An update on my TIG inverter project

I'd go one step further and opto-isolate the logic- level stuff from the driver and output stages. Then he could use his Wavetek and only let the smoke out of the opto-isolators.

I agree with you. I think that I will do my project in stages. (just like I do computer programming). First, I will do the power part, that is, IGBTs (which I received today), and the gate driver. I will drive the gate driver with a DC power supply, then with the wavetek and use a small battery or some such as input to the H bridge, instead of using the welder.

As far as timers or other drivers for the gate driver, I want to regulate pulse width as % of the cycle (positive vs. negative as % of cycle).

It could be done with two 555 timers, I guess, but I am not sure if it is easy to control them naturally, using two pots (one for frequency and another for percentage).

I was looking at the MAX038 chip, which can make square wave, can be controlled by simple pots, etc. The only minus of it that I see, is that it requires different voltage than the gate driver IR2109. Gate driver needs 16 volts power and 3-5 volts logic input, whereas MAX038 needs +/- 5V power and makes +-1v output.

As you can see, I am quite confused.

Very nice. I saved that pdf.

Thanks. I am looking at it right now. I can see how pulse width is controlled by 50k pot, but I have to figure out how to make the frequency separately controllable. I will read more on it.

Yes, it is a great page, I read it a little bit before and will read it more. My concern is that the duty cycle cannot be very well adjusted (the range is not that big). Maybe I am wrong.

I definitely agree with your larger premise, that I need to use a chip to drive the gate driver. What I am figuring out now is what chip to use, and how. I want to have two pots, Frequency and Cleaning cycle percentage (percentage of DCEP vs DCEN).

Yes, I will put something, but I think that the IGBT manual specified a "snubber circuit". Which is a resistor and a capacitor between source and drain. Table 5-3 on page 5-9 of the fuji manual.

I am not sure how to calculate or measure the inductance of the welder.

Thanks. Please forgive me if I say something stupid, but I thought that gate drivers such as IR2109 provide isolation between outputs and the rest...

You remember a while back, I suggested you use a 555 timer instead of your Wavetek? Well I still recommend that. Just imagine that the first time you switch it on, lots of smoke comes out of the Wavetek. That is quite likely to happen. I have let the smoke out of a Wavetek and it upset me. Believe me, it is not hard work at all to build a circuit which will be just as satisfactory as a Wavetek for your use, and much easier to repair. Start with figure 4 in the following document:

Also, you really do need to put at least varistors across the welder output and probably the load terminals too (I suggest the big sort which you connect with nuts and bolts). The inductance within the welder will easily produce more than 1200V when the arc goes out, and I am not talking about HF starting here, just the inductance of the transformer coils.

Good luck, Chris

How cheap? I might be interested ...

Andy

Actually, you really can't turn off HF for the application you are suggesting. You could turn off HF during DC welding, or do without HF altogether. But for AC TIG welding (i.e., for aluminum TIG), you have to have continuous HF to maintain the arc.

Note that you can do some limited aluminum TIG welding using DC, according to Ernie ...

I am not a welding expert, by any means. My understanding is that HF arc stabilization is required for regular 60 Hz sinewave AC, since the period when voltage is low, during transition through zero, is too long and the arc extinguishes itself.

What I am trying to accomplish is to build a square wave inverter, where transition through zero, from high absolute value of voltage to the opposing high absolute value of voltage, happens very quickly.

The difference is that for sinewave, we are talking about a few milliseconds, whereas with square wave and my IGBT, the transition would take about 1 microsecond or less.

Welding books that I read suggest that HF stabilization is not necessary for square wave AC. Miller went as far as suggesting to not use HF even for starting, and created a lift-arc arc starting technology.

Yes, that's right.

An excellent point.

This is way beyond my electronics level. I would rather try to find some simple PWM chip or some such.

I will check the IGBT application notes and look for that.

Thanks. I am definitely open to this idea.

Another issue that concerns me is dead time. If the H bridge is turned off for, say, 500 ns 100 times a second, could it possibly confuse electronics of the welding machine in some way that is destructive to it?

Adding capacitors and snubber circuits could alleviate this issue,. but I would like to hear your thoughts on it.

I have schematic of my welder, in fact.

I doubt you get any isolation from most MOSFET drivers, and I would even question the level of isolation you get from an opto-isolator after the smoke has already departed. In my opinion, you would be better off separating valuable things from the welding circuit by a distance significantly greater than the distance between your tungsten electrode and your piece of metal that you are welding, and to be certain that an opto-isolator is good enough might require quite a bit of investigation. Also there are very few opto-isolators that are decently fast. I vaguely remember some part beginning with QT.... that was ok, but much more expensive than a 555 timer!

Chris

Ignoramus19740 wrote:

I think the MAX038 has a SYNC output or something like that which might have a higher swing, but I don't know if the duty cycle of that output is controllable. Also needing -5V is a pain.

The simplest way of changing the frequency that I can think of is to add a rotary switch with s few different possible tuning capacitors. That isn't very elegant but it's pretty simple.

One option that I have used in the past is to make a triangle - wave oscillator using an op-amp as an integrator (non-inverting input to a fixed mid-rail voltage (such as resistive divider between the supply rails with bypass cap), and a capacitor from op-amp output to inverting input, and a resistor from the inverting input to the signal to be integrated which is the output of a comparator mentioned next...), and a comparator wired to have hysteresis (resistor from output to non-inverting input, and resistor from non-inverting input to op-amp output, and fixed mid-rail DC voltage on the inverting input) to set the upper and lower extremes of the triangle wave. You can then vary the frequency of the triangle wave using one pot, which would be in series with the resistor between the comparator output and the op-amp inverting input. You can set up a second comparator with the non-inverting input connected to the wiper of a second pot between the supply rails, the other comparator input is connected to the triangle wave as produced previously at the op-amp output. When the triangle wave is above the threshold set by your second pot, then the output is high, otherwise it is low. By adjusting the second pot you set the duty cycle. I have used this kind of circuit before but don't have a schematic ready to post. It is a bit of a tricky circuit to get working because of the hysteresis calculations and also the many ways of getting the feedback in the wrong sense. For that reason I would use the 555 as a first step.

I think you would need the varistor(s) too.

I don't know either, but I would expect that the main thing that limits the output current of your welder when there is a short circuit would be the inductance present before the rectifier of the welder. The thing that might save you (if the HF doesn't get you) is that they probably had to put some protection in the welder already to stop it blowing up its own rectifier. Still I would suggest adding the varistors, they are not that expensive, and once all of your IGBTs are blown, you might not be able to get more at a good price and that will fit the circuit, bolt holes on the heatsink, etc. Better not to blow them in the first place.

Chris

Actually you can do alot of Al with He. Thick or thin, just change the tungsten and polarity for the purpose. The cleaning and ease of AC just makes it a better choice for most.

Possibly, but note that I found an XR2206 chip (which was, possibly, mentioned earlier), which seems to do all that I need, has a convenient supply voltage, etc.

I will definitely check on that. I would hate to burn up these IGBTs.

That said, open circuit voltage of my welder is 85 volts, and IGBTs are rated for 1,200 volts, or 14 times more than OCV.

So, when you say get a varistor rated for 2 times OCV, I have to wonder what purpose would it serve.

Yep, you are 100% right.

I have these capacitors, 92 uF, rated for 535 volts:

Sounds like my 92 uF 535 VAC capacitors ought to handle it well.

They may be more for the protection of the welding machine, rather than for protection of the IGBTs.

I think that some dead time will be unavoidable. Using a cap should be good...

I got it. Should I pick the amp rating to match my welder's output?

ie, 150V, 200 amp varistor?

thanks Chris. I am still days away from plugging anything into my welder, but like you said, I do not want to blow up my IGBTs.

Actually it would appear simple if I could be bothered drawing a proper diagram. I'll have to download one of those ASCII-ART programs one day.

I think it is more a characteristic of the welder that makes you need the varistors, so the IGBT manufacturer may not have foreseen this need. Anyway I would just look in your preferred component catalogue for a varistor rated at maybe 1.5 or 2 times the open circuit voltage of the welder, and which looks stout enough that it won't melt on the first little spike. I have used Siemens varistors in the past but they are called Epcos now. I think Harris used to make them too.

I would say dead time is OK, provided you connect some varistors and preferably also some sturdy capacitors across the output of the welder where it supplies the IGBT bridge. If all the IGBTs go off for a microsecond, then 200 Amps would charge 1uF to 200 Volts which your IGBTs could handle. On the other hand I would suggest using more like 10-100uF across the supply, because physically bigger caps might handle the current pulses better. The capacitor will have to deal with whatever ripple voltage is on the welder output, it will not normally be able to smooth the ripple due to the rectified mains frequency. If you were to put an electrolytic of 10000uF across the welder, the electrolytic would probably be unable to handle the ripple current and would explode. On the other hand if you put a polypropylene motor run capacitor of maybe 10uF instead, then that should handle the mains ripple without too many problems, because even though the ripple voltage would be similar, the high impedance of the capacitor to mains - related ripple would limit the current in the capacitor to a level that would not overheat it. If the capacitor is sturdy it should be OK with the 1us pulses of 200A whenever the IGBTs all turn off, as long as the frequency of the IGBT drive is not too high.

If you decide to use no dead time then you should probably not put a capacitor across the welder output, but you should still include the varistors.

Chris

Well, if you ever played with old telephone relays when you were in school, you will know that if you connect one to a 6 Volt battery, and then disconnect it from the battery, you can get a spike of hundreds if not thousands of volts across the coil at the moment when you disconnect it from the battery. The sensation is unmistakable. The high voltage is developed in the inductance of the relay coil.

Similarly, 85V from your welder can, where inductors are involved, lead to voltages high enough to blow your IGBTs. The caps might prevent this, but the varistors will be extra insurance, and cheap compared to IGBTs.

You could choose the varistor voltage anywhere from 85V up to 1200V, but I would not go lower than 100-150V in case the varistors get hot, and I would not go above 400-500V in case there is some more inductance after the varistors. If you could guarantee that the varistor will be directly across the terminals of the IGBT module, (closer than 2 inches) then you could go for 1000V varistors, but this would bring no advantage over 150V varistors. It will probably be easiest to get varistors rated for the AC mains voltage in your location. These should work fine.

Chris

I think varistors are usually rated for the number of Joules they can take in a single hit, as well as a peak (one - off) current rating which will be thousands of amps even for very small devices. They are not meant to be used for repetitive surges like you will be giving them, but as long as they don't get too hot, I would not expect trouble. As a first step, I would choose a size which is affordable but doesn't look flimsy. This isn't very scientific, but without knowing the frequency you will be using, the dead time, the supply capacitance and also the inductance within the welder, a guess is all I can come up with, and even given all that information I probably would have trouble to do any better! If during use, you notice that the varistors smell like they are hot, then obviously they were either too small or you need to change one of the aforementioned parameters. Also, if you choose varistors rated for too low a voltage, lower than the welder voltage, then they will promptly get too hot and fail. In any other situation I would expect they would give some warning before they fail since surges are what they are meant for, and they would only fail if the average power dissipation were excessive for their size. The varistor makers would probably have some useful application notes.

Chris

How much did you pay for the elephants?

$1 each... I bought 5 for $5, used 2 in my rotary phase converter, and have 3 remaining ones. Do you think that I overpaid?

I think you answered about the caps. I was asking about the elephants. Page title is "Stuff that I bought at Pioneer Industrial Services" and the first picture is of two elephants. :-)