm wondering if I could replace those with a lithium battery pack made from a lot of 18650 sourced from post-consumer recycled laptop batteries.
To start with a might use the pack I built for my 24V electric go-kart.
I have a solar charge controller that can manage lithium batteries on order. The pack already has a balancing BMS.
I donated my inverter to a worthy cause so am waiting for another one to arrive.
Once that's all working I could build another pack identical to the electric go-kart pack and run them in parallel. Or build up a larger pack just for the main solar setup - which could be designed in the style of a https://www.teslamotors.com/powerwallTesla power wall.
Worth a look: