I am building a micro TIG power supply (and probably torch too), maximum
current about 35 A.
I have done a little full-size TIG on a basic machine, but that's the
only TIG experience I have (I can braze and stick weld reasonably well,
have used a MIG but only a few times)
My question is about the controls needed. Looking at full scale TIGs, I
have tentatively planned to include:
pulse low current
argon flow rate
Is that about right? Is there a difference for micro TIG rather than
It isn't particularly hard to add more controls, as the output will be
controlled by a microprocessor.
-- Peter Fairbrother
Sounds like a fascinating project. Please post your results.
I think you will find it nearly impossible to hand feed filler wire
that small so you should also look into building a tiny cold-wire TIG
feeder. I have worked with TIG down to 5 amps, and manipulating wires
0.024" or smaller is very very tricky.
With pulser and sequencer this will be a unique little machine.
I might buy one from you.
I wasn't planning on selling any, more of a junkbox project. However I
could probably post the basic circuit design and the microprocessor code
at some point if you like.
Probably be built from a 750-800W server PSU and an atmega
microprocessor chip (think arduino). HF comes from a boiler ignition
transformer (driven by transistors - no spark gap).
I might build it starting from a 750w standard ATX PSU, which would be
easier - I can reuse all the mains voltage part of an ATX PSU as-is, but
would need to collect the mains voltage parts from the server PSU and
However the parts from a server-grade PSU would be much better, and more
robust. Still thinking about that.
Maximum current would be about 35A at 18V, say 30V no-load.
I could add a low current voltage source to increase the open-circuit
voltage to maybe 60V, would that be necessary or desirable?
Would an AC frequency control be useful?
AC would most likely be square wave, not sinusoidal - does that matter?
How about a finish current timer? That seems to make some sense.
An initial current timer?
Hmm, too many controls? might need a PhD or a kid to operate it ...
I'm still trying to get my head around Martin's post about an off timer.
Can't see the need - the sequence is switched on manually at preflow
beginning, and switched off manually at downslope beginning.
-- Peter F
On 14/02/14 21:51, Ernie Leimkuhler wrote:
One other question - would it be worth while having a control to vary
Most TIGs have a steeply dropping characteristic, ie in the operating
range the current only changes a little as the voltage (and arc length)
My initial idea was to have constant current control, which would
correspond to a vertical characteristic - the current stays exactly the
same as the voltage varies.
However I thought it might be useful for the current to drop slightly as
the arc length increases beyond a basic minimum.
Does anyone know anything about this, from the welding perspective? With
digital control t shouldn't be hard to simulate any characteristic.
-- Peter Fairbrother
On 22/02/14 09:00, Ernie Leimkuhler wrote:
On Mon, 24 Feb 2014 00:34:01 -0800, Ernie Leimkuhler wrote:
Slope, characteristic, droop, all used to describe the variation of
current with voltage in a CC supply or variation of voltage with current
for a CV supply. Nominally zero for a perfect supply but rarely if ever
for any real welding power supply, for reasons of stability and
usability. (How good is your PID controller tuning?). In the old days
of transformer based welding machines a fairly large amount of droop was
unavoidable due to the relatively low gain of the magnetic regulation
used, and AFIK no one ever had any problems with this characteristic.
Modern inverter machines allow for much stiffer regulation, but this is
not necessarily best in use, and I suggest you make droop programmable,
and be prepared for stability issues if you set it too close to 0.
Very low current TIG welding of small parts is usually best done semi-
automatically with some sort of motorized fixture which maintains arc
length and travel speed, and for this use droop is not important. If
however you intend to weld by hand, you will not be able to hold arc
length exactly, and the arc voltage will increase with increasing length.
Since power is volts x amps, with amps exactly constant but volts varying
with arc length, power in the arc will increase significantly with
increasing arc length. This is probably not what you want; you probably
want enough droop to approximate constant power with small variations in
I don't know why a slope adjustment would be used on a MIG supply, but it
might be to reduce variation in arc power with changes in the speed at
which the wire approaches the work from an unsteady hand. While the arc
length adjusts automatically for constant voltage, moving in faster ups
current and power and moving out decreases it, according to the speed of
the motion rather than distance.
I will second dcaster's suggestion for an open circuit voltage of 60 to
80 volts. I am fairly sure all of the Miller machines I have used for TIG
(Gold Star and Dialarc HF) use 80 VDC open circuit, but you might not
need quite that much in a small inverter machine.
Regarding HF start, I believe most inverter based welding machines just
use the high frequency from the H-bridge switches (bypassing the output
inductor directly or with a shunt capacitor) rather than building a
separate HF generator of any sort, not sure if that is the best way but
you might take a look at the schematics of some commercial inverter
welding machines to see how they manage the HF.
-- OT --
Not relevant to your use, but the easiest to use stick welding supply I
ever used had a lot of droop. It was a shipyard gang welder with (IIRC)
48 sections in an 6 x 8 array, each section about 12" square with a 4 x 4
array of quarter turn quick connector sockets wired to a resistor bank
directly behind, in a screened convection cooled box (no fan). Half of
the sections covered the range of 60 amps to about 250, and the other
half covered 150 to 400 with steps interleaved for 32 possible current
settings at mostly 10 amp steps. The resistors were powered with 80 VDC
from a separate large "12-phase" transformer-rectifier unit, with the
negative lead bolted to the hull. I used mostly 80 amps with 3/32" 7018,
on the end of 3 or 4 hundred feet of 4/0 cable. So for 25 volts at 80
amps at nominal arc length, the resistors needed to drop 55 volts at 80
amps, and the 80 amp tap connected to the 80 volt supply through about a
0.6875 ohm resistance (14 of the 16 resistors in series), short circuit
current therefore 116 amps.
The resistors were all heavy nichrome wire on ceramic bar supports pretty
much identical to what is used in high power load banks. When someone
slammed a 2" thick armor plate hatch on my cable for a nice dead short,
some of the resistors glowed a dull red for a few hours with no ill
effect. The complete lack of any moving parts (not even an on/off
switch; to turn it off you had to go out on the pier and cut the supply
power) and design for continuous dead short on any output resulted in
pretty good reliability; built in 1940's and still in service in '74,
with what looked like original factory paint on all fasteners like it had
never been serviced.
The complete lack of any deliberate inductance (fast, easy arc start) and
the high droop, providing more current on a short arc and tapering off
current as the arc is drawn out at the end for minimum crater size
(craters must be ground out on pipe, they always have a crack in them),
made for the easiest stick welding I have ever experienced, although the
Gold Star came close.
I wasn't planning on using a PID, just a proportional controller -
though of course the time constants etc make that work in a PID-like
fashion, but only the proportional signal is variable.
Sounds like good advice.
I'm a unsure what you mean here - could you be more explicit please?
The HF is used eg when starting an arc in DC mode, when the bridge
switches are not changing state. Also I am planning for arc detection in
DC mode, such that if the electrode voltage approaches the open-circuit
voltage the HF should kick in again.
If you mean the main PWM transistors driving the main transformer, they
work at 75 kHz, not the few MHz preferred for HF. The flyback pulses
could I suppose be placed in the MHz range by a suitable LC constant,
but I'd be worried about frying the diodes and H-bridge FETs.
Ah, if only :(
It may be that my Google-fu is weak, but I haven't been able to find a
single circuit diagram for a modern commercial TIG inverter welder which
goes to component level on the 'net.
Yeah, that's the kind of welder I'm hoping for - and why I'm thinking
about a droop control.
-- Peter Fairbrother
On Wed, 26 Feb 2014 06:22:54 +0000, Peter Fairbrother wrote:
I think we are envisioning different designs. Based on what I recall of
welding machine block diagrams, I was expecting your output current to be
produced by PWM control of the output H-bridge at 100 kHz or so, from a
DC bus higher than output open circuit voltage, with a small value
inductor on the output to filter out most of the 100 kHz while allowing
the output to slew fast to produce a decent low frequency square wave.
The switching edges of the H-bridge PWM will contain energy out into the
MHz range required to start an arc (consider the FFT of a square wave),
and this energy could be sent to the output by bypassing the output
inductor at least in theory. You will have a lot more high frequency
energy from the fast switching edges of MOSFETs compared to IGBTs, and
using the minimal snubbing required to protect the switch would maximize
that. Mind you I am just tossing out ideas for you to consider, I
haven't tried it and you should also be considering the capacitor
discharge HF start mentioned by
I suppose I should not be too surprised that schematics have vanished
from owners manuals with the switch to inverter technology. In the days
of transformer machines a schematic would tell the owner everything
needed for service without giving away the design details of the
transformer, but with inverter machines, mostly assembled from readily
available parts, the schematic makes it too easy to copy, and the
manufacturers would rather have owners buy $$$ circuit boards rather than
buy a two dollar part from Digi-Key. But I have seen articles in Power
Electronics magazine discussing inverter welding machine design, with
block diagrams, and some manufacturers of microprocessors have useful app
notes and sample code - recent advertisements in Power Electronics should
turn up the newest of these.
Back issues are available online.
Yep, I guess so.
I am thinking of using the mains voltage parts from a computer PSU, ie
the mains inlet filtering, rectifier, main storage caps, power
transistors, helper transformer, main 75kHz transformer etc.
The output from this would be smoothed with an LC, then the current
would be measured and this measurement used as feedback to the mains
voltage power transistors (via about half of the PSU existing control chip).
Then I was thinking of replacing the half-wave rectifier in the PSU with
a full bridge rectifier so as to get a nominal 24V, probably about 35v
O/C. However if a higher O/C is needed then I may change this, not
decided which option to choose yet.
I reckon on getting a maximum of about 30A at 20V from a 750W computer
PSU, that should do for a miniTIG.
I'd also be monitoring the electrode voltage so as to allow arc and
short circuit detection, and to control droop. A microprocessor would do
that. It would also output desired current signals for the sequencer,
and H-bridge (see later) control signals.
The main current control feedback loop is too fast for the atmega micro
I have in mind, so I would use the existing PSU chip instead - analog
read on the micro is about 10 kS/s, which is too slow for that (and for
analog short circuit detection).
However the micro is fast enough for droop control, supplying an analog
desired current level, digital short circuit protection, gas and
H-bridge control, driving a display, and everything else.
This current controlled output would then be fed through a FET H-bridge
so as to make variable polarity and AC operation available as well as
DC; and after that the hf would be added when so ordered by the micro.
I was thinking about a ~1MHz transistor oscillator driving a ~20x stepup
transformer, powered from the ~300V on the main psu caps. I don't like
using spark gaps here (or anywhere else for that matter, too noisy).
Power would necessarily be a bit high (!), but on times would be low.
That way, with a bit of filtering, I could get the HF to be fairly pure
1MHz - so you could eg touch a 5kV HF electrode without getting a
[ for any interested lurkers, the HF in a TIG start circuit stands for
high frequency, not high voltage.
It is of course high voltage as well, in order to make a spark, but at
1MHz frequency a high voltage doesn't conduct like it would at a low
frequency or DC - that would be more like a stun gun. The high frequency
current travels on the very outside of a conductor, so it doesn't affect
the nerves which are under the skin. That's why you should not get a
jolt from a well-adjusted HF unit - unfortunately, many of the available
HF units seem to have low frequency signals mixed in with the HF, and
often you do get a jolt
The other reason for getting a jolt is that the 80V main O/C voltage is
well enough to give you a jolt, but it doesn't usually do that because
the skin is too insulating - however the HF can make the skin conduct a
bit, and you then get a noticeable jolt from the O/C voltage, not the HF.
Or both, of course. ]
Ain't that the truth.
Thanks I'll have a look.
And thanks to both you and Ernie for helping.
-- Peter Fairbrother
I'd call it a mini TIG myself. Micro is tiny.
Looks like (in engineering concept) you have a pulse width modulated
initial current is minimum or baseline current.
Upslope - rise time of pulse -
Main current - top of the waveform or maximum current.
downslope - fall time of pulse
finish current - exit
Gas control exit.
What I see missing is off time. Which makes the repetition rate.
seems like there is a need to have the main current width and the finish
The repetition rate determines the heat.
seems to me the main current needs a width of time (before downslope)
and the finish current needs another width of time (before another upslope).
On 2/14/2014 8:14 AM, Peter Fairbrother wrote:
Yep. Taken from an adapted computer PSU.
Maximum DC current is about 35A at 18V, 30v no-load. I can add a low
current voltage top-up to get the open circuit voltage up to 60V or so,
This is fed through a H bridge (for square wave AC), a current detector
for the PWM feedback, then a choke (to slow down and limit the current
in any short circuits - the Schottky diodes are a bit sensitive to
overcurrent - and with a capacitor to ground to isolate the HF) then a
transformer to add HF.
There will be a zero-voltage detector on the torch lead to detect short
circuits, controlled by an interrupt in the main microprocessor, which
will set the current level, H-bridge status and HF status. Otherwise I
don't think I need any interrupts, though maybe the main current control
will need one. we'll see.
The rest is software. I know _how_ to write the software, but I may need
some advice on _what_ to write, eg when HF should be applied.
I don't quite follow this. I see the reason for a finish current
width/time, but not for a main current width or an off time - won't
these be manually controlled?
-- Peter Fairbrother
On Sunday, February 16, 2014 2:58:09 PM UTC-5, Peter Fairbrother wrote:
-- Peter Fairbrother
Having an open circuit voltage of 60 to 80 volts will make it a lot easier to use. When you let the tungsten get a bit away from the work, the higher voltage will keep the arc going.
Square wave AC is better than sine wave. With sine wave the arc tends to stop when the voltage is reversing.
Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here.
All logos and trade names are the property of their respective owners.