PWM to give a constant voltage

[ninjalicious]

Alright, so I want to convert a Constant Current welding supply (GMAW, TIG) into a Constant Voltage supply(MIG).

A little background. I built an arc welder and added a rectifier/choke section to give me DC. It works well. I am going to add TIG attachment which will work fine with CC nature of the welder. I would also like to build a MIG attachment. MIG requires a CV supply. My research into MIG welders shows that this is either accomplished the old fashioned way using a specially wound transformer which has more of a CV characteristic, or the new fashioned way using a switch-mode welding supply. My stop-gap measure is a wire-speed controller for the MIG which measures output voltage and adjusts the length of the arc by varying the speed of the wire feed into the molten puddle. That is an industry trick but has fallen into disfavor in manually operated welders because it is difficult to adjust wire speed fast enough to compensate for shakey-handed weldors. I have built this circuit using opamps and it works in my mock test but have not tried it yet with the actual welder (since I have not built the wire feed mechanism yet).

What I am thinking of doing is merely chopping the DC welding output using a beefy IGBT (600V, 200A) with a PWM controller (like the SG3525) that is voltage controlled at the output. I was thinking of using a large capacitor to smooth the chopped wave some, and including a large load resistor to always provide some load. Is this idea even feasible or possible to do, or am I wasting my time? I am guessing this is DC-DC converter, but are there any common circuits/diagrams I could adapt from for this type of application? I am mainly trying to avoid having to pipe it through a transformer since that component would be very expensive (it would need to pass upwards of 10kW). Also it would kind of make the idea of using my welder as a power supply redundant.

I had a look at the SG3525 (http://www.datasheetcatalog.org/datasheet/motorola/SG3525AN.pdf) and fig 8. shows a single-ended configuration that looks sort-of what I would like to do. Is this crazy? I've built plenty of linear supplies but in this application the power requirements make a linear design completely unfeasible.

If anyone could give me some pointers that would be great. I am an EE but I don't get to use that part of my education as much as I want, and that was a long time ago

Thanks!

 

[Resqueline]

You need an inductor before the capacitor on the output of a pwm stage to get its average voltage. A capacitor only will give you the peak voltage only.
I don't know how the weld will turn out with raw unsmoothed pwm. You do get to adjust the power of the arc however, and maybe this is enough, Idk..

 

[ninjalicious]

I did some more in-depth reading last night and your right, I do need an inductor.

This appears to be one of the more fundamental blocks of switching designs, a switching regulator. The topology requires a commutation diode, an inductor, and then a smoothing capacitor.

The diode is relatively easy to source, and I assume it would need to be a schottky variety to give fast switching capability since the switching frequency would probably be around 20kHz, or more. I have seen 100-200amp diodes of this type, for example, on ebay.

I think the primary governing factor here is the inductor I would need to select. It would need to be ferrite core, I am guessing "gapped", and have considerable current capability ( ~200 amps).So I am not necessarily gaining much here on savings with an inductor as opposed to a full blown high-frequency transformer of the required power rating, I would assume.

I currently(har har) adjust the power to the arc manually using a simple light-dimmer type circuit on the mains side of the welding supply. I have a double SCR with a few other components and a variable resistor which I can roughly set the output voltage, but the supply is effectively constant current in nature since there is no regulation. These components however are significantly cheaper since they only need to be able to pass 50 amps, instead of the 200+ amps at the welder output.

Maybe I need to build a circuit which varies the turn-on time of the SCRs before the welding transformer as a function of the output voltage of the welder to provide voltage regulation. Any ideas of what that scheme would be called? I'd rather use some off-the shelf chip than try to reinvent the wheel. I can't use pure PWM in this case to modulate the AC input because..well...I had thought that you can only turn off an SCR/TRIAC on the zero crossing (although you can simply skip whole cycles of the waveform, which seems like it may present a very offensive lower frequency to the 60 Hz magnetics, which I think might saturate the core and cause other problems).

 

[Resqueline]

Schottky diodes are nice for their switching speed & low power loss but they can't be made to withstand much more than 80V so you'd have to use an avalanche type.

Yeah, the inductor would likely be as big/costly as the transformer.

It might be ok for the transformer if whole cycles are skipped (zero-current switching), or if every third/fifth/seventh half-period is passed (reducing frequency & voltage).
On the whole I think a triac solution will be hard to make good, but it might be the cheapest solution. How about using an optical feedback loop to what you got?
I see a small lamp connected to the output, enclosed with & shining on an LDR in the triac circuit, thereby stabilizing the output voltage. I cant say if it'd work good enough.

An IGBT might used in a primary switchmode circuit to lower the voltage to the transformer, having a feedback sampled from the output to enable regulation of the voltage.
But although it may be easier to design a lower current primary regulator, it still has to handle the same amount of power (= same size/cost).
I'm not sure how to best design it but I have an idea that the phase compensation circuits often used in modern PSU's could be a favourable configuration.

 

[ninjalicious]

That's an interesting idea with the LDR in the triac circuit. It'd take care of isolation too. I am not sure a lamp would work well since they have variable resistance based on current (not sure about their linearity with current, is what I mean) but an LED would work great since I know their output intensity is directly related to current through them. I could use a resistor dividing network on the welding output to bring the welding voltage down a bit and feed that into an op-amp configured as a transconductance amplifier which would give me LED current proportional to the voltage, and shine that on the LDR as you mention. I could configure an adjustable gain on that to get it "tuned" and working smoothly in application...probably.

"I'm not sure how to best design it but I have an idea that the phase compensation circuits often used in modern PSU's could be a favourable configuration."

Are the phase compensation circuits you refer to used to prevent terrible power factor, or to just resolve the problems associated of using a triac to control an inductive load (Or both) ? I'll look into them when I get home.

Thanks for the great ideas!

 

[Resqueline]

I don't think you need absolute linearity, that would only be important for the scale of the control, but using a lamp is bad from a durability viewpoint anyway.
A lamp increases its output progressively with input voltage while a LED actually does the opposite - its output does not increase proportionally with input current.
But at low current levels I guess a LED would be linear enough, and it sure is easier to drive and adjust with electronics. Temperature dependency? Idk..

The answer is both, on the power factor thingy. I'd try the optical triac control first though, absolutely cheapest & quickest thing for you to try out if it works.

You're welcome, glad to have been of any help!
I hope I don't lead you out on a wild goose chase, but you sound competent enough to figure out the good from the bad.