Voltage Booster For Pulse Arc Welder

[backXslash]

Hey guys!

Long time lurker, first time poster.

First, a little back story - I'm a network security engineer turned bench jeweler. I built a capacitive discharge spot welder based on a design by Fritz Geyser. It's an awesome unit, and works great, but it's not exactly what I need it to be. It's a resistance welder, which means that the resistance of the workpiece generates the weld heat. Gold and silver (the metals I work in), are very very conductive to both heat and electricity. I need to convert the unit to a pulse arc welder. For reference look at the Orion welder line by Sunstone Engineering, the PUK line by Lampert, or the Pulse Arc line by ABI.

http://www.orionwelders.com/
http://www.lampert.info/en/products/industry-115/puk-u4
http://www.abiusa.net/Pulse.htm

So, on to what I'm hoping someone can help me with. To convert the unit to pulse arc, all I need to do (in theory) is boost the voltage above the breakdown voltage of air. It has a 3 farad main bank at 25 volts making it's an 800Ws welder when charged to it's safe max. The main bank can be charged to anything from 0 to 25, with 2 output pulses anywhere from 0ms to 35ms each. The breakdown voltage of air is about 370 volts, so I'm thinking I'd need some kind of step up transformer or coil to boost the output from 2 volts to 400, which would make the max output a little over 2600 volts when charged to it's safe max (23.2).

The other option would be some kind of high frequency start. In all honesty, that's the more attractive option to me I think, because I could position the arc where I want it on the piece and then "inject" the weld pulse into the plasma streamer which should nicely create a small weld puddle.

I know that the ABI unit is probably the closest to my intended goal. It's got a main cap bank that's charged from 10 to 90 volts DC, which still puts it UNDER the 370 volt breakdown of air for a plasma arc. Looking at the video though, it DOES arc off a tungsten needle at 45 volts, which means that even IF the operator was using Argon shielding (which he isn't) the breakdown voltage is still above 75 volts. Essentially, there HAS to be some kind of boost stage that I don't have.

Does anyone have any ideas on how to accomplish either one of those? I'd REALLY appreciate some help.

[backXslash]

As a visual example, here's the video I'm referencing by ABI.

WARNING - The video is WMV encoded, and also crap, it MAY prompt you to download it - WARNING

http://www.abiusa.net/SteeceH.wmv

You can clearly see that the arc occurs in air, and he's not using argon (he says so), and he's not using the ceramic shield the weld stylus comes with.

Also, the unit in question uses and analog DC voltmeter to show you the main bank charge, and has a three position switch to select weld energy, which would seem to indicate that it's just a pulse timing selector. Low would be a very short pulse, medium slightly longer, so on and so on.

My unit has digitally controlled outputs, with a uC measuring cap voltage to within 0.01 volts, and digitally controlled output pulse timing to within 1Ms. It uses 18 IRFP2907 75v @ 209A mosfets in parallel to ensure fast and accurate pulse timing. So those are... analogous-ish. I think my only hurdle is arc ignition.

[backXslash]

Shameless self-bump

 

[peter.mitchell]

Would something like this work:

a very simple way is to get a HV,HF transformer, say 2-5kv, low current is fine, then get a large toroidal core, maybe 120 size or so, put two turns of your output through it and then put two turns of the hv transformer through it. this will couple the HV into the output to strike the arc.
you will need some protection on your circuit so you dont cook your caps or whatnot, so for this put some 28 or so volt TVS diodes antiparallel accross the output, these will clamp the voltage seen across your caps.
to drive the transformer a simple 555 and a darlington should do. put some decent series caps on the output to limit the power to the "transformer" [15kv rated x2 capacitors are pretty common and cheap and good for the job] an adjustable spark-gap on the output of the transformer too to limit the voltages produced to what is necessary for safe operation.

Operation:
the cap bank charges up to the desired voltage/charge
you position the tungsten electrode
to tap the "button" you have controlling the HF/HV transformer
the transformer couples energy into the output
the TVS diodes conduct keeping your caps safe and completing the internal part of the circuit
an arc forms between the electrode and the work piece
immediately the caps begin to discharge through the arc and the work piece
you have released the momentary button and the transformer is not producing an arc anymore
the plasma continues to conduct, welding your work peice
the votlage on the caps drops, the plasma dissipates
the arc breaks
repeat

[T3sl4co1l]

Transistors in a spark discharge path seems awfully dubious to me... have you even tested it under peak fault current conditions?  What other components are in the path to limit or control current (inductance, resistance, including ESR of capacitors, stray inductance of all wiring..)?  What's across the transistors to prevent them from exploding when commutating such a current?

Ignition should be pretty reasonable, using inductance to your advantage.  For example, add some extra inductance in series with the output.  Couple the inductance to a primary winding, forming a step-up transformer.  Pulse the primary, either from a modestly rated transistor (transistors over 1kV aren't practical, nor is a large voltage ratio, like spiking to >1kV from a <25V source), or as a resonant circuit, like with an SCR capacitive discharge circuit.  You have to beware that the added inductance doesn't slow down the spark current too much, and that the EMF delivered by it doesn't fry the ignitor.  But that shouldn't be too hard to arrange.

Tim

[tautech]

Let Google be your friend.
https://www.google.co.nz/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#q=welder+HV+arc+start

I used a furnace ignitor transformer and air coupled windings to induce the HF HV into the welder lead, however it was an old-school transformer type welder with zero semiconductors to worry about. Worked OK but mighty scary to have 10KV bussing away across a spark gap.

All the major welder manufacturers used to use a similar HF HV induction method but I'd doubt any do now as the EMI and RF generated was horrific. At a place I worked at decades ago when the TIG HF was turned on, all radios, RT's and TV's were affected for 100 meters.

Interference generated will be your biggest problem, especially if the HF HV is run continuously.

[Richard Head]

I would use a series connected injection transformer and dump a HV capacitor into the primary. If 5kV secondary voltage is adequate then use about a 10:1 turns ratio with say 20t on the secondary and 2t on the primary. In the past I've used  small gas arrestor as the switching element on the primary side. You can get some 5kA units from Siemens in different voltage ratings. They are very robust as switches, far better than semiconductors. To get them to switch you simply increase the voltage across them until they break down thereby dumping the charge on a cap bank into the primary of the step-up transfomer. Just remember that excellent isolation is required between primary and secondary. I've used mylar tape in the past.
V= nAe dB/dt can be used to calculate the output pulse width before the core saturates. I would aim for about 100ns.

Dick

[HAL-42b]

I never understood how they protect the electronics from the HV. Do they just make everything 5kV proof?

[backXslash]

Holy wow! Thank you for all the replies!

I didn't expect it, but I actually had a chance to physically touch and even operate one of the commercially produced models by the Orion people.

After having done so, I've got a MUCH better handle on how they work. The weld sequence goes thusly:

1 - The weld probe makes contact with the grounded workpiece
2 - A HF start circuit engages
3 - The probe retracts, pulling a streamer of ionized gas along with it, similar in principle and function to a Jacob's ladder
4 - The main bank is fired, discharging DC weld energy through the now existing plasma channel, completing the weld

The main board uses a 5v line to control the mosfet board. I'd like to design an interrupt of sorts that will receive the 5v pulses from the main uC and just overlay the HF waveform over the main weld pulse. That seems to me to be the simplest way to get it done.

That means, all I need is an HF source, and some way to couple that source to the weld probe and workpiece and allow the interrupt board to control that source. Does that make any more sense to anyone?

[backXslash]

Is there such a thing as a filter that will allow low voltage (<25 volts) to pass, but stop high voltage from passing?

Some kind of.. inductive filter?

[T3sl4co1l]

With some kind of HF (pulse or CW) start function in series, you want a low impedance from the low voltage source (capacitor), not a high impedance.

Tim

[backXslash]

True. The mosfet board essentially just functions as a fast switching and controllable method for shorting the caps through the workpiece. The impedance of the entire "output" section is as low as I could get it.

I found a high voltage step up transformer that takes 3.6V - 6V and outputs 400KV - 400000V. That solves the high voltage source problem, as there's a dedicated 5v rail in my circuit.

That just leaves the coupling, I.E. how to stop the HF from damaging or interacting with the control or mosfet board. The rest is uC's and relays, right?

[LaserSteve]

Oh, This is not so hard.  I work on three phase ion lasers, with 35-60 amp plasma currents to generate the lasing.
The current is regulated by 18 to 36 NPN transistors (depending on model)  ran common emitter in parallel. There are also some smaller plasma tubes that need 9-10 amps and run off 110 AC. These are regulated by a high side  buck switcher using a few big mosfets as a constant current source. I'll admit that a 1600 watt 160 VDC  constant current buck is not that small a PSU.   

The small ones use a 600V or 1200V 60 amp blocking diode, an inductor wound on a ferrite bar, and a metal can mica return capacitor (600 WVDC, 0.1 uF) to keep the ringing wave from the HV transformer out of the PSU.  The cap provides a LF return path for the igniter, whether the tube fires or not.  These use series injection triggering, and the secondary of the igniter carries the full plasma current.  The small ones have a boost cap ~1 uF charged to 600-800 volts and coupled using a blocking diode to have a little extra voltage to ensure the plasma forms a "cathode spot" on the tube  cathode that can emit electrons.  Welding is also dependent on forming a "cathode spot" of intense electrons to start and sustain the plasma.



The igniter is high side series injection triggering up to 36 KV.   The trick is to use a decaying, ringing wave, superimposed on the DC.  Typically 100 KHz to 500 KHz. This way a simple LC  low pass  filter and a metal can, mica "return capacitor" can be used to keep the HVDC out of the transistors and filter caps. 

For small systems, the igniter is a big toroid with 30 turns of # 8 on the secondary and a one to two turn primary. 400 VDC is dumped into the primary via a industrial grade SCR from a 1-10 uF cap, depending on model.  A reverse diode across the SCR protects it from frying. A small ceramic cap across the secondary resonates it.

On the big systems brand "A" uses a series injection transformer designed for xenon flashlamps and capable of carrying 60 Amps thru the secondary.  We call the return capacitors "bathtub" caps, because they look like a little metal bathtub, with a lid soldered on.   Brand A has a large inductor in the hot side, the bathtub cap to ground, and then the igniter coil.  The plasma tube serves as a spark gap if it does not light the plasma.


Brand B does things a little differently,   They have a ferrite rod inductor that has about 15 turns of # 10 copper magnet wire to keep the HF out of the DC from the PSU on the hot side.  Right after that they have a spark gap (air gap) that couples the igniter pulse from a classical flashlamp style HV transformer into the tube anode.  The bathtub cap is on the high or PSU side of the inductor for brand "B" as a secondary means of protecting the PSU.

Moral of the story, igniter pulses need a "AC" return to ground independent of the load. Usually a small return cap sized to  conducts RF but almost nothing else is across the load to close a loop around the ignite pulse.  key word here is LOOP, you do not want to shunt the whole HV ringing wave to ground.

Classical Arc and Tig  Welders with HF ignition have a sort of tesla coil and a LC filter network  and spark gaps to keep the igniter waveform out of the rectifier diodes and control system.

Wima makes or made the mica caps used in all brands of the lasers. While they are only rated 600 WVDC, they comfortably shunt 20-30 KV pulses to ground, provided there is a plasma tube there to break down.

A 1" arc length  xenon strobe tube from a disposable camera breaks down at ~1200 VDC in my bench tests.  I've used them as spark gaps for that reason, and you can also get gas discharge overvoltage protectors for a few dollars to protect your caps from rogue pulses.

Make sure your igniter has a ringing wave so you can filter it out.

Have fun designing the filter  inductors, however keep in mind the ignite pulse should be 5-10 cycles of ringing, decaying wave, so your not too worried about the inductors saturating if the plasma does fire and start welding.
This should take a load off your mind about having a large inductor messing up your double pulse waveforms.

If your weld plasma fires, you can pretty much bet that will shunt the ignite pulse. It is what happens when it doesn't fire, that concerns you.  In that case the ignite spike wants to break down the insulation in your storage caps.

I wish I could show you the inside of one of the PSUs I work on. For a 20,000$ laser, they are quite simple, just banks of water cooled NPNs controlled by a few op-amps, and the newer ones are 7-10 kW   buck switchers.. 

We have a Sunstone upstairs, but there is no way that department is going to let me pop the cover for pictures.
I've used it, I see no evidence of a huge HV ignite spark if we don't get a weld.  They may be doing things differently, such as flowing a "Simmer" current right before they pull back the electrode. The simmer current would be just enough to start a pilot arc between the tungsten point and the metal to be welded.  Quite honestly, they have some circuitry to prevent it firing if conditions are not perfect.  We often get movement of the rod but no weld or even a spark  when holding the probe freehand.

 It will not fire unless the tungsten tip is FIRMLY in contact with the object to be welded. I think they measure the resistance of the contact before firing.  As I said, if things are not perfect, it will pull back the tip when you hit the pedal, but there will be no attempt at a weld.

I'd also think Sunstone would warn of us of a potential "sting" if they used a free running start pulse. There is no such warning.

Steve

PS some reading:

http://laser-caltech.web.cern.ch/laser-caltech/report/Flash%20lamp%20Eg&G.pdf

http://www.repairfaq.org/sam/532hsch.gif (T1 is the igniter toroid, note the blocking diode)

http://www.excelitas.com/downloads/dts_triggertransformers.pdf

http://pulsearcwelder.blogspot.com/ (mentions lift start and HV pulse start)

[HAL-42b]

This was very interesting to learn. Thanks for writing it up 

[backXslash]

Steve

PS some reading:

http://laser-caltech.web.cern.ch/laser-caltech/report/Flash%20lamp%20Eg&G.pdf

http://www.repairfaq.org/sam/532hsch.gif (T1 is the igniter toroid, note the blocking diode)

http://www.excelitas.com/downloads/dts_triggertransformers.pdf

http://pulsearcwelder.blogspot.com/ (mentions lift start and HV pulse start)

The last link in your list there is actually the same guy that designed the unit I've got right now. He's been helpful to a point on this quest of mine, though he seems justifiably distracted with his new design. His new design is a lot closer to a Sunstone unit. My end goal is basically the same unit from ABI.

If anyone here could help me replicate and implement the start circuit, I'd be very grateful. I'm... a bit lost with it all honestly.

[LaserSteve]

See Attached...

Component Values are left as an exercise for the student..

Please do not get hurt when working with this..

The trap inductor is best wound on a large ferrite rod.
The LC lowpass may be needed if you can't get the series resonance right on the trap inductor.
The trap should have a resonant frequency much lower then the ignition coil's ringing wave by a factor of two or three.

If the ignition coil is built correctly, it will ring, and that is easily seen on a oscope with weak coupling to the coil. Typically a single turn loop of wire far from the coil will see the waveform. Once you know what the loaded ringing frequency is, you can sketch out the trap inductor.

In the ABI picture, they are using a TV  Flyback (HOT)  coil, which is kind of a crude choice for this task. So that coil is ringing at 10 KHz or so..

Steve

[T3sl4co1l]

Gee, 470 ohms, that'll rather take the sting out of it!  (To get some ringing, I'd think on the order of 10 ohms would be preferred; much of which is probably in the inductors as-is, so nothing extra might be needed.)

With the LC ("bathtub" and "trap") dimensioned properly, there's no need for the SG1 or other TVS device (the "bathtub" should've been chosen properly to keep the voltage sane in the first place!), and the additional filtering probably isn't needed either (the OP's dumping a big ass capacitor, so anything would be rather ineffective compared to what the actual source is, in this case!).

That said, controlling the balance of RF/transient energy will be important to anything nearby or connected.  As shown, it's assumed ground is "ground", which means big, solid, unbroken ground planes, throughout the build.

That's probably a little unrealistic, so perhaps it should be designed for balanced operation, out at the "hot" end: "trap" is split into two series halves, one on +V, one on GND.  That allows both + and - to vary with respect to ground (you still want the rest of the circuit solidly grounded though!), which they probably will regardless, due to the big current.  So, might as well make it easier, thus reducing stray currents going backwards into the supply circuit.

Some additional common mode filtering (i.e., where the LC lowpass is, instead make it a common mode filter -- big wires wound around a ferrite toroid in the same direction) might help with that, too, but differential filtering probably isn't a big deal.

Tim

[LaserSteve]

For safety reasons, the hand held welders have the return lead at Earth and Case ground...

    On the other hand, depending on the plasma, the optional LC filter does need to be there for some plasmas that generate large amounts of RF, independent of the ignite pulse.  Typical examples are very short arcs near or above atmospheric pressure such Xenon short arc lamps. Since a Xenon lamp has a very similar arc in terms of voltage and current curves to the Argon pulse welding arc, it is very safe to assume a fairly large amount of conducted  RF is generated.  It is a very good idea to keep that broadband  RF out of the control system.

The gas spark gap/clamp tube is there for what happens when you have  a partial arc ignition failure.  I can't go into details due to a NDA. Suffice it to say you want a clamp around there just in case. Overkill, maybe, then again maybe not..

I've both designed, and maintained these types of PSUs.  Including a 400 watt handheld, battery powered,  Cermax lamp based device for a security  application.  That application had to pulse the lamp and restart  at 30 Hertz.  Long term testing of the prototype indicated that additional filtering to protect of the drive circuit FETs was highly warranted.

As I have no idea what the OP will use for a ignition transformer, calculating Xc, Xl, gap size,  and everything else   is just academic at this point. He may get away with a very gentle spark, if his system is capable of very fast timing. If that is not the case, he will need a very beefy ignition system, especially if he uses air as the working gas.

Looking at the low energy pulse welding web literature, he has perhaps 20 to 40 microseconds to initiate the ignite pulse after the tip is moving.  If he can't do that, he needs a very hot spark to initiate welding.


Steve

[backXslash]

Well, the micro in use is a PIC16F877A. The timing on it is pretty good. I'd honestly figured I'd use a ceramic collar or tube on the probe to prevent arc over, and then essentially superimpose the main weld pulse on a continuous arc. I figured that'd be the most reliable method.

Either that, or start the continuous arc before and only end it after the main weld pulse.

[LaserSteve]

My employer's unit does double pulses with both pulse widths adjustable.  At the same time I can very delivered "Watts" from ~10 to 140.  First pulse conditions the weld area and starts the bonding. Second pulse is the weld. Pulsing is so fast that to the eye it is one short burst. Current drops to near zero in-between pulses.

Steve