kWeld - "Next level" DIY battery spot welder

[tatus1969]

Wired up 5 x BCAP0310 in series now, and the server PSU also had its first task today.

I didn't mod the PSU's output voltage yet, by default it has 12.5V. Even though this is not enough yet for the capacitors (5x2.7=13.5), this setup can do 110J weld energy into 1mOhm.



I don't have balancers on the capacitors yet, but they already show that they want that.

Next will be stress testing them, in order to see if they can handle this current without damage or overheating. But I think that they will need some forced air cooling.

The PSU is doing very well, I'm impressed. With my setup below, it trickle charges the capacitors because it goes into overcurrent shutdown at ~100A every 25ms. At ~9V, it succeeds to start up, but that is partly because of that high impedance lab wiring. I definitely don't recommend doing this, because I fear that the PSU may die on that the sooner or later.

But the good news is that the PSU has a remote sense input that supposedly allows voltage control from 10.5V to 13.5V, which is spot on for the capacitor bank. The idea is: resistor limited pre-charge up to 10.5V (adjustable). Then using the remote sense input (plus an opamp plus a current shunt resistor) to charge with 70A (adjustable) current. Then keep the capacitor topped at 13.5V (adjustable).

At 70A, the PSU is theoretically able to recharge from a 50J / 1mOhm weld pulse in 0.7 seconds. That should be more than fast enough to be usable.

I'll breadboard this and make capacitor torture testing with that setup...

[tatus1969]

If it is not too rude a question, why are people so keen to use the supercapacitors instead of battery power sources ? Lower ESR ? Higher current tolerance ? , it certainly isn't low cost

Certainly not a rude question :-) My objective is to provide a stationery alternative to the portable Lipo supply. It should be cost effective, and I think I can even beat the price of the suggested Lipo plus the neccesary charger (at least 100€). The five Maxwell capacitors are 7.74€ per 100 each, the control board may be around 40€. I want to design it specifically for the HP DPS-800GB server PSU, there are lots of used at ebay for around 20€.

Best Wishes to all my fellow kWeld owners and especially to Frank, the father of the kWeld 

Thanks, I go a long with that wish!

[SirJMD]

I've started testing the kWeld - with sparks going everywhere.

At 50J I burned straight through the nickel strips. At 14.6J I get strong welds but still sparks going everywhere, even though I press down on the electrodes rather hard. Peak current was just over 1400A.

14.6J
6,74ms
1.03 mOhm
1428A

With the calibration it set an offset of 66 and a resistance of 1.95 mOhm.

What kind of nickel strip do you have, is it 0.1mm? And is it maybe nickel plated steel? (You can test it with a Dremel with cutting disc, if it is pure nickel then you won't see any sparks.)

The numbers are consistent though, as 1428^2 * 1.03e-3 * 6.74e-3 = 14.15J. Current and weld spot resistance are also typical, and if you can verify the pulse time for correctness then I think that the system works as it should.

It should be pure nickel, but probably isn't. I bought it long ago on eBay, so it's probably plated steel. Sadly I don't have a Dremel or an angle grinder at hand, to test it with.


What kind of pulse is the kWeld making? Single large pulse?
I know that commercial spot welders utilizes a dual pulse. Would that be possible with the kWeld?

[tatus1969]

What kind of pulse is the kWeld making? Single large pulse?
I know that commercial spot welders utilizes a dual pulse. Would that be possible with the kWeld?

kWeld is single pulse - I currently keep my position that it doesn't need double pulse technique, because the implemented energy metering is superior. Of course it is very simple to change the behaviour in the firmware... I'm happily up to the discussion.

EDIT: I think that double pulse is necessary in conventional welders because when pulse time is kept constant, then delivered weld spot energy depends on the weld spot resistance. That can vary when there is bad contact due to corrosion or weak contact force. I made the constant energy approach to get rid of these factors.

[tatus1969]

i have been trying to learn some of the background "science" of this nice project by tatus1969 from googling
and here is also a version of LTspice simulation, i added a C2 n some diodes to deal with the kick back which i assume is what was mentioned in 1 of the posts as being a danger? it appears due to the very high current, the simulation seem to say the kick back could go over 1kV? not sure about this but maybe someone else can comment on this.

the cable specs are "stretched" to simulate what if lengths.
the simulation is somewhat fun, its a somewhat "non-destructive" way to find cable parameters that fit a certain current/weld output profile
(the txt file needs to be renamed to *.asc for LTspice)

Inductive kickback has no theoretical limit. If you manage to instantaneously interrupt the current in an inductor, then it produces -infinity volts.

The switch and diode are not wired in the same way as in kWeld (not really essential, but it may help in our further discussion if you change this): the power transistors are in the ground return path. You need to add a big 50V zener diode across them, because that's their behavior when given too much voltage (MOSFET avalanche). The TVS diode that is in the system as well goes from the + output lead (=cathode) to the - output lead =(anode). The voltage across the swtich at turn off should look like in the picture (standard 1.0m system lead length + 0.2m on the battery, 1400A current).

EDIT: one more comment: the capacitor across the switch is not a good idea, as it stores charge that is dumped into the switch at turn on. The charge at 10V would be 10^2 * 47m / 2 = 4.7J. This would destroy the power switch.

[SirJMD]

What kind of pulse is the kWeld making? Single large pulse?
I know that commercial spot welders utilizes a dual pulse. Would that be possible with the kWeld?

kWeld is single pulse - I currently keep my position that it doesn't need double pulse technique, because the implemented energy metering is superior. Of course it is very simple to change the behaviour in the firmware... I'm happily up to the discussion.

EDIT: I think that double pulse is necessary in conventional welders because when pulse time is kept constant, then delivered weld spot energy depends on the weld spot resistance. That can vary when there is bad contact due to corrosion or weak contact force. I made the constant energy approach to get rid of these factors.

I've heard that the first pulse's purpose is to remove oils and other impurities, so the second pulse can perform the actual weld.

I'm by no means an expert in that area, but it sounds plausible. I'm not sure how to test it in practice.

[SvanGool]

... It trickle charges the capacitors because it goes into overcurrent shutdown at ~100A every 25ms ...

I got halogen 12 V/100 W projection lamps in: they measured 100 mOhm when cold and will be about 1.5 Ohm when normally on.
I assume that if you would put such a lamp in between one supply power lead of your power supply and the capacitors, the power supply would operate normally: at the start the lamp would be turned on at 12 V (caps at 0 Volt) and current at app.8 A, while charging the voltage over the lamp will decrease, the resistance of the lamp will decrease and the current will increase after that. Can't test it myself yet 

[joddi]

I'm curious if I can use this LiTiO (6S):
https://www.ev-power.eu/LTO-technology/Lithium-Titanate-Oxid-Battery-Cell-LTO-2-4V-30AH.html?cur=1

Price may not be competitive but my idea is:
- use 6S LTO battery pack in my car as a secondary battery (should have good cold temperature performance for cold starts)
- make this battery removable from car and use it as a occassional portable powersource (for kWeld...)
- charging would be done in car from alternator

[Kjelt]

Removed my post since it was my own stupid mistake. The taps were placed back in different order so I found just out while putting everything back that it was an M6 tap instead of an M5 tap I used. 
Since now one piece of the electrode brass pipe is damaged I have to find another way to construct the electrodes.

[Fraser]

We all make mistakes ! At least you discovered why you were having a problem 

I was considering other probe designs myself and one that I was going to investigate was using a pair of the XT-150 (or similar bullet connectors) at the probe tip. The female part would be soldered to the brass tube that is filled with a 4mm copper rod. The welding tip would be made from shaped copper rod that is soldered into the male XT-150 connector. The XT-150 plastic casing would not be used. This would make it easy to change between different welding tip shapes and profiles. For example, straight and 30 degree angled tips, or offset/stepped diameter tips for confined space working. I have lathes so can easily experiment with tip shapes.

The drawback of using the XT-150 connectors for the probe tip mounting may be the connection resistance that it may introduce. They are gold plated and oxidisation resistant though.

Just a thought at the moment as I have no time to play for a while.

Fraser

[SeanB]

Simple fix there Kjelt is to retap with a recoil tap and put in a M5 insert, which uses almost the same diameter and which will recover most of the diameter and most of the strength. Not as good conductivity wise as it is stainless steel as opposed to brass, but so long as tyhe ends which are good conductors are in good electric contact it will work well enough just with slightly higher resistance.

Been there done the same, and did a similar fix.

[Fraser]

Something else to consider. The brass tube does not need to be very long as it is just a coupler. You need just enough length for the crimped section and a short threaded section for the tip mounting. You can insert the 'coupler' into a nice handpiece made from an insulating material, such as a pens shell. You could also solder a M5 nut to the end of the damaged brass tube and use that as the threaded tip retainer. A brass nut would be best for this.

The brass hollow rod is pretty commonly available on eBay. It might be better to just buy a length and start again ?

Fraser

[Kjelt]

Simple fix there Kjelt is to retap with a recoil tap and put in a M5 insert, which uses almost the same diameter and which will recover most of the diameter and most of the strength. Not as good conductivity wise as it is stainless steel as opposed to brass

Thanks but the wall of the tube is 0,2mm left so it is to fragile to put anything in, I will visit the hardware junkyard or hardware store somewhere this week to find some alternative. I am not in a hurry since my main powersource is also still under development.

The brass hollow rod is pretty commonly available on eBay. It might be better to just buy a length and start again ?

If the store does not have or I can not find any alternative I will take a look. The problem with Ebay is that it takes sometimes 4 weeks before things show up 
I do like your quick replacement idea though with the connectors, I do not know if I will need that since I won't be doing production type of welding anyway but good thought. I have no idea what the quality of those 150 connectors are, as long as it is cupper under the gold you won't have any issues.

[tatus1969]

I've heard that the first pulse's purpose is to remove oils and other impurities, so the second pulse can perform the actual weld.

I'm by no means an expert in that area, but it sounds plausible. I'm not sure how to test it in practice.

Yes, and another rationale is to soften the nickel strip and let it sink to the battery tab through the tip pushing it down.

My experience is that I get consistent welding results from kWeld, no matter how hard I push down the electrodes. I also made welding tests on batteries from I which I had dremele'd away existing strips, which is dramatically changing their surface structure, and I still get the same results. I made some tests with solder flux in the path, and I already though about artificially aging/oxidizing some material. I'm most curious of the feedback on this topic from other kWeld builders.

Removed my post since it was my own stupid mistake. The taps were placed back in different order so I found just out while putting everything back that it was an M6 tap instead of an M5 tap I used. 
Since now one piece of the electrode brass pipe is damaged I have to find another way to construct the electrodes.

That was my first question because I was surprised that you said that there would only have been less than 1 mm remaining wall thickness. An M5 tap drill could not have caused that, because the tubes are 6mm outer diameter. Maybe you can source/make that brass tube yourself? Another builder has taken M8 brass threaded rod from his local hardware store. Anyway, you don't want to know all the mistakes (that are not on video) that I went through when starting this

using a pair of the XT-150 (or similar bullet connectors) at the probe tip

Interesting idea! When considering alternative probe systems, there is one unresolved thing that I have on my todo list. Even when all contact resistances and cross sections are perfect, the probes still heat up too quickly when doing higher energy welds like 100J. This is because copper is a very good thermal conductor and it sucks a lot of heat from the glowing weld spot itself. Thermal mass (larger structures like my initial probe design) only helps to delay this problem. Better insulation would help, for example a surrounding teflon tube, or better heat dissipation capability - a heat sink? Fan cooling?...

The drawback of using the XT-150 connectors for the probe tip mounting may be the connection resistance that it may introduce. They are gold plated and oxidisation resistant though.

During my torture tests, I continuously checked several temperatures. The XT150 did not get any hotter than the AWG8 wire.

[tatus1969]

for anyone trying to buy the Turnigy nanotech battery. Hobbyking doesn't seem to be well organized. When searching globally, they say that it is out of stock. But the EU warehouse does have them: https://hobbyking.com/en_us/turnigy-battery-nano-tech-5000mah-3s-65-130c-lipo-pack-xt-90.html

[Fraser]

With regard to the thermal issue with prolonged use of the copper probes, I can see why this can become a proble.

I used to use Portisol Gas soldering irons and these utilised a clever axial heatsink at the base of the heated tip. This significantly reduced the heat to which the handle assembly was exposed. Pictures attached.

Sadly such a radial heatsink can make the probes bulky and harder to use.

Fraser

[Fraser]

There are many reansistor heatsink that may be pressed into service as a probe shaft heat disipator.

TO5 (8mm) is the most common fitting but smaller diameters are available.

Fraser

[Fraser]

Heatsink with 5mm hole down the centre. The outer diameter is a little too large though. You get the idea though 

[tautech]

A very useful (to me) 8mm brass rod resource is broken ballcock arms. Farmers with stock water drinking troughs should have several in their scrap metal pile. Sometimes used in toilet cisterns too (if not plastic) and they usually have a 5/16" UNC thread on one end and are forged into a hockey stick shape on the other to act as a fulcrum on the plunger to shut off the water. ~10' (250mm) of 8mm brass rod in each piece.
All plumber and rural suppliers should have new shiny ones in stock that can be cut into the lengths you need.

Thinking outside the square can be useful. 

[tatus1969]

Heatsink with 5mm hole down the centre. The outer diameter is a little too large though. You get the idea though 

Thanks! I have these in mind as well, but I'll have to see how that looks in reality and how usable that arrangement will be.

p.s. that was my first ever soldering iron ages ago:

[tatus1969]

I'm curious if I can use this LiTiO (6S):
https://www.ev-power.eu/LTO-technology/Lithium-Titanate-Oxid-Battery-Cell-LTO-2-4V-30AH.html?cur=1

Price may not be competitive but my idea is:
- use 6S LTO battery pack in my car as a secondary battery (should have good cold temperature performance for cold starts)
- make this battery removable from car and use it as a occassional portable powersource (for kWeld...)
- charging would be done in car from alternator

I checked the specs, they say <1mOhm per cell. Whatever that means. If the reality is close to 1mOhm, they would be fine. You would probably charge them to <13V. Current would then be 13/(6+3.3) = 1400A.

[SirJMD]

I've heard that the first pulse's purpose is to remove oils and other impurities, so the second pulse can perform the actual weld.

I'm by no means an expert in that area, but it sounds plausible. I'm not sure how to test it in practice.

Yes, and another rationale is to soften the nickel strip and let it sink to the battery tab through the tip pushing it down.

My experience is that I get consistent welding results from kWeld, no matter how hard I push down the electrodes. I also made welding tests on batteries from I which I had dremele'd away existing strips, which is dramatically changing their surface structure, and I still get the same results. I made some tests with solder flux in the path, and I already though about artificially aging/oxidizing some material. I'm most curious of the feedback on this topic from other kWeld builders.

The strips and cells I've tested with are 3+ years, so there you have your age test
I've ordered some new strips that I hope are pure nickel. Where did you get yours? I like the slot and dimples.

One limitation I've noticed is that the high current makes it difficult, if not impossible, to weld thin wires. They.. vaporize. I'd really like to be able to weld fuse wire, like Tesla uses.

[tatus1969]

One limitation I've noticed is that the high current makes it difficult, if not impossible, to weld thin wires. They.. vaporize. I'd really like to be able to weld fuse wire, like Tesla uses.

The minimum pulse width is 50us, which would result in ~ 0.1 joulses (into 1mOhm). I'll make some tests, but I have the hope that it is just an issue with the minimum adjustable energy (2J). I already had received the suggestion to make the dial knob logarithmic, which would easily enable lowering that limit.

[tatus1969]

ops, this is now re-arranged.
thanks for the correction
now even the short battery lead is making quite a ton of voltage spikes

To me this clearly shows the limitations of SPICE when using components this far beyond their ratings. Both voltage peak level and shape aren't even close to reality as shown in the scope screenshot. Thanks for the simulation, and let's continue this way, but at the moment I see my doubts confirmed.

[SirJMD]

One limitation I've noticed is that the high current makes it difficult, if not impossible, to weld thin wires. They.. vaporize. I'd really like to be able to weld fuse wire, like Tesla uses.

The minimum pulse width is 50us, which would result in ~ 0.1 joulses (into 1mOhm). I'll make some tests, but I have the hope that it is just an issue with the minimum adjustable energy (2J). I already had received the suggestion to make the dial knob logarithmic, which would easily enable lowering that limit.

Sounds great.

I was actually thinking that a rotary encoder, or simply up/down buttons, would work better than the current dial.