Making a 300 Amp square wave AC inverter

[Ignoramus965]

I am considering a project to build a 300 amp, about 40V inverter that
would invert DC into square wave AC. That's for converting a DC
TIG welder into a square wave AC TIG welder. I do not yet have the
welder, but may get it in a few days.

I am coming here after some preliminary discussions on
rec.crafts.metalworking.

I have some basic knowledge of physics and some knowledge of
electricity to start with, but I am not well familiar with solid state
electronics. I built a well functioning phase converter, and the only
solid state piece that I ever used was a crydom relay with zero
crossing.

The objective is to make the inverter cheaply but in an effective manner.

With that introduction in mind, let me describe what is my current
plan.

The plan is to set up an H bridge, using 4 arrays of MOSFETs. These
mosfets would be controlled by variable frequency AC signal. To
generate this signal, I want to use an HP 204C oscillator (which I
have), coupled to an appropriate amplifier (which I do not have).

HP 204C produces 2.5 volts and needs to be amplified about 16 times.

Signal from amplifier would control mosfets, although, possibly, I
would also use diodes to prevent negative voltage from being sent to
the gate of mosfets.

The H bridge would produce inverted, almost square wave AC.

The mosfets that was recommended would be something like this

http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?Ref=199933&Row=338490&Site=US

For gate input, I would need to produce about 40 V AC (if I am not
mistaken), that sort of determines what the amplifier should be. Looks
like wattage requirements for the amplifier are minimal.

Is that a sensible plan, and if so, what would be your suggestions. I
know how to solder, use a multimeter and use a drill press. I am a
moderately handy person and am open minded and welcome suggestions.

Thanks!

i

 

[Rich Grise]

I am considering a project to build a 300 amp, about 40V inverter that
would invert DC into square wave AC. That's for converting a DC
TIG welder into a square wave AC TIG welder. I do not yet have the
welder, but may get it in a few days.

I am coming here after some preliminary discussions on
rec.crafts.metalworking.

I have some basic knowledge of physics and some knowledge of
electricity to start with, but I am not well familiar with solid state
electronics. I built a well functioning phase converter, and the only
solid state piece that I ever used was a crydom relay with zero
crossing.

Then bite the bullet, and buy the AC welder. You couldn't possibly
design and build anything like that for less than the difference in
cost of the two models - parts alone would do it, and building
something like that is definitely not a project for a beginner.

Sorry )-;
Rich

 

[Ignoramus965]

Then bite the bullet, and buy the AC welder. You couldn't possibly
design and build anything like that for less than the difference in
cost of the two models - parts alone would do it, and building
something like that is definitely not a project for a beginner.

You are, quite possibly, right, but I would like to know why that is
so, what are the technical reasons.

i

 

[Chris Jones]

You are, quite possibly, right, but I would like to know why that is
so, what are the technical reasons.

i

I agree that it isn't worth doing, but you might enjoy trying.

Try building a little one first - e.g. 5 Amps, 40 Volts. When you have that
working (with light bulbs as a load), you can think about building the
bigger one. This way, you won't blow up quite so many MOSFETs in the
learning phase. You will need an oscilloscope, I suggest a Tek 475 or
similar, and an isolated adjustable power supply for testing your miniature
version.

You need to work out the proper gate voltage you need for each MOSFET. The
bottom ones are easy, because the gate either goes to the negatve rail of
the H bridge, or 12 Volts above that. It's the top MOSFETs of the H bridge
that are hard to drive, because you will have to use all NMOS FETs in the
bridge even at the top, and the gate drive of the top FETs has to be
controlled relative to their sources, and the source terminal is going up
and down too. Really you should probably opto-isolate the upper two gate
drive circuits from the clock source, but then the gate driver circuits
will need their own isolated powre supplies too.

The HP oscillator isn't really ideal for the signal source since you want a
square wave, try using a 555-timer oscillator circuit instead. Look for a
circuit that allows you to alter the duty cycle - this could be useful.
Then you could feed the 555 output signal into a gate driver IC such as a
TC4421 to drive the MOSFET gates. You still need to figure out the opto
isolators and the extra power supplies though, so that you can drive the
top FETs. You probably also want to alter the logic signals a bit so that
the top and bottom FETs cannot momentarily conduct at the same time as this
is bad for them.

Also beware the inductance of the welder output. This could cause large
voltage spikes if the load gets disconnected, which will blow up your
MOSFETs. You will want to put capacitors and varistors (MOV) across the
incoming DC.

If your welder has HF starting that will blow up your MOSFETs too.

Chris

 

[Ignoramus965]

I agree that it isn't worth doing, but you might enjoy trying.

Try building a little one first - e.g. 5 Amps, 40 Volts.

I agree 100%. I will start witth a full powered control circuit, but
with a lot less mosfets and a small load.

When you have that working (with light bulbs as a load), you can
think about building the bigger one. This way, you won't blow up
quite so many MOSFETs in the learning phase. You will need an
oscilloscope, I suggest a Tek 475 or similar, and an isolated
adjustable power supply for testing your miniature version.

I already have an old trusty Tek 475, as well as a dual HP DC power
supply (handy for a control circuit vs. load). Plus an HP 204C
oscillator that I mentioned.

You need to work out the proper gate voltage you need for each MOSFET. The
bottom ones are easy, because the gate either goes to the negatve rail of
the H bridge, or 12 Volts above that. It's the top MOSFETs of the H bridge
that are hard to drive, because you will have to use all NMOS FETs in the
bridge even at the top, and the gate drive of the top FETs has to be
controlled relative to their sources, and the source terminal is going up
and down too. Really you should probably opto-isolate the upper two gate
drive circuits from the clock source, but then the gate driver circuits
will need their own isolated powre supplies too.

I have to admit that I am a little confused regarding this.

Why is there any asymmetry between top and bottom MOSFETs.

The HP oscillator isn't really ideal for the signal source since you want a
square wave, try using a 555-timer oscillator circuit instead. Look for a
circuit that allows you to alter the duty cycle - this could be useful.
Then you could feed the 555 output signal into a gate driver IC such as a
TC4421 to drive the MOSFET gates. You still need to figure out the opto
isolators and the extra power supplies though, so that you can drive the
top FETs. You probably also want to alter the logic signals a bit so that
the top and bottom FETs cannot momentarily conduct at the same time as this
is bad for them.

Would it make sense to try to use "comparators" to convert sine wave
into square wave with some breaks in between.

Also beware the inductance of the welder output. This could cause large
voltage spikes if the load gets disconnected, which will blow up your
MOSFETs. You will want to put capacitors and varistors (MOV) across the
incoming DC.

Makes perfect sense. I hope that someone could suggest a good formula
for computing needed capacitance. I have plenty of caps (60VDC, IIRC).

If your welder has HF starting that will blow up your MOSFETs too.

it does not...

i
--

 

[Ben Bradley]

I have to admit that I am a little confused regarding this.

Why is there any asymmetry between top and bottom MOSFETs.

A probably too-short answer: because they are all N-channel (or
NMOS as stated above).

FET's have three connections labeled source, drain, and gate. You
turn it on by applying approximately 10 volts positive to the gate
relative to the drain, and this effectively connects the source to the
drain. The bottom FETs in the H-bridge have the drain grounded, so
you just "apply 10 volts to the gate" which is understood as a voltage
relative to ground.
In the upper part of an H bridge, the FET's source is connected to
some voltage source, apparently 40 volts in your case. When turned
off, the gate and drain are both at zero. To turn on this FET, you
must put 50 volts into the gate to make the gate 10 volts above the
drain, because the drain rises to 40 volts when the FET turns on.
Oops, your power supply is only 40 volts. Well, you can use charge
pumps and/or other stuff to get around that, but however it's done,
it's messier and needs more "support circuitry" than does switching
the low-side FET's in an H bridge.
Also keep in mind when you get this working (a 5-amp circuit will
only need one power MOSFET in each leg of the H-bridge) that a circuit
that drives a single power MOSFET might not drive 10 or 20 or 50 power
MOSFET's in parallel, and you'll need somewhere around that many in
each leg to switch 300A. See Don's response about gate capacitance.
with N devices in parallel, you have N times the gate capacitance, and
you need to supply and remove N times as much current through the
gate(s) to swith N MOSFETS as fast as the one MOSFET.
You have to switch them fast so that the spend as little time as
possible in the "linear range" where they turn electric power into
heat. If they get too hot, the magic smoke will come out (this can
happen in an explosive way, so don't be having your face over
something when you first power it up) and they won't work anymore.

Would it make sense to try to use "comparators" to convert sine wave
into square wave with some breaks in between.

Yes, you could send the HP's signal through one comparator in a
LM339 quad comparator (or use the single-comparator in a package, I
forget the part number), but you're using a benchtop device and a 50
cent part smaller than your thumb to do the work that can be done by
one or two 50 cent parts smaller than your thumb (plus a few resistors
and capacitors at a few cents each). If you want to do something fancy
that the HP oscillator lets you do, like vary the frequency or the
duty cycle, a couple of potentiometers wired into the right places can
allow you to do that, and they're only a dollar or two each.

 

[Ignoramus965]

A probably too-short answer: because they are all N-channel (or
NMOS as stated above).

FET's have three connections labeled source, drain, and gate. You
turn it on by applying approximately 10 volts positive to the gate
relative to the drain, and this effectively connects the source to the
drain. The bottom FETs in the H-bridge have the drain grounded, so
you just "apply 10 volts to the gate" which is understood as a voltage
relative to ground.
In the upper part of an H bridge, the FET's source is connected to
some voltage source, apparently 40 volts in your case. When turned
off, the gate and drain are both at zero. To turn on this FET, you
must put 50 volts into the gate to make the gate 10 volts above the
drain, because the drain rises to 40 volts when the FET turns on.
Oops, your power supply is only 40 volts. Well, you can use charge
pumps and/or other stuff to get around that, but however it's done,
it's messier and needs more "support circuitry" than does switching
the low-side FET's in an H bridge.
Also keep in mind when you get this working (a 5-amp circuit will
only need one power MOSFET in each leg of the H-bridge) that a circuit
that drives a single power MOSFET might not drive 10 or 20 or 50 power
MOSFET's in parallel, and you'll need somewhere around that many in
each leg to switch 300A. See Don's response about gate capacitance.
with N devices in parallel, you have N times the gate capacitance, and
you need to supply and remove N times as much current through the
gate(s) to swith N MOSFETS as fast as the one MOSFET.
You have to switch them fast so that the spend as little time as
possible in the "linear range" where they turn electric power into
heat. If they get too hot, the magic smoke will come out (this can
happen in an explosive way, so don't be having your face over
something when you first power it up) and they won't work anymore.

I see. So, an appropriate chip would solve these issues by combining
voltages, supplying adequate power quickly, etc etc. Is that correct?

Yes, you could send the HP's signal through one comparator in a
LM339 quad comparator (or use the single-comparator in a package, I
forget the part number), but you're using a benchtop device and a 50
cent part smaller than your thumb to do the work that can be done by
one or two 50 cent parts smaller than your thumb (plus a few resistors
and capacitors at a few cents each). If you want to do something fancy
that the HP oscillator lets you do, like vary the frequency or the
duty cycle, a couple of potentiometers wired into the right places can
allow you to do that, and they're only a dollar or two each.

That makes perfect sense. I kind of like the idea of using a signal
generator to control frequency, but I could give it up. I now am
trying to find something to understand how to drive these chips and
what it is that they do to convert input into output.

i

 

[ehsjr]

You are, quite possibly, right, but I would like to know why that is
so, what are the technical reasons.

i

Rich *is* right. Out of courtesy to your request,
I'll mention a few technical issues below. First,
let me say I think experimentation is great and I
encourage it. But what you are talking about should
be undertaken only by those who would not need to post
as you did. This is not a knock on you. Screwing around
with 300 amps is not a "give me the schematic and a few
pointers and I'll cobble it together" project.

Some technical issues:
1) You want a 40 volt, 300 amp inverter. Thats 12,000
watts. Twelve THOUSAND watts. Have you considered
this? How will you determine what to use to connect
your inverter to the DC? I'm assuming you already
have the DC source - do you?

2) Assuming you do have the DC source, how do you
know it is capable of driving the inverter? How
do you intend to protect the source?

3) Can you predict the possible failures within a
circuit and take steps to prevent them, and, more
importantly, do you understand what will or could happen
if one (or more) of those failures does occur?

Ed

 

[Ted Edwards]

importantly, do you understand what will or could happen
if one (or more) of those failures does occur?

It does tend to be rather spectacular! :-(

Note, also, that welding is a rather nasty load.

Further, you said you didn't intend to have a high frequency start. I
assume you are interested in this because you wish to weld aluminum. Is
this for TIG? I've not seen or heard of an AC welder without HF start.

Ted

 

[Ignoramus965]

Rich *is* right. Out of courtesy to your request,
I'll mention a few technical issues below. First,
let me say I think experimentation is great and I
encourage it. But what you are talking about should
be undertaken only by those who would not need to post
as you did. This is not a knock on you. Screwing around
with 300 amps is not a "give me the schematic and a few
pointers and I'll cobble it together" project.

Some technical issues:
1) You want a 40 volt, 300 amp inverter. Thats 12,000
watts. Twelve THOUSAND watts. Have you considered
this?

Well, I am aware that it is 12 kW, but, quite possibly, not
aware of all implications of that.

How will you determine what to use to connect
your inverter to the DC? I'm assuming you already
have the DC source - do you?

Not yet, but if I acquire a CC DC welder on Monday, I will have a DC
source. If I do not acquire such a welder, I will not go ahead with
this project.

2) Assuming you do have the DC source, how do you
know it is capable of driving the inverter? How
do you intend to protect the source?

I suppose that a DC welder has internal protection.

3) Can you predict the possible failures within a
circuit and take steps to prevent them, and, more
importantly, do you understand what will or could happen
if one (or more) of those failures does occur?

Well, I would attempt to do that.

i

 

[Ignoramus965]

It does tend to be rather spectacular! :-(

Note, also, that welding is a rather nasty load.

Further, you said you didn't intend to have a high frequency start. I
assume you are interested in this because you wish to weld aluminum. Is
this for TIG? I've not seen or heard of an AC welder without HF start.

Well, check out "arc stabilizers" on ebay, they are add on devices.

i

 

[Ian Stirling]

That makes perfect sense. I kind of like the idea of using a signal
generator to control frequency, but I could give it up. I now am
trying to find something to understand how to drive these chips and
what it is that they do to convert input into output.

You've got no hope.
At least not cheaply, you might get there after a few dozen iterations,
with blown output transistors in ma of those.
And in many cases you won't understand what went wrong, as you don't have
adequate measurement equipment (at minimum probably 8 channels of digital
scope).


You're trying to design a competitive rally car, and at the moment you're
saying "Ok, so I need a gearbox between the engine and the wheels. I'm now
trying to understand what sort of duct tape I should use to secure it to the
frame".

You are looking at several weeks of study, before you even know the questions
to ask.
A not-bad first step would be to get "The art of electronics", and read
it cover-cover several times, omitting the chapters on microprocessors and
low power stuff.

The only chances that you'd save money would be to essentially learn
everything in AoE, probably something going into FET pathology a bit more,
and learn how to accurately work a simulator.

I'd say the best part of a year, and even then I'd say that it's almost
certain that it'll die either immediately, or a few hours of welding later.

 

[Chris Jones]

A probably too-short answer: because they are all N-channel (or
NMOS as stated above).

Yes. You could use PMOS devices at the top if you really wanted to but if
the main supply voltage is much more than 20 Volts, it's pretty much just
as complicated, and you need at least twice as many FETs for a given
current if they're PMOS therefore the cheapest and best design will use all
NMOS.

FET's have three connections labeled source, drain, and gate. You
turn it on by applying approximately 10 volts positive to the gate
relative to the drain, and this effectively connects the source to the

You have source and drain mixed up. The gate voltage is relative to the
SOURCE, for both NMOS and PMOS.

drain. The bottom FETs in the H-bridge have the drain grounded, so

you just "apply 10 volts to the gate" which is understood as a voltage
relative to ground.
In the upper part of an H bridge, the FET's source is connected to
some voltage source, apparently 40 volts in your case. When turned Again....

off, the gate and drain are both at zero. To turn on this FET, you
must put 50 volts into the gate to make the gate 10 volts above the
drain, because the drain rises to 40 volts when the FET turns on. Again...


[snip]

Would it make sense to try to use "comparators" to convert sine wave
into square wave with some breaks in between.

Yes, you could send the HP's signal through one comparator in a
LM339 quad comparator (or use the single-comparator in a package, I
forget the part number), but you're using a benchtop device and a 50
cent part smaller than your thumb to do the work that can be done by
one or two 50 cent parts smaller than your thumb (plus a few resistors
and capacitors at a few cents each). If you want to do something fancy
that the HP oscillator lets you do, like vary the frequency or the
duty cycle, a couple of potentiometers wired into the right places can
allow you to do that, and they're only a dollar or two each.

And things connected to welders tend to get blown up, better use the 50c
chip. I would not use the comparators with a sine wave input, I would use
a monostable chip of some sort, or maybe just some R-C delay circuits and
74HC series logic e.g. AND gates to ensure non-overlapping pulses.

 

[Chris Jones]

Well, check out "arc stabilizers" on ebay, they are add on devices.

i

You want to be a bit careful here, the HF start on the TIG I used to use was
basically a Tesla coil with a thick secondary winding in series with the
electrode wire. It would tend to blow your MOSFETs and probably your scope
too while you're building it. Avoiding this happening might involve
expensive trial and error.

 

[Ignoramus965]

Thanks. I just ordered this book. If anything, it will help me make
more money on ebay (I sell electronic test equipment as a hobby) and
be able to tell which equipment can be profitably taken apart and
which parts to salvage.

i

You've got no hope.
At least not cheaply, you might get there after a few dozen iterations,
with blown output transistors in ma of those.
And in many cases you won't understand what went wrong, as you don't have
adequate measurement equipment (at minimum probably 8 channels of digital
scope).


You're trying to design a competitive rally car, and at the moment you're
saying "Ok, so I need a gearbox between the engine and the wheels. I'm now
trying to understand what sort of duct tape I should use to secure it to the
frame".

You are looking at several weeks of study, before you even know the questions
to ask.
A not-bad first step would be to get "The art of electronics", and read
it cover-cover several times, omitting the chapters on microprocessors and
low power stuff.

The only chances that you'd save money would be to essentially learn
everything in AoE, probably something going into FET pathology a bit more,
and learn how to accurately work a simulator.

I'd say the best part of a year, and even then I'd say that it's almost
certain that it'll die either immediately, or a few hours of welding later.

--

 

[Ignoramus965]

You want to be a bit careful here, the HF start on the TIG I used to use was
basically a Tesla coil with a thick secondary winding in series with the
electrode wire. It would tend to blow your MOSFETs and probably your scope
too while you're building it. Avoiding this happening might involve
expensive trial and error.

I just said that they are available, I do realize that adding HF to an
inverter may present problems...

i
--

 

[JosephKK]

You've got no hope.
At least not cheaply, you might get there after a few dozen iterations,
with blown output transistors in ma of those.
And in many cases you won't understand what went wrong, as you don't have
adequate measurement equipment (at minimum probably 8 channels of digital
scope).


You're trying to design a competitive rally car, and at the moment you're
saying "Ok, so I need a gearbox between the engine and the wheels. I'm now
trying to understand what sort of duct tape I should use to secure it to
the frame".

You are looking at several weeks of study, before you even know the
questions to ask.
A not-bad first step would be to get "The art of electronics", and read
it cover-cover several times, omitting the chapters on microprocessors and
low power stuff.

The only chances that you'd save money would be to essentially learn
everything in AoE, probably something going into FET pathology a bit more,
and learn how to accurately work a simulator.

I'd say the best part of a year, and even then I'd say that it's almost
certain that it'll die either immediately, or a few hours of welding
later.

Damn, Ian, don't be so snotty. You remind me of some no real experience
professors i have had. Moreover, and in particular, SPICE cannot handle
the problems inherent in modeling an arc welder at all.

 

[Ignoramus21002]

Damn, Ian, don't be so snotty. You remind me of some no real experience
professors i have had. Moreover, and in particular, SPICE cannot handle
the problems inherent in modeling an arc welder at all.

I am not terribly worried. I will try to implement my project the
right way, with gate drivers, snubber circuit etc. I already have some
parts working. Before putting it to the toughest test, I will get it
through a series of nicer, milder tests.

If, in the end, it all blows up, I will lose perhaps $200-300 and will
gain valuable experience. The experience would help me make more $$ on
ebay and will be generally enlightening.

i

 

[JosephKK]

Ignoramus21002 wrote:

If, in the end, it all blows up, I will lose perhaps $200-300 and will
gain valuable experience. The experience would help me make more $$ on
ebay and will be generally enlightening.

i

I like this part of your attitude.