The idea behind resistance soldering is to stuff a tremendous amount of current through a relatively low-resistance joint, thus producing enough heat to melt the solder and bond the parts. Because power dissipation varies as the square of the current, more current is much better!
Commercial resistance soldering units have relatively small transformers with small windings that impose a low duty cycle: you must let the transformer cool off between joints. A 50% duty cycle seems about the norm for hobbyist-grade units; you can obviously pay more to get more.
Given that the transformer is free, getting one that can supply a kilowatt for half an hour without breaking a sweat seemed like a Good Thing. It hasn’t warmed up appreciably during any of the relatively short projects I’ve used it for… although the electrode holder can get pretty toasty.
Find a discarded microwave oven: the bigger, the better. Our ancient Sears ‘waver went casters-up at an opportune moment; IIRC, the magnetron filament failed.
I harvested the transformer, line fuse, triac, snubber, and a handful of other odds and ends. The magnetron shell became a really nice desk tchotchke.
Transformers really don’t care, at least to a good first approximation, whether the secondary is producing a huge voltage at a low current (which is what ‘waver transformers do for a living) or a piddly voltage at a huge current (which is what we want). The core flux is proportional to the ampere-turns in the primary (a kilowatt is about 8 amps at 120 V) and we can rewind the secondary to get what we want.
This transformer had separate primary and secondary windings, which made life easy. The secondary is on top, the primary below. The primary has about 120 turns of stout wire, giving a turns-per-volt of about 1; that seems to be common with relatively cheap transformers. Remember that number…
With the transformer in hand, apply your least-favorite wood chisel to the secondary winding: chop the entire winding out. Pay very careful attention to not damaging the primary winding. You’ll probably find a few turns of relatively heavy wire for the magentron filament; chop that out, too.
After a bit of beating, the secondary should come out in two hunks that you can toss in your copper recycling box. It’s pretty much solid metal:
This transformer had a pair of flux shunts between the primary and secondary winding that stabilize the output power. We don’t care about that, so pry them out; they’re the small laminated steel blocks just above the primary winding. I also removed the cardboard liner around the core opening. The primary terminals are 1/4-inch quick-disconnect tabs, aimed straight at you so they’re hard to see.
The primary has about 1 turn per volt, so the secondary will, too. You can use nice floppy 4 AWG silicone insulated wire, but I went with four parallel strands of 10 AWG wire stripped from a length of house wiring to get the same cross-sectional area. Five turns produces a 5 volt secondary, which may be a little high; it seems commercial units run at about 3-4 V.
I terminated the secondary in heavy copper crimp lugs, but, not having the proper crimper, I made an open-top clamp to support the sides of the lug. Applying a punch to the top did a satisfactory job of making the lug one with the wire. I also soldered the joint, less for electrical goodness than simply excluding oxygen and improving the mechanical rigidity.
Caution: you must use a crimped joint, because the whole point of this exercise is to put enough heat into a soldered joint to melt the solder. Think about this: if you have two soldered joints in series with the one you’re trying to make, what could possibly go wrong?
I put some thin cardboard around the opening to prevent insulation scuffs, but, frankly, that’s not needed: this wire has really tough insulation and can take care of itself.
And then it looks like this…
The components are what you’ll need to measure & plot the B-H curve, as described there. For what it’s worth, here are three curves with 20, 60, and 120 VAC on the primary.
The core is pretty much saturated at 120 VAC, which is a simple form of power regulation: small voltage changes won’t make much difference in the power output. The peak flux density is about 20 kG, out there on the limbs of that hysteresis curve.
Next: put a triac in the primary circuit…
[Update: Herewith, the oven schematic. The transformer core is bonded to one end of the secondary winding. The other end is capacitively coupled to the halfwave rectifier that drives the magnetron. Note that the three-turn filament winding is attached to the hot side and is floating at 4 kV off ground.
None of that matters here, because you’re chopping those windings out and replacing them with five turns of husky wire.
Don’t you wish all consumer electronics came with schematics?
end update]
The Smell of Molten Projects in the Morning 
I’m glad you’re writing this up: now I can rebuild my homebuilt spotwelder on a more rational basis.
I’ve been told that a lot of microwaves have the core attached to the high side of the secondary, which is a good reason to not use them. Obviously since you’re removing the secondary that’s not a concern, but did you happen to notice if this was the case before the windingectomy?
the core attached to the high side of the secondary
The secondary was definitely bonded to the core; the tab is visible above the right-hand opening on the picture of the gutted core.
*sound of rummaging*
OK, I just added the schematic to the post. One end of the secondary is connected to the core, which is firmly bonded to the grounded cabinet.
I can’t imagine floating the core at 4 kVDC, though. That makes no sense at all. Maybe somebody got confused by looking at the magnetron voltage, which is negative with respect to the case?
I’m guessing the person was just confused — I don’t recall the source being that knowledgeable about electronics, and just saying not to use microwave transformers. In any case, it’s great to know this because it expands the horizons of my own “ya gotta have stuff” mindset.
Looking at that schematic, I briefly wondered why the logic board transformer offered 2.6VAC. Then I thought “Hey, that would be a useful array of voltages to run some logic and a VFD!” Then I thought “Well, duh!”
Those Sears engineers, they were no dummies… [grin]
Actually, Sears doesn’t build anything, they just rebrand stuff. You can find out the actual manufacturer from the part number. At the top of the schematic, I see that it’s 564.something, which crosses to Sanyo. So it’s Sanyo engineers who weren’t dummies!
It’s “Sears” until it fails, then they give you this song-and-dance about their manufacturer having problems…
Is this setup capable of welding too? I know you didn’t solder the lug for the sake of electrical continuity, but I would still be tempted to seal it with some epoxy or similar goop (high temperature for safety)rather than solder.
Be vewy vewy quiet I’m hunting micwowaves
Welding requires enough energy to melt ordinary metals, rather than solder. Although this setup has a reasonably high current, the open-circuit voltage is only 5 V: too low to start a good arc.
Perhaps you could weld very, very small workpieces, but my feeble attempts to weld nickel strips to AA cells didn’t work out at all. Real battery contact welders use a large capacitor bank to provide plenty of power during a short time to get very high power.
seal it with some epoxy or similar goop
Yeah, I do that a lot, don’t I? In this case, I figured doing it with solder made more sense: a positive bond that can’t work loose as the copper wire flexes.
My simpler and far more gothic homemade spotwelder — an arcwelder transformer rewound to produce about 10V at about 150 amps — does an *excellent* job of welding stainless steel, and a pretty good one with mild steel. They’re full-on welds, too, as strong as the base metal. The proviso is the electrode contact patch has to be really small to get a weld between two sheets of steel. I built it for welding wire together, like thermocouples, and it does a great job of that all the way up to 14 gauge stainless wire. In my case, to avoid switching over a hundred amps, I have a relay on the input side, which means it’s not what you’d call precise timing. But the whole thing was an exercise in ugly, so the timing issues are on par.
contact patch has to be really small
That’s the key; the carbon gouging-rod electrode I’m using is far too chunky.
I should grind down a tungsten electrode into a really small, really small probe, then clobber a nickel strip on an AA cell again. The original electrode was a bit over 2 mm in diameter: much larger than your wire joints.
Do you have a mechanical clamp holding everything together during the weld cycle? I’m certain that’s the right way to do those joints: more pressure, less motion.
I offer with considerable hesitation my arcwelder design to explain what I’m doing. Keep in mind I made this 12 years ago and it’s really primitive: I’ve learned a lot about construction since then. But the point is: there’s a threaded rod with a counterweight, allowing me to vary the pressure holding stuff together from almost nothing up to maybe 10 kilos of force, and also allows the electrodes to move extensively during the weld — which was useful for me because I was welding thick wires to other thick wires. The guy whose design I was generally imitating used a hold-down clamp he’d modified so he gets a lot of pressure, with almost no movement, and that’d be way better for sheetmetal spotwelding.
I grovel & abase myself at your feet: you originated the “orc engineering” concept! Wonderful term, use it all the time… uh, this means I owe you royalties, right?
Anyhow, I read your writeup long ago, long before I actually built my gizmo, back when I was trying to figure out what was known, what I had to do, and how to make it happen. It helped me push my doodles into order and get the right order of magnitude for voltages & currents.
Needless to say, I was overawed by the 5bears mechanical design, too…
Many thanks!