TIG Welder and Power Control

Block Diagram Arcstarter Power Control Construction Schematic Diagram More Drawings References Return to Home


This project started from seeing pages describing arcstarters that could be constructed for practically no cost from stuff just laying around in the typical tinker's shop. It has been a fun project, but I have a basement crammed full of stuff and I still had to spend over $600 to build the thing.
My arcstarter section runs at 300kHz, uses Miller arc points, a homemade ferrite core coil/transformer, homemade multi-capacitor modules for the filter and tank capacitors, and generates a tremendous amount of radio-frequency interference. I guess you would say it is a typical homemade arcstarter.
On the other hand, the power control section is not typical of homemade circuits. Power control is linearly contolled from zero to full power at no, partial, and full current. The circuit uses zero-crossing detection and multi-pulsed gate triggering to ensure symetrical firing of a reverse-parallel SCR module.


Block diagram
This block diagram includes the original arc welder that is connected to
the Arcstarter and Power Control with a plug-in cable.
Back to Top


Making a decent looking case has always been the slowest part of a project for me. I use to have access to a metal working shop, but now I can only do it at home. So my case started as a computer case that I cut down to a size close to the size of my arc welder.
Bare painted chassis Front of finished case Rear of finished case
The arcstarter section is on the left. I wanted to put the arc points at the rear so I could adjust them by removing a panel. Due to the available space, I had to move the arc points to the side. This location also provides a way to incorporate the ceramic tile insulator into a plenum to cool the points and keep the UV light from the point arcs away from eyes.
Points-side view Back-of-points plenum Aft view Forward-view Left-right side view SCR-side view
A fan at the front cools the tank coil secondary and the MCM. The cooling air then is divided, with part flowing past the arc points, and part flowing over the heat sink for the power control SCR module.
Control of the arcstarter and the welder is on the right. A relay turns the arcstarter on and off as needed. Shielding gas is turned on and off by a a solenoid valve, and the SCR module controls the arc welder power level. The metal screen between the left and right sections is to keep the rf energy confined to the arcstarter section. The arcstarter generates a lot of radiation that can mess up the operation of the power control board and any electronic device on the house's electical wiring if not massively suppressed.
Back to Top


There are four portions of the arcstarter that required research, design, and tinkering. They were the spark points, the filter capacitor, the tank capacitor, and the tank coil.
Two spark point sets (four spark points) are needed to to reduce the input voltage to each set below the quench voltage at the end of each 60 Hz half-sine cycle. Spark points have to dissipate about 30 watts of heat from the four spark points. Attempts to use 1/8" tungsten rods were not successful because the heat could not be removed from the rods to a heat sink fast enough. The result was that the gap decreased as the rods heated and the tips of the rods deteriorated. Commercial Miller spark points (PN020603 from Airgas) were too expensive, but I had to bite the bullet and purchase them. I mounted them in heat sink holders made from 1/2" thick aluminum plate.Each point holder has two pieces 2" x 1-1/2" with three 1/2" fins cut into the 1-1/2" side. I get 27 square inches of surface for each spark point
I used a 6 kv oil-ignition transformer for my high voltage source. The peak voltage on the filter capacitor should therefore be 8484 volts. I wanted at least 500 pf of capacitance and I was planning on using 1600 vdc rated capacitors. Six, .003uf caps in series would give me 500 pf @ 9600 volts. I had planned to keep the capacitor voltage rating better than 1.5 times the peak voltage because the capacitor rating is a vdc rating. An opportunity to very cheaply buy six, .0056 pf capacitors rated at 2kv induced me to build a 930pf,12kv filter capacitor. The filter capacitor and the tank capacitor both are constructed on polycarbonate frames with 1/2" spacing between capacitors and 8.2 megohm bleeder resistors across each individual capacitor.
MMC capacitor
The capacitor and resistor leads were twisted together and soldered, then inserted into holes in the polycarbonate housing. Epoxy glue was used to cover the soldered wires in the housing holes to prevent corona.
The peak-to-peak voltage on the tank capacitor is double the 8484 volts on the filter capacitor so a voltage rating of over 24kv should be the target. However, I compromised and used 14, .05uf @1600v capacitors in series for a 3570 pf @ 22.4kv capacitor. This was necessary to keep the capacitance low enough for the tank circuit, meet space requirements, and allow using surplus capacitors to keep spiraling costs down. The 22.4kv capacitor still has a peak-voltage to dc-voltage multiplier of 2.5.
Selection of the size of the tank coil has been a subject of discussion on many of the forums where arcstarters are discussed. I used a piece of 1-1/2" PVC pipe as my coil form and wound the primary on the inside and the secondary on the outside. Fourteen turns of #22 wire on the 1-1/2" inside radius, spaced at a 1/4" pitch, gives a 5" long 1.94 uh free air primary. With a 3570 pf capacitor, the resonant frequency is 1.9 mhz before a ferrite core is added. A ferrite bar was constructed from seven line cord interference suppressors bought on ebay. Each suppressor has two, 1.3-inch-long pieces of ferrite. One piece was cut in half and each half used at the ends, then the rest were put together offset by one-half the pieces length. The resultant 9" long ferrite bar with a hole in the center was wrapped with four layers of Kapton polyimide tape. The final coil inductance is about 30 uh and resonates with the 3570 pf capacitor at about 300 khz.
RF tank primary Tank coil and ferrite side view
The primary coil is wound on the inside of a pvc form. After spacing was perfect(?), the ends of the wire were epoxied in slots cut in the pvc and coil dope was applied to the windings."
The coupling coil is 14 turns of #6 bare copper wire wound 5" long on the outside of the PVC pipe. The tank coil is mounted immediately behind the front-panel cooling fan, so #6 wire is big enough to handle intermittant 175 amp welding current.
Another subject of discussion on many of the forums where arcstarters are discussed has been the welder bypass capacitor values. Based upon the experimental data by John Phillips, I finally settled on 9.4 uf of capacitance in parallel with a 10 kilohm, 10 w resistor. I had to put the bypass network above the input "stick" terminal and connect to a bus connected to the input "work" terminal. The network leads should be kept as short as possible, and are far too long in this position.
Back to Top

Power Contol

A double-sided circuit board holds the power-control and function-selector circuit. There are a lot of discrete components used in this circuit because I have them. It also made layout of the board easier. The function selection circuitry may not be to everyone's liking and could be stripped out and replaced by a rotary or other switch. I like the gas time-delay timer. It is very linear and uses a unique negative-slope linear ramp generator. But the best circuit on the board is the power-contoller. It uses a positive-slope linear ramp generator for phase timing, and the power-selection potentiometer is compensated to provide linear power selection. A simple oscillator in the gate driver gives multiple trigger pulses so the power SCRs will fire symetrically regardless of the phase lag variability due to an arcing or open-circuit welder. Some of the seemingly un-necessary components are the result of SwitcherCAD III simulations.


The mode selector is a single pushbutton clocking a shift register. As the logic 1 is moved through the register, it activates different functions through a diode matrix.
The following table illustrates the operation:
Reset, 0 1 0 1 1 1 0 0 0 0 0
Push-1 0 1 1 1 0 1 0 0 0 0
Push-2 0 1 0 1 0 0 1 0 0 0
Push-3 0 1 1 0 0 0 0 1 0 0
Push-4 0 1 0 0 0 0 0 0 1 0
Push-5 1 0 1 1 1 0 0 0 0 1
Reset, 0 Y Y Y (1) D17, U4D, U4C, U1C, Q23, Q17, Q16, D16, Q6, Q7
Push-1 N Y Y (2) D21, Q21, D27, U4D, U4C, Q16, D25, Q6, Q7
Push-2 N Y N D22, Q21, D28, U4D, U4C, Q16
Push-3 N N Y (2) D23, Q21, D26, Q6, Q7
Push-4 N N N D24, Q21
Push-5 Y Y Y (1) D17, U4D, U4C, U1C, Q23, Q17, Q16, D16, Q6, Q7
(1) Variable power with TIG torch plug or foot control plug, else panel power control.
(2)Variable power with stick plug or foot control plug, else panel power control.



Power supply diodes provide a positive half-sine signal to a comparator. The comparator output goes to +12 volts when the half-sine goes above 0.7 volts. A constant current is supplied to capacitor C2 by Q8 and thus a linear ramp is applied to the second comparator. Power-panel control or torch control sets a reference voltage on the second comparator. When the charging voltage across the capacitor exceeds the reference voltage, the comparator output goes positive and starts the trigger oscillator. A two transistor oscillator,Q1 and Q2, has the trigger transformer as a collector load for Q1 and generates short duration pulses during all of the SCR on time. This ensures that a lagging current will not go to zero after the trigger pulse and cause the SCR to skip a part of a cycle.
Transistors Q9 and Q10 are held off during the half-sine period. When the output of the first comparator drops low, both transistors are turned on. Q9 discharges the ramp capacitor, and Q10 pulls the output of the second comparator to ground. Without Q10, the hysterisis from R12 of the second comparator would not allow power control to near zero levels.


Flip-flop U3 is a bounce and interference suppressing circuit for the torch or foot control switch. It enables the the starter relay through U4D, the gas relay through U4D, and the SCR trigger oscillator through Q23 when allowed to do so by the mode selector.
With no foot or torch control plugged in, U3 is reset, Q23 is turned off allowing full power contol, and U4D is enabled allowing gas relay and starter relay contol to be determined by the state of U2. The absence of a ground through plug pins 7, 8, or 9, causes Q5 to turn on and Q3 and Q4 to turn off. This isolates the external power reference voltage control network of R74 and R75. It also connects the panel power reference voltage control network of R6, R72 and R73.
Back to Top


Schematic diagram
Back to Top

More Drawings and Pictures

Back to Top


Several of my references have disappeared. If anyone knows where any of the Dead Links have been relocated, please let me know.
Dead Link: http://www.geocities.com/CapeCanaveral/Hanger/6160/welder/
Bill the Arcstarter's page. This is where I got the idea I could build a cheap arcstarter. Not recently updated and has some inaccuracies. Definite must-read for someone thinking of building one of these. Check the rest of the References for some alternate ideas on some of the construction details.
Dead Link: http://www.geocities.com/CapeCanaveral/Hangar/6160/welder/arcstarter/batch1/
A disagreement between Bill the Arcstarter and Bill the Arcstarter on what the tank coil should be.
Dead Link: http://forums.diywelder.com/forum/index.php
Forum discussions by John Phillips, Brian Tiedermann and many more on arcstarter construction.
Dead Link:http://forums.diywelder.com/forum/viewtopic.php?t=365
Discussion of John Phillips' bypass capacitor experiments. Dated Oct. 02, 2002 (posted Sept. 22, 2002).
On Semi AN1048/D, Fairchild AN-3004, Teccor AN1003,
Application Notes: general SCR characteristics, triggering transformers, Snubbers that may not be needed, etc.
ST Semi AN308,
Triac control applicaton note showing a multi-pulse trigger (figure 7, figure 11). May be simpler than mine. Also illustration of the inductive load control problem.
Pictures of Miller arcstarter construction.
Dead Link: http://users.tpg.com.au/john_c/
Pictures of Lincoln arcstarter construction.
Capacitors for building a MMC.
Tesla coiler forum with info on MMCs.
SwtcherCAD III
A really good circuit simulator. It has a large library of Linear brand devices (they are giving the program to you after all) and it is easy to add existing Spice models or make your own. Free
Back to Top

Copyright Dale Thompson.
Last revised: 16 August 2006.