Some questions about my SMPS TIG Welder build...


New member
Dear Friends:
I am currently in the planning/designing phase of building a Switch-Mode TIG welder and I have some questions. Before anybody asks, I have built several lower-power <500W SMPSs in the past, as well as one 1000W SMPS(The miniature version of this welder). I've built a couple really tiny <10W SMPSs as well, but this welder will be my largest power supply I've ever built so I am taking many precautions. I do have a scope and a home-brew differential probe as well as several 100X probes and as soon as my next paycheck comes in I will be purchasing a real differential probe to help with the building of this beast. I am fully aware of the high voltages involved in such a power supply and the lethality of such voltages and current. Anyways, on to the project.

Input voltage will be supplied from 240VAC mains voltage, rectified, then filtered by 8x1000uF/400VDC capacitors. I ABSOLUTELY MUST HAVE a very robust soft-start circuit to charge these so that the inrush current is minimal and the voltage surge to the H-Bridge is minimal as well. Any ideas here? This will give approximately 340VDC across the H-Bridge.

Output voltage will be ~30-40VDC with a maximum of ~250-300A.

Output regulation will be variable voltage/constant current as is required by a TIG or Stick welder.

Current feedback will be supplied by TWO current transformers:
One on the secondary side to provide feedback for the normal operating range.
And one on the primary side to provide a shutdown mechanism should the transformer's primary current exceed the maximum limit.

To prevent a high OCV(Open Circuit Voltage) and to also prevent the PWM controller from running at max duty-cycle when idle, I am also incorporating an opto-isolated voltage feedback circuit on the output to limit the duty-cycle to the minimum when there is no load on the output.

The main switching transformer will be wound on double-stacked Ferroxcube EE80/38/20 cores.
The secondary side will be wound with 1.0mm X 54mm copper strip. 2-Turns.
The primary side will be wound with 30 strands of 0.75mm wire. 14-Turns.
This was calculated using the program from here(

The output inductor will be wound on a single EE80/38/20 core using copper strip. How do I calculate the amount of turns for this?

Output rectification will be a full-bridge provided by 8xSTTH200L06TV1 ultra-fast diode packs. 2 in parallel for each of the 4 diodes.
The STTH200L06TV1 diode packs contain 2x120A ultra-fast diodes(so 240A per package) and can easily be bolted to a very large heat-sink.
Does anybody know of an ultra-fast diode or diode pack that will handle more power so that I would only need 4 instead of 8?

Topology is a full H-Bridge running at 25KHz in hard-switching. Pulse-Width-Modulation will be provided by a TL494(Because I'm almost a master at using this chip and its VERY easy to control.) giving a maximum duty-cycle of 44%.

Gate drive will be provided by IR2110s that will be opto-isolated from the TL494 using an SFH6732:
This chip provides two isolators per chip and features a totem-pole output to ensure fast triggering.
I'm using the opto-isolators so that the TL494 and its associated circuitry can be earth grounded to provide isolation for the user controls since the IR2110 will be directly hooked to the H-Bridge.

Switching transistors will be IGBTs. My original plan was to use 8 of these: IRG4PF50WDPbF
I have since changed my plan and I intend to use 8 of these instead: HGTG30N60A4D
I decided on the change of device due to the fact that the latter IGBT(according to the datasheet) will handle it's full 60A@110*C rating below 30KHz whereas the former IGBT will only handle its full rating of 28A@100*C below 1KHz where the current capability begins to fall off sharply and is down to 10A at 25KHz.

According to the transformer calculator the peak primary winding current will be just under 50A at peak output current.
I calculated the transformer using a 300A output current but if I am able to squeeze out 250A of usable current I will be happy with this unit. If I can get a peak output current of 300A, I'm only shooting for a 20% duty-cycle at that amperage. This means that the power supply should be able to sustain 300A of output current for 2 minutes, then have a cool-down period of 8 minutes. This is how the welding industry rates the duty-cycle of their welding power supplies. At 250A I'm shooting for at least a 40% duty-cycle. At 200A I'm shooting for 70% DC. And below 175A I'm shooting for 100% DC, meaning that my power supply should be able to supply 175A or less continuously without overheating or damaging components.

I will incorporate a thermal overload protection using several temperature sensors placed as close to the switching IGBTs and the output diodes as possible to get the most accurate temperature readings.

As for the final output stage, after the output from the switching transformer is rectified and filtered it will then pass through another large IGBT H-Bridge using these devices: CM300-12NF from Powerex. These are dual IGBT modules so I will only need 2 modules to create a full-bridge. The datasheet says these devices can handle 300A DC and much more when pulsed. This output stage is so that I can provide a simple logic-level toggle switch to change the output from DC+, DC-, or AC output. When the output is in AC mode the frequency will be variable from 20-250Hz and the duty-cycle of the output stage will be fixed at the maximum amount allowed by the devices, with some added dead-time for insurance, to ensure that the maximum amount of current is passed through the output stage. The output current will still be controlled by varying the DC of the switching transformer.

The output IGBT modules will be driven by these gate drivers, also provided by Powerex: BG2B-5015
These attach directly to the modules and are the recommended drivers for these modules.

I think this is all of my plan, but if I've forgotten anything or have any more questions I will be sure to post a follow-up. Sorry for such a long post. I just wanted to detail out my project as much as possible so you guys can all get an idea of what I am undertaking here. I also hope everyone understands that this project will be very slow-going as I am not made of money and many of these high-current devices are very expensive, especially the IGBTs for the chopper stage and the output rectifier diodes.

Here is the design of the switching H-Bridge circuit board. I've tried to make it as compact as possible and I am planning on getting the board made with as thick of traces as possible(4oz). The board measures 4in by 3in and is double sided, although the only traces on the top side are for the gate drive. The switching transformer will be mounted directly below this bridge board to keep the leads between the transformer and the bridge as short as possible. They should be < 1/2inch long. The gate driver board will then mount directly above(on top) the bridge board to keep the gate drive leads as short as possible as well. Also the gate-drive leads will be made with double-shielded wire if possible. The IGBTs mount on top of the board with their backs facing outwards. The heatsinks will consist of 4in x 4in square tube with the fins on the inside of the tube. I will utilize two tubes, one for each side of the board, and the tubes will run the entire length of the welder's case with forced-air cooling blowing through the insides of the tubes. The air will intake at one end of the welder and blow out the other end. Anyways, here is the bridge board, green is the bottom layer of copper and brown is the top. If you don't quite get my idea for cooling, I will soon be posting a drawing of how I envision it.

View attachment Bridge Board.pdf


New member
Something I wanted to add. Since the transformer calculator put the peak primary winding current at ~48A. I would like to be able to run this power supply full-tilt off of a 60A circuit breaker. Will I be able to do this without any PFC? Also, does anybody have any recommended reading for building a high-amperage power-factor-corrector that I can incorporate into the mains-input/rectification stage. I know for this amount of power I will be needing an active PFC but none of my power supplies have ever used PFC so I'm wondering if the added complexity would be worth the effort in the first design. I can always build the welder without the PFC but leave space in the case to incorporate a PFC at a later time if I find that the current draw from the mains isn't ideal (ie: To much harmonics and too little power-factor).

Edit: One more thing. Would it be worth the effort to increase the switching frequency and make this a fixed frequency, ZCS, Quasi-Resonant design? My reasoning for choosing a low switching frequency(25KHz) and hard switching was just because that's how I've built all my other supplies, except the others were half-bridge where-as this welder will have a full-bridge. I've never built any type of resonant SMPS before and I have no idea how to calculate the components for such a power supply so I'm not sure if the effort would be worth it if I am able to stick to what I know and achieve my desired power level with a lower frequency(25KHz) and hard-switching.
Last edited:


New member
Well I managed to get my hands on Altium Designer. I needed a program that would produce Gerber files so that I could have the PCBs professionally made. It has a pretty steep learning curve but I'm fairly good with software and managed to get a basic handle on it within a few hours. I've redrawn my bridge board in Altium and got it to pass all of the design rule checks. Here are some screen shots of the board...

Bridge Board Layout.jpg
Bridge Board 3D.jpg
Bridge Board 3D Bottom.jpg

So far I'm liking this program. I think the board turned out well.


Staff member
Altium is very powerful software, Hope you will get access to the proper libraries.


New member
I downloaded 305MB worth of add-on libraries for Altium from here:
These add-on libs had almost everything I needed from tons of different manufacturers including, IR, Fairchild, ONSemi, ect, so these libs had my power supply PWM controllers and my IR gate drivers and all kinds of neat stuff. The C5 & C6 capacitors I made those components myself and compiled my own integrated library that I will use to add custom components or anything else that I need to add that isn't in the libraries. I'm really loving the Altium software, it's fairly straight forward to learn, and I'm happy that the guy from EEVBlog suggested it to me. The only downside to it is the very high price-tag, but well worth it. I won't need to buy Schematic/PCB software ever again in my lifetime I wouldn't think. Haha.

PS - Whats your thoughts on the board MicrosiM?

PSS - Really love the 3D views that Altium provides and the fact that when I define my components, I can set a height with them so I know what my vertical clearance is on the boards as well.
Last edited:


Staff member
The PCB looks neat, I am not sure if the PCB traces will handle the current you are dealing with, Also you need to design a proper gate drive circuit to drive this system

I really like your idea making a welder by yourself from scratch, this is what talented people do.

I wish to see some progress soon, I can suggest you some PCB manufacturers when you need.



New member
I'm going for a 4oz copper board to get the heaviest traces possible. If that won't handle the current(which I'm doubting it will), I will get some copper sheet and some plain PCB material(no copper) then use non-conductive epoxy to affix the copper sheet to the board backing and have it CNC machined into a functional board. That's the only other way I know of to get a board capable of handling 50A. Even though the 50A only needs to be handled for about 2 minutes then an 8 minute cool-down period will follow. I have to agree though, I don't think even a 4oz board will handle the 50A for even 2 minutes(20% Duty Cycle). This board design will be the first high-power model I'm building. I'm also going to assume that I will need to add more IGBTs to keep the heat in the devices under control. The devices I've chosen are said to handle 60A at 100*C but I'm not sure if only 2 in parallel will handle the power while keeping the heat in check. The heat-sinks will be massive and have as much forced air cooling as possible, but I'm trying to get all of this to fit into as small a space as possible.

Gate drive will be with multiple IR2110 devices as I've had excellent success with them when driving IGBTs(So long as the in-rush voltage is limited when the system is powered on with uncharged storage caps). PWM signal will come from a TL494CN. I know there are probably better PWM controllers out there, but I just love the way the TL494 works as far as being able to control the dead-time and duty-cycle. I also like the dual error amplifiers as well. I've all but mastered using the thing as I've used it in every PSU(>100W) that I've ever built just for its simplicity.

My main reasons for building this welder are for a learning exercise, and because the welder I really want(Miller Dynasty 350) costs just under $10,000 and there's no way I can afford that. I think I can build this beast for $1000-$2500 once I factor in all the prototyping and failure components. I'm sure there will be some failures.

I'm currently working on the mains storage capacitor charging circuit. With 8 or 10, 1000uF caps to charge, my charging circuit has to be VERY robust. I also have a 50A PerPole, 4-Pole, 200A Total, contactor that I will be using to bypass the charging resistors once the caps are charged above 300VDC. This contactor was pulled from an old 400A transformer based welder that required 240VAC@100A service to operate, so I think it should handle the input of this machine.


New member
Well after doing a bunch of research on the subject(I managed to find some good explanations and papers), I've decided that PFC is an ABSOLUTE must for this project. There is just simply no way for me to draw 48A worth of primary current on the switching transformer without a good power-factor-correction. The original calculations was running the switching bridge directly from the rectified/filtered mains power(~340VDC) which gave a 48A primary current on the switching transformer. Now that I'm going to be incorporating a boost-type PFC stage on the input, this will raise the bridge voltage to 400VDC and according to the transformer calculator, this will drop my primary switching current to around 40A.

Now, I've come across a few different standalone PFC controllers that use external switching devices, MOSFETs or IGBTs. In the datasheets for each of the controllers, there are example schematics, the most powerful ones being only 450W(400VDC@1.125A). I've emailed the component manufacturers to see if these controllers can be used for much more powerful PFC and I'm currently waiting for the replies from their engineering departments. If these devices can be scaled up to ~16kW then I will order some samples and build some small-scale PFCs for testing purposes to see how each of the controllers operate.


New member
The last welder I had was an old(1983) model that was an entire constant current linear supply. 16KW/23KVA transformer, ran off the 240VAC mains from a 100A breaker. Output on AC was 405A, DC was 315A, current was controlled by a massive saturable reactor. Really cool setup but alas the high-voltage(3500V)/high-frequency arc starter by-pass caps failed open and the HV that is coupled to the output to provide non-touch arc starting got back-fed into the reactor and transformer and broke down the wire enamel and caused a short inside the reactor and ruined the welder. I ended up scrapping the copper and iron out of the unit because I couldn't find parts to fix the thing anywhere in the US and even if I did, shipping would have been insane. Transformer was >250lbs, Reactor >300lbs.

The SMPS that I'm building, I'm only looking for a maximum output of 250-300A. I figure if I do enough research and studying I can take this knowledge and make a working unit. I've seen an example of a 100-150A Output SMPS stick welder online using a 2-Switch Forward Converter and the guy had no PFC and an absurdly terrible layout of his circuitry. I think if I put quite a bit of thought into my design, it will be doable. They make much larger SMPS Welders than what I'm trying to do, I think I should be able to get this. It might take me a really long time to get it all working, but be damned I'll figure it out eventually.

Output of my machine will be ~10.5kW: 300A@35VDC or 300A@35VAC(Square-Wave Chopped)

My quote of 16kW input is based off the 40A@400V primary current of the switching transformer calculator at the max output. My goal is to get this thing to run comfortably from a 50A@240VAC circuit breaker here in the USA(Other machines of this output rating can do this). This would be completely impossible w/o any sort of PFC. I ordered some chip samples of a couple different controllers, both CCM(Continuous Conduction Mode), and are supposedly able to be scaled up to a much higher power level than their example schematics. I'm running though all of the formulas in the application datasheets and calculating all the component values and will be drawing up a prototype schematic/board tomorrow. I'll be designing a small-scale, 1000W(400V@2.5A) PFC as a test-bed then if that all works out, I'll be going for the full-scale and see how it all turns out...Here's hoping! Cheers!


Staff member
That sounds intresting to me. I cant wait to see results.

PFC I am working at now is 1kw for audio SMPS. Your project will become reality soon. I am sure about this.

waiting for good results


New member
While I'm waiting for the PFC controllers and 450VDC bulk capacitors to arrive I'll be drawing up the schematic and circuit board for the PFC controller. The controllers are SMD so I will be having some prototype boards professionally made for testing purposes. I still have to get through all of the calculations for the boost inductor for the PFC and all the current calculations for the IGBT switches and the forward diode after the inductor. I think the hardest part about building this PFC is going to be load-testing it. I don't know what in the world I could use to load-test 400VDC@40-50A worth of power. The PFC has to be able to handle this power for at least 2 minutes to meet my duty-cycle requirements of the welder at the full 300A so I'm going to have to figure out a way to load-test the PFC as a stand-alone system so that I can be absolutely sure that the PFC is up to providing enough power to my switching stage.

Thank you MicrosiM for the kind words of encouragement. It's good to see that somebody is interested in this project and I will be publicly publishing all of my schematics and design notes as this project moves along so that other people can experiment with my power supply.


New member
Calculations are done for the PFC stage. I used all the formulas provided by the datasheet and application notes and here is what I came up with.

I'm going to try using an On-Semi NCP1654. It offers a current limiter for the switching inductor, Under-Voltage Shutdown, Brown-Out Detection, Over-Voltage Protection on the output bus voltage, Over Power Limitation on the input, and a Thermal Shutdown.

These are the values I used for the calculations:
VIN-Normal = 240VAC
IOUT = 50A
FSW = 65KHz
Eff. = 92% @ VINLow-Level
Vp-pMAX = 20VDC @ Max Load
POutMAX = 20KW

(1) - Coil Selection
LCOIL = 25.367uH

Now this is with a 20KW output(400VDC@50A) which is about 10A more than I will need for the full 300A output current of the welder. I just wanted some overhead built into this thing as a fudge factor. I'll most likely lose some efficiency but I'd rather lose some efficiency than build the thing and find out the PFC won't make enough power for the switching stage.

CBULK = 10,000uF@450VDC

I still have to figure out how many and what size(uF) caps I'm going to need so that I can get the 10,000uF bulk capacitance while making sure I don't draw too much current from the caps and overheat them.

(2) - Feedback Arrangement
RFBU(Upper Feedback Resistor) = 3.7M (1.5M + 2.2M)
RFBL(Lower Feedback Resistor) = 23.2K

These resistors form a voltage divider for the voltage feedback system.

CFB(Feedback Filter Cap) = 100pF

CP = 0.22uF
CZ = 2.2uF
RZ = 12K

These components (CP, CZ, RZ) form a compensation loop to set the feedback bandwidth. I used the given values and this will probably need to be tweaked during testing.

(3) - Input Voltage Sensing
RBOU(Upper Brownout Resistor) = 18.2M (9.1M + 9.1M)
RBOL(Lower Brownout Resistor) = 84.5K

These resistors form a voltage divider to set the brownout shutdown voltage.

CBO(Brownout Filter Cap) = 0.47uF

This cap filters the voltage from the brownout V-Div and prevents the controller from oscillating should the power flicker very quickly.

(4) - Current Sense Network
RSense = 0.00075R
RCS = 649R
RM = 110K
CM = 680pF

These components monitor the inductor current as well as the input load and controls the MOSFET/IGBT switching based on the current level of each.

I know all of these values probably mean nothing unless you're familiar with this controller. I'm just putting my design values out there just for the sake of sharing them. These values are for the full-power model of this device. I will also redo all of these calculations for a 1000W PFC that I will build first and use as a test bed to make sure I fully understand all the functionality of this controller before I try to build the full power model.

Also, I'm going to try to use the HGTG30N60A4D IGBT for this PFC as well, it seems like it should work well, and a few of them paralleled together should give the needed current capacity. The gate driver in the NCP1654 can source/sink 1500mA so it should be able to drive a couple of these IGBTs, if not(I haven't done the calculations), I also have a gate drive IC that can be used in a PFC application which I can use. The NCP1654 is in an SO-8 package. I've never used SMD components before so this will be a new design task for me, and for the 1000W prototype, I might just mount the SO-8 onto an adapter board to convert it into a DIP-8 chip so I can build the prototype as a through-hole board.


Staff member
Very nice.

That IGBT is very good choice, I have used it into my previous PFC 1.7KW using UCC28019 chip, and results was good.

I am sure that you will have to use a Gate drive IC, its a must into this project, with proper PCB design, things should be fine.

Working with SMD parts wont be a hard task, all my prototypes now are SMD.

Also a proper soft start circuit will be needed here.

Very nice work really.

Good luck


New member
Yes a very robust soft-start/cap-charging circuit will be on the input, once the caps are fully charged to the max line voltage(~340VDC) then a large high-power contactor will short the charging resistors, then the PFC will take over and finish charging the caps to the full 400VDC. The PFC controller does have a inrush current detector and soft-start built in. I'm not dead-set on this PFC controller(It's just one of the more simple and I was told it was easily scalable to higher power levels), one thing it is lacking is a shutdown pin to shut it down remotely but I have made a simple add-on to my start-up sequencer to work around this. The switching stage itself will have a soft-start, not too soft because the output current must ramp up quite fast(<1 Sec) when the welding arc lights off and this will have a hard-set value. The output will also have a ramp-up/ramp-down control in the weld sequencer.


New member
Buy the way what is a PFC - Private Frist Class ?:confused:

Hi Guys

I'm am following this tread, nice to see people DIY things out on a steep learning curve, gives me hope with my little project.

Buy the way what is a PFC - " Private First Class ?" :confused:



Staff member
Yes a very robust soft-start/cap-charging circuit will be on the input, once the caps are fully charged to the max line voltage(~340VDC) then a large high-power contactor will short the charging resistors, then the PFC will take over and finish charging the caps to the full 400VDC. The PFC controller does have a inrush current detector and soft-start built in. I'm not dead-set on this PFC controller(It's just one of the more simple and I was told it was easily scalable to higher power levels), one thing it is lacking is a shutdown pin to shut it down remotely but I have made a simple add-on to my start-up sequencer to work around this. The switching stage itself will have a soft-start, not too soft because the output current must ramp up quite fast(<1 Sec) when the welding arc lights off and this will have a hard-set value. The output will also have a ramp-up/ramp-down control in the weld sequencer.

To tell you truth about these simple chips for PFC, I really like how simple they are to deal with, but it wont be easy to control it ON/OFF for example or soft start. and this is not good into my opinion. I like the chip that once you apply power to it, it starts slowly, not direct ON. this is not good, and I dont like it. I like soft start. like the SG3525 SMPS controller, it has SS PIN so you can select the time using one capacitor.

I recommend, on later stage of improving your PFC stage to select a better chip with SOFT START, so you will be able to shutdown the PFC when the unit gets hot for example.

Otherwise the PFC will keep running. also another choise, and it will be more complex, by going to interleaved PFC option.



New member
Wow, thats a MASSIVE PFC. Those IGBTs are almost the size of my output chopper devices. I really like the interleaved idea, I could make two smaller PFCs which might be more manageable. These NCP1654 chips have a built in soft-start. The way I understand it, they use the signal from the current sense resistor network to monitor the inrush current and the start-up current and slowly ramp up the duty-cycle to keep this current in check so things don't get out of hand. The remote shutdown wasn't too hard to implement with one of the pins on the controller. When it's over a certain voltage threshold it shuts down the switching device and it has a hysteresis on it as well to keep it from oscillating. Like I said, this is just the first PFC chip I'm trying, if it won't handle what I'm doing I have no problem stepping it up to a more advanced chip which I have my eyes on a couple.

Thank you very much for that link to that other guy's project, that is pretty sweet and I can learn a good bit from that. Thanks for all your help so far. I can't wait till these parts get here so I might have something to show you guys. Haha.