Pictures -- My TIG inverter plugged into the actual welder

[a small introduction, I am working on a solid state advanced square wave inverter to add to my DC tig welding machine]

I want to say a big THANKS and a BIG HUG to all those who helped me so far. I would not be where I am, without you. I am not quite at my destination yet, but the light at the end of the tunnel becomes more and more apparent.

To readers of s.e.d., I would appreciate if you could please scroll down and answer my question about the need for a fast recovery diode for my RCD snubber that I made today. I think that it is needed to prevent voltage spikes that are still seen.

formatting link

Today was a big day for me. I changed my snubber and plugged my prototype into my actual TIG welder.

I was very afraid and wore protective goggles all the time, but nothing terrible happened.

Here you see pictures of my DC -> AC inverter working with my actual Hobart Cyber Tig welding machine.

On the pictures you see the welder running the bridge at 65 amps. The outputs are simply shorted with a 1 gauge cable. No actual welding was taking place. HF arc starting was disabled.

It is quite apparent that whatever I have, in terms of bridge shorting time, as well as the snubber, is working adequately (more on this below). The 150V varistor that I placed across outputs, did not even get warm at 135 amps.

I went all the way to about 135 amperes. The only problem that I has was that the snubber diode got quite warm soon and I had to discontinue the test (reduce amperage to 65 amps).

The bridge seems to be functioning perfectly in a stable manner at 65 amps.

RCD Snubber

After much experimentation, I realized that the resistor on a simple RC snubber was overheating very quickly and that the capacitor was dumping too much energy when the bridge shorted. I replaced it with an RCD snubber, which you see on the picture 4.

(I may have gotten the schematic wrong, it is late) DC + -------------------------------+-------| \ --- Diode R / /\ BRIDGE \------| --- --- Cap DC - ---------------------------------|

The diode that I used in the RCD is NOT fast recovery. I just do not have a fast recovery diode. That, I think, explains spikes that I still observe. I will probably plop down $25 and buy a 240A diode:

I want to buy this diode: Digikey information:

Technical/Catalog Information 497-4408-5-ND Standard Package 100 Category Discrete Semiconductor Products Family Diodes/Rectifiers Vendor STMicroelectronics Diode/Rectifier Type Fast Recovery Voltage-Rated 600V Current Rating 240A Package / Case ISOTOP Packaging Tube Lead Free Status Lead Free RoHS Status RoHS Compliant Other Names 497-4408-5

Let me know if you that that it is wrong for my application.

(Click on the thumbnail images to enlarge)

Front panel of my cybertig. Shows 65 amps of current.

Prototype of inverter. Screwed down to a wood board so that nothing falls off.

Top -- timing signal going into the Semikron 23 drivers Bottom -- voltage across the DC rail. You see spikes. at 65 amps, they are about 90 volts.

Snubber Circuit -- a big diode, capacitor, and 20 ohm resistor.

Reply to
Ignoramus10725
Loading thread data ...

First, a 20 Ohm resistor is a flea compared to the 65 or greater amps you will be flowing. You should be using a resistor of a couple of Ohms, I think. Where is this snubber, at the output of the full-bridge switch? it might be better to have a separate snubber at the output of each half bridge, and one across the DC input. As the welder already has a huge inductor at its output, you need to keep the input current to the bridge from fluctuating. These spikes are ocurring right after the bridge is commanded to switch, if I'm reading the scope right. This may be the welder's response to the current momentarily rising during the overlap time, then dropping afterwards. Is there a similar negative-going spike during the shoot-through? (If you really have both top and bottom transistors conducting simultaneously, then the DC input to the bridge SHOULD dip substantially.) Hmm, but running into a dead short, the voltage can't drop much, can it? OK, so then these positive transients must be the result of the overall resistance from DC+ to DC- rising a bit during the switching. If the current from the welder stays at a steady 65 A, then the apparent resistance must be going up to R=E/I =90/65 = 1.38 Ohms.

I don't think the RCD snubber you've drawn is right. The diode seems to be backwards to clip positive transients. It would work well to clip negative transients, though. If the diode is really as drawn, you might want to turn it around.

Jon

Reply to
Jon Elson

John, do not forget, this is an RCD snubber. The resistor and the diode are in parallel. When there is a spike on the DC rail, current flows into the capacitor through the diode, not through the resistor (ok, through the resistor too, but minimally).

Now, when the potential in the capacitor is greater than in the DC rail, the current flows OUT of the capacitor and into the DC rail, thtough the resistor, since the diode is not conducting in that direction.

At the input. On the DC rail.

DC + -------------------------------+-------| \ \/ Diode R / -- BRIDGE \-------| --- --- Cap DC - ---------------------------------|

I agree with that. I have some snubber boards that were removed from the same machine as the IGBTs that I bought. The seller first sold the IGBTs, and then they sold the snubber boards separately (kind of silly, IMHO). 16 snubber boards cost me $9.99.

they are pictured here

formatting link
and I would like to hear your opinion which ones are better suitable.

Agreed.

You absolutely right.

You are 100% on target here.

No. The voltage across the DC rail drops momentarily, but to zero, not to negative values.

It does.

No, it was all messed up. The diode itself is reverse polarity, I think that usually diodes conduct the other way. Then my picture in the post was also wrong due to me being tired/not thinking clearly. I corrected it. The diode is wired properly, I first tried the wrong way and it was clearly bad.

Things became much better after I reversed the diode.

DC + -------------------------------+-------| \ \/ Diode R / -- BRIDGE \-------| --- --- Cap DC - ---------------------------------|

The short duration spike is, I think, due to reasons that you mentioned (short condition being abruptly terminated). Now, that the spike is not fully clamped, is, I hope, due to the snubber diode not being fast enough (slow recovery).

As soon as the diode starts conducting, the spike disappears in an instant. That was not the case without the diode.

That's why I asked if a fast diode like Digikey item 497-4408-5-ND would be more acceptable. What's your take on it?

i
Reply to
Ignoramus10725

Look right across the cap. If you see the spikes there as well, the diode is working OK.

Reply to
Don Foreman

Good idea, I will do that tonight. The cap is a CDE 940C, made for snubbing, low ESR, blah blah. But who knows.

The spikes were extremely narrow and ended abruptly, to a perfectly flat line, that's why I suspected diode recovery time. I will try to get some measurements of the spike duration too. It is nice that I can set my oscilloscope to trigger on the timing signal on channel 1, while looking at the bus voltage on channel 2. That way the trigger is not confused by all the oscillations in the DC bus, and I can see a clear and not shifting picture.

In any case, I need a heatsink mountable diode, since this one dissipates quite a bit of energy. I could mount this one also, I suppose, although it would not be easy due to threading, need for isolating the heatsink, etc.

What do you say Don, am I on more or less the right track so far?

i
Reply to
Ignoramus29530

Kind of a meandering track, but yeah, you seem to be gaining on it. Going to a proper gate drive was a big step forward. Having the bridge short rather than open during switchover was also a good move, since you're switching a current source.

Reply to
Don Foreman

Thanks. My approach was just like in computer programming, to go through a series of successive working prototypes, and testing as I went along. Rather than trying to assemble the perfect device from the first time.

I will report my findings on the diode recovery time issue. I know that fast recovery diodes are required for RCD snubbers. I used the slow diode because that was all I had at home. Perhaps too soon, but I already ordered a 240A fast switching diode with isolated contacts that is heat sink mounted. Another $26 down the drain... But I hope that it will be good. Digikey item 497-4408-5-ND.

i
Reply to
Ignoramus29530

formatting link

I can't really say just from looking at pictures.

Yes. it is not recovery time, specifically, but TURN ON time, which is a darn hard spec to find in any diode data sheets. I was using some reasonably fast diodes in a circuit to prevent inductors from putting negative voltages on the lower transistors of half-bridge sections and saw a similar effect. It took the diode a microsecond or more to start conducting the 20 A that was flowing in the inductor. I saw voltages as large as 12 V across the diode in the forward biased direction. I switched to an ULTRA-fast diode, and the forward voltage was reduced to about 4 V, which the driver chips could tolerate.

Hey, they actually GIVE a forward recovery time spec - and it isn't so good at 700 ns! Might look around a little more, but this may be as good as it gets in silicon. You might check Silicon Carbide diodes, but they are probably prohibitively expensive.

Jon

Reply to
Jon Elson

Yep. Similar effect indeed.

I see. The diode that I just ordered today, STTH200L06TV1, is an ultrafast diode. The spec sheet says "especially suited in switching power supplies and welding".

formatting link

I figure, even with my old regular rectifier diode, my setup is not exploding at 135 amps and the spikes are quite limited. Some improvement may just be good enuf. That's all to be verified using an oscilloscope.

i
Reply to
Ignoramus29530

I believe the most effective method of inverting after a large inductance is to short the current flow during the change of phase.

This is most simply achieved by turning all the bridge switches on; by enforcing a drive overlap on the phase switch drive circuit.

There will be an overshoot of voltage as the arc has to be re-established on each switching cycle, but there is no need to absorb energy or clamp this voltage, until the load is removed completely. When this happens, you'd best short the output and turn off the source.

RL

Reply to
legg

That's exactly what I am doing, yes.

Well, that's square wave output, and supposedly, the arc would not have enough time to extinguish during about 5 uS when the bridge is shorted. At least that is my home.

The overshoot is somewhat acceptable, but I would like it lower for safety considerations. The spec for my manual says that highest voltage is 150v, and I would like to keep all voltage below that, so that I do not damage the welder..

i
Reply to
Ignoramus29530

Put a capacitor in parallel with the diode?

Reply to
kell

That's a great thought. I can use a little cap of some sort, I think. I will think a little bit about that. That somewhat defeats the purpose of the diode that prevents dumping of electricity into the shorted bridge. I can use a little Wima 0.01 uF cap that I have soldered onto some board that I have.

i
Reply to
Ignoramus29530

Try to measure rise time of the pulse - the rising edge - or falling - that will give you an idea of the frequency components - and might lead to an RC filter could be developed.

Martin Eastburn @ home at Lions' Lair with our computer lionslair at consolidated dot net NRA LOH, NRA Life NRA Second Amendment Task Force Charter Founder

Ignoramus29530 wrote:

Reply to
Martin H. Eastburn

Well, following this thread throughout, it's clear you've been having a lot of fun, Igor. Good on ya and congrats on getting it working under load!

By the way, do you live here? :)

formatting link

Reply to
John Husvar

Thanks John... It definitely was fun, although now I begin to question the cost vs. benefit issue. Just how often would I weld aluminum, is not clear. But once I started, I could not stop. And it was very educational and not too expensive, although the little electronic and electrical trinkets cost a lot more than expected.

I will try to find a computer that can play this...

i
Reply to
Ignoramus12834

Should work with Windows Media Player. It's apparently what happens when an EE has too much idle time. :)

Reply to
John Husvar

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.