spot welder


I needed a new transmitter pack for my old JR X347. I went to Fry’s and got a pack of 8 2750mah NIMh cells to replace the old 600mah nicad pack. I don’t fly anymore, but these packs are for testing Victor motor controllers that I rebuild for the Combat Robotics crowd. In a couple of weeks Combots will be at the San Mateo Fair Grounds again!!

Here are the first 2 cells getting spot welded together. I had no fixture so I used this block and some scotch tape. I cut up some .010″ nickel strips to spot weld the pack together. The settings on the welder worked fine for this thickness of nickel.

Fold them in half.

Push them together.

Now I’m spot welding pairs of cells together.

I added the next pairs of cells to spot weld.

Here is the finished JR X347 pack.

I took the old JR X347 cells found the dead one, added a good one to them and made this my Futaba AM transmitter pack. Note that the settings on the spot welder are lower. These nickel tabs are recycled from the old Futaba pack.

Here is the new cell being charged by itself to bring it up to the same 100% state of charge that the other cells are at.

I wonder if it’s even possible to spot weld aluminum with my spot welder? It is not as conductive as copper, but is far less expensive. Copper has at least tripled in the last couple of years. So I will start with trying to spot weld some aluminum foil from the kitchen and see how that goes. It’s cheap and easy to use for proving the concept, one way or the other. I will still be trying to spot weld braid as well.

Here is the finished spot welder. I sure like the way the LCD display is shining through the smoked acrylic. When it’s off it’s hidden. With the foot switch jack is in the rear and the LCD behind the smoked acrylic, the front bezel is kept very clean. The knobs blend in nicely. I will be this is the smallest chassis this welder has even been built into. The tight confines did create some EMI issues however.

Here is the waveform across the fet source and drain with the probes held firmly against the work, making a good weld.

Notice the 40.8v spike on this waveform. It was caused by the probes not being held tightly against the work. The fets have a 75v internal diode that should protect them. But it has been recommended that a schottky board be installed per the note from the designer, to attenuate these spikes to help better protect the fets.

The spot welder is finally working now! It was the fet driver signal from the processor that was not going back to ground after it sent the dual pulses. I added a couple of different values of resistor and placed them from the fet driver signal on the processor’s pin 27 (terminal block position 17) to ground. I ended up soldering the 2.2k resistor onto the right side of R27 and the left side of R22 for a permanent installation. 2.2k worked the best and draws only an additional 2ma. The I/O pins are rated at 25ma on the processor data sheet.

What also helped a lot was not connecting the ground lead for the fet driver to any of the grounds on the terminal block. When it was attached to any of the terminal block grounds, the spot welder kept sticking on. So I soldered it to the ground pad of C9, which is one of several parts not installed per the manual. It worked far better after that. This fet driver ground lead location was always contributing to the sticking on issue.

I also added a 1k resistor in series with the fet driver input signal so that when the fet driver fails shorted, it won’t take out the processor as well. When the first fet driver failed it was forcing 3.4v into the fet driver input circuit. Which forces pin 27 of the processor to 3.4v when it’s trying to stay at ground.

Now it’s time to get accustomed to using the settings to make good welds! All I have right now is .010″ thick nickel for testing. My aim is to spot weld braided cable for the combat robotics folks.

When I disconnect the fet driver wires from the control board, the fet driver pulse works fine. It no longer locks in the on state. Also when the ground lead for the fet driver output is attached near the grounds where the pots are attached, positions 6 & 7, it causes that pot to wander all over the place. So I soldered it to the ground hole near position 1 on P3. That corrected that wandering issue. Still the output gets stuck on. At least one press of the foot pedal corrects it. It almost looks like a ground loop.

Here is what the fet driver input signal looks like on the control board when the spot welder stays on. Trace #2 shows that it’s just stuck full on at 5vdc. This holds the fets in the on state and that’s why the probes are energized. The power supply is straining to keep up with the constant load instead of the milliseconds long pulses that it normally sees.

This is what the fet driver input signal looks like when operation is normal.

I found that the scope shot in the previous post represents when the spot welder is working correctly. When it’s not working correctly and the probes are constantly energized, the scope traces show 15vdc on the fet driver output. The entire time current is being moved into the cap using Q4. Almost as though the foot pedal was sticking closed if you will. But it isn’t, and it wouldn’t matter anyway since it just sets in motion some code to fire the fets. After the foot pedal is pressed just once more, the problem is corrected. It’s done this since it was first built. I wonder if the EMI from the cap charge leads are interfering with the system some how. The charge leads are running right past the terminal block and the 3 sets of pot leads. I’ll probe around with the scope to see if I can find something not doing it’s job.

The parts to fix the spot welder showed up this afternoon. I was thrilled.

After putting in a new fet driver, I put my DVM in series with the power supply
to the fet board. The fet board draws only 28ma ac. So it’s basically in fine
shape, so far.

The first time I took the probes and pushed them up against a piece of nickel
strip to test the spot welder, there was a large flash and pop! The probes had
voltage across them!! Almost like the foot pedal was being pressed ahead of
time. Yet it wasn’t. I put the scope across the output of the control board that
drives the fet driver. I couldn’t get the scope to trigger right. So I put the
probe on the output of the fet driver. The scope was triggering on something
kind of random. I adjusted the scope and there is a pulse on the fet driver
output that is 18.6v p-p. But only 500 nanoseconds long. Something around
2.5mhz. Wow. The normal pulses are between .1-35 milliseconds. But now I finally
got the scope to sinc up nicely to this odd pulse. So I adjust the other scope
channel to the signal that triggers the fet driver on the control board, and
it’s the same frequency, but only a 4v p-p pulse that I have not identified the
source. So the fet driver is actually being triggered to fire the 4 farad cap by
this micro tiny pulse. When I hit the foot pedal I get the correct pair of
pulses. So now I have found out why my spot welder probes are hot before I even
hit the foot pedal. This happened yesterday too! Some how this must have smoked
the fet driver. I whipped out my EMI probe and am now sniffing over the system
for a source that fits this pulse. The processor is 20mhz. A 4:1 harmonic
perhaps? I disconnected the fan and the lcd hoping they were involved, but no
such luck.

Oh brother, is this how I am spending my Friday night?

Party on Garth!

Here is the phantom pulse. Ch 1 is the output of the fet driver. Channel 2 is the input of the fet driver.

I did order some fet drivers from Newark. But there is not much of a chance that they’ll get here by Saturday USPS. So I called TI and they are overnighting the fet drivers to a local office. So I should have them in my hot little hands soon. I spent the evening drawing up a schematic for the fet board so I have a better view of what caused the fet driver to smoke.

The one thing I will modify is the output that goes directly from the processor to the fet driver should have a 1k ohm resistor in it in case the fet driver melts down like it did. I also learned this with motor control hardware. The versions with the resistor withstood much more abuse than the versions that had no resistor. This was particularly true of the self balancing scooter project. When I discovered that the fet driver was very hot, I measured the voltage on the enable line and it was 3.5vdc, coming from the fet driver! It had shorted internally and was pushing voltage out it’s input, to the processors output pin. The normal signal is only an enable line, so it’s very low current anyway. I also saw some noise on this same line with my scope. A little more isolation for the processor is a good idea I’ve learned from projects past.

I also ordered some 45v schottky’s for the protection board that Fritz designed. He calls out 100v parts and then says it provides 60v limit. So I ordered 45v parts at $1 each.

I painted the leaking light sources on the lcd tonight. A few more areas and the stray light will not show through the bezel any more. This cleans up the look of the lcd behind the smoked bezel. I just used Tester’s flat black enamel.

Using some 6ga silicone wire from my combat robotic days, and some nice electrical connectors from Home Depot, these probes came out great. This set is 23.5″ long. I have enough wire to make another set at 40″ long.

I finally decided on taping the 1/4″ copper probes with an 8-32 thread and screwing the connector to it with a brass screw. This makes the probes universal so I can change them out for a different configuration of probes. The connector is plated, solid copper. So these connections should not be a bottleneck for the high amperage current flow.

Here are some practice spot welds. The first couple blew holes in the .010″ thick nickel strip. I dialed down the power and time. With more practice I could make strong welds what took pliers to tear apart, and only then the material would fail, not the weld. Then I started to see a vertical trail of smoke so fine, I thought it was my vision was messing with me. It was real smoke coming from the fet driver. It and the voltage regulator were blazing hot. I grabbed my IR temperature probe and the driver read over 400F. The input to the fet driver was at 3.4vdc. It should be at zero. The processor output was working, but not putting out any voltage. My DVM confirmed that the fet driver was shorted internally. So I ordered more drivers from Newark. Probably won’t get here until Monday. There is a fellow here in San Jose that made one of these welders too. Maybe I can borrow a driver from him.

Next Page »