TaK's Battery Build - 18650 w DIY CD Welder- Lessons Learned

[TaK]

Hello,

This is my first contribution to this forum.

I've been working on my second ebike battery pack. This one is made with 18650 cells in a 14S4P configuration. I decided to make a capacitive discharge welder to properly assemble the pack since I already had most of the necessary components and I only had to spend $10 or so additional (the SCR). The pack is almost finished at this point, and I figured it would be good to write out some lessons learned for others who might want to do similar builds. I don't think anything in this build is unique, but there is definitely some things I could do better a second time... I've taken a lot of pictures so I could probably write out more stuff if there is interest.

Here's a few pictures of what's going on:

Welder:

Battery:

Welds:

So anyway, in no particular order, here's some things I recommend:

1. Don't use a car audio cap

Some people have been successful with them, but I tried one and it was garbage. I recommend a real cap. I used this Nippon Chemi-Con electrolytic rated for 80V and 0.1F. I think it was about $50 when I got it years ago. Actually, I used two of them, but that's just because I had two sitting around. It was working fine with one.

2. Put a resistor across your probes

If you're going above ~20V to charge your caps and you're using an SCR like I did, then you should add a parallel resistor across your probes. You can see the resistor (the ceramic, high wattage) in the pictures above dangling between the caps. If you don't, your welder can/will discharge as soon as you touch your probes to the nickel strips even without you commanding a discharge (foot pedal in my case). It makes a huge pop and a lot of sparks (and is very exciting) and it damages your battery.

The issue is that the SCR is triggered by current on the gate, but can also be triggered by a high voltage slew between the anode and cathode. When you touch the probes to the metal, they are shorted and you have very quickly slewed the voltage and it will immediately trigger. To fix this, I just had to basically short the probes with a resistance which will keep them at approximately the same voltage when the probes are floating so there is no change when you short the probes together. Typically there will be no current going through this resistor since the probes make a lower impedance path, but I made them large power resistors to save them in case I triggered the welder without the probes making good contact in which case this resistor would take all the energy.

Here is a crude schematic. The resistor in question is the one that effectively shorts the probes.

3. Insulate your cells better before building

If I could go back, I would put another heat shrink sleeve on each cell before building. I don't want the insulation rubbing and the cans of the 18650 cells from adjacent rows to make contact. I would probably also but sheets of Kapton insulation between adjacent rows.

4. Nickel strips only

The first nickel spool I got was cheap and it was actually nickel plated steel. I didn't know to look out for this. If the spool is cheap, its probably nickel plated steel which is no good if your bike pulls good power.

5. Cut Vs in the strips

The welds are a lot stronger if you separate the weld dots by removing a strip of material. If you image search the commercial battery packs, you'll see that most have a notch taken out between the welds. It wasn't too hard to implement with a simple V cut as you can see in the above images. Remember you want to take all the energy stored in your cap and burn it at the interface of your nickel strip and your battery case; anywhere else is wasted and makes problems.

5. Keep the probe tips clean

The biggest contributor to making consistent welds was to clean the probe tips often and only touch new copper. The nickel actually melts onto the copper probes. You can see it if you look under a scope. This adds a lot of resistance to your probes and soon you're getting a lot of sparking and scarring. You want to make sure that the area of the probe tip that is touching the nickel strip is fresh.

I got in the habit of only making two welds (4 dots) and then reconditioning the tips. I used a Dremel tool with a little sanding wheel to recondition the tips quickly. Also, I would make one weld then switch the probes between my hands which would basically force me to use a fresh section of the probe tip make the second weld. This welding schedule was tedious: one weld, switch probes, one weld, sand, repeat... but since I had no sparking once I started that schedule. Since this battery pack is a one-off, it was worth the time.

Here's a pic of the reconditioned (sanded) probe tips:

Happy battery pack making.

TaK

 

[redilast]

None of the pictures are working

 

[TaK]

How about now?

 

[Punx0r]

Those are good looking welds, thanks for the write-up Am I right in thinking you use no discharge pulse-control?

 

[tmho]

If possible, please provide more details in making the DIY CD Welder as well as using it.

 

[okashira]

Well done!
Yep, when I was looking at welding at first, I tried three different car audio caps. All of them tested to 1/10 or so their rated Farads.
I love frugal/diy efforts like this

 

[TaK]

Those are good looking welds, thanks for the write-up Am I right in thinking you use no discharge pulse-control?

Thanks. Correct, there's no pulse control included here. This is definitely a KISS build. My understanding is that if you want to make a larger weld that needs a lot more energy, then the pulse control becomes critical. Also, I'm sure it would help keep my probe tips in better shape.

On the subject of the size of the welds... The questions come up:
How big should my welds be?
How many should I make?
How do I check to see if I made the right choices.

Here's my thoughts. As far as the welding goes, you want good electrical connectivity from the battery to the strip and you want good mechanical connectivity so the welds don't degrade with the shock and vibration associated with an electric vehicle. I went with fairly small diameter welds, each about 0.025" diameter because I was able to make them easily and consistently and they have good connection.

These welds were made with 40J:

It's not the best pic but you can see that when I pulled the tab off, the weld stayed behind so it is adhering well. I decided to go with 6 welds per battery terminal because my 0.375" x 0.005" nickel strips have about the same cross sectional area as 6 of these 0.025" welds. When this pack is done, I'll probably do a full load discharge and look at the welds with a FLIR camera to see if anything is getting hotter than it should. This should be a reasonable test to see if 6 welds is enough electrically.

For the mechanical side of the problem, I also think 6 welds will be sufficient. The tabs take a lot of force to pull off and I need to use pliers to get a good grip. I used low temp hot glue between every cell to try to keep them from vibrating. The hot glue is not really to keep the battery pack together, I'll reinforce the outside later, it's just to keep the individual cells from moving relative to their neighbors, and thus keep fatigue off the welds. I don't want the welds to be structural in any way for the battery back (except the weight of the nickel strips themselves). I think the only way to check this is going to be periodic inspection though.

If possible, please provide more details in making the DIY CD Welder as well as using it.

I can write something up later.

Well done!
Yep, when I was looking at welding at first, I tried three different car audio caps. All of them tested to 1/10 or so their rated Farads.
I love frugal/diy efforts like this

Ya, I tried this SoundQuest SQCAP2M 2Farad POS from Amazon. It doesn't work. I was using it at 12V which would theoretically be 144J, but it cannot make even very small welds. They're trash. It serves me right for buying something without an actual datasheet.

TaK

 

[TaK]

Progress update:

I've welded the parallel nickel strips to the pack:

Since there will be almost zero current going through this strip, I cut the 3/8" nickel strip in half to make it a little easier to work with. I'll solder wires onto the overhanging tabs to go to the BMS and then fold the tabs flush with the edge.

Those little, white scratch marks are actually just fibers from an alcohol wipe. They are a PITA to remove it turns out...

Next steps are to attach the BMS and wire the outputs.

Edit: These welds were done at 50J. I had to bump it up from 40J to get a nice weld from the nickel strip to nickel strip without the V gap in the upper layer.

TaK

 

[999zip999]

How much easier was it with the v gap ?

 

[okashira]

How much easier was it with the v gap ?

slitting the nickel makes a huge difference. Can mean the difference between no weld and good weld on thicker nickel with an underpowered welder

 

[TaK]

How much easier was it with the v gap ?

slitting the nickel makes a huge difference. Can mean the difference between no weld and good weld on thicker nickel with an underpowered welder

I agree, and another thing is the consistency of the welds. In my testing, when I was not using a V, differences could be seen from weld to weld which I believe were from how hard I was pressing and how far apart the probe tips were from one another. The V gap helps make your welds invariant to small differences like those, although I did try to keep approximately the same pressure and distance between welds anyway.

 

[okashira]

Here you go, if you may find this useful. I made this calculator for a mosfet type welder and for batteries, but you could tweak it to work with a cap welder (hint, major change in the % peak current column is needed - you need to account for the falling voltage of the capacitors based on coulombs discharged) let me know if you need help with that. I spent about 15 minutes giving it a start... added calculated voltage drop @ the cap, now you need to use this to compensate for new voltage..
I put a good amount of work into it, so please give credit and keep my contact on there if you modify it. And distribute this read only link.
https://onedrive.live.com/redir?resid=6A929FF99605B0EA!12003&authkey=!ABMgmXsIdQapFfw&ithint=file%2cxlsx

Make sure you update the blue values to match your setup. Green values i carefully tweaked based on lots of testing and data... but if your setup is different keep that in mind

 

[TaK]

okashira-

Thank you for sharing that document. I read through it and I can see a lot of time spent. It's nice to see a community with genuinely helpful members.

That said, I think it would be very challenging to adjust this methodology to a capacitive discharge type of welder. The problem is that the total inductance makes a huge difference. In your setup it looks like you have a battery or other high current source. This means (as I'm sure you know, but for other readers...) the system inductance really only affects your current ramp rate, but not your peak current which will be limited by the loop resistance. Of course this is a very big deal if you're using short pulses that don't really get to the steady state current, but I presume you're pulsing long enough to get into the resistive limited region. The longer you dwell in this region, the less the error from the inductance estimation will matter.

In the CD welder though, the pulse is fast enough that the rise, peak, and fall of the current waveform will all be dominated by the inductance. I doubt it would ever get into a resistive limited region of operation. Thus, the estimation of inductance will make dramatic differences. It would be very challenging to estimate that inductance for some random setup. The capacitors I'm using don't even specify the ESL. One nice thing at least is you know your total system energy is fixed by the energy stored in your capacitor(s) E=0.5*C*(V^2)

So all that sucks if you're trying to plan a welder from scratch. One nice thing though is that if you've already made the welder, you can always make measurements I've done a bit of work here so if anyone wants to copy parts, they will have a better idea what's going on.

I measured the current for my welder using a 1mOhm precision shunt and 500Mhz differential probes. It's not perfect, but it's pretty good and it was quick. Here's a picture.

For scale, in case anyone tries to duplicate, that capacitor is 5.7" tall. The wiring is mostly 12AWG. The probe tips are 3" of 1/8" diameter copper.

Here's the current waveforms:

The two screenshots are the same but with different timescales. The taller of the pulses is a 50J discharge and the lower is a 40J discharge. I accidentally switched the trace colors between the two images (sorry) but the taller is 50J and lower is 40J for each.

Continuing to work backwards... It looks like my setup has about 600nH of inductance and about 7.2mOhms of resistance. These discharges were done with the probe tips shorted together, but included the resistance of the shunt.

As you can see, these values make a theoretical waveform very similar to the measured waveform:

The next challenging part, which I have not done, would be to determine the amount of energy put into the weld. That would be seriously hard since you need to estimate the resistance of the weld even as it is being melted and fusing together.

Anyway, I'm short on writing time right now. I'll update on progress soon. Also, I will make a post talking about the welder in more detail sometime.

TaK

 

[Lurkin]

Tidy, Simple, Straightforward.

Quick query - what are you using to ultimately power the caps themselves?

 

[chessir]

Can I use a 50v Nippon which I happen to have , New to capacitors , how do you charge the capacitor ? Can I use my Meanwell 24V 450watt power supply? Where can I order the probes and what is that positive cable comming off the capacitor with the stainless connector on it?

 

[okashira]

okashira-

Thank you for sharing that document. I read through it and I can see a lot of time spent. It's nice to see a community with genuinely helpful members.

That said, I think it would be very challenging to adjust this methodology to a capacitive discharge type of welder. The problem is that the total inductance makes a huge difference. In your setup it looks like you have a battery or other high current source. This means (as I'm sure you know, but for other readers...) the system inductance really only affects your current ramp rate, but not your peak current which will be limited by the loop resistance. Of course this is a very big deal if you're using short pulses that don't really get to the steady state current, but I presume you're pulsing long enough to get into the resistive limited region. The longer you dwell in this region, the less the error from the inductance estimation will matter.

In the CD welder though, the pulse is fast enough that the rise, peak, and fall of the current waveform will all be dominated by the inductance. I doubt it would ever get into a resistive limited region of operation. Thus, the estimation of inductance will make dramatic differences. It would be very challenging to estimate that inductance for some random setup. The capacitors I'm using don't even specify the ESL. One nice thing at least is you know your total system energy is fixed by the energy stored in your capacitor(s) E=0.5*C*(V^2)

So all that sucks if you're trying to plan a welder from scratch. One nice thing though is that if you've already made the welder, you can always make measurements I've done a bit of work here so if anyone wants to copy parts, they will have a better idea what's going on.

I measured the current for my welder using a 1mOhm precision shunt and 500Mhz differential probes. It's not perfect, but it's pretty good and it was quick. Here's a picture.

For scale, in case anyone tries to duplicate, that capacitor is 5.7" tall. The wiring is mostly 12AWG. The probe tips are 3" of 1/8" diameter copper.

Here's the current waveforms:

The two screenshots are the same but with different timescales. The taller of the pulses is a 50J discharge and the lower is a 40J discharge. I accidentally switched the trace colors between the two images (sorry) but the taller is 50J and lower is 40J for each.

Continuing to work backwards... It looks like my setup has about 600nH of inductance and about 7.2mOhms of resistance. These discharges were done with the probe tips shorted together, but included the resistance of the shunt.

As you can see, these values make a theoretical waveform very similar to the measured waveform:

The next challenging part, which I have not done, would be to determine the amount of energy put into the weld. That would be seriously hard since you need to estimate the resistance of the weld even as it is being melted and fusing together.

Anyway, I'm short on writing time right now. I'll update on progress soon. Also, I will make a post talking about the welder in more detail sometime.

TaK

TaK, if you look closely I am using inductance to estimate current based on pulse time. It's actually easier to estimate inductance then you think. You can assume the ESL of the caps you're using is pretty negligible, and all of the inductance is in your board and electrodes, which is pretty easy to estimate.
Of course, you can also determine inductance of your setup since you've measured your current vs time.

On measuring weld energy, just put a probe ground and tip on the very tip of each electrode and log voltage, while logging current with another probe

You measured 600nH from your setup, which looks about right for the short wires that you are using.
I calculated mine to about 2.1 microhenry, which is pretty spot on for the length of electrode's i'm using.

 

[that1guy]

Tak, nice work! And thanks for the pics and explanations.

With 40-50J welds, that back calculates to 30V. Prior to your post, I would've guessed 30V welds would result in blow outs. But looks like you're having good success.

Check out this for calculating/guesstimating inductance of parallel wires. I would zip tie those wires together to reduce inductance.
https://www.eeweb.com/toolbox/parallel-wire-inductance/

 

[TaK]

Battery update first. Answer questions second.

Update:

I attached the BMS wires to the tabs on the metal strips that parallel each bank of cells. Then I folded them over and Kapton taped them down. I taped the under the tabs first for some structure and insulation then I taped over the tabs to insulate again. The main positive and negative connections are still missing, but it's getting close to a test ride.

I have a game plan for attaching the main power wires. Hopefully I'll be able to get that done tonight.

TaK

 

[TaK]

Tidy, Simple, Straightforward.

Quick query - what are you using to ultimately power the caps themselves?

It's a bench power supply. I'll make a post shortly on the welder itself since several people have asked.

Can I use a 50v Nippon which I happen to have , New to capacitors , how do you charge the capacitor ? Can I use my Meanwell 24V 450watt power supply? Where can I order the probes and what is that positive cable comming off the capacitor with the stainless connector on it?

The voltage rating of the cap just has to be high enough for whatever voltage you're charging it to. The voltage depends on the amount of energy you want to store as well as the capacity of the caps. What is the capacitance of the cap you have?

For the power supply, you really want to use one that is happy running in a constant-current mode. I had a few thoughts on inexpensive ways to do that. I'll elaborate in the post on the welder itself.

The probes are the 3" long, 1/8" welding tips from Sunstone.
http://sunstoneengineering.com/products/accessories/electrodes/
They are not very expensive, but if I were to do it again, I would first try going to Home Depot (or similar) and getting some thicker household wiring (maybe 10 AWG), cutting it into a couple small sections, and filing it to a suitable point. Remember to make the tip sort of roundy (see the pics from my first post). The contact area determines the size and quality of your weld.

I'm not sure what positive cable you're referring to. Can you circle it in the picture and attach it?

TaK, if you look closely I am using inductance to estimate current based on pulse time. It's actually easier to estimate inductance then you think. You can assume the ESL of the caps you're using is pretty negligible, and all of the inductance is in your board and electrodes, which is pretty easy to estimate.
Of course, you can also determine inductance of your setup since you've measured your current vs time.

On measuring weld energy, just put a probe ground and tip on the very tip of each electrode and log voltage, while logging current with another probe

You measured 600nH from your setup, which looks about right for the short wires that you are using.
I calculated mine to about 2.1 microhenry, which is pretty spot on for the length of electrode's i'm using.

Haha. I might try the weld voltage measurement if I can find a sacrificial probe. I think I can hear my expensive probes shaking with fear though.

I now agree with you about being able to take a good guess at the total welder inductance. Originally I thought a lot of the contribution would be from the caps themselves, but that seems to not be the case. I took a swag at my wire loop area and the theoretical inductance for that comes to around 470nH, which is quite close to my measured 600nH.

Tak, nice work! And thanks for the pics and explanations.

With 40-50J welds, that back calculates to 30V. Prior to your post, I would've guessed 30V welds would result in blow outs. But looks like you're having good success.

Check out this for calculating/guesstimating inductance of parallel wires. I would zip tie those wires together to reduce inductance.
https://www.eeweb.com/toolbox/parallel-wire-inductance/

I'm actually using 2 of those big caps that are each 0.1F. It's hard to see that in some of the pics. So these welds in the pics are made at 20V and 22.4V. That being said, I did some tests with the one cap only at 30V and it worked pretty well also.

I was having blow outs (pops) in my early testing, but once I went to an aggressive tip reconditioning schedule (see the first post) to keep the tips free of melted nickel, I haven't had any more blow outs. The blow outs are a complicated function with many variables and voltage definitely is one of those variables. I wouldn't recommend much above 30V since it was getting touchy.

That's a great inductance calculator tool. I would definitely agree with you on keeping the wires bundled together and thus the loop small, but these wires are so short that I have a hard time maneuvering them to the locations I want. Some people tie the probe tips together but I was struggling to make good contact on both tips in that configuration.

TaK

 

[TaK]

OK. The welder.

There were some questions on the details of my welder. Here is the schematic:



You can make it really easily if you have a programmable bench power supply. If you do not, you can make it very cheaply with a charge pump doubler. I'll explain in detail later.

There is nothing unique going on here. Lots of people have done this before me and I read several build logs from Google searches. I made this design to be as simple, easy, and cheap as possible.

There are a few main parts: the main capacitor, the copper probes, the SCR, a bench power supply to charge the capacitor, and a controller. I've talked about the main capacitor in my first post in this thread and the copper probes at another point, but I'll elaborate on the other items here.

The SCR I chose is a KP100A from eBay, new, for around $11. It has a ton of current handling capability and is't stressed for this design at all.

I'm using a fancy controller for my design because I already had it, but you can use almost any micro-controller you want like and MSP430 or an Arduino. It needs 2-4 DIO and possibly RS232 or similar to communicate with the bench supply (GPIB in my case).

The most critical thing you need is a power supply to charge the capacitor. Ideally you would have one that can work in constant-current mode, but at minimum it needs to continue to work through current limit and not just shutdown. Charging these huge caps looks like pumping current into a short as far as the power supply is concerned so the power supply needs to be happy doing so.

If you have a programmable bench power supply then its very easy. If you have a manually adjustable bench power supply then you also need a relay in series to turn it on and off from the controller. If you only have a simple ATX supply or simple AC/DC brick then I have another option for you at the bottom of the post.

The reason the power supply needs to be programmable is that it's in parallel with the capacitor. The SCR will not reset until the current going through the anode and cathode has dropped below some critical threshold (check the datasheet) so if your power supply is still connected when you're making a weld, the SCR can never turn off and reset. To fix this, you need to command the power supply output off, or command the output to 0V (assuming its a normal single quadrant PS), or open a relay so the power supply becomes disconnected from the cap.

Here's the way it works:
1. power supply turns on to the set voltage (say 20V) and is limited in current (say 1A)
2. when the cap is charged, take the probes and touch them to where you want welds and make good, strong connection
3. push the foot pdeal
4. the controller commands the power supply OFF (or opens the relay)
5. the controller triggers the SCR for a short time (the SCR latches on so the trigger time can be short)
6. the weld is made
7. the controller commands the power supply to charge the cap again... repeat

OK, what do you do if you don't have a fancy bench power supply. Here is something that should work... check the bottom right corner of the schematic attached...

If you have a 12V supply and a micro-controller, it would only take a few cheap components to make a switched capacitor doubler (or tripler, depending on what voltage you're going for). The down side is that it will take a bit longer to charge the main capacitor, but if you're looking for simple this is it. There are 4 components and an input channel from the micro-processor. Just make sure you're using components sized to handle the current, depending on the size of the caps chosen and the frequency you choose to operate at, there can be some serious current here. I have not tested this, but I've used doublers like this plenty of times professionally, and I don't see any issues using it for this application.

If you want to get fancy and adjustable with this sort of circuit, you could do something like a tripler off of 12V. Then you can use an ADC (through a divider) of your micro-controller to stop switching when you've reached the target voltage. It's fairly slow to charge so it should be easy to control. The free version of SIMPLIS or LTSpice is your friend.

I've already made a lot of recommendations in this thread but here's a few:
1. try to stay below 30V
2. keep the probe tips clean
3. have a few extra cells for testing purposes
4. wear safety glasses

I took a video of the welder working, but when I try to upload it, it resets my post (this is the 3rd time I'm writing this).

In other news, I've finished the electrical construction of my battery pack so its time to use it a bit before I button it up.

TaK

 

[wannesd]

Clean job, but you might want to use isolation washers next time on your positive ends.
The main body of a cell is negative, and if one of your tabs eats through the thin plastic, instant sparks...

 

[\/ampa]

He hot glued all cells together. There will be no eats or rubbing there.
Also the cells have from factory the white plastic washers which are not as thin like the plastic sleeve.

 

[wannesd]

He hot glued all cells together. There will be no eats or rubbing there.
Also the cells have from factory the white plastic washers which are not as thin like the plastic sleeve.

Still, cheap insurance.