ARC WELDER FROM A SINGLE TRANSFORMER
Having the need
for a really powerful electric arc welder and not the ready cash to run out and
buy a new one, we considered constructing one, since there seems to be little
inside a welder other than a transformer and occasionally a fan. Below, is pictured the inside of a 120 Amp AC
hobby welder with its 22lb transformer.
Several such projects are to be found on the net, using such
things as a bank of 8 MOT’s. Seeking a simpler approach and having a
supply of several large (40lb.) transformers salvaged from old IBM 5360
mainframe computers and x-ray control consoles; we decided to make this a snow
day project.
The transformer
chosen seemed ideal for several reasons.
First, this 41 lb. transformer was wired with a stout 240 VAC primary,
large shunts, and multiple secondaries (some wound with #8 wire and even 3”
wide copper sheet). These were intended
to deliver between 5 and 12 volts at upwards of 100 Amps for the huge disk
drives and power hungry boards of these old computers.
|
|
Considerable
space was evident between the secondaries and the frame. The removal of the secondaries, however, is
every bit the noisesome a chore as reported by
others. We first spent a good hour letting
a metal band saw eat down through the windings protruding from one side. This got us within a bout ½” from the bottom
of the secondary windings when tension broke the band saw blade! Not to be daunted by minor setbacks, we spent
the next hour or so with a hammer, chisel, screwdriver, wire clippers, and even
a large bolt cutter. Ultimately, this
brute gave up its last winding, leaving sore fingers and a considerable mess on
the shop floor. The huge vacant space
revealed will allow the use of #4 or even #2 THHN for the secondary
rewind. The magnetic shunts were then
tapped out.
|
|
|
|
This was the main
base transformer for the 5360. This 1985
era mainframe was wired for 240 VAC 20 Amp service for the base unit with no
expansion cabinets. Assuming about 15
Amp average current draw, the primary was around 3600 VA. At 100% efficiency, this would support up to
a whopping 720 Amps at 5 Volts on the secondary (360A @ 10V, 300A @ 12V, 180A @
20V, and 150A @ 24V). We decided to
shoot for the 10-12 Volt no load secondary at 300-360 Amps with the knowledge
that the heavy load of arc welding would drop this to between 2 and 4 Volts
nominal. The primary consists of about
150 turns of #12 magnet wire.
Approximately 6-8 turns on the secondary would give the appropriate
voltage reduction.
Desiring to use
the largest AWG possible due to the output current and the fact that magnet
wire of extremely heavy gauge is out of the question from an expense
standpoint, our choice was between #4 and #2 THHN. The cost difference of 10¢ per foot made the use of #2 AWG at 59¢ per foot the obvious choice and 15 feet
was more than enough. Rewinding with #2
THHN proved a considerable challenge given its size and lack of flexibility,
but half an hour of strenuous effort closely resembling alligator wrestling
resulted in 6 turns of wire in place – after a fashion. I doubt there is any way to get a ‘tight’
wind with something this big. The heavy
insulation was not damaged but the thin shiny dress coating was shredded in few
places from the unavoidable friction of pulling the heavy wire through the
frame.
Output voltage
was measured at a little over 10 volts at the secondary with 246 Volts in at
the primary. A mild hum was produced
when the unit was powered up with no load.
Shorting the secondary instantly melted the end of the copper wire and
almost brought the transformer up off of the floor; so it appears we have a
winner here – easily capable of sustaining welding level currents. The relatively light 30 Amp fusing on the
circuit was unharmed. Due to the
considerable excess of THHN, 2 more windings were later added, which brought
secondary voltage to about 13.5 Volts.
The thought behind this was that windings can always be removed if the
additional 3 volts proves problematic, but adding more once the #2 was cut off
would be next to impossible.
As a sort of
preliminary test to insure that we were heading down the correct path, battery
jumper cable clamps were installed on the ends of each output lead. A 9” length of standard steel coat hanger was
secured into one clamp and the other was clamped onto a piece of ¼” plate
steel. The unit was energized and rod
brought into contact with the work briefly as a single tap. The resulting spray of sparks was quite satisfactory. The rod, however, had turned incandescent red
over its entire length. A second tap
bought the rod to a bright yellow-orange color, persisting even after contact
was broken. An attempt at that point
strike an arc and draw a short bead across the work surface immediately brought
the steel wire to white heat followed by complete liquefaction of the entire
steel rod - the resulting puddles of liquid steel running across the work
surface. Inspection after de-energizing the unit found the transformer and
cables not even warm.
With the addition
of a heavy duty SCR and a control circuit for its gate, we hoped to gain the
ability to choose between AC, +DC, or –DC type welding. An NTE 5588, 1600 Volt PIV,
355 Amp average forward current, 5000 Amp peak surge current, SCR was in one of
the parts bins. Since SCR’s emit 1.5 Watt as heat for every Ampere conducted,
up to 600 Watts may need to be dissipated. For this reason, it was mounted on a
large heatsink with a 240V Muffin Fan directed across it.
Gate trigger
voltage for this SCR is between 0.25 and 3.0 Volts. Preferring an isolated
control circuit, about 5-6 turns of #14 AWG THHN were wound around the
transformer center leg, giving about 8 Volts that dropped to 6 VDC when pushed
through a full wave bridge rectifier. A
2-stage filter was then connected to smooth the DC ripple. This circuit was connected between the gate
and cathode of the SCR to provide a flat and stiff DC current that keeps the gate
triggered to the on-state. Below, is the schematic.
FAILURE!
Although the
foregoing design and experiment seemed promising, the results did not bear out
the theory as expected. We were simply
not able to generate high enough current levels to be useful despite the
theoretical potential. Postmortem
analysis reveals the inability to get the heavy wire wound snugly against the
transformer frame resulted in tremendous efficiency loss. The magnetic field drops off sharply by the
Law of Inverse Squares and the field lines were just not strong enough an inch
or so out from the frame to induce significant current.
The second factor
was not so apparent and only surfaced after speaking with a local authority on
welding. The output voltage chosen turned
out to be way too low. It was based on
my reading of several hobby websites which seemed to indicate that welders
should output around 10 volts with no load and drop to below 5 volts during
welding. We obtained the use of a small
arc welder for comparison, and to our considerable surprise, its output no load
voltage was 45-50 Volts which dropped to 18-20 Volts during welding. This corroborated what the welding specialist
had said. No doubt it was time to return
to the old drawing board.
One possible solution
was considered and rejected for cost reasons.
It seemed fairly obvious that removing our secondary and rewinding
tightly with very large magnet wire and 4 times the number of windings would
probably solve the problems. However, a
cursory examination of sites offering magnet wire revealed that the cutoff is
around #12 for most companies though we did find one site that offers
#8 wire.
Further, the cost per pound is high and a pound of the big stuff is only
a few feet.
In the interim, a
novel solution presented itself. We have
a number of large to very large (30 to 150 lbs.) autotransformers lying around
collecting dust. I examined one in the
55-60 lb. range to find that it was entirely wound with stout magnet wire of at
least #8 AWG. Consulting the schematic,
I discovered it was designed for input voltages between 220 and 260 Volts to
give an output of 0-300 VAC full scale and handled a couple hundred Amps in the
equipment from which it was removed. It
was provided with 17 output taps over that range.
Since output
voltage of an autotransformer is directly proportional to its coil length, we
postulated that it should be possible to calculate the results of converting it
to a traditional type transformer by splitting the coil into 2 coils – a
primary and a secondary - at one of the tap positions. With 245 Volts as the intended primary input
voltage, a transformation ratio of 1 in 10 would yield a secondary out put of
24.5, which is exactly half of the 49 Volts desired. Therefore, a turns
ratio of 2 in 10 is then what is needed.
Assuming the
autotransformer to be relatively uniform in winding, then the tap with a
designated output voltage closest to 80% of the full scale voltage is where the
coil should be severed. Since 80% of 300
is 240, the tap at 238 Volts was chosen. The protective covering was removed and the
tap loop cut at its center. Continuity
check with the DMM resolved the identities of the new primary and
secondary. The primary was then
energized and the voltage ramped up slowly with a Variac
until it reached 120 VAC. The output of
the secondary was exactly 24.5 Volts! It
followed then that when we switched to a 240 VAC source, we would have the 49 Volts
we were seeking.
The only
noticeable drawback to this new design was that the autotransformer had no place for
the insertion of shunts to control the output current. Two solutions came to mind. The first would have been to have designed some type of
electronic modulation circuit – interesting but complicated. Another approach would have been to have rewinded our first
transformer with magnet wire, shorted the secondary, installed large adjustable
shunts, and put the primary in series with the input line to this new
transformer. This would have been a variable
inductor to control input current draw.
And since we know that power in on a transformer equals approximately
power out, this would in turn have regulated the output current.
As it turned out,
we certainly needed a heavy duty current control method for this brute. Full power testing was done by shorting the secondary
and applying 240 VAC to the primary. The
first source circuit had 30 Amp max and the breaker instantaneously tripped
before the current clamp on the secondary could give a reading. The primary was then energized using a 50 Amp
circuit. It took about 1-2 seconds
before the tremendous heat melted a soldered lead and broke the circuit, but
not before the current meter displayed 585 Amps. Wow!!!
If VApri = VAsec at 100%
efficiency (probably more like 92-95% in reality but we will ignore this), we have
240 (X) = 50 (585). Primary draw may
have been on the order of 100-150 Amps and would most certainly have tripped
the breaker if the circuit had not failed first. This, of course, is way beyond useful welding
current and had to be throttled back. The
first step was to place a current limiting inductor in the input line
capable of withstanding the savage power draw.
We have a 22 lb. dual scale adjustable inductor that has a reactance
range of 66 to 71 Ohms on one scale and 7.5 to 28.5 Ohms on the second and more
useful scale. Using this second range,
we able to limit the current draw to between 8.4 and 32 Amps thus allowing
comparison with output current to find the exact relationship. Once this was determined, a precision inductor
was designed to allow adjustment of current in the useful welding range
(50-250 Amps.).
With the
resistance of the work and electrode substituted for the dead short above, the
maximum secondary current was reduced to about 360 Amps. The results of testing with the inductor
showed that even at 7.5 Ohms reactance, the current reduction is still too
great. A maximum output current of
around 60 Amps was obtained. By placing
a second inductor with a reactance of 4.5 Ohms in parallel, we obtained a low
end output current of 120 Amps, which has no trouble welding plate steel.
Now that we see
that we have a successful welder, we will replace our experimental inductor
with a permanent current control. A
small rotary tap autotransformer might give a workable solution but we were
seeking a solid state solution to hold down the weight. Unfortunately, multiple attempts to design a
front end adjustable current control using a large dual SCR power module in a
circuit resembling a light dimmer failed.
The circuit would control voltage in the same manner as a light dimmer
to a small resistive load (actually a light bulb) but had no effect on current
draw to the primary of our welder other than to act as an off/on switch at the
limit of the potentiometer.
Multiple examples
of inductive ballast were tried with the expected range of variation in current
control. We finally settled on a small 10-tap
autotransformer that resembles a tiny version of the power transformer. The rotary tap switch was still connected to
the taps and the weight and footprint are small. Under actual welding conditions, the measured
secondary current for the 10 settings ran from about 30 amps to about 190 Amps.
The external
cables for welding were constructed of #1/0 AWG welding cable of about 12 foot
length each. To one was connected a 600
Amp ground clamp:
To the other, a
300 Amp electrode holder:
the AC, DC, and Common connections to the
unit were made via removable interlocking LC-40 type connectors:
This fun and
instructive project set out to build a useful welder with a single salvage
transformer. We ultimately had to use a
second much smaller transformer as ballast but in essence we have accomplished
what we set out to do.
The 60 Hz half wave pulse from the SCR did not give smooth
results for +/- DC welding. For this
reason, we to replace decided this rectification method with a full-wave
bridge. The SCR, heatsink, and computer
power supply 5 Volt source were removed.
Our inventory of diodes/thyrisors included a half dozen or so GE A 71PB diodes. These
are basically equivalent to JEDEC # 1N3296 and we have both standard and
reverse polarity types. These have a
average forward current rating of IAF = 125A and therefore an RMS or continuous
on-state tolerance of IRMS = 250A, which is also a
Two large solid aluminum heatsinks were tapped at each end
for the threads of the diode studs and a pair of standard
polarity diodes were inserted into the ends of one and a pair of reverse
polarity units into the other heatsink.
The heatsinks were mounted side-by-side about 2 inches apart. The braided cathodes of the two diodes on one
side were secured to the braided anodes of the diodes opposite them, creating
the bridge. AWG #2 THHN was run from
either side of the transformer secondary to one of the afore-mentioned
connections, providing the AC input to the bridge.
This then makes the heatsinks themselves the positive and
negative output terminals. Additional #
2 THHN was run from each heatsink to the LC-40 female output connectors on the
front plate and labeled + and – respectively.
The 3rd LC-40 female was connected back to secondary of the
transformer with #2, bypassing the bridge, to provide one permanent AC welding
connection. Should AC welding become
desirable for some reason, a shunt is placed across one of the DC lines to also
bypass the rectifier and provide the other AC leg.
Both DCEP (
In high-power AC mode at around 360 Amps, the metal fairly
explodes when the rod is brought into contact.
Sheet metal is cut completely through even with a surface rod like
E6013. The eighth inch rod, however, is
unable to stand the tremendous heat generated.
After about 1 foot of cut, the rod becomes cherry red and will melt
completely if an attempt is made to use further before it cools. Obviously, a heavier rod or gouging setup is
indicated at these current levels.
We later came into possession of several monster diodes that have an average continuous forward current rating of 300 Amps (nearly 600 Amps RMS) and a maximum repetitive peak current of 1400 Amps!. We could not resist rebuilding the the rectifier bridge to incorporate these into the design
The control circuit was added on a smaller second level board. The 'OVERDRIVE' switch supplies 240 volts to the 20:1 step down transformer. Its 12 Volt output is rectified and supplied to the coil of the high current relay to switch the adjustable inductor loop in or out of one leg of the primary input line. We had a 600 Amp to 5 Amp current transformer but our meter was 300:5. Using these 2 together necessitated changing the scale on the meter. As a temporary measure, the numeric decals seen in the photo were placed meter cover.
The above welder has been such a successful project and enormously useful tool that we decided to make another unit and go as large we possibly could with materials on hand. Our largest autotransformer weighs a little over 150 lbs. Wound with AWG 6 magnet wire, it was intended for up to 400 Amp service. We split the winding at a tap that gave 72 Volts open circuit on the secondary at a maximum of over 450 Amps output when 240 Volts is applied to the primary. The primary current draw is less than 100 Amps. Below, is a photo of the transformer and secondary output curcuit set into the steel frame made up with the welder contructed above.
In the next view below, the secondary circuit is seen from above. The secondary output is taken from the transformer using heavy aluminum block connectors with set screws. The wiring to the aluminum blocks on the board is AWG 4 bare copper. A current transformer will be placed around one of these and run to an ammeter on the front panel. The output is then taken to the rectifier bridge using AWG 2 stranded THHN.
Below, is a closeup of the rectifier bridge. We were able to acquire some very high current diodes at a price so low that we couldn't pass them up. They are Ruttonsha 300 UM 120 diodes − 2 standard and 2 reverse polarity types. These have a 1200 Volt PIV and forward current average of 300 Amps (about 450 RMS Amps), with a repetitve forward current maximum of 1400 Amps. Two standard polarity diodes were inserted into one heatsink and 2 reverse polarity diodes into the other. The braided terminals were connected between pairs to form the bridge and the input is to these connections. This arrangement then renders the heatsinks themselves as the positive and negative output terminals for the unit.
A second tier will be added to contain the control circuitry. In this unit, output current is controlled with a solid state control curcuit on the primary side. It was built around a high current dual SCR block with adjustable gating. This circuit is pictured below.
