DC choke design for an arc welder

[jul059]

Hello,

I am thinking about converting an old AC arc welder to DC. These welders are very simple: A very hefty transformer with the primary winding on 240v and the secondary around 70-80 OCV (open-circuit voltage). when the arc is struck, voltage drops to ~25-30 V. The current is coarsely controlled in various different ways depending on the design: multi-tap selection on the secondary or some form of variable coupling between the windings (usually, movable shunts or movable windings). For the welders I'm thinking of modifying, the current can go up to 230A. Arc welding is generally thought of as being "constant" current.

I have attached what the basic DC circuit would look like. The big question is L1. from my preliminary calculations, an inductance in the range of at least a few mH is needed to have a significant smoothing effect on the current. I don't see how that's possible without reaching core saturation or being absolutely enormous. As a comparison, I have attached a picture of the inside of the old Lincoln Idealarc 250, regarded as the best design of the time. I think the choke can be seen in the upper part of the picture. They seem to be made of standard laminated steel, and have ~36-42 turns. This machine is big, so the core cross-sectional area is probably around 15x10cm. Even then, I don't see how it wouldn't saturate below 100A.

In any case, current is never smooth and there is significant ripple. Arc welding is very tolerant in that regard. Still, the smoother the better.

Any ideas how to proceed?

EDIT: I have attached an old schematic for the Idealarc 250

[Siwastaja]

Inductance is not just for smoothing. Due to Paschen's law, open circuit voltage of a welder does not do much anything; it's simply too low to strike an arc. High voltage pulses are needed, way beyond the open-circuit voltage. Shorting the workpieces charges energy into series inductance, which, when the short is released, tries to keep current flowing, generating an "infinite" voltage; exactly like a boost converter. This high voltage pulse starts the arc, and after enough energy going in ionized air, then the low voltage of the transformer is enough to keep the arc going. Some welders implement this series inductance as an explicit component (sometimes adjustable), some as leakage inductance of the transformer.

Saturation itself may not be a problem. The question is, do you still have enough energy stored in that inductance to start the arc, despite saturation.

To reduce ripple, add capacitance to the DC bus, before the inductor discussed.

[jul059]

Good point.

Do you know why most designs don't include sizable capacitors? I have seen old MIG welders with capacitors of 80 000uF, but not arc welders. I've even heard of people attempting the same conversion and having to remove the capacitors they had installed because the welder was unusable. Perhaps the circuit needs to be tuned according to the inductance of the choke?

[Terry Bites]

L1 is used to charge a cap that usually sits across the load.
That limits the available current to recharge the cap and this arrangent is not for smoothing per se.
L1 might be rated at 25mH @50A with a 0.1F capacitor bank.
Assuming the cap is charged, before the arc is struck there is current flowing to charge L1 so no energy is stored in it.
Its the capacitor energy that provides the energy for the strike.

Stick welding died out with the dinosaurs.

[T3sl4co1l]

Design of a choke-input rectifier requires the minimum load current desired for continuous output current.  For welding, that will be the least current for which you want smooth DC.  Which should maybe be lower than usual, in case you want to run at lower currents, or have more leeway swinging around the stinger (more inductance means more time at higher distance as you adjust the arc length, or shift position or whatever, before it extinguishes).  For say 50A, at 120Hz, that would be around 265uH.

You can make a "swinging choke" by air-gapping the choke in a distributed manner: a section with low airgap to give high inductance up to some margin, then the remainder of the gap of normal length.  The close-spaced section saturates first, turning it into effective airgap*, and inductance decreases to a 2nd nominal value, until current goes high enough that the whole core becomes saturated.

*Albeit still loaded with metal, and the metal doesn't stop being a conductor; so the fringing field in this area will increase losses somewhat.  This is really just for point of interest; it's irrelevant for a welder -- you have tons of losses everywhere, and a couple, or maybe dozens, of watts of core loss around the airgap isn't going to affect your operating duty cycle.

Alternately, you could make the choke with a variable airgap, so you can shrink it to increase inductance for low-current operation, or vice versa.

A microwave oven transformer might make a good start here.  A typical old/large one is around 2500 mm^2 cross section, and ~1500 mm^2 winding area.

To handle 100A at 265uH, you need to store 1.3J.  1.2T is 0.57 mJ/mm^3 so you need 2270 mm^3; at 2500 mm^2 area that's a spacing of 0.91mm, not bad.  This gives AL = 3.2 uH/t^2, so N >= 9.

Let's see if the current rating is reasonable.  At reduced duty cycle, current density of 10 A/mm^2 is a reasonable starting point.  9t * 100A = 900At is required, and fits into 1500 mm^2 maximum or 0.6 A/mm^2.  I mean, that'll be lower due to winding factor (if you're using welding cable looped around there, you'll lose a lot to the bulky insulation), but that sounds very chill.  Maybe an MOT is an overestimate -- or you can get even more inductance (more turns and air gap) from one, or use a smaller one probably more typical of compact/modern-ish MOTs.

Or I've made a gross mistake, which, is entirely possible.

L1 is used to charge a cap that usually sits across the load.
That limits the available current to recharge the cap and this arrangent is not for smoothing per se.
L1 might be rated at 25mH @50A with a 0.1F capacitor bank.
Assuming the cap is charged, before the arc is struck there is current flowing to charge L1 so no energy is stored in it.
Its the capacitor energy that provides the energy for the strike.

Stick welding died out with the dinosaurs.

A large enough capacitor makes a relaxation oscillator (spark discharge, recharges thru inductor, repeat).  Strike energy isn't needed, at least if your scratch technique isn't awful.  Stick welding remains a useful, and highly affordable, process.

HF start can also be added; it's basically a spark gap Tesla coil without the long resonant secondary.

Such a power supply could be used for TIG as well, if one is so inclined.  The lack of ready adjustment will be, probably rather annoying to use though.

But with DC output, there's potential to add a switching power stage to make it adjustable, implement pulse modes, etc.  A lot easier than making a whole-ass offline supply (if still a challenge to build from scratch!).

---

As for the rectifier, if you don't have one yet -- you can get diodes that big, and that's the preferred solution, but you may find it attractive to build from smaller parts instead.  You can't simply wire these in parallel, but you can parallel them with ballast resistors.  Put one resistor in series with each AC terminal of each rectifier, and size them to give, say, at least 0.2V drop at rated current.  Also, put them all on the same heatsink, preferably a very thick one (or an even chunkier aluminum plate, with heatsinks bolted to that, all joints greased with thermal paste of course).  So that would be, say 0.4V at 35A is pretty much 0.01Ω, (35A)^2 * (0.01Ω) = 16W so choose 20-50W rated resistors*, and, for say a 200A rating, six GBPC3502 (or similar) in parallel should suffice.

*You may find aluminum-case resistors attractive here, but mind they must be heatsinked.  The bother of putting everything on such an even massiver heatsink may prove less worthwhile; I would prefer metal-link or wirewound style resistors, personally.  You can also make your own from thick nichrome wire, if you have the means to join it (i.e. crimping).  Or ribbon, for that matter... (Stainless steel is also acceptable, having not much more tempco, not much different resistance, from nichrome.  Neither one you want to operate anywhere near glowing temperatures, for safety, anyway, so the stainless-ness of it, or the maximum operating temp limit, doesn't matter here, both are plenty good enough.  Stainless can be joined by brazing (use appropriate flux) or bolting.)

Tim

[coppercone2]

eh I am pretty sure I read the real chokes measure some where in 1-10 mH on forums some where before, for DC sticks.

i assume they are extremely heavy.

I ASSUME what you get with that old washing machine sized pipe line welder is... enough inductance for a really smooth DC. The smaller units have more ripple. I think that people usually make comments about the higher quality industrial units being 'smooth'.

I think that basically not all DC transformer units are equal.

And what they like to do is use DC generators, or 3 phase, to reduce the ripple and filtering requirements.

[jul059]

Design of a choke-input rectifier requires the minimum load current desired for continuous output current.  For welding, that will be the least current for which you want smooth DC.  Which should maybe be lower than usual, in case you want to run at lower currents, or have more leeway swinging around the stinger (more inductance means more time at higher distance as you adjust the arc length, or shift position or whatever, before it extinguishes).  For say 50A, at 120Hz, that would be around 265uH.

You can make a "swinging choke" by air-gapping the choke in a distributed manner: a section with low airgap to give high inductance up to some margin, then the remainder of the gap of normal length.  The close-spaced section saturates first, turning it into effective airgap*, and inductance decreases to a 2nd nominal value, until current goes high enough that the whole core becomes saturated.

*Albeit still loaded with metal, and the metal doesn't stop being a conductor; so the fringing field in this area will increase losses somewhat.  This is really just for point of interest; it's irrelevant for a welder -- you have tons of losses everywhere, and a couple, or maybe dozens, of watts of core loss around the airgap isn't going to affect your operating duty cycle.

Alternately, you could make the choke with a variable airgap, so you can shrink it to increase inductance for low-current operation, or vice versa.

A microwave oven transformer might make a good start here.  A typical old/large one is around 2500 mm^2 cross section, and ~1500 mm^2 winding area.

To handle 100A at 265uH, you need to store 1.3J.  1.2T is 0.57 mJ/mm^3 so you need 2270 mm^3; at 2500 mm^2 area that's a spacing of 0.91mm, not bad.  This gives AL = 3.2 uH/t^2, so N >= 9.

Let's see if the current rating is reasonable.  At reduced duty cycle, current density of 10 A/mm^2 is a reasonable starting point.  9t * 100A = 900At is required, and fits into 1500 mm^2 maximum or 0.6 A/mm^2.  I mean, that'll be lower due to winding factor (if you're using welding cable looped around there, you'll lose a lot to the bulky insulation), but that sounds very chill.  Maybe an MOT is an overestimate -- or you can get even more inductance (more turns and air gap) from one, or use a smaller one probably more typical of compact/modern-ish MOTs.

Or I've made a gross mistake, which, is entirely possible.

L1 is used to charge a cap that usually sits across the load.
That limits the available current to recharge the cap and this arrangent is not for smoothing per se.
L1 might be rated at 25mH @50A with a 0.1F capacitor bank.
Assuming the cap is charged, before the arc is struck there is current flowing to charge L1 so no energy is stored in it.
Its the capacitor energy that provides the energy for the strike.

Stick welding died out with the dinosaurs.

A large enough capacitor makes a relaxation oscillator (spark discharge, recharges thru inductor, repeat).  Strike energy isn't needed, at least if your scratch technique isn't awful.  Stick welding remains a useful, and highly affordable, process.

HF start can also be added; it's basically a spark gap Tesla coil without the long resonant secondary.

Such a power supply could be used for TIG as well, if one is so inclined.  The lack of ready adjustment will be, probably rather annoying to use though.

But with DC output, there's potential to add a switching power stage to make it adjustable, implement pulse modes, etc.  A lot easier than making a whole-ass offline supply (if still a challenge to build from scratch!).

---

As for the rectifier, if you don't have one yet -- you can get diodes that big, and that's the preferred solution, but you may find it attractive to build from smaller parts instead.  You can't simply wire these in parallel, but you can parallel them with ballast resistors.  Put one resistor in series with each AC terminal of each rectifier, and size them to give, say, at least 0.2V drop at rated current.  Also, put them all on the same heatsink, preferably a very thick one (or an even chunkier aluminum plate, with heatsinks bolted to that, all joints greased with thermal paste of course).  So that would be, say 0.4V at 35A is pretty much 0.01Ω, (35A)^2 * (0.01Ω) = 16W so choose 20-50W rated resistors*, and, for say a 200A rating, six GBPC3502 (or similar) in parallel should suffice.

*You may find aluminum-case resistors attractive here, but mind they must be heatsinked.  The bother of putting everything on such an even massiver heatsink may prove less worthwhile; I would prefer metal-link or wirewound style resistors, personally.  You can also make your own from thick nichrome wire, if you have the means to join it (i.e. crimping).  Or ribbon, for that matter... (Stainless steel is also acceptable, having not much more tempco, not much different resistance, from nichrome.  Neither one you want to operate anywhere near glowing temperatures, for safety, anyway, so the stainless-ness of it, or the maximum operating temp limit, doesn't matter here, both are plenty good enough.  Stainless can be joined by brazing (use appropriate flux) or bolting.)

Tim

Thank you for the thoughtful reply.

How did you come up with 265 uH?

Using 2 or even 3 microwave oven transformers in series might be a simple solution then.

For the diodes, I have found Vishay VS-UFB250FA60 for a good price.

[coppercone2]

description of inductor in 200 amp welding machine
"The inductor in a aead200 is mounted on top of the generator right ahead of the fuel tank under the cover. It's wound with flat sheet instead of copper wire. It's about 10 inches square and probably won't fit inside your thunderbolt. "

https://weldingweb.com/vbb/threads/141331-Inductor-to-smooth-out-DC-stick/page2

the comments seem to agree that this will nearly always be a bus bar/ sheet transformer, not wire.

[coppercone2]

I think this is the choke from a DC lincoln 225 welding machine (considered bottom of the barrel I think)

from


the bottom of the unit has the AC transformer with taps (standard, now made with aluminum, rather hot running transformer)

[langwadt]

Good point.

Do you know why most designs don't include sizable capacitors? I have seen old MIG welders with capacitors of 80 000uF, but not arc welders. I've even heard of people attempting the same conversion and having to remove the capacitors they had installed because the welder was unusable. Perhaps the circuit needs to be tuned according to the inductance of the choke?

mig is ~constant voltage, stick/tig is ~constant current

[langwadt]

Stick welding died out with the dinosaurs.

what makes you think that?

[T3sl4co1l]

How did you come up with 265 uH?

For the diodes, I have found Vishay VS-UFB250FA60 for a good price.

Ah good, that will do.

L >= Vout / (6 pi Imin F), and say 30V 50A 120Hz.  Note that for a FWB, the ripple is twice the mains frequency.

More inductance is fine, at least up until it's so large that the time constant becomes large fractional seconds, and the arc will feel... a bit laggy, less responsive, something like that?

Tim

[jul059]

I think this is the choke from a DC lincoln 225 welding machine (considered bottom of the barrel I think)

from


the bottom of the unit has the AC transformer with taps (standard, now made with aluminum, rather hot running transformer)

Thank you for that. I looked at another of this channel's videos () where he restores an idealarc 250. I see the same choke I posted in the first post. it looks like around 38 turns, on a core of around 11000 mm2. Either these chokes are completely saturated, or they have a 8-10mm air gap. It gives me an inductance of 2-3 mH.

I'm using a Bmax of 1.3T and relative permeability of 2000. Does this makes sense for laminated steel?

[Benta]

Stick welding died out with the dinosaurs.

what makes you think that?

I wondered about that myself.

[jul059]

How did you come up with 265 uH?

For the diodes, I have found Vishay VS-UFB250FA60 for a good price.

Ah good, that will do.

L >= Vout / (6 pi Imin F), and say 30V 50A 120Hz.  Note that for a FWB, the ripple is twice the mains frequency.

More inductance is fine, at least up until it's so large that the time constant becomes large fractional seconds, and the arc will feel... a bit laggy, less responsive, something like that?

Tim

Thank you, I have just read a bit more on choke input here: http://www.r-type.org/articles/art-144.htm

I kept thinking that adding a few capacitors before the choke would be a good idea...
It is still unclear to my why core saturation doesn't matter much above the minimal chosen current.

[coppercone2]

well that is your % ripple

[coppercone2]

https://groups.google.com/g/rec.crafts.metalworking/c/d8tRl0HT6YA?pli=1

also some patent says that inductance should be inversely proportional to current for arc welding. So just having a big choke on the output is already considered a 'budget move'. So for whatever reason, a good machine probobly has a adjustment for the output choke? or there should be a ganged switch between the transformer taps and the inductor taps

MIG Have the ability to adjust inductance (i guess simulate). It has something to do with the V/A curve slope. Metal transfer mode regieme. Don't know if it applies to stick welding. Can you spray stick weld?

This setting of inductance on STICK welding might be called DIG but I am not sure. I don't like stick welding because its so fucking messy. I think its a tool of the desperate, and for economic implementation of very large scale projects. Nothing wrong with that but no use for me personally IMO, at least not for fabrication/prototyping.

But anyway, I think this project to make a DC stick welder is foolish, people like stick welders because they are considered fool proof and trust worthy. But there is still things to know about the rods etc. Once you start getting custom welding equipment, combined with the flux/plasma chemistry/dynamics, I feel like selecting rods and knowing if you work is worth a damn its gonna be a question.

Unless you got a real stick welder that has the control, and you can compare the oscilloscope graphs and VI curves to see if they are comparable, you will have a untrustworthy machine. Just putting this out there incase you are thinking about repairing farm equipment or something important.

if dig and inductance are the same control paramter for different processes, then
https://www.reddit.com/r/Welding/comments/htiiha/im_confused_about_what_the_arc_force_and_never/

So I understand it this way, maybe

1) the inductance of the reactor is generally set to put the welder in a stable DC operating REGION
2) within this REGION the inductance can be trimmed in order to effect what is called the DIG parameter (or arc force, or whatever.. in different generations of machines, which will not actually adjust a inductor but simulate one with the control feedback loop).

So you go into the ball park AND then you can play ball.

So a welding machine with a non adjustable reactor on the output puts you in the DC welding region, at some random 'allowable inductance' point that may or may not be optimzied for your process. If you pay extra you get to adjust where you are in this region by the bounds of the machine controls so you can optimize your arc.

Is this right? and lol usually the problem is if your not a professional welder you don't have enough feed stock and time to actually figure out what is optimum.

I think the idea is that if there is ripple, each ripple has the chance of (higher power) to break a short circuit of the rod. So is this DIG parameter really a RIPPLE ADJUST for a DC  process?

[T3sl4co1l]

I kept thinking that adding a few capacitors before the choke would be a good idea...
It is still unclear to my why core saturation doesn't matter much above the minimal chosen current.

Perhaps I glossed over that too quickly: saturation should include peak operating current.  I picked 100A as a typical example, but as it seems the example core will handle plenty more, you should get 200A just fine with suitably sized wire (or rectangle/sheet), adjusting air gap and turns to accommodate the higher energy requirement.

Which would be... well 2x current is 4x energy or 5.2J, at same Bmax is 4x the gap or 3.6mm, a quite reasonable gap for something that size.  It'll be a little wider than that even, actually, because fringing is beginning to be noticeable at this point.  Which also reduces |B| in the gap area, allowing a little bit more amp-turns before total saturation -- nice.

1.2T is probably an underestimate as well, albeit a safe one.  1.5, even 1.8T might be typical for MOT cores, granted they normally run deep saturation at the peaks, but so can you; and you don't really care if the inductance saturates on the peaks, it just means less energy stored on the backswing i.e. the minimum current through the arc (during a cycle) doesn't increase as DC current increases, but that's still fine because you're more concerned with the arc staying ignited than it being smoothed out substantially (but that is of course an added benefit, and makes a larger inductor desirable).

Also a reminder that, if you're shimming an E-core, the shim counts for double -- there's one gap around the outer legs and another to the center leg.  So a 1-2mm spacer would do ably for that.

Probably up to 3-4mm spacer thickness is where winding losses start to get unwieldy (needing too many turns of too-thin wire for given current), so somewhere around here should be fine.  Or ~6mm if you want to mill down the center leg.

Tim

[jul059]

Yes that makes sense.

I think the simplest solution for the air gap is to separate the E core at the weld, insert a separator of constant thickness across the 3 legs, and put everything back together (epoxy maybe? compressed with a clamp until completely hard).

Maybe I could create a kind of equivalent stepped choke by winding 2-3 different cores with different gaps and simply wiring them in series, although the advantage of this doesn't seem obvious when the cores can be had for free.

[David Hess]

The big old AC stick welders I have seen have a movable section in the core between the windings so that the leakage inductance can be adjusted to control the current output, so the transformer core provides the inductance.  Converting them to DC stick welding just requires adding a big bridge rectifier to the output.

Long ago some friends of my father had a welding shop and wanted to do bigger projects like truck frames, so they somehow got a pole distribution transformer.  They took it apart, unwound the secondary, doubled the secondary up a bunch of times, and then rewound it.  The first time they touched it off, the welding rod just exploded because there was no series inductance.

[T3sl4co1l]

Yes that makes sense.

I think the simplest solution for the air gap is to separate the E core at the weld, insert a separator of constant thickness across the 3 legs, and put everything back together (epoxy maybe? compressed with a clamp until completely hard).

Maybe I could create a kind of equivalent stepped choke by winding 2-3 different cores with different gaps and simply wiring them in series, although the advantage of this doesn't seem obvious when the cores can be had for free.

Yup. You can even weld it back together if you like-- just not all the way across, make a couple tack welds instead, just enough to hold it in place without shorting things out.  Otherwise, epoxy is fine, and phenolic board, fiberglass (various kinds), etc. are fine spacer materials.  Even wood, or hard cardboard (if you don't mind that it's Class A insulation).

Bonus, the winding window increases with airgap, assuming you use cut bits of spacer (the dimensions of each leg face).

You can do cores "in series" all in one stack if you like -- get two or three cores of identical size, stack them up to make one much taller stack of laminations, and set the 'I' pieces at whatever gaps you like.  You could make one short that saturates quickly, others with more.  Like for two cores about this size, 1mm and 4mm might be an option, and just wind as much wire as you can fit.

Stacked cores saves the resistance of the "turnaround" parts of the winding -- instead of one turn going along the core stack, bending around to the other side, going back up the stack, and across, you have two goings-along-the-core-stacks in a row with no bend-around-and-back.  Overall resistance is higher than if the core itself were larger, but if you just don't have a larger core, or you just don't care too much about losses, yep, nothing wrong with that.

Tim

[johansen]

I have a tombstone style 225 amp ac only Lincoln welder. The kind with aluminum foil transformer, and an E core inductor welded on the the side of the transformer, (which uses the side of the transformer as the inductor return path*) with maybe 8 more taps on it to generate the lower 40-60-75-90-105-120-135 amp range. From the lower end of the 150-175-195-205-225 range (which does not use  the inductor

For a rectifier i use the 4 aux diodes from 4x 100amp rated 600 volt darlington transformers. They have survived years at questionable duty cycles at 150 amps and higher maybe someday i will make a pulse generator to make it start an arc.

I made an inductor from 4 identical E cores from mots. The I part of the core was discarded, two cores face each other so the cross sectional area is like 3cm by 12cm The coils is just 8 or 9 turns of 1/0 aluminum wire.. (60hz saturation would be 2 volts per turn) but the dc ripple is at 120hz. So in theory the inductor can block 32 volts ac ripple at 120 hz
 
Air gap is like 1mm. It works pretty well even as low as 40 amps

Above 75amps the current is pretty continuous.

[coppercone2]

do you find it welds comparably to a real machine with the same electrodes at the same settings?

[jul059]

I have a tombstone style 225 amp ac only Lincoln welder. The kind with aluminum foil transformer, and an E core inductor welded on the the side of the transformer, (which uses the side of the transformer as the inductor return path*) with maybe 8 more taps on it to generate the lower 40-60-75-90-105-120-135 amp range. From the lower end of the 150-175-195-205-225 range (which does not use  the inductor

For a rectifier i use the 4 aux diodes from 4x 100amp rated 600 volt darlington transformers. They have survived years at questionable duty cycles at 150 amps and higher maybe someday i will make a pulse generator to make it start an arc.

I made an inductor from 4 identical E cores from mots. The I part of the core was discarded, two cores face each other so the cross sectional area is like 3cm by 12cm The coils is just 8 or 9 turns of 1/0 aluminum wire.. (60hz saturation would be 2 volts per turn) but the dc ripple is at 120hz. So in theory the inductor can block 32 volts ac ripple at 120 hz
 
Air gap is like 1mm. It works pretty well even as low as 40 amps

Above 75amps the current is pretty continuous.

That's an interesting design for the transformer, never seen it. That's a later model I guess?

So you're basically creating 2 E-E cores with 4 MOTs? Have you by chance tried welding aluminum?

Does the choke get very hot? I could get a good deal on 1/0 aluminum wire, but the insulator is soft PVC, rated for 176 deg F.

Have you measured the maximum current it can output? Surely below 225ADC.

[jul059]

Yes that makes sense.

I think the simplest solution for the air gap is to separate the E core at the weld, insert a separator of constant thickness across the 3 legs, and put everything back together (epoxy maybe? compressed with a clamp until completely hard).

Maybe I could create a kind of equivalent stepped choke by winding 2-3 different cores with different gaps and simply wiring them in series, although the advantage of this doesn't seem obvious when the cores can be had for free.

Yup. You can even weld it back together if you like-- just not all the way across, make a couple tack welds instead, just enough to hold it in place without shorting things out.  Otherwise, epoxy is fine, and phenolic board, fiberglass (various kinds), etc. are fine spacer materials.  Even wood, or hard cardboard (if you don't mind that it's Class A insulation).

Bonus, the winding window increases with airgap, assuming you use cut bits of spacer (the dimensions of each leg face).

You can do cores "in series" all in one stack if you like -- get two or three cores of identical size, stack them up to make one much taller stack of laminations, and set the 'I' pieces at whatever gaps you like.  You could make one short that saturates quickly, others with more.  Like for two cores about this size, 1mm and 4mm might be an option, and just wind as much wire as you can fit.

Stacked cores saves the resistance of the "turnaround" parts of the winding -- instead of one turn going along the core stack, bending around to the other side, going back up the stack, and across, you have two goings-along-the-core-stacks in a row with no bend-around-and-back.  Overall resistance is higher than if the core itself were larger, but if you just don't have a larger core, or you just don't care too much about losses, yep, nothing wrong with that.

Tim

You would stack them so that the Es are all superimposed on one another? And then wind around all cores in a single turn of the wire? So that essentially makes one bigger core with 3x the cross sectional area?

What would be the advantage of a stepped design? Is it to have better filtering at low currents, making sure the arc never extinguishes? For the same amount of core steel, there seem to be a tradeoff between more inductance at low current and worse inductance at high current (stepped) vs a constant, lower inductance (fixed, large gap preventing saturation).

Also, isn't it possible to wind around the 2 side legs instead of the central one, allowing essentially 2x more turns? Or is this inefficient? Of course I need to consider saturation, but it might be useful if I use large aluminum wire for example.

[soldar]

I have attached what the basic DC circuit would look like. The big question is L1. from my preliminary calculations, an inductance in the range of at least a few mH is needed to have a significant smoothing effect on the current. I don't see how that's possible without reaching core saturation or being absolutely enormous.

You are fundamentally misunderstanding how that works (or I am misunderstanding you).

That circuit works essentially the same as a fluorescent lamp with magnetic ballast choke. The purpose of the choke is to (1) establish the arc and (2) limit the current once the arc is established and resistance falls way down. The choke always sees AC and it does not notice or care if you put a rectifier after it, it still sees AC.

That circuit is very old. The next step was the machines that incorporated transformer and choke into a single transformer with adjustable flux link between primary and secondary with an adjustable magnetic shunt so that the value of the choke could be varied.

But it is all in AC.

[jul059]

I have attached what the basic DC circuit would look like. The big question is L1. from my preliminary calculations, an inductance in the range of at least a few mH is needed to have a significant smoothing effect on the current. I don't see how that's possible without reaching core saturation or being absolutely enormous.

You are fundamentally misunderstanding how that works (or I am misunderstanding you).

That circuit works essentially the same as a fluorescent lamp with magnetic ballast choke. The purpose of the choke is to (1) establish the arc and (2) limit the current once the arc is established and resistance falls way down. The choke always sees AC and it does not notice or care if you put a rectifier after it, it still sees AC.

That circuit is very old. The next step was the machines that incorporated transformer and choke into a single transformer with adjustable flux link between primary and secondary with an adjustable magnetic shunt so that the value of the choke could be varied.

But it is all in AC.

I'm not quite sure we are talking about the same thing. You are describing the adjustable transformer in an AC arc welding machine. There are AC only machines, and AC / DC machines. I'm talking about rectifying that AC output, and then installing a choke after the bridge, not before. And I agree I might have misunderstood it's function in my original post.

[T3sl4co1l]

You would stack them so that the Es are all superimposed on one another? And then wind around all cores in a single turn of the wire? So that essentially makes one bigger core with 3x the cross sectional area?

Exactly.

Imagine a wire looping around one core.  It goes in one side, hooks around, and out the other, making a 'U' shape around the 'E' middle leg.

Extend the wire and loop it around another core: you have two cores acting in series, one turn each.

Bring the cores closer.  Turn them so that the wire 'U's are facing, and snug up the wire.  You have a pair of leads (ends) close together, and a nearly circular loop enclosing both cores.  Thread the pair of ends through one core (to add and remove, respectively, a "half turn" around one core).  Now there is nothing between the cores and they can be butted together.

Cores in contact, but otherwise extending their own symmetry (in this case the stack of sheets), makes essentially no difference to the fields, which are 99.9..% within the core itself, and the only major external difference is in the space around the airgap.  We're adjusting the airgap with shims anyway, so we don't care that the required air gap may change in the process.

The cores can also be placed "in series" (two 'E's face-to-face, making a figure-8 split in half), which does change the field structure (the magnetic path length is longer, and the airgap is symmetrical rather than facing a flat 'I' piece to one side), but not by very much (for similar reasons) and the basic effect is to double the winding area, instead of the core area.

What would be the advantage of a stepped design? Is it to have better filtering at low currents, making sure the arc never extinguishes? For the same amount of core steel, there seem to be a tradeoff between more inductance at low current and worse inductance at high current (stepped) vs a constant, lower inductance (fixed, large gap preventing saturation).

I wouldn't bother making a swinging choke, personally, and just make it good enough to handle the full desired current range as-is, which clearly is easy enough to do; more to say, it's an easy option, simply by using different airgaps on sections of the core.

Also, isn't it possible to wind around the 2 side legs instead of the central one, allowing essentially 2x more turns? Or is this inefficient? Of course I need to consider saturation, but it might be useful if I use large aluminum wire for example.

How exactly would you double the turns with the center peg still wound?

Oh, or you mean wind just the outers?  But each winding encloses half the core area. So you're doing the same thing but you've increased wire length (more lateral/bent segments, aye?).

Between the higher resistance, and the thick jacket, you may find the figure-8 arrangement more beneficial for aluminum: more winding area means more room for the wire.  The general downside of this arrangement is, the core path is longer, i.e. core volume (and thus losses, heating) is larger for a given capacity.  It's not by a large amount, give or take 2x say (for stacking vs. figure-8-ing two cores), again, losses hardly matter here.

Tim

[johansen]

I may have used 7 turns of 4 awg copper. Air gap is about 1mm actual. So 2mm  real.

I have welded at 40 and 60 amps using a spool of 200 feet of 10 gauge wire as an inductor. Its too much. The arc wanders around on you.

[DavidAlfa]

T3sl4co1l, do you work with inductors all the time  right?
Anything I've not used for the last 5-10 years is simply gone!
Amazing writeup there!

[T3sl4co1l]

I combat that by remembering connections.  Magnetics (with a well defined core and all that) is just proportions, dimensional analysis with unity coefficients.  Forgot how to go from core dimensions to inductivity?  Magnetic constant is in H/m, and that's a volumetric "per length" like resistance so multiply by area and divide by length.  Remember it's air gap equivalent with mu, or use mu_eff (look up a formula if you need a reminder), or add up the reluctance of core and gap, or other more complicated geometries if they're still amenable to simple magnetic circuit analysis.

Tim

[jul059]

Thank you all for your input, and especially you Tim.

I will slowly gather MOTs, try to fit this project in my schedule somehow someday, and post back here when I have some developments.

[coppercone2]

I strongly suggest if you put that much effort and time into it to just get a broken TIG welder and fix it. They are hard but it does stick too usually.

I say this because you will learn much more 'electronics' things. I just can't shake the feeling that your result will be dodgy. Depending on how broke it is you might get a really good deal.

i also think this transformer you are making is gonna shake like a mother

And I am still really curious about anyone who did this mod, how it compares to a commercial DC transformer arc welder. If you have one, I would love to see a cross section acid etch of the weld made with the same rod at same settings on same material. (slice polish and dab with pcb etchant)

and beware of turning this into a TIG, I just watched a youtube video that is showing if you plug a tig torch into a DC arc welder the impedance is too low and it consumes too much reactive power or some shit that makes it require PFC to stay sane

[jul059]

I strongly suggest if you put that much effort and time into it to just get a broken TIG welder and fix it. They are hard but it does stick too usually.

I say this because you will learn much more 'electronics' things. I just can't shake the feeling that your result will be dodgy. Depending on how broke it is you might get a really good deal.

i also think this transformer you are making is gonna shake like a mother

And I am still really curious about anyone who did this mod, how it compares to a commercial DC transformer arc welder. If you have one, I would love to see a cross section acid etch of the weld made with the same rod at same settings on same material. (slice polish and dab with pcb etchant)

and beware of turning this into a TIG, I just watched a youtube video that is showing if you plug a tig torch into a DC arc welder the impedance is too low and it consumes too much reactive power or some shit that makes it require PFC to stay sane

Thank you for your concern!

The way I choose and go forward with my projects (very few lately I must admit) is based on my interest, and an idea that sparked for some unknown reason. The goal is never to learn the most electronics (or woodworking, or cooking skills, or programming), that would be quite boring and akin to going to school. The reason why I get involved in a particular idea instead of a million other alternatives will remain a mystery, but that's part of why I love those projects and it's what makes them feel like play.

[T3sl4co1l]

I strongly suggest if you put that much effort and time into it to just get a broken TIG welder and fix it. They are hard but it does stick too usually.

I say this because you will learn much more 'electronics' things. I just can't shake the feeling that your result will be dodgy. Depending on how broke it is you might get a really good deal.

Pffbt, you're just mad it doesn't incorporate heavy mu-metal plates or something.

Tim

[coppercone2]

more like I am concerned a DIY trailer filled with practice scrap metal is gonna detatch itself on the high way, or a motorcycle trailer

I think you really should want to see comparitive data between a modified machine and a certified one, considering the heft that MMA is typically associated with being used for. With a MIG welder you assume a chicken coup roof might flood, with a stick its usually something heavy. And it goes for the ten other people that read this and never post.

I hope someone posts a proofing test, not just 'its buttery smooth and looks great'

I started getting a little nervous when your initial calculation said something like 200uH based on calculations, then the guy comes up with multiple mH (which agrees with hearsay, but who knows how this random inductor data was gathered by people who I take it are not familiar with impedance measurement). Factor of 10 disagreement. IDK whats correct but that makes it seems like  its a guessing game and that there might be some missing info. It has to work in spec with the rod flux chemistry. And whatever trades person concensus is that the weld quality is some how poor if insufficiently filtered. Metrics for this poorness? no clue.

Maybe at least someone can take a inductor out of a machine, clean it up, post picutres from numerous angles. Cuz whats going on right now is reverse engineering based on grainy photographs. area51 bull shit. Extremely amusing but you know someone is gonna assume this works for 600 pounds because there was some napkin math done with no verification.

[T3sl4co1l]

Bruh, weld quality is a separate issue. You can do it with AC, doing it with pulsating DC isn't going to be any different... give or take technique.  Weld quality is controlled by preparation, technique and inspection.  It might be easier to do with a machine of certain capability, that's all.

Tim

[jul059]

An arc welder can, at it's core, be an extremely simple device. The Empire State building was built with (most likely) AC arc welding in 1931. A simple rectified AC welder without any additional choke is basically an alternating current at twice the frequency (120 Hz vs 60 Hz) with a superimposed DC component. There is no reasonable reason why that would affect weld quality vs pure AC. People have simply rectified an AC machine and have been able to weld properly, without a choke (although only with AC-compatible electrodes). The arc will behave differently, the arc strike will be different, you will have to weld differently. But the device itself will have no impact on weld quality. That seems really pretty obvious.

A welder who can't tell if his welds are any good should not be welding anything mildly important until he can.

The addition of a choke is part of an optimization process: to make welding easier, more enjoyable, perhaps with less spatter, to be able to consistently use DC-only electrodes.

In any case, I think that's slightly off-topic!

Cheers

[johansen]

a friend bought a miller shop master 300. It has 2 transformer coils. One for  80v open circuit stick welding and the other for 30vdc mig.

If anyone is interested i can take the cover off and photograph it

[jul059]

Thanks for the offer! I think I have enough information, seeing as I will likely have to build it out of MOTs since I don't have access to anything else cheap.

I came back to this picture and I think winding this way would be best to fit in more turns, and aluminum sheets should be relatively inexpensive. I'm just wondering how they insulated each turn from one another? It looks like some red paint.



Any major downside to connecting 3 aluminum "rolls" like in the picture in series on the same core instead of winding one layer across, then back, etc. ?