Battery sandwich - different approach to DIY powerwall (based on 18650s)

roman2

New member
Joined
Mar 17, 2021
Messages
13
Hello, at the beginning I'd like to thank all the contributors in this place. (and sorry about my english in advance)

According to DIY powerwalls which are build on PCBs, I saw couple different designs (on aliexpress, etc) but none off them was 100% compliant with my requirements ;) So I decided to design and build my own solution. After couple tries and tests I designed "Battery Sandwich v1.1", which looks like this:
20210314_141320.jpg
(I've put just a half of 18650s to show on the pics how it looks inside).

One set to assemble 8 batteries 18650 looks like this:
20210312_170920.jpg
20210312_171006.jpg(You need just standard battery holders and make holes in that, in place where the screw comes throw).

Advantages - in short sentences:
- It's very easy and fast to assemble, no soldering nor welding as well, just a screwdriver
- Very solid and stable. Special spacers under the PCBs make them good positioning and stability
- Easy to maintenance (to disassembly, to replace some 18650s, to rebuild etc). Important especially when you're building from recycled/salvage batteries.
- Each of 18650s is connected to bus bar (to connector at the center of pcb) via fuse. I'm using axial, glass fuses. They are not so great, but not terrible. Easy to replace if the fuse blow up.
- Scalability, you can build i.e. 7s8p batteries, but also big powerwall packs with it as well.
- Build-in balancer. One set, one PCBs has two ICs of passive balancer with power of 0,5A each. Balancer is active from 4.2v and drop voltage to 4.19v, each one is switchable (on/off) by jumper on PCBs. During assembling/disassembling switching off to prevent short circuit on the screw.
- The whole design is simply - no complex ICs, no programmable logic. Easy and simple as like each of the best solution should be.

Design considerations:
- Fuses. I don't like fuses based on PCB traces or so called 'tesla' fuses. Both solutions are pretty unpredictable and could start the fire. Besides 'tesla fuses' are often made so wrong (on the yt there are many videos where the fuses are made and each of them is different in length, which make them working on very different working parameters). Cheap, glass axial fuses are kind of average of its quality, but I think they are good enough.
- The connection spring/leaf. Made from steal, nickel plated. 5mm in wide. So they make good connection to top and bottom side of 18650s (if of course the whole set is properly assemble).
- The balancer. Active above 4.20v (and releases at 4.19v). Interesting thing is if you building big packs, which are build i.e. from 80 x 18650 (all of them in parallel) you also increasing the power of balancer, because you are using 10 sets (10 pcbs with balancer), so you will get power of balancer: 10 x 2 x 0.5 = 10A. Despite the power of balancing is significant, still the additional protections is recommended. Another interesting thing with the balancers in the big packs is that they never be active on the same voltage (4.20v), i.e. one of them will active at 4.20v but another on the 4.21v. It's kind of good because there is no need to all of the balancers should active at the same time. i.e. if the voltage is 4.2 some of them will be active, but the rest of them will turn on only if the voltage is still rising.
- the "plus" wire is connected to "minus" side of the sandwich by the screw. There is no dangerous during the mounting, because there is no power until is connected by the jumper (on the "plus" side). After the wire is connected, it's protected by the fuse. Besides there is enough space (between the screw inside and the 18650 batteries) that there is no risk to short the circuit, but overall the heatshrink on 18650s is a must. The plastic nippels (sorry, I don't know is it called) on the end of the screw protect against accidentally short circuit. But, if you even make in some way the short circuit throw the screw it will blow only one fuse no matter how big is your module (rest of the balancers will work).
- The balancer generates heat by the large resistors on PCB, but space from pcb to the bottom of 18650s is aprox. 5mm, so the heat does not affect on the single cells (during the balancer is active).
- Nominal power. I'm still before tests, but based on my calculation one set (8 of 18650 batteries) should handle continuous 16A at least with no power losses (or minor ones - the fuses cause some minor power losses for sure), which I believe is applicable for most of DIY powerwalls based on recycled/salvage 18650s. This gives 2A per 18650 battery. In details: The PCB is design to handle 2A per one 18650s, the PCB traces are wide moreover this could be improved in project by 2oz copper PCB if needed. The bus connector is made from 1mm width brass (tinned). I didn't found what is the rating given by the vendor, but i believe it could handle much more power then that. (Currently I'm working on the bus bar, these ones shown on the picture are 2x4mm^2 copper). The connection leaf are made from steal (nickel plated). The steal has poor conductivity (comparing to copper), but it should handle 2A each as well. Improving this could be much costly (beryllium-copper leaf would be needed especially made for this project).
- I've try to design this project coherent with all good practices, simple and inexpensive as much as possible.


So... what do you think? :) If someone would like to test it I have couple sets and I can send it (in Europe for sure, the rest of world - I don't known what is the costs of delivery so I'll check individual).

More pictures:
20210314_141237.jpg
20210314_141330.jpg
20210314_141336.jpg
20210314_143100.jpg20210314_141434.jpg
 
Last edited:
Very very nice Roman, good effort and good result man(y) I like the jumper stuff to keep circuit open; I was wondering if jumper could become loose, sometimes jumpers tend to "open up" a bit losing their strong grip. I already thought of myself mounting your board (lucky man!) and doing some maintanance sooner or later, did you consider using "socketed" fuses, something like this to avoid unsoldering a blowed fuse:

ATTiny IR.jpg

Keep it up, good job!

jes
 
Thanks.

Well, you are putting up or off the jumpers only during assembling/disassembling, so I suppose, it's once for a long time, but if they get loose, there is easy trick to make them tight again. You need just to squeeze a little (but very little) the pins with the plier or sth.

"Socketed fuses" are great, but unfortunately they are too big (the length is 20mm). I thought to use them and i.e. make 4 of them on the "plus" side and the other 4 on the "minus" side, but I don't like it that way. The good choice could be smd socket fuse, but it would be costly. Another easy way is to design tiny pcb traces as a fuses (which I don't like!) and (as a spare) make place that additional axial fuses that could be added to "replace" the blown up one.
 
Bus bar. I see 2 ways to do it:

1. To buy a flat copper bar and make holes in that, sth like:
PLASKOWNIK-MIEDZIANY-20X5-10-CM-MIEDZ-M1E.jpg

2. To use standard wire. On the photo in the first post I used 2x4mm2 wire. It could not be enough for big packs, so I found a way how to use much more wire. On the picture below there is 10x2,5mm2 wire, which gives 25mm2. I use spacial washer (this is connector actually but upside down) to hold wires on the terminal. Everything is thigh screw:

20210330_090653.jpg20210330_090705.jpg20210330_090726.jpg20210330_090739.jpg

Which way is the best.
The 1. It's good solution and one piece of metal looks fine. The weakness are: flat bar could not be easy to buy, and there is a need to use bench drill.
The 2. Seems better, because: easy to buy, standard wire, easy to do (no special tools needed, there is no need to make the wire super straight even, you can use the wire as is). There is one design consideration. The wires are rounded (It's not flat), so the contact between the wires and the top of terminal on pcb is not so great. On the oder hand, the current in this place is only from/to one set (8 x 18650), so it's limited, the current is not from all pack. And, in my example, 4 wires are in contact with top of terminal, so this should be fine. Or... to use a hammer to make some wire a little less rounded in those points ;)

If somebody of you thinking about building battery based on my 'sandwiches', just let me know, I can send you samples. I'm going to make bigger order from my suppliers in the future.
 
Last edited:
Nice work Roman.
For the wire holder system, yours is already quite good.
Maybe have a look at a typical circuit breaker. They grab cable just like this all day in every house switchboard....
Also maybe consider the busbar they use in switchboard shaped like E design in a strip eg
E
E
E
etc
Maybe put plastic tube on wires/busbar for insulation?
 
Can you post the schematic for the balancing circuit? Just professional curiosity.

I imagine the resistors can potentially get very hot. Placing them farther apart could avoid hotspots. Or placing them near the (-)Terminal - that the big piece of metal and copper wires can soak up and dissipate the heat much better

Most people on the forum only charge to 4.1V or even less to increase longevity of the cells. So perhaps some solder bridges to allow adjustment to the balancing voltage would be useful:
1617240020699.png
 
Nice work Roman.
For the wire holder system, yours is already quite good.
Maybe have a look at a typical circuit breaker. They grab cable just like this all day in every house switchboard....
Also maybe consider the busbar they use in switchboard shaped like E design in a strip eg
E
E
E
etc
Maybe put plastic tube on wires/busbar for insulation?
I know what you mean with the "E" like bus bar, I was looking them, but all of them have different resolution (not 20/40mm).

Plastic tube is a great idea! It's so simple. As soon as the stores will be open again I will try it.
 
Can you post the schematic for the balancing circuit? Just professional curiosity.
Sure:
schematic.png

Placing the resistors near to terminal is good idea, but this place is 'forbidden'. There should be only the wide copper area, this is crucial place, because of big current transmission from pcb to terminal. I don't want to place there anything what could make the transmission be weaken. I'm thinking about placing the resistors very close the edge in one line. I thought that placing the resistors near the end of the screw is good idea, but there is a plastic nipple, so distraction of the heat I suppose is not that great. But, on the other hand no matter where the resistors are, they getting cooled very quickly.

I like the idea with "switch-bridges" which allow to adjust the balance threshold. Of course it's not easy, the circuit would be different. Maybe in the next version... What options do you see? 4.0v, 4.1v, 4.2v? And what would be the default one, 4.2V?

Thanks.
 
While I studied this stuff long time ago, I didn't pay too much attention back then since I thought I'd never use it. The last couple of years, I've been using it more and more at work, but still re-learning many of the things. So please take the following as a mix of advice, and questions to advance my re-learning.

Placing the resistors near to terminal is good idea, but this place is 'forbidden'. There should be only the wide copper area, this is crucial place, because of big current transmission from pcb to terminal. I don't want to place there anything what could make the transmission be weaken.
Very sensible . I was only thinking of my use case of 200mA/cell, but obviously most people will want to use much much higher currents.
I assume having 2 separate balancing circuits is for the same reason, to not obstruct the current "nexus"?


HY2213: very interesting component! Though detection voltages are not adjustable :(
But why did you choose the HY2213-BB3A to drive a 2N7002 N-Channel to drive a P-Channel SI2301?
Would it not make more sense to use a HY2213-BB3A to drive something like the N-Ch IRLML6344? Far less components and simpler design?
Or a HY2213-AB3B to directly drive the P-Channel SI2301?

If I understand correctly, the function of C2 and C3 are to: allow SI2301 to quickly turn ON, but only slowly turn OFF? What is the purpose of that?
If to slow down the rise/fall time of SI2301, would it not be simpler to just use a higher value R9?
The HY2213 already has a built-in 250ms detection delay, so reducing switching frequency is probably superfluous.


Do you solder the PCB yourself? Are those 0805? Or even 0603? Looks very good!!
 
Last edited:
While I studied this stuff long time ago, I didn't pay too much attention back then since I thought I'd never use it. The last couple of years, I've been using it more and more at work, but still re-learning many of the things. So please take the following as a mix of advice, and questions to advance my re-learning.


Very sensible . I was only thinking of my use case of 200mA/cell, but obviously most people will want to use much much higher currents.
I assume having 2 separate balancing circuits is for the same reason, to not obstruct the current "nexus"?
You're right. There is one more reason for 2 circuits. I design the solution to be "dummy-proof" as much as possible. For example: I'm imaging like someone forgets to "un-jumper" the circuits before disassemble and makes short-circuit. It's nothing dangerous, the fuse just blow up and that's it. But he could not even noticed that, and won't be aware that the balancer is not working. That's why there are 2 (each one with independent fuse).

HY2213: very interesting component! Though detection voltages are not adjustable :(
But why did you choose the HY2213-BB3A to drive a 2N7002 N-Channel to drive a P-Channel SI2301?
Would it not make more sense to use a HY2213-BB3A to drive something like the N-Ch IRLML6344? Far less components and simpler design?
Or a HY2213-AB3B to directly drive the P-Channel SI2301?

If I understand correctly, the function of C2 and C3 are to: allow SI2301 to quickly turn ON, but only slowly turn OFF? What is the purpose of that?
If to slow down the rise/fall time of SI2301, would it not be simpler to just use a higher value R9?
The HY2213 already has a built-in 250ms detection delay, so reducing switching frequency is probably superfluous.
You are almost right, but... it's all about reducing switching frequency ;)
The HY2213-BB3A is usually used in cheap, chinese BMSs with balancing current like... 20mA. And they work ok, because when the balancer is set to active, the voltage drop is minor. For example: when the voltage is 4.2v on particular cell, the balancer will set to active and the voltage will drop immediately on this cell to (i.e) 4.195v. This won't make the chip to turn it off, because the falling threshold of HY2213-BB3A is below 4.19v, so it's still above it. That the right way to work.

My use case is different due to larger bypass current. When the balancer is set to active the voltage drop immediately to (i.e) 4.18v, which is below the threshold, and the chip is trying to deactivate the bypass. Yep, there is a window of 250ms time activation, but it's not enough. The result is that the balancer is active for quarter the second, then deactivates, and activate again, and so on. I found 3 reason why it's bad idea (that are my opinions):
1. (I'm not an expert about li-ion battery, but I suppose) It could not be so healthy for battery
2. What is the really (the average) bypass current? Less then nominal for sure. 0,3A? 0,2A?
3. How to check the voltage with multimeter when the balancer is active? There is no way that simple multimeter show you the correct voltage during the balancers working.

That's why there is a need to expand the activation window and that's what the capacitors are for. And I needed to use 2 transistor, because the output of the HY2213 is low/high state. (Depending which one: for N-mosfet or P-mosfet) when the balancer is active the output is in high state and the capacitor will charge immediately, but when the balancer is not active the output of HY2213 is in low state so it would discharge the capacitors immediately if they would be connected directly to the HY2213 output. So I've put there additional transistor, and 1M resistor as pullup (which "turns off" the last mosfet after the capacitors slowly discharge).

The result is that the activation window depends on the capacity of the capacitors. Actually 2x10uF, what gives like 15second continuous bypass, after that there is very short time when HY2213 checks again the voltage. And so on... But if the voltage is on very high level (i.e. 4.25v) all the balancers will be active in continuous way without pause.
Do you solder the PCB yourself? Are those 0805? Or even 0603? Looks very good!!
All of them are 0603, but no, I didn't soldered by my hand. I soldered first version and after that I said to myself "Oh no no no no..." :)

(Sorry about my english, I've done my best, I hope you'll understand;) )
 
All of them are 0603, but no, I didn't soldered by my hand. I soldered first version and after that I said to myself "Oh no no no no..." :)

(Sorry about my english, I've done my best, I hope you'll understand;) )
My minions complained when I told them to solder 0805 parts!
It's perfectly understandable!
 
I like this project very much, I have been thinking about something like this too. The idea about not welding or soldering on the batteries is good. Where did you get the battery clips / springs from? I searched both ebay (Germany) and Aliexpress, maybe you can give me a hint...
 
I like this project very much, I have been thinking about something like this too. The idea about not welding or soldering on the batteries is good. Where did you get the battery clips / springs from? I searched both ebay (Germany) and Aliexpress, maybe you can give me a hint...
Thanks,

You can buy the battery contact on ali: aliexpress.com/item/4000055282831.html
I'm sure I bought it much cheaper, try to look similar auctions.
 
Ok, thank you very much. Do you have an estimated price on your boards? At the moment i'm building a small powerwall with the JEHU boards, because I like that the cells can be changed. Your design give the same possibility, but with less work (I guess).
 
Ok, thank you very much. Do you have an estimated price on your boards? At the moment i'm building a small powerwall with the JEHU boards, because I like that the cells can be changed. Your design give the same possibility, but with less work (I guess).
I really don't know right now what would be the final price. What I can say right now is that I am aware that the boards have to be cheap and I try to design this with acceptable costs in production, but without loss on quality.

My goal is that the price should be like max. 10Euro per one kit (with all the periphery as on the photos in first post) , or even just 1 Euro per single 18650 cell (8 Euro), but as I said - I don't know. I will try to optimize some production things, i.e. right now I'm trying to talk to chineese fellows, but the answer is usually the same: "Ok, my friend, but minimum order quantity is..." and here comes some big numbers :)

Anyway, what's for sure, if the project will succeed the very first orders will be processed at cost or even below that, so the price could be much lower at the beginning.

For now, I decided to give a shot myself and design new version with additional feature like switchable voltage threshold (even with option for lifepo4, but I don't know if somebody will ever use it;) ). But it takes some time.

What I can offer to you right now is sample kit for free, you will see it by your hand, will try it, and any opinions would be appreciated. If you are interested contact me on priv.
 
Have you thought about keeping heat from the bypass resistors away from the cells?
Probably not good to be heating the cells at all & especially unevenly, eg one end gets hotter because of board/resistor position?
 
Have you thought about keeping heat from the bypass resistors away from the cells?
Probably not good to be heating the cells at all & especially unevenly, eg one end gets hotter because of board/resistor position?
I'm gonna to do the tests so we'll finally find out ;) Stay tuned.
 
Hello, at the beginning I'd like to thank all the contributors in this place. (and sorry about my english in advance)

According to DIY powerwalls which are build on PCBs, I saw couple different designs (on aliexpress, etc) but none off them was 100% compliant with my requirements ;) So I decided to design and build my own solution. After couple tries and tests I designed "Battery Sandwich v1.1", which looks like this:
View attachment 24228
(I've put just a half of 18650s to show on the pics how it looks inside).

One set to assemble 8 batteries 18650 looks like this:
View attachment 24229
View attachment 24230(You need just standard battery holders and make holes in that, in place where the screw comes throw).

Advantages - in short sentences:
- It's very easy and fast to assemble, no soldering nor welding as well, just a screwdriver
- Very solid and stable. Special spacers under the PCBs make them good positioning and stability
- Easy to maintenance (to disassembly, to replace some 18650s, to rebuild etc). Important especially when you're building from recycled/salvage batteries.
- Each of 18650s is connected to bus bar (to connector at the center of pcb) via fuse. I'm using axial, glass fuses. They are not so great, but not terrible. Easy to replace if the fuse blow up.
- Scalability, you can build i.e. 7s8p batteries, but also big powerwall packs with it as well.
- Build-in balancer. One set, one PCBs has two ICs of passive balancer with power of 0,5A each. Balancer is active from 4.2v and drop voltage to 4.19v, each one is switchable (on/off) by jumper on PCBs. During assembling/disassembling switching off to prevent short circuit on the screw.
- The whole design is simply - no complex ICs, no programmable logic. Easy and simple as like each of the best solution should be.

Design considerations:
- Fuses. I don't like fuses based on PCB traces or so called 'tesla' fuses. Both solutions are pretty unpredictable and could start the fire. Besides 'tesla fuses' are often made so wrong (on the yt there are many videos where the fuses are made and each of them is different in length, which make them working on very different working parameters). Cheap, glass axial fuses are kind of average of its quality, but I think they are good enough.
- The connection spring/leaf. Made from steal, nickel plated. 5mm in wide. So they make good connection to top and bottom side of 18650s (if of course the whole set is properly assemble).
- The balancer. Active above 4.20v (and releases at 4.19v). Interesting thing is if you building big packs, which are build i.e. from 80 x 18650 (all of them in parallel) you also increasing the power of balancer, because you are using 10 sets (10 pcbs with balancer), so you will get power of balancer: 10 x 2 x 0.5 = 10A. Despite the power of balancing is significant, still the additional protections is recommended. Another interesting thing with the balancers in the big packs is that they never be active on the same voltage (4.20v), i.e. one of them will active at 4.20v but another on the 4.21v. It's kind of good because there is no need to all of the balancers should active at the same time. i.e. if the voltage is 4.2 some of them will be active, but the rest of them will turn on only if the voltage is still rising.
- the "plus" wire is connected to "minus" side of the sandwich by the screw. There is no dangerous during the mounting, because there is no power until is connected by the jumper (on the "plus" side). After the wire is connected, it's protected by the fuse. Besides there is enough space (between the screw inside and the 18650 batteries) that there is no risk to short the circuit, but overall the heatshrink on 18650s is a must. The plastic nippels (sorry, I don't know is it called) on the end of the screw protect against accidentally short circuit. But, if you even make in some way the short circuit throw the screw it will blow only one fuse no matter how big is your module (rest of the balancers will work).
- The balancer generates heat by the large resistors on PCB, but space from pcb to the bottom of 18650s is aprox. 5mm, so the heat does not affect on the single cells (during the balancer is active).
- Nominal power. I'm still before tests, but based on my calculation one set (8 of 18650 batteries) should handle continuous 16A at least with no power losses (or minor ones - the fuses cause some minor power losses for sure), which I believe is applicable for most of DIY powerwalls based on recycled/salvage 18650s. This gives 2A per 18650 battery. In details: The PCB is design to handle 2A per one 18650s, the PCB traces are wide moreover this could be improved in project by 2oz copper PCB if needed. The bus connector is made from 1mm width brass (tinned). I didn't found what is the rating given by the vendor, but i believe it could handle much more power then that. (Currently I'm working on the bus bar, these ones shown on the picture are 2x4mm^2 copper). The connection leaf are made from steal (nickel plated). The steal has poor conductivity (comparing to copper), but it should handle 2A each as well. Improving this could be much costly (beryllium-copper leaf would be needed especially made for this project).
- I've try to design this project coherent with all good practices, simple and inexpensive as much as possible.


So... what do you think? :) If someone would like to test it I have couple sets and I can send it (in Europe for sure, the rest of world - I don't known what is the costs of delivery so I'll check individual).

More pictures:
View attachment 24231
View attachment 24232
View attachment 24233
View attachment 24235View attachment 24234
I think your design is great! I have a similar design of 14s16p times 2 that is powering my house like a champ. I will share mines after I clean up the protype design. The beauty of this type of design is if a cell goes bad, open your "sandwich" and dump the bad pickle. With my 14s16p design I can get to the bad pickle and the battery is still operational on the other 15 in parallel. No down time ever. Downside is I have not put a BMS to my battery. 244 cells is a lot for me. I have developed better a better BMS then those from China using an Ardruino. Now I'm too lazy to do all the analog multiplexing for 244 cells. I think I'm going to monitor cells in pairs. Anybody you know doing this?
 
I think your design is great! I have a similar design of 14s16p times 2 that is powering my house like a champ. I will share mines after I clean up the protype design. The beauty of this type of design is if a cell goes bad, open your "sandwich" and dump the bad pickle. With my 14s16p design I can get to the bad pickle and the battery is still operational on the other 15 in parallel. No down time ever. Downside is I have not put a BMS to my battery. 244 cells is a lot for me. I have developed better a better BMS then those from China using an Ardruino. Now I'm too lazy to do all the analog multiplexing for 244 cells. I think I'm going to monitor cells in pairs. Anybody you know doing this?
How are you running your house off of 2x 14s16p?? That's not a lot of power, unless you are only powering lights and maybe fans.
 
How are you running your house off of 2x 14s16p?? That's not a lot of power, unless you are only powering lights and maybe fans.
My solar system starts at 8:30am with the sun and my battery packs. My house is powered just like the utility company would do it. When the sun goes down we are watching a little TV or on the computer still on battery. When I'm in bed between 9 and 10pm I hear the relay click. Back to utility. We are sleep and everything is off so this little utility used is nothing to me. Even the refrigerator goes to sleep after 10pm in my house I found that it does not need to be powered. I gave away a Sony 55" TV because it draws to much energy. The cheap TCL brand at Best buy looks great and use very little power. Yes everything in my house is ultra low power consumption. Dish washer, Washing machine run on structured times. (11am) I'm just playing to see how little energy I can use and still have a house that functions. At the flip of a Transfer switch I'm back on Utility power if need be. Everybody I know see my bill is really low but can't believe it is so. Only thing that matters is I am happy and I am not going to lie to myself about my system. It is what it is.
 
Back
Top