6Y880 Spot Welder Controller and Various Other Components

Not long ago, I published a review of a board for building a simple spot welder powered by supercapacitors. Today, I’m continuing that series: I’ll tell you about a much more capable device and, along the way, build a welder based on it.


First, I’d like to say a big thank-you to Vladimir, who sponsored this review and bought all the necessary components for me on Taobao. Without him, this review probably wouldn’t have happened. We spent quite a while choosing a reasonably optimal configuration, and below you can read what we ended up with.
The review will be fairly brief, but I’ll do my best to make it as informative as possible.

In addition to the supercapacitors and the handle with contacts from the previous part, I needed a controller, a power board, and some other components. This is what I received.


I’ll start with the controller. It’s the so-called CPSMC 6Y880, listed on the product page as a fifth-generation controller. It costs $22.72 — link, also available on AliExpress
There’s a version without the fan and a few other minor parts, but it’s easier to buy everything at once.

The package includes:
1. Control board
2. Balancer resistor board
3. Encoder board and a pair of connectors
4. Fan and cable tie.
5. Encoder with knob
6. The necessary wiring for connecting everything.
7. Standoffs for stacking the boards



The screen’s visible area is about 50x40 mm. I think many of you have seen it in other devices, too. Under it, on the power board, you can see a couple of capacitors and a buzzer. The buzzer had an interesting effect: I usually work at night, and to stop it from annoying me with its periodic beeping (because the supercapacitors were disconnected), I tried covering it with electrical tape as usual. That only made it beep even louder.


The power board houses the charger, the balancer transistors, and the voltage regulator. The encoder is also installed here.
The board can be powered with 12 to 19 V. Its current consumption depends on the settings; in practice, it’s the power that matters, not the current, since the required current depends on the power and supply voltage. The lower the voltage, the higher the current.

The charger is built around an RT9214 synchronous converter, so its efficiency is quite high.


The display board also carries the microcontroller, which controls the entire device and measures the voltage across the supercapacitors and at the board input.


The gate driver for the power switching board is also mounted here. The circuit is simple, using a few transistors and no dedicated ICs, which is perfectly normal.


The resistor board has two groups of five resistors. The resistors within each group are connected in parallel, and the groups don’t share a common wire, so four wires are needed for the connection.


There are plenty of connectors.
1. The power input uses a 5.5x2.1 mm jack. To its right is a two-pin fan connector, and farther right is the spot for the encoder connector. The encoder can be installed either on the main board or separately. That’s why the package includes both an encoder board and a pair of connectors, as well as the matching cable.
2. A three-pin connector for the supercapacitor power switching board and a two-pin connector for the control button.
3. On the left is a three-pin connector for measuring the voltage across the supercapacitors; on the right is a terminal block for the resistor board, and farther right are three terminals for the power wires to the supercapacitors.
4. The resistor board’s terminal block is here, too. There’s also a warning symbol indicating that the board can get very hot.

Note that the supercapacitors are connected with six wires: three carry the charging/discharging current, while the other three measure the voltage. This arrangement improves measurement and balancing accuracy, and also speeds up charging the supercapacitors. The wire ends are joined at the supercapacitor terminals, or as close to them as possible.


The package includes a small 50 mm fan, all the wires except the power wires, standoffs, and the encoder board. Overall, the kit is quite well thought out. I especially liked that the encoder has its own board, connectors, and ribbon cable, so you can install it either on the main board or anywhere on the front panel.


I’ll show the wiring diagram later. For now, let’s look at how to operate the controller.
When powered on, the manufacturer’s logo appears, then the controller switches to the main control menu, which is in Chinese.
To change the language to English, open the settings menu in the top row on the right. Turn the encoder to navigate and press it to select.
In the settings menu, go to the second-to-last item, click it, select English, click again, move down one more item, and exit the menu. That’s it.



Here’s an overview of the main menu.
1. Set the duration of the first pulse, from 1 to 35 ms. Press the encoder to select it, turn the encoder to set the time, then press it again to confirm. This and all the other settings are saved when the power is turned off, which is very convenient.
2. Set the duration of the second pulse, from 0 to 35 ms. It’s set to 0 by default, so no second pulse is generated.
3. Set the pause between pulses, from 0 to 99 ms. At 0, the pulses merge into one, so you can set a single pulse duration of 1–70 ms.
4. Automatic current-feed mode. By default, current is controlled by the microswitch in the handle, but if there isn’t one, you can enable activation when the electrodes touch. This setting controls the delay between contact and current delivery, from 0.1 to 5 s.
6. Voltage adjustment. In practice, this is equivalent to adjusting the current, since current depends directly on voltage.
The voltage can be adjusted from 0.1 V up to the battery’s maximum voltage.

The screen also displays:
1. The battery voltage and the voltage of each cell, both numerically and graphically.
2. The overall charge level (the green bar at the very top of the screen). There’s a catch: the bar shows the range from 0 to the set value, so it adjusts dynamically, while the fill level in the battery icons is based on the preset battery voltage. For example, if the battery is at 5.4 V but the setting is 2.7 V, the bar at the top will be full, while the battery icons will show only 50%.
3. The board’s supply voltage.
4. The temperature, although it’s not clear what it’s measuring....
5. The battery charging current.


Additional settings menu
Buzzer switch - turns the sound on or off. The sound is only used when turning or pressing the encoder, or in the event of a fault, so there’s no real need to turn it off.
Auto - Not set/set. An unclear setting, since automatic power delivery works without it.
Max charging current - the charging current for the supercapacitors, adjustable from 0 to 10 A. It’s set to 8 A by default. At 0, charging is disabled (for example, if you’re using an external charger).
Single capacitor voltage - the board is designed for a 2S configuration. This setting specifies the maximum voltage per cell, from 2.3 to 4.2 V, so you can use either supercapacitors or batteries, from LiFePO4 to standard Li-Ion. The setting increments in 0.01 V steps. Once set, the main menu lets you choose a voltage from the minimum up to the value set here, so there’s no risk of overcharging the battery.
Clear count - resets the activation counter, i.e. the number of welds made.
Continious welding - continuous welding. Unlike the standard 1–70 ms range, you can set a maximum duration of up to 3 seconds.
Language - Chinese/English


Just as an experiment, I hooked up the board with a makeshift connection.


Charging works without any problems and stops at the set value, but of course there’s no balancing because the resistor board isn’t connected. I set the charging current to 2 A because the connected power supply has a low power rating.

Speaking of the power supply’s power rating.
As mentioned above, the input voltage range is 12–19 V, the charging current range is 0–10 A, and the battery voltage range is 4.6–8.4 V (two cells in series).
So calculating the input current is very simple. Take the battery’s maximum voltage, say 5.4 V, and the maximum charging current, say 8 A. Then, at an input voltage of 12 V, the current will be:
5.4 × 8 = 43 W
43 × 1.2 (efficiency) = 52 W
52 / 12 = 4.3 A

First, we calculated the charging power (voltage × charging current), then added the conversion losses (actual efficiency is closer to 85–90%; I used a conservative estimate), and finally divided the resulting power by the power supply voltage to get the current.
Since charging usually takes relatively little time, there’s no need to get an oversized power supply. I actually went the other way later, but more on that later.


A new version of the board is already available, with more settings and charging current up to 20 A. It costs $33 — link.


I also ordered:
Power transistor board
Copper busbars, included with the transistor board
The transistors themselves.
Copper terminals.



Copper screw terminals in two sizes, SC25-6 and 16-6 — link
The busbars and their mounting bolts came with the board — link.



Power board.
These boards are sold for different numbers of transistors, with or without transistors, with or without the power switches split into separate sections, with through-hole or SMD components, and with just the transistors. In short, you can choose whichever board you like. Here’s what sets them apart.
1. Board without transistors. This lets you install better transistors, but soldering them in is quite tricky. There are versions for TO-220 transistors, but they usually have worse specs.
2. Board with transistors. No soldering needed: just buy it and use it as supplied...
3. Board with split switches. Some boards have separate copper areas for connecting the power busbars. In theory, this should make it easier to diagnose blown transistors. In practice, I think a transistor will simply split in half, leaving nothing to diagnose. Still, you can choose this type of board; it won’t hurt.
4. Number of transistors. In theory, the more the better, but controlling them becomes a problem. In my opinion, 12 is optimal, although there are monsters with 32–52 transistors.
5. Additional components on the board. It’s better to have them; you can always remove or short them out. The problem is that the control circuitry is often extremely basic, while a board with components gives you at least something to work with.



We chose a board for 12 transistors, with additional through-hole resistors and diodes, and copper busbars.
For some reason, though, they included one extra resistor and one fewer diode. It costs about $5.75 with the busbars — link


The circuit is primitive: the transistor gates are driven through diode isolation, and the gates themselves are connected to ground through 10 kΩ resistors.
In other words, the gates charge quickly and discharge through 10 kΩ. That seems like a lot to me; if I’m not mistaken, with a gate capacitance of 22 nF, the total turn-off time will be about 1 ms.
I chose this board with the idea of modifying it by adding drivers.


Power transistors.
Two types are commonly used: D2PAK and TO-220. The latter are less widely available and have worse specs, so I chose IRL40SC228, among the lowest-resistance options, with a specified on-resistance of 500 µΩ, or 0.5 mΩ.
They’re sold on tape, but I wasn’t expecting them to be genuine and new. And after they arrived, I noticed that even at a glance there were slight differences between individual units. I bought 24 for $0.62 each — link.


The dimensions all match, but I found a minor design flaw. The copper busbars and six bolts came with the board: five of one size and one a size larger. There’s a hole on the board where the vias between the two sides are set farther back from the bolt hole, but the hole itself was made the same diameter as the others. The other board photos show everything done correctly, so apparently there was a mistake in the specified diameter.


Next, I cut off a piece of tape with 12 transistors. The other half is “spare” or for any modifications...


They look like ordinary new transistors, but there are small chips and scratches on the cases that definitely shouldn’t be there on transistors supplied on tape.


What’s more, even if you take 12 in a row, you’ll find that they differ in appearance, too. One even had a fingerprint on it, though I can’t be sure—maybe I left it myself, but that seems very unlikely.


The tricky part about mounting these transistors is that you really need a hot-air rework station with bottom preheating. I reviewed one a while back.


But before soldering all 12 transistors in place, I decided to test one first. The test was simple, of course, especially since testing it “to the limit” is very difficult, and its specs are quite impressive.
The only downside, in my opinion, is the high gate capacitance typical of transistors like these.


For the test, I applied a current of 12 A and a gate voltage of 9 V. The results were:
1. Transistor resistance near the leads coming out of the case: 0.493 mΩ
2. On the copper pours near the case: 0.626 mΩ
3. Flange-to-pour section: 18 µΩ
4. Lead-to-pour section: 92 µΩ.

As you can see, the resistance matches the stated value. The lead-to-pour section contributes the most to the total voltage drop, so I used a little more solder there than necessary.
The board’s calculated total resistance is 55 µΩ, which at 400–500 A gives a voltage drop of 22–27 mV and just 9–13 W of power dissipation—though that’s for steady-state operation.


But besides the power section, I also needed a way to connect the handle with its contacts. There were two options:
1. Run wires out of the case
2. Install a connector.

The first option is straightforward, but the second is a bit more complicated. People usually use welding machine terminals, but “we don’t take the easy route,” and I wanted something a little more elegant.


First, just a wire: super flexible, with silicone insulation, 7 AWG or 10.5 mm² by our standards. I bought it in case something went wrong and I needed more current.
I didn’t even bother counting the strands. There’s a clearly visible core and outer layer, which may be twisted in opposite directions.
This cable costs $3.16 per meter for a single color (red or black) - link. The page also offers wires in different colors and gauges up to 2 AWG (33 mm²)!

It’s a really great wire. I recommend it.


But as I mentioned above, I wanted a neat, detachable connection, so I ordered a set consisting of a male/female connector pair and a cable.
This connector is called the Amass AS150U. This configuration costs $6.04 + 4.46 = 10.5 - link



The wires here are thinner, 8 AWG or 8 mm², and the ends are tinned. One useful feature is that, in addition to the power contacts, there are four auxiliary contacts that can be used to connect a microswitch and possibly some lighting.

The wire is also very flexible, and the insulation is silicone, of course.


Why did I choose this option instead of the usual welding machine terminals? The connector is rated for 140 A continuous, and its stated resistance is 0.4 mΩ, comparable to that of welding machine terminals.


But this option looks neater and is more pleasant to use, especially since it lets you connect six contacts at once: two power contacts and four auxiliary ones.
It’s sold in versions with a plug or socket on the cable, and a corresponding plug or socket for the device. The cable version has a flexible lead.


The contact arrangement is different: one of the contacts has a plastic insert. I assume this is the positive contact, so the common wire will always connect first.
The housing is marked with the manufacturer and connector type, as well as the contact polarity. Unfortunately, yellow on yellow is hard to see.
The 8 mm² wire fits the contact perfectly; 10 mm² fits with difficulty, literally a tight squeeze.

I ran into a few issues while working with the connector. The terminals are meant to be pressed into the plastic part after soldering, which is why they come separately, although there is also a version with them already installed. But getting them in there isn't easy, so when soldering, leave a little room to reach closer to the middle with something you can use to drive the contact in. Otherwise, if you hit the edge, the contact may go in crooked. I'll show the installed contacts later.



I'm sure many of you are familiar with the popular XT60 connector. I've put one next to the AS150U for comparison; I think it speaks for itself :)


At first, I planned to fit everything into the enclosure I'd already used for two of my devices, an LCR meter and an adjustable power supply. Unfortunately, everything would have been packed in too tightly, and I'd have had to build the power supply separately anyway.
In the end, I bought a larger enclosure from one of the local online stores. Besides the size, I needed it to have a stand so I could prop it up while working.

I settled on the relatively expensive (about 11–12 dollars) KH-34-4, measuring 240 × 210 × 100 mm.
I liked the enclosure at first: screws to join the two halves instead of self-tapping screws, plenty of standoffs for circuit boards, and even special slots. But during assembly, I found that the top and bottom were both bowed outward, and stayed that way after assembly. The whole thing also felt rather flimsy, and the stand was very hard to pull out. Overall, I don't recommend it.

I also bought a power switch with an indicator LED, a fan grille, and transistors and diodes for modifying the power-switch driver.


Surprisingly, the biggest challenge during assembly was installing the supercapacitors, especially since the bottom of the enclosure isn't flat. The manufacturer even made the standoffs different heights to account for this. The middle section is relatively flat, while the sides are slightly raised as a design feature.
In the end, I used a scrap of fiberglass board that had been lying around unused for ages. I cut it to fit, screwed it to the enclosure, degreased it, and attached the capacitors with soft double-sided tape. They seem firmly in place, but I plan to secure them further with cable ties later.

Warning: to reduce the risk of problems, I recommend discharging the supercapacitors before working on them.


I assembled it according to the connection diagram intended by the board's manufacturer. Just in case, I drew it separately, since the one on the product page is incomplete and is for a different board revision. I'm no artist, of course, but I think it's clear enough.


Here's what it looked like partway through assembly. I didn't photograph the complete build because it depends on the type of enclosure, so everyone will have their own layout. I'll just point out a few things:
1. I attached a heatsink to the resistor board, but first I had to straighten the resistors. I applied thermal paste between the heatsink and the resistors.
2. Since the resistor board gets hotter than anything else, it's best to position it so that airflow from/to the fan passes over it.
3. I placed the power supply along the same line. I used a 40 W, 12 V unit from Sanmim - link, although I wanted to use a different one. This one had mounting holes, though.
4. The controller board also heats up, so it shouldn't be "in the shadow" of the airflow.

I'll jump ahead and say that cooling is excellent; the stock fan is enough in this configuration.


For the power wires, I used 6 mm² for the positive lead and 4 × 2.5 mm² for the negative. I could have used 8 mm² wire with silicone insulation, but I couldn't bring myself to cut that kind of wire into pieces.
The third photo shows the contacts being pressed in. I had to drive them in with a small hammer, using a screwdriver. Ideally, I should have pressed or struck them right in the center, but I managed without doing that. Of course, this should be done after soldering the wires, so you don't melt the connector's plastic.
Then I soldered the signal wires and the wires for the backlight LED, but haven't installed it yet because I haven't figured out the best way to mount it on the knob.


At first, I wanted to glue the connector to the enclosure. I cut off only some of the stops, then decided it would probably be better if it were removable. In the end, I made a U-shaped retaining plate from a piece of plastic left over from cutting the display opening.
It worked out great. Now I can remove the connector without disturbing anything else, and it holds very firmly.


Finished assembly.


The power supply board ended up with a cover plate. I wanted to place it closer to the side wall and farther from the front panel, but the enclosure's design wouldn't allow it. As a result, it ended up too close to the power switch.


I desoldered the controller board's power connector and soldered a terminal block in its place. All the wires connected to the terminal blocks are crimped with ferrules. By the way, I really didn’t like the terminal blocks on the controller board. I’d have replaced them with something better, but there weren’t enough suitable ones nearby.


I didn’t drill out the larger-diameter bolt hole in the board; I just used a different bolt. I also connected the wires diagonally, although the design calls for the blue wire to go to the middle hole.


And here’s the finished result. The front panel feels a bit too empty, and the overall look somehow reminds me of a small TV from the late ’70s.


There’s just a fan and a power connector on the back. For convenience, I used a standard “computer” power connector; almost all my devices have the same kind.


Time to see how the whole thing works.
Since the supercapacitors were discharged before assembly, the first thing to do was charge them. I initially set the charging current to 2 A, left over from previous tests. Then, figuring the new power supply could handle more, I raised it to 8 A, which noticeably sped up charging.
I didn’t time the full charge; it took about 15 minutes. According to the board’s readings, the temperature peaked at 51 degrees, though that’s probably not accurate.
As the voltage approached the setpoint, the current started to drop. Near the end of charging it fell to 0.2–0.4 A, and the board balanced the cells at about that current.
The photo shows that the difference was about 15 mV near the start of charging, but had dropped to zero by the end. Both supercapacitors were charged to 2.5 V.
I later experimented and charged them to 5.4 V, but for now I’ve decided to stick with a lower voltage.

The fan starts running at a current of about 5–6 A and switches off at the end of the balancing process.

I expect someone will ask how long the startup preparation takes. Since it usually only needs a top-up charge, it’s actually pretty quick, especially with the charging current set to 8–10 A. When the device is switched off, the supercapacitors discharge slowly. For example, after two days switched off, the voltage dropped from 5.4 V to 5.28 V.


Since I may need to adjust the welding current during use, I checked how well the board can discharge the battery. This is more of a side effect than a dedicated function: the balancer does the discharging. Lowering the voltage also changes the balancer’s activation threshold, so it acts as a load until the battery is discharged to the set level.
I set it to 4.5 V. The fan started, and the board began discharging the battery. A slight imbalance appeared during the process, but was eliminated by the end. I didn’t measure the discharge current, but discharging from 5 V to 4.5 V took a minute and a half.

The fan also starts during discharge and switches off at the end.


Thermal performance.
When charging at 10 A, from an initial voltage of 3 V to a final 5.4 V, the control board gets noticeably hot. Surprisingly, most of the heat comes not from the converter, as I expected, but from one of the balancer transistors—though it’s more likely the transistor that switches power to the supercapacitors.
That transistor reached 80 degrees. The converter inductor was next, at around 65–70 degrees. But charging at 10 A across that voltage range is unlikely; besides, I’d already tested it from 1 to 5 V at 8 A, and it charged without any problems.

At first I was more concerned about the balancer resistors overheating during charging/balancing, but it turned out they stay almost cool in that mode (second photo)—unlike during discharge.
When discharging the battery from 5.4 V to 3 V, the resistors reached 105–110 degrees, so I think my idea of adding a heatsink wasn’t unnecessary after all. The control board itself was just warm.


In my previous review, I complained that I could only weld steel strips, and had trouble with nickel strips even at the maximum settings and with short leads. I won’t repeat the previous tests here, since I was specifically interested in welding nickel strips.
For the test, I initially used strips 0.12-0.15 mm thick, but quickly realized that they were just toys for this machine, so I started experimenting with 0.2 mm strips, especially since thicker strips are rarely used for welding cells.

It turned out that the machine could weld strips like these without any problems—or almost without problems, to be more precise, the problems were of a different kind.


I tried welding in different modes. Usually 20-30 ms is enough, but when I “wanted more,” I increased it to the maximum 70 ms, although that was already a generous margin. In this mode, heat tint appears on the strip.


I was just playing around, welding the strip to some used batteries, when at one point I noticed a strange smell. It turned out I’d accidentally welded right through the battery’s negative terminal. In the second photo, you can even make out a small bubble of liquid that had started to seep out of the battery.
So it turns out you can get considerably more power than you need for normal use.


Since it turned out there was plenty of power to spare, I moved on to more extensive tests—or, to put it simply, started trying to weld all sorts of things.


A 0.65 mm thick piece of galvanized steel that came with some fan.
I welded on 0.15 mm thick strips and can say that it also welds with plenty of current to spare. The photo also shows an attempt to weld two nickel strips together. For that, you need to set the pulse duration to around 10 ms and the voltage to 5 V.


It heats things up so much that the marks show through on the other side, but I overdid it a little during the tests.


This is an attempt to weld a 0.55 mm steel plate to another plate like it. Unfortunately, the machine couldn’t handle it, though this is extreme testing anyway.
The photo shows both sides of the plate. I think if I’d tried welding from both sides of the “sandwich,” I might have managed it.


Another 0.15 mm thick nickel strip, but this time I tried welding it to a massive base—in this case, a pair of side cutters. This test went great: it took quite a bit of force to tear the strip off, and in the end the strip itself ripped.


Next, a few oddities I didn’t quite understand. In fact, I’d noticed all of them earlier, but they became more obvious in this test.
The thing is, in many of my welding tests, one of the two contacts got much hotter than the other. I’d place the contacts, current would flow, and one contact point would glow red-hot for a short time while the other stayed cool. And it was almost always the same contact, usually the negative one.
At one point this caught my attention. I thought it was due to the polarity, so I switched it, but the welding results were the same. I rearranged the wires to make them more convenient while keeping the polarity correct, and after a while things went back to how they were: the negative contact heated up more during welding.


But the problem actually goes deeper. During welding, it often happened that the strip being welded by the negative contact would stick firmly to the base, while the other point, connected to the positive contact, would also stick firmly—but to the welding contact...
I tried shaping the contacts in different ways, both as a trapezoid and as a hemisphere, but after 7-10 welding attempts, the contact would return to a flat shape anyway. And when I pried a stuck contact off the strip, copper was left on the strip.

This was especially clear with a nickel strip and a piece of DIN rail. The contacts barely stuck to the strip here, but it was obvious that one point in each pair was quite different from the other.


Here I marked the “negative” weld points in blue and the “positive” ones in red. I welded them almost one after another. The “negative” points got a perfect score, while the “positive” ones could be pulled off by hand without any effort.


I can’t say anything definite about the polarity, since I don’t have enough experience and an internet search didn’t turn up anything conclusive. As for the contacts themselves, the ones supplied with the handpiece work fine with batteries, but if you exceed the current limit, they start to overheat.
The quality of the contacts may also be partly responsible for them “sticking” to the strip.
I started looking into what people actually use, and while searching on Taobao I came across some expensive contacts. Some cost $23 a pair (first and second photos), while others apparently go for $40 a pair (third photo)...


But for the handpiece I use, there are also much more affordable options. For example, this listing offers a whole bunch of options: finished and unfinished (I assume that means tapered at the tip), Chinese and Japanese, 100 and 500 mm long — link.


While searching, I also came across a flux that makes working with the copper strip that has recently started appearing easier. A friend who works with batteries had mentioned it and complained that it was more “delicate.”


After getting discouraged, I decided to experiment some more a while later. I’d seen people online say it’s better to weld with higher current and shorter pulses, and that’s more or less what I’d been trying to do.
But then I thought, why not try the opposite: lower the current and increase the pulse duration? I set the voltage to 4.0 V and the pulse to 35 ms, and found that it welded really well. Both contacts fused properly, and they didn’t stick.
I set the pulse sequence to 10, pause 50, then a second pulse of 35. It worked even better, but when I raised the voltage to 4.5 V, it welded just as well, though the contacts started sticking.

Red — five contacts, 4.0 V, 1/99/35 ms (I’ll explain later why the settings are so unusual)
Green — four contacts, 4.0 V, 10/50/35 ms
Yellow — six contacts, 4.5 V, 10/50/35 ms.


The welds were solid. The strip tore, but the weld spots held perfectly.


Before wrapping up and describing the machine’s features, I want to answer a question I’m sure people will ask: how much did all this cost in the end?
Let’s add it up.
1. Supercapacitors — $5 × 2 = 10
2. Handpiece with welding contacts — $13
3. Controller — $22.7
4. Transistor board with copper busbars — 5.75
5. Transistors — $0.61 × 12 = 7.32 (cheaper options are available)
6. Power supply — $10 (about $5 if bought used)
7. Wires and connector — $10.5
8. Terminals — about $0.3
9. Enclosure + power switch + fan grille — $12
10. Small hardware, wires and other odds and ends — let’s say $0.5
Total, excluding shipping — $92

Shipping is complicated: both the actual cost and the choice of sellers. It’s cheaper to buy everything from one seller because shipping within China costs less, but sometimes another seller has something cheaper, so you have to work out which option is better. That was the case with the transistors, for example: one seller had them for $0.61 each, another for $0.54, but buying them at $0.61 each worked out cheaper.
Shipping the supercapacitors adds the most to the cost. They’re heavy, and you can get them for $5 used or $6.5 new. I use used ones.

A few things to note.
Above, I mentioned some quirks with setting the pulse and pause durations. In the last test, I used the 1/99/35 mode.
The problem is that the microswitch triggers too early, when the clamping force is still low, and the contacts weld together because they “scrape” against each other. During testing, I tried disabling this function and using the automatic contact-pressure detection mode, with a 1-second delay before current is applied. It worked much better: clamp the contacts firmly, and only then does the welding current turn on.
But then I realized how to make it even better. Set the first pulse to the minimum, then pause, then apply the full pulse. That way, the first pulse tacks the contacts together slightly, and only then does the actual welding happen.

And, of course, the verdict.
I really like the controller and the machine I built with it. I had some trouble at first, which I described earlier, but it turned out the problem was simply that I lacked practice and needed to get used to the machine. I also recommend keeping spare tips on hand—don’t skimp on them.

What I liked:
A wide range of settings: you can adjust both the welding current and pulse duration over a very broad range. In addition, you can use the microswitch control and automatic mode triggered by closing the contacts, with a pause delay.
The controller also has a powerful charger, a balancing function, the ability to force-discharge supercapacitors, support for working with batteries, and the option to disable the charger.
The fan is controlled properly. It isn't annoying during operation, since it rarely turns on and only when needed.
The transistors I bought meet the specified on-state resistance, and the heavy-duty connector also turned out to be excellent. It's convenient to use, and there are enough contacts even to connect lighting.
Now there's more than enough welding current. To work properly, I had to use 4-4.5 V instead of the maximum 5.4-5.6 V.

What I didn't like:
The power board has the transistors, but no driver. Although it stays cool during operation, I don't think discharging the gates through 10 kΩ resistors is a very elegant solution.
There's no menu option on the board to switch between external-contact and automatic modes. Make sure to install a switch that disables the handle microswitch.
I had problems with welding: one contact welded properly, but the other didn't. It turned out the welding current was too high.
The terminal blocks on the controller board should be replaced with better-quality ones.

All in all, I can say that this machine welds, and it welds quite well.
Of course, you could say it's easier to build one using a microwave oven transformer, but the unit reviewed here has at least one advantage: it only needs 12 V at up to 5-6 A. It will even work with a 12 V, 1 A power supply, though the charging current will be no more than 2 A. That means you can use it out in the middle of nowhere with a power bank, or in a car.

That's all for now, but I'll keep experimenting.