A software engineer's failed attempts at creating a spot welder

[mehmedbasic]

Hey everyone

I'm a software engineer by trade and have had electronic engineering as a hobby for a couple of years. Most of the stuff I've worked with has been low voltage, low amps and very digital. This project is a little outside of my comfort zone and I need a little help understanding what went wrong.

I've been attempting to build me a cheap spot welder for 18650 cells to be able to build battery packs for my power tools and ebike. I'm using a car battery as a current source, 470 CCA. It's fused with a 300A fuse. All tests were conducted welding nickel to steel plates (not batteries) wearing safety gloves and glasses in a well-ventilated workshop with a fire extinguisher at hand. The timing is controlled by an Arduino and the timing control has been verified using an oscilloscope.

Attempt 1)
Initially I tried controlling with a starter relay from a car driven by an ULN2003. This worked pretty nice for short pulses but at ~25ms or more the current overwhelmed the relay and it jammed in the open state resulting in a runaway weld. The couple of long pulses in the second-range resulted in very melted nickel and funnily enough not a broken fuse. The relay was not reliable after this.

My hypothesis is that it either sparked internally or the magnetic field generated by the couple of 100 amps held the relay contacts in the open state. Feel free to correct me if I'm wrong.

Attempt 2)
Next attempt was using 10x FQP30N06L N-channel MOSFETs  (from AliExpress)  in parallel mounted to a not-completely-flat thin sheet of copper-plated aluminum using steel nuts. The electrode was connected to the the aforementioned aluminum plate.

The gates were all wired to each other and pulled down with 10k.
The gates were driven directly by the Arduino at 5v with a 680 ohm resistor in series.

A photo of the not-so-glorious attempt at handling 300A is attached.

The first test was with 10ms pulse and it was a MOSFET massacre. All of the FETs are dead and only 1 of them didn't break. Three of them literally blew their tops off. The project ended with me ordering a DIY kit from https://malectrics.eu, but I would like to understand what went wrong.

So far I've concluded/learned a couple of things during my research:

1) Every MOSFET needs their own resistor to prevent ringing between them
2) 5V logic pin does not have enough kapow to drive 10 gates, use a gate driver
3) The transistors in question may have been on the weak side rated at only 32A sustained
4) Mounting pressure is critical and any imperfections will lead to current being drained there
5) Both circuits need flyback diodes, especially the relay one
6) There's a risk that cheap chinese components are fake and may or may not conform to the datasheet

So my questions to you are:
1) Would this have worked with more beefy FETs, more consistent mounts and a proper gate driver?
2) Why does one need a gate driver when Vgs is well within the range of the MCU logic level?


Thanks in advance

[BrokenYugo]

The gates have capacitance, this is inherent to the design of a MOSFET. The MOSFET is turned fully on and off by fully charging and discharging this capacitor. More in parallel means more capacitance and a slower on/off time at a given drive current. As you've learned mosfets don't much care to flow a lot of current while halfway on, hence the using a gate driver to snap it to the on state (and back off again) as fast as possible.

[RichC]

30A per mosfet with them only rated for 32A is pretty marginal. I'd want a good deal more headroom on the specs I think.

A slight inconsistency and you're over rating for one of the fets, once one blows the rest will follow as they're then all over spec.

[Benta]

Erm... how do you expect the 18650 cells to look after this kind of weld technique? You might as well use an acetylene torch.
Cell spot welding is done with ultrasound to have a very localized heat spot that won't damage the cell interior.

[james_s]

Ultrasonic? The commercially available battery welders I've seen have all been capacitive discharge. There is also the K-weld which is an open source spot welder more similar to the concept the OP here tried, my friend has one and reports it works well.

I use a capacitive discharge welder I built about 10 years ago. It consists of a bank of around 1F worth of 22,000uF 25V capacitors in parallel charged by a very simple LM317 based power supply and the largest stud mount SCR I could find on ebay at the time to switch the current. I've welded a few hundred battery packs with it without any real issues. I think the commercial units typically use a higher voltage with correspondingly less capacitance but I used capacitors I was able to get a good deal on.

You place the electrodes about 5mm apart, press down firmly and then step on the foot switch, if all goes well it isn't very exciting, just a little thunk sound and you end up with two tiny weld spots.

[pqass]

...
Attempt 2)
Next attempt was using 10x FQP30N06L N-channel MOSFETs  (from AliExpress)  in parallel mounted to a not-completely-flat thin sheet of copper-plated aluminum using steel nuts. The electrode was connected to the the aforementioned aluminum plate.

The gates were all wired to each other and pulled down with 10k.
The gates were driven directly by the Arduino at 5v with a 680 ohm resistor in series.

A photo of the not-so-glorious attempt at handling 300A is attached.

The first test was with 10ms pulse and it was a MOSFET massacre. All of the FETs are dead and only 1 of them didn't break. Three of them literally blew their tops off. The project ended with me ordering a DIY kit from https://malectrics.eu, but I would like to understand what went wrong.

So far I've concluded/learned a couple of things during my research:

1) Every MOSFET needs their own resistor to prevent ringing between them
2) 5V logic pin does not have enough kapow to drive 10 gates, use a gate driver
3) The transistors in question may have been on the weak side rated at only 32A sustained
4) Mounting pressure is critical and any imperfections will lead to current being drained there
5) Both circuits need flyback diodes, especially the relay one
6) There's a risk that cheap chinese components are fake and may or may not conform to the datasheet

So my questions to you are:
1) Would this have worked with more beefy FETs, more consistent mounts and a proper gate driver?
2) Why does one need a gate driver when Vgs is well within the range of the MCU logic level?


Thanks in advance

A single 680R gate resistor driving 10 MOSFET gates to only 5V will be slow to turn on; each popping like Chinese firecrackers!
See (with playback set to 2x):
Your MOSFETs Vgs limit is +/-20V.  You should use a gate driver (chip or half H-bridge) to switch between at least 10V and 0V. 

BTW, Googling "battery spot welder diy" yields many implementatons usings MOTs and starter solenoids (vs relays).

Oh, I see you took the "Measure with a micrometer, mark with chalk, cut with a chainsaw." mantra to heart.   

[james_s]

Oh yeah, 5V is way too low, they even sell special "logic level" mosfets designed to work with 5V but they are typically not super high power types. For a typical power mosfet you want to drive the gate with 15V and use a driver capable of several amps. For an array of mosfets like that you'll want either multiple drivers or a very beefy one. Don't forget to have ample decoupling capacitors placed as close to the gate drivers as possible and be mindful of construction techniques, keep all wiring as short as possible. Use a separate resistor on each gate to prevent ringing, 4.7R is a reasonable starting point, then you'll want to look at the gate drive on a scope.

[jpanhalt]

This works great for me (with minor modifications) : https://www.philpem.me.uk/elec/welder

Or try the KWeld approach: https://www.eevblog.com/forum/projects/guesses-on-what-i-am-attempting-here/msg1232857/#msg1232857

[james_s]

That's very similar to the design I came up with, except he used 5 huge capacitors and I used about 80 smaller ones.

Regarding his comment on the 1F car audio capacitor, I tried that initially and it worked, for about 10 welds then the capacitor failed.

[jpanhalt]

My design uses a pack of 12 parallel "computer"capacitors.  Works well at about 17V to 18V for everything i have wanted to weld.

[amyk]

Spot welders are usually either capacitive discharge, or transformer based. They both have in common the fact that you don't need to worry about breaking the welding current; capacitors will discharge to 0, while a lower-current high-voltage switch on the transformer primary suffices.

[jmelson]

Get a 100 A or bigger SCR, 60 V PIV is fine.  Use a decent size, maybe 50,000 uF 50 - 75 V electrolytic capacitor, charged through a wirewound resistor from a variable DC power supply.  Provide a little R-C circuit to deliver a pulse of a few hundred mA to the gate of the SCR to fire it.

I built one of these for a "flash volatilzer" in an analytical lab about 40 years ago, it worked great.

Jon

[mehmedbasic]

I looked into using a MOT but it was out of my comfort zone and thought that a 200A rated relay would be able to handle it.

I've watched the linked video from bristolwatch.com and it accurately describes my issues.
One of them probably turned on faster and gave up and the rest followed.
So I made an effective and easy to assemble firecracker simulator

I've ordered a bunch of appropriately rated parts along with a pre-made PCB all of which should  be arriving next week.

Again thank for all the help and stay tuned for even more melted copper and/or electronics.

[jpanhalt]

I second rejection of the MOT for a battery tab welder.  I used a MOT powered by a Variac and solid-state relay in my first attempt.  It was too difficult to control heat.  That is, rather then a nice dimple from a CD welder, the nickel strip became overheated.

[ledtester]

Look on youtube for reviews of the "mini spot welder" devices that have been available on aliexpress recently.

They are all very simple and this video goes over their schematics, problems and fixes:

All "Mini Spot Welder" versions: circuit Analysis, differences and faults
https://youtu.be/RSrlXqFxhp8

[james_s]

I looked into using a MOT but it was out of my comfort zone and thought that a 200A rated relay would be able to handle it.

I've watched the linked video from bristolwatch.com and it accurately describes my issues.
One of them probably turned on faster and gave up and the rest followed.
So I made an effective and easy to assemble firecracker simulator

I've ordered a bunch of appropriately rated parts along with a pre-made PCB all of which should  be arriving next week.

Again thank for all the help and stay tuned for even more melted copper and/or electronics.

It's not just that one turned on before another, the problem is that they were all turning on FAR too slowly and probably not fully. Look at the Vgs value on the datasheet of the mosfets you used, it's probably greater than 5V, meaning that driving from 5V is never going to turn them fully on, they will always have higher than the specified on resistance. A microcontroller IO pin is also orders of magnitude too weak in terms of current to properly drive a bank of mosfets. For a high power and high speed application like this you should aim for at least an amp of gate drive capability per mosfet. It's a very short pulse so you don't need a huge power supply to power your gate driver(s) but you need enough decoupling to supply a large pulse.

[Doctorandus_P]

It's all explained quite well (but a bit long in the tooth) by Youtube user "Luca", linked to from the posts above.

Long story short:
1. Buy a EUR 20 spot welder. It's cheaper as the parts. (and your time)
2. Old versions had a good MOSfet driver (5A) but it's "optimised away" in later versions.
3. The discrete fet driver replacement is "adequate" for the few pulses a spotwelder makes.
3a. Some explanation of scope images and the "miller" effect, and why it destroys your MOSfets with a bad driver.
4. Add a resistor to limit inrush current through the input diode.
5. Battery pack output voltage can sag to below 4V during a 300A spotweld surge.
6. Input diode and buffer cap is mandatory to prevent the gate voltage from sagging during a spotweld.
7. Such current spikes are probably not good for your battery packs.

[james_s]

One of those car audio capacitors might actually work in that application to supplement the battery. 300A pulses I'm sure they could handle, the thousands of amps resulting from a capacitive discharge welder kills them, at least it killed the one I tried.

[bob91343]

This is overkill.  I have two methods to accomplish this.

1. I solder to the coin cell.  Yes, it's dangerous but I use a hot iron, flux, and get off the cell in less than one second.  Never had a problem.  It helps to scratch the cell surface to make sure the metal is clean.

2. I use a cheap cell holder.  No welding required.

[Alex Eisenhut]

https://www.aliexpress.com/item/1005001799079203.html?spm=a2g0o.productlist.0.0.461642e4zeTL06&algo_pvid=null&algo_expid=null&btsid=0b0a555816193087051358920e7b99&ws_ab_test=searchweb0_0,searchweb201602_,searchweb201603_

[james_s]

This is overkill.  I have two methods to accomplish this.

1. I solder to the coin cell.  Yes, it's dangerous but I use a hot iron, flux, and get off the cell in less than one second.  Never had a problem.  It helps to scratch the cell surface to make sure the metal is clean.

2. I use a cheap cell holder.  No welding required.

The first method is an incredibly sloppy hack, it is not only dangerous but will damage many types of cells and it looks terrible, cells should *never* be soldered directly, the fact that you haven't had problems (yet) doesn't make it a good idea. The second method is utterly inadequate for any situation where high current is required and anything that needs to fit inside an existing battery pack housing. Did you miss the fact that he's interested in rebuilding battery packs for power tools and e-bikes? This means high capacity, high drain lithium ion cells that pose a serious fire hazard if they are damaged. There's a reason all commercially made battery packs are spot welded, it's the correct way to do it. There is nothing overkill about doing something properly. It's not like a spot welder is especially difficult to build or acquire.

[bob91343]

That is true.  My comments apply to the tiny button cells that are so common.

[Gyro]

I second rejection of the MOT for a battery tab welder.  I used a MOT powered by a Variac and solid-state relay in my first attempt.  It was too difficult to control heat.  That is, rather then a nice dimple from a CD welder, the nickel strip became overheated.

It can work well for bigger stuff though -  DiodeGoneWild did a video of a very basic one recently. No switching and he kept the magnetic shunts in. He even did some sums! 

[bob91343]

That guy's speech is impossible to comprehend.

[SeanB]

That guy's speech is impossible to comprehend.

I have no problem with it. US speech, on the gripping hand, is often unintelligible, and you need to have a translation of it, despite it being nominally English speech. Let us hear you speak Czeck, or even Hungarian, or Polish, where you would think it was written on a keyboard missing all the vowel keys.