I've been needing one of these for a long time and finally put one together.

The design came from this website http://www.philpem.me.uk/elec/welder/

The author there used 5 qty 120,000uF 25VDC caps for his capacitor bank. I scoured Ebay and missed out on a few auctions for similarly rated caps but then saw a good deal on 100,000uF 15VDC caps. Thinking that 15VDC max would do it, I bought 8 qty off Ebay for ~$40 delivered. Hopefully by minimizing the resistance of all the conductors, 15VDC will work.

Caps came and the were BIG and HEAVY! Pretty impressive.

I decided to mount them as you can see in the pictures below. With 8 capacitors, if they were all linearly arranged in a single level, the pack and the bus bars would need to be pretty long. Stacking them shortened the length of the pack and the bus bars and also minimized the electrical losses. I extended the bus bars 1" past the last capacitor so I could bolt on my cable. Bus bars were bought through McMasterCarr and are 1/16" thick, 1/2" wide copper. A Whitney punch was used to punch clean, perfect holes into the bar for mounting the caps.

The SCR purchased (IR 50RIA20) is the same as used by the author and was bought from Newark for ~$18. I originally wanted to get an SCR with a heavy flexible cathode cable but those were nearly 3X the price. The 50RIA20 has a continuous amperage rating of 50A, but a surge peak of 1480amps.
Here's the datasheet
http://www.irf.com/product-info/data...data/50ria.pdf

Per the original author, I wired in the SCR into the negative bus bar of the capacitor bank. A flexible connection was made to the cathode of the SCR by cramming 2 qty 12 or 14ga super flexible silicone battery cable wire into the terminal and soldering it. The wire was cut short (to minimize resistance losses) and the wire was crimped and soldered into a heavy gauge ring terminal and bolted to the negative bus bar. The reason for having this flexible connection is to prevent mechanical damage to the SCR.

Spot welding cables were made from some 4 gauge marine power cables I bought surplus over 10 years ago. It's like Deans Noodle wire but 4 ga!! The construction is impressive, many fine strand and woven like a rope to keep it flexible. I crimped heavy ring terminals on the cable and cut them each to 2 ft in length. Long enough to use, short to minimize resistance losses.

Welding tips were made from 1/4" copper rod (also bought from McMasterCarr). I bought 101 alloy copper since it was cheaper. Almost got 185 alloy (better suited for spot welding) but the price was 4x. Figure I can reshape the tips for a quite a while before I run out of material.

I originally tried to belt sand the probe to a point but didn't like the results. Then I tried to cut tips on the lathe but copper doesn't cut well (very gummy). In the end I did use the lathe but just to spin the probe and used a bastard file by hand to shape the tips.

I made the connection from the 4ga cables to the probe tips by using a heavy crimp on ring terminal, then bending the ring around the probe tip and soldering them together. Came out well and is linearly arranged making it easy to use.

Nickel strips were purchased from Sunstone Engineering. I bought their sampler pack of 0.003" and 0.005" thick, 1/8" and 1/4" wide strips, 10' of each. The pn from Sunstone is SNSSP - see here http://sunstoneengineering.com/site/pages/sales

Power supply is an adjustable lab supply. I have the current max'd out at 5.5A, voltage is adjusted to whatever energy I need to make the weld but has typically been 14V for 0.005" strip.

Results? It works! 15VDC is plenty of energy to even weld 0.005" stick to cells. Takes some practice but my welds are coming out nice! I can also weld button top cells so now I can make packs from store bought NiMH cells.

What would I change? It would have been nice to get higher voltage caps. The energy dumped is 0.5 * C * V^2
So the energy goes up by the half of the capacitance but the square of the voltage.

I am also still trying to find a bigger SCR. Although the peak current rating on the one I am using is good, the voltage drop at high current is pretty big. If I can get a higher amperage SCR, I can cut my voltage drop across the SCR by half or more.