DIY SMAW welder (stick welding)

[RocketScienceSmurf]

I sometimes have a need to weld stuff and I always have to turn to people for help since I don't have the knowledge/skill or equipment for it. The latter can be remedied by getting a welder and what better way than to build one, right 

What I have that may or may not be useful

* A large DC power supply capable of 150A continous @ ~45VDC
* Some film capacitors taken from different welders
* A couple of BIG stud diodes (I think they are rated at 300A)
* Two HUGE chokes, one at 25mH @ 50A and one at 35mH at 50A. They are built on what I believe to be silicon steel sheets so maybe not suitable for high frequency.
* A lot of IXYS IXFN520N075T2 mosfet trench gate modules rated at 480A each

From what I understand a stick welder works best with a constant current power supply so I was thinking that I could build some kind of low frequency PWM controlled constant current curcuit and attach it to my big power supply. Can this be done in a crude manner with the parts I have together with some kind of simple microcontroller to monitor the current and adjust the PWM duty cycle?

[rikkitikkitavi]

You have EXACTLY the same plans as I do, apart from that  I have three 48V(nominal) -51 V effective 3000W Server supplies (teh infamous ESP-120)

My plan was exactly the same- build a constant current step down converter (ie buck converter) , but with analog control. It is only an idea sofar and not highest priority though. So two ideas independandtly producing the same idea, what could be wrong ?

Some design ideas:

When idling - the voltage should be close to 100% and therefore arc striking should be quite easy @50V penetrating rust and impurities of the work piece.

As for the output choke I would parallell a number of iron core toroids  alvaged from old PC power supplies or purchased (tips, se på Tradera där det finns en säljare som säljer 20 pack, pma för detaljer) of operating above the audible frequency regarding swithcing frequency. I have had the pleasure of standing close to a frequency converter operating within the audible range in which parts resonated. High pitch wining has a new meaning... (120 dB @ 8-10kHz) so I doubt that is feasible. The output capacitor has to be huge too.
 
As for the caps, parallelling a huge number of electrolytes is the way to go to achieve sufficient ripple current. I recently sold of 2 pcs of 25V 470 mF electrolytes as they had a ripple current of only some 20Amps each. Parallelling simple 4700 uF caps will give a ripple current capacity in the hundreds of amps, and that is with natural cooling. As for how to parallell them, MMC for Tesla coil design is probably a good start.

Perhaps your Mosfets have a little to low voltage rating in case of parasitic inductances causes overvoltage surges. That would care for some careful physical design and perhaps fast overvoltage protection. Again Tesla coil deisng is good input.

Parallelling smaller semiconductors has the advantage of heat spreading but the disadvantage of increasing wire length.

For simplicity I would design with a dumb diode output and not a synchronous rectifier solution. Beefy diodes necessary though.

Sensing output current- it has to be fairly accurate (within 5% ) and fast, especially when it comes to the welding stick shortening to teh work piece and getting stuck kind of problem.

The supply must be protected from transients, and adding a hughe output capacitance (storage capacitor at the input for the current converter ) would cause startup and stability problems, therefor some decoupling is necessary.

Just a few ideas...

PS if wanted I have four 600Amp , 200V hockey puck diodes.

[RocketScienceSmurf]

I have thought about replacing the choke in my power supply which is the original ESAB choke that I have measured to about 1,8mH with the 25mH and perhaps use the smaller one to the actual welder. I think that will give less transients back to the mosfet modules.

I thought a lot about the current sensing yesterday and although I have a few shunts from the welders I have scrapped I feel that ~50mV is too low voltage. I would like closer to 2,5-3V @100A but the heat losses would be a pain. Then I thought about using copper brake lines (the same I use for the induction heater) and circulate water through it I haven't figured out a suitable dimension and length yet though.

Since the chokes I have available are not suitable for HF my plan was to use 50-400Hz PWM frequency.

When it comes to caps I could perhaps use one or more of the film caps I bought for my power supply but I really like them where they are. Wouldn't the required capacitance be dependent a lot on switching frequency? Higher frequency cold perhaps get away with fairly low capacitance?

[rikkitikkitavi]

Let's begin from the back...
 A higher frequency means lower capacitance.

Higher capacitance means lower ripple, but for large supplies its is permissible ripple current and ESR (causing ripple voltage and internal heating) that is most sizing important.

Larger inductance also means lower ripple current at the expense of more loss, more money and slower response due to the LC output filter formed by both.

Compare to an AC welder where current regulation I perforde by some leakage or series inductance of large dimensions, and line freuqency. Frequency response of an welder is not an issue, it is just dumb steel being welded

Clever protection of switching semis is important though.

I have looked at Mads teardown of a welder for layout inspiration. There are more such teardowns on the Web.

[RocketScienceSmurf]

Larger inductance also means lower ripple current at the expense of more loss, more money and slower response due to the LC output filter formed by both.

I believe I have enough "juice" to put in to compensate for the losses. And you are correct about the cost but since I have two VERY big chokes just laying around in the workshop I thought I might as well put to good use.

[rikkitikkitavi]

Look on my Choke ye Mighty, and despair!

I dont think that the loss would impair the welder such that the output would be insufficient, but the losses can be to much for the acutal coil to handle without overheating.

The coil you showed will not work above a few hundred Hz. That is for 50/60Hz filtering or wereabouts.

[RocketScienceSmurf]

I have thought a lot about what you have written and I think I begin to understand.. A very low frequency circuit probably wont work well and my chokes are, as we both say, suitable for ~50Hz up to perhaps a few hundred Hertz. However they are rated for continuous use at around 50A so as far as welding current is concerned I think they will work well.

Would it be feasible to use linear regulation instead with a ton of NPN-transistors? I realize I need to cool of a ridiculous amount of heat loss in the transistors but I don't regard that as a problem... yet...

[rikkitikkitavi]

you could use linear regulation as a constant current source by all means, but the disadvantage is that you need a higher voltage striking the arc, say 50-60V and a normal arc is about 20-25V @ 50-150Amps. The advantage of a regulated power is that the current will be almost constant regardsless of arc conditions, ie voltage drop.


BOne advantage is that since it is a linear supply the need for careful layout is greatly reduced compared to switching hundreds of amps. Your 480Amps MSOFETS could probably be used then as linear  elements, as tehy dont suffer from secondary breakdown as BJT(NPN)  , and can easily handle 50V.

But just for calculation- 100amps, 50V in, arc voltage 25V gives 2,5kW power loss. It demands a huge heat sink. Or more likely water cooling. Some type of domestic heat pump radiator with water cooling perhaps?