Capacitive-discharge spot welder

Lately, I’ve been seeing quite a few reviews where people either build spot welders or use them in their projects to assemble battery packs. I’ve got one of these devices myself now, though it’s quite different from the ones I’ve seen before.


First, I’d like to thank my regular reader, who sponsored this review by ordering everything I needed to build a small spot welder. And, as usual, I’m here with the review.
I’ll tell the story in chronological order, since things turned out to be more complicated than I’d initially expected.

Most spot welders are built around microwave oven transformers. It’s cheap and relatively simple—and with a controller by Yurok, relatively simple too. I’ve helped a friend build a couple of these welders myself.
But building one of these welders comes with a couple of challenges: the transformer needs to be rewound, and the finished unit is heavy. That’s why I wasn’t in a hurry to build one for myself. Besides, I wouldn’t need to use it very often, so I’d have to find somewhere to store it. There are also options based on a powerful inverter—that is, a power supply with low voltage and high current—but they’re difficult to build.

Then, for a while, I started seeing compact welders based on high-current batteries. A little later, powerful supercapacitors began replacing the batteries, and that really piqued my interest.

So I ordered a relatively inexpensive, simple kit for building one of these devices. I bought it on Taobao for about $20 ($11 for the board and $4.50 each for the two supercapacitors), not including shipping.


Of course, this isn’t actually a welder, just the components needed to build one: a board, nuts, screws, wires, a buzzer, welding electrodes, a couple of heat-shrink tubes for them, and a small piece of nickel strip.


The kit includes four 27 cm wires. One pair is already terminated at both ends, while you’re expected to terminate the other pair yourself; terminals are included for that.
The wires are 10 AWG (5.26 mm²), with silicone insulation and stranded conductors, so they’re very flexible. I won’t count the strands, but there are a lot of them.
I wasn’t impressed with the welding electrodes, though. They’re basically just copper tubes, with a wire crimped into one end and a contact at the other. One of the contacts had gone all the way inside the tube, making it impossible to use as supplied.


And, of course, there’s the controller: a small board with a handful of components.


You can see the controller chip, voltage regulator, mode-selection button, and, of course, the power transistors—there are four of them.

A single button selects the mode, while an LED and a buzzer provide feedback. Each press cycles through four modes: 1, 2, or 3 pulses, or off.
There’s no charger or balancer on the board. The product page says that’s the user’s problem.


The transistors on the board are quite interesting. According to the datasheet, their on-resistance is 0.77 mΩ (0.53 mΩ typical at 100 A), and their maximum continuous current is 300 A. So the four transistors together should have a resistance of about 0.2 mΩ and handle up to 1200 A.


The supercapacitors are no less interesting. This welder needs at least two of them, or another even number.


They’re impressively large: each measures 155 × 55 × 55 mm, excluding the terminals, with a capacitance of 3000 farads and a voltage rating of 2.8 V. The pair weighs 1.36 kg with the busbar, so each one weighs about 650 g.
Judging by the description and appearance, these are used supercapacitors removed from some kind of electric vehicle, but they look to be in excellent condition.


The busbar is also substantial. It’s bolted on and sealed with silicone in places.


1, 2. There was a slight voltage imbalance when they arrived. By the way, I strongly recommend using balancer boards, because even if you charge them to the same voltage, they’ll drift apart over time.
3, 4. The ESR at 1 kHz is 0.22 mΩ for one, or about 0.47 mΩ for a pair.


Naturally, I couldn't resist testing them—or, more precisely, trying to measure their equivalent capacity. For that, I connected them to an EBC-A40L battery tester, which can charge and discharge at up to 40 A.
Connecting them with alligator clips is very awkward, so I had to attach one of the measurement leads directly to a power lead.


When discharged from 2.7 V to 0.8 V (the load's cutoff), the supercapacitors delivered roughly the same capacity: 1355 and 1365 mAh, or 2294 and 2304 mWh.


The capacity was slightly higher when discharging from 2.8 V. You can see that the load only maintains 40 A down to 1.4 V; below that, the current drops.


The charging curve is also linear, with a sharp drop at the end of charging. Here's the graph for 2.7 V.


But if you charge them to 2.8 V, things look quite different: the current at the end of charging didn't always fall below the set 100 mA, so charging never stopped. That's why I recommend limiting the charging voltage to 2.7 V. You won't gain much capacity, but the supercapacitors will last longer and be more reliable.


Before I could do anything else, I first had to repair the contact assembly. I took a pair of fine tweezers and side cutters and carefully pulled out the pushed-in contact. It was very tight at first, but then came out easily; the second contact came out even more easily.
At this point I realized it would be difficult to secure it firmly back in place. Besides, holders like these are very inconvenient to use: you need both hands, and somehow you also have to hold the battery itself.


After talking it over with the person who sent me the set, we decided to order a proper holder—spring-loaded, sturdier, and more convenient.
There were plenty to choose from, but we narrowed it down to three options:
The first, for $12.5 — link
The second, for $7.6 without the case or wires — link
The third, for $12.76 — link



In the end, we went with the third option because it was more convenient. We ordered the basic set; there's also a version with wires for $18.15, but I already had wires, albeit thinner ones.


The package included a spare switch, a connector socket, and a coil of nickel strip.


The handle is noticeably heavy at 272 grams. I weighed the coil of strip just in case; it weighs 20.6 grams.


The handle turned out to be genuinely convenient, and it comes in several colors. It's hard to see in the photo, but the plastic part is 3D-printed. In my opinion, though, the contact assembly is what really matters.


There are two M8 bolts for connecting the wires, with the welding contacts on the other side.
Inside are solid copper busbars, 15x3 mm. A wire with a connector also comes out of the handle; when the contacts are pressed together, a limit switch is activated. The connector then plugs into the spot welder to activate it when pressed. In this case, the switch can be removed because the board doesn't support this feature—which is a shame.


1, 2. Two springs and a limit switch are visible near the welding contacts. The switch activates after 3 mm of travel; total travel is about 5 mm.
3. The contact holder is more sophisticated than it looks: it has a slot with a clamping screw, so the contacts can be replaced as they wear out.
4. The contacts have been ground down, but the ends are rather roughly finished. I eventually ground them down a little more, shaping them into trapezoids. There’s no need to grind them down much or try to make them needle-sharp; just make them a bit more even.


The product page offers an option that includes wires, but it’s noticeably more expensive and heavier, so shipping will cost more too.


While I was at it, I ordered some terminals from another store, since I wouldn’t have had enough of the original ones after replacing the handle. They cost 3 cents apiece.


The screw hole is 6.5 mm in diameter, and the wire hole is 3 mm. I got this wrong: it turned out that the hole was too small for my wires. Still, I liked the terminals; they solder very well using just the flux in the solder.


I don’t have anything to crimp wires like these with, so I used a powerful soldering iron to solder them on, then crimped the terminals with pliers. I also found that the wire insulation is good quality: it didn’t just resist melting from the soldering iron, it didn’t even try to come off. All in all, the wires get 5 out of 5.
Next, connect the wires to the board first, then to the handle, as shown in the photo. The input wires are red (positive) and black (negative).
It turned out that the screws for attaching the wires to the supercapacitor terminals were very long. I couldn’t be bothered to look for shorter ones, but I had some suitable nuts at hand, so I screwed them on to shorten the threaded section.


There are a few rules to follow when connecting the wires.
1. Avoid routing a wire over the controller components. Electromagnetic interference at these currents is significant and can disrupt the controller’s operation.
2. Also, don’t use terminals directly opposite each other. With this kind of connection, most of the load will fall on the transistor closest to the connected terminals, and it may fail.
3, 4. To distribute the current evenly between the transistors, either bring the wires together in pairs, or connect them diagonally—for example, to the top-left and bottom-right terminals, or vice versa.

In short, the top shows what not to do, and the bottom shows the right way.


Before connecting everything, I soldered on the buzzer. It seemed like the manufacturer had got the spacing between the contacts slightly wrong, and the buzzer didn’t sit all the way down.
Connect it with all due precautions—or better yet, do this with the supercapacitors discharged. A short circuit in this setup will simply vaporize the traces on the board, never mind what could happen if you short the supercapacitors themselves.

Once power is applied, the LED should light up. Press the button to select 1 to 3 pulses; the fourth press puts the board into sleep mode. The board also goes to sleep on its own after a few minutes, whether or not you’re using the welder.


First I decided to practice “on cats”—or, more accurately, on 0.22 mm² wire—and found that the board can detect when the contacts are shorted. In other words, the pulses aren’t delivered at random; they’re applied about 0.7 seconds after the contacts are shorted, which turned out to be quite convenient. Here’s how it works: press the terminals against the strip, and current flows after a short delay. To get another pulse, separate the contacts and press them together again.

Well, the welder burns through 0.22 mm² wire without even breaking a sweat.


I think we’ve got the machine figured out, so we can move on to some practical experiments. But first, I think it’s worth taking a closer look at the strip they sent me.
I received four rolls of strip. The small one on the left came with the first kit, the one on the right was included with the heavy-duty handle, and the two in the middle are test samples I bought from another store.


1. The one included with the first kit is 6 mm wide and 0.1 mm thick.
2. This one came as an extra with the handle; it’s 7 mm wide and 0.12 mm thick.


These were bought here. The link offers lots of options, from 6 to 50 mm wide and 0.1–0.5 mm thick.
1. Strip, 7 mm wide and 0.125 mm thick.
2. Strip, 8 mm wide and 0.15 mm thick, as advertised.


1. The one included with the first kit is 63 cm long—they were clearly cutting costs.
2. The handle came with a decent 3 m roll of strip, and the strip has an ideal cross-section. Nice.
3, 4. The test strips purchased separately are exactly one meter long.


I'll check what the strips are made of by measuring their resistance. Before testing, I checked the milliohm meter's accuracy.



I described a similar measurement method in this review. Pure nickel should have a resistance of 87 mΩ for a 1 mm² cross-section and a length of 1 m. The other common material used for these strips is steel, which has noticeably higher resistance.


1. The first strip measured 122 mΩ over a length of 60 cm, with a cross-section of 0.6 mm². For a 1 mm² cross-section and a length of 1 m, the resistance would still be 122 mΩ. This isn't nickel, but coated steel.
2. The second strip measured 135.9 mΩ over a length of 99 cm, with a cross-section of 0.84 mm². For 1 mm² and 1 m, the resistance would be 115 mΩ—clearly steel.
3. The first strip from the store measured 92.2 mΩ over a length of 99 cm, with a cross-section of 0.875 mm². For 1 mm² and 1 m, the resistance would be 81.5 mΩ, which looks more like nickel, although it's lower than the resistance of pure nickel.
4. The second strip from the store measured 70 mΩ over a length of 99 cm, with a cross-section of 1.2 mm². For 1 mm² and 1 m, the resistance would be 84 mΩ, which is consistent with pure nickel.

The first two strips are clearly steel. The ones bought from the store are most likely pure nickel, although the measured resistance was slightly lower. There may be some measurement error, but in any case, they're much closer to the 87 mΩ specified for nickel. They also bend differently: steel is springier.


The first test was done with the strip supplied with the handpiece: 7 × 0.12 mm.


I tried welding one strip to another using different numbers of pulses.
1–4. With 1 or 2 pulses, the strips barely welded at all; the pieces could be pulled apart by hand with no noticeable resistance.
5, 6. With three pulses, it started to look a little more like a weld, although the bond was still very weak.


I forgot to measure the voltage on the supercapacitors before the tests, and only remembered at this point. They had been charged to about 2.5 V per cell 12 hours before the experiments; after the preliminary tests, the total voltage was 4.86 V.


Since I wasn't at all happy with the results, I decided to increase the current a little by connecting the wire from the holder directly to the supercapacitor terminal instead of routing it through the terminal on the circuit board. In other words, I effectively shortened the total wire length by 1/4.
With this connection, the positive wire going to the circuit board can have a smaller cross-section.



All the following experiments used three pulses, and I can say that after changing the terminal connection, the unit started welding properly. I joined two strips without any problems.


It took quite a bit of force to pull them apart, and you can see in the photo that the strip was even burned through in one spot.


Now for some practical tests with batteries. I used Sanyo NCR18650GA cells left over from other experiments, so I didn't mind risking damaging them.


I started with steel strip, and it welds very well. I could only tear it off with pliers and quite a bit of force; in some places, the strip tore at the weld.



The results were identical in both experiments: it was really hard to tear the strip off.



After that, I tried 0.15 mm nickel strip, but unfortunately the unit barely welds it. One spot was welded properly, but most of the time I got a "cold" weld.



The last piece of nickel strip, 0.12 mm thick, did a little better, though not by much. I pulled it off by hand, but it took some effort.
In this case, the battery contacts may have been affected by the previous experiments.



After the experiments shown above, the voltage on the supercapacitors dropped from 4.86 to 4.45 V.


Since the supercapacitors had discharged quite a bit, I decided to try again, this time charging them back up to 5 V.
I couldn't really feel any difference between 4.45 and 5 V, or it was very slight. It welds at the edge of the strip, but a little farther in, it no longer holds.


Since the current depends on the wire length, I decided to shorten the wires further by connecting the board so that the negative lead was connected directly to the supercapacitor terminal.

A simple calculation shows that with 10 AWG wire, which has a resistance of 3.28 mΩ per meter, every 27 cm reduces the circuit resistance by 0.9 mΩ, not counting the drop across the terminals. So, very roughly, the original setup had:
0.47 mΩ ESR for the two supercapacitors
0.22 mΩ for the transistors
3.5 mΩ for the wires
I reckon another 1–2 mΩ or so for the various terminals and connections

That puts the total resistance at no less than (and probably more than) 5–6 mΩ. At 4.5 V, the short-circuit current should be around 800 A. In reality, a lot is lost across the contacts, the board, and the actual welding point. But in any case, reducing parasitic resistance in the cables should increase the welding current.


And indeed, it welded a little better, at least strip to strip, but welding to a battery wasn't much better.


Feeling thoroughly discouraged, I remembered I still had some strips left over from a review, including 0.1 mm nickel strip. While I was visiting a friend, I also picked up some old batteries left over from battery pack replacements.


But that didn't seem like enough, so I thought, "in for a penny, in for a pound," and charged the supercapacitors almost to their full voltage.


When I tried welding one strip to another, it turned out the unit could easily burn right through it. Welding to a battery also went much better, though a different problem cropped up: when the strip burns through, it sometimes sticks to the welding contacts, so they need to be cleaned.


Next, I tried the 0.12 mm strip again, the one I'd experimented with earlier.
I was able to burn right through it when welding strip to strip, too. Welding to batteries was much like in the previous experiment: near the edges, the strip might even burn through, but there are spots that don't weld properly, although there's clearly been progress.


Now for some add-ons, details, and other odds and ends.
First, balancers are sold as separate boards, but they're only rated for up to 2.5 V. In my opinion, because of their increased leakage current above 2.7 V, the supercapacitors balance themselves reasonably well anyway. Also, with these balancers you'll only get 5 V instead of 5.4 V, which could be significant.


Controllers.
The one in the review is the simplest controller option, with very few features. The power transistors are built in, though, and there are much more advanced options available.


These already include a charger and a balancing circuit, although the power board has to be bought separately.


Power boards also come in different versions, for different numbers and types of transistors. You can usually buy them from the same sellers as the controllers.


The same sellers also offer supercapacitors, usually salvaged ones. Sometimes the price varies even from the same seller—for example, $5 or $6.50.

They also offer copper busbars for connecting supercapacitors in a 2S2P configuration. This can both extend runtime and increase the output current; for systems like these, the Chinese sellers claim 1500–2000 A.


Now for the main thing: my conclusions and various thoughts.
1. The unit does weld, although calling this board by such a grand name is a stretch, since it's just a board and some supercapacitors.
2. The stock welding contacts are junk. Better to look for a board without them and buy a decent set separately.
3. It handles steel strips with ease, welding them so firmly you can't pull them apart with your teeth.
5. Nickel strips are a problem: it only welds thin ones, 0.1–0.12 mm, and only after removing the extra wires and with the supercapacitors almost fully charged.
5. Supercapacitors are really great—much better than batteries.
6. The current-feed automation behaves strangely. Apparently, it uses a simple RC circuit, since the delay can be 0.5 seconds or 1 second, depending on the pause between touching the contacts.
7. I think the circuit could be modified so the delay starts when the limit switch closes, rather than when the contacts touch.
8. To be able to weld pretty much anything, you need a more powerful board, different wires to the welding contacts, and possibly four supercapacitors instead of two.
9. The board's features are basically sufficient: three modes and a sleep mode. But if I were buying one now, I'd probably look for a full-fledged controller.

In short:
This welding method has potential, but needs some work. One advantage is that you can weld just about anywhere, though not for very long. You can charge it from either a power supply or a power bank, but for normal operation I'd recommend keeping a power supply connected permanently—or, more precisely, a charger with CC mode, rated for 5–5.5 V and 1–2 A. That's more than enough.

I'll keep experimenting for now, but I have a feeling the board under review is already operating close to its limits. At any rate, I don't dare replace the wires, since they partially limit the short-circuit current; making them thicker could easily burn out the transistors if the terminals themselves are shorted.