Some questions about my SMPS TIG Welder build...

kees54

kees57
Hi There


I have played some with LTspice, after some learning (special that convergence trouble thingx-()
I have now simulate a fase shift LVI soft switching version, and a normal resonance version.

It do nicely, give 220 amps, but this is afcourse to much for me, I use 160 amps.

however the model for fase shifting is or not good because I have no shift making no voltage but this
I do find later, maybe the model of the uc 3875 is not well.

Maybe I go build such soft switcher, it give a lot less grid harmonics and more efficienty with higher switch
frequentie. output voltage with 220amp was 70 volts, load resistor 0.3 ohms.

regards

kees

ScreenHunter_532 May. 05 20.30.jpg
 

kees54

kees57
Hi All


The softswitch welding inverter do well, I am busy with the LVI version who has a coil that automatic adjust with current.

The ltspice sim did let see that it works well, using IGBT with diodes in them. See the blue line covering soft switching,
it last from 30 amps to max, but it is sim, not real yet because I need the LVI coil with core perma 1290 earea 4.32 cm2,
magnetic length 19.2 cm saturable two ring cores
so I have to search..

ScreenHunter_539 May. 09 18.54.jpg

ScreenHunter_541 May. 10 12.57.jpg



regards

kees
 
Last edited:

kees54

kees57
After find a site http://diy-welder.com I have try his idea of a feedback, credit to him..

Thanks to him I have with changes get a pretty good feedback who looks stable
RC time is much smaller in the feedback because this is not a alternator.

It looks like the most I have seen, there are so much possibillitys.

LT spice do complaine a lot when simulating, Maybe there is a much better simulator
somewhere? wh can do complicated sims.

I have some pictures how the plot did, it look very good for me, very stable fase shift also.
I have no hot arc, because it is for MIG welding.

I do see that fase shift zsv have not so much min max range, a quasi resonant UC386*
with a VCO has more, using a resonant tank. Pictures has the min and the max amperage
plotted.

regards

kees

ScreenHunter_591 Jun. 16 17.04.jpg

ScreenHunter_592 Jun. 16 17.30.jpg

ScreenHunter_590 Jun. 16 16.59.jpg
 
Last edited:

KX36

New member
All simulators beyond the most simple throw lots of errors unfortunately. There are things you can do with spice for switching converters and those are to add "uic" on the end of the execution line which skips dc operating point solution. Spice struggles to find the operating point in these non-linear models. You can set initial currents in inductors and voltages on capacitors and wires when necessary.

Also it can help to put a high value resistor in parallel with capacitors when the capacitor would have meant no dc connection from a wire to ground although LTSPICE is not so bad for this.

Also set a maximum time step of well under the switching period. I usually do about 1% of the switching period to avoid the sampling causing an oscillation which isn't there.
 

blasphemy000

New member
I've been once again working on my TIG Welder design in my spare time. I determined back when I originally attempted to undertake this project that I just wasn't knowledgeable on SMPS design to be able to create this powerful of a supply. After over a year of research and simulating, I think I'm ready to start settling on a design that I will be able to prototype here. This will still be a hard-switched full-bridge(driven with GDTs) with CMC and the output will be constant-current regulated. I've recently stumbled upon some very fast, high power, low loss, IGBTs from Fairchild that should make this design much easier to implement from a DIY standpoint. I'm still a little bit unsure on the specific controller IC that I would like to use for the prototype, I've been considering the UC3846 as my initial selection and it seems as though it would be adequate for the task, although I'm open to any suggestions on this. I do have one main question about the controller and the bridge driving technique. I've done a little bit of research about Phase-Shift controllers for use in resonant SMPSs. One of the application notes I've found on this subject used a UC3875 controller driving a resonant full-bridge with GDTs. One interesting thing I noted about how the Phase-Shift controller works is that the GDTs are always driven at a near 50/50 duty-cycle(there is a little dead time to avoid shoot through on the bridge), but due to the way the phase-shift works, the main power transformer can see anywhere from 0 to 50-4xTd for it's duty cycle. Td being the turn-on delay for each of the IC's outputs driving the GDTs. My question is, since these converters often only achieve resonance above a certain load threshold, and below that load they usually run in either quasi-resonance or hard-switching; Is there any reason that a Phase-Shift controller could not be used for a strictly hard-switched application(The GDT 50/50DC could make the drive circuitry slightly easier to implement)? I've ran some simple, manually generated, simulations of just the bridge itself, and it seems that it's possible, but I know simulations are sometimes way off from how things really work. My main concern involves the fact that in a regular full-bridge, the dead-time that the main transformer sees is essentially an open circuit for the transformer, in that all four of the switches are off. This doesn't happen in the phase-shifted bridge except at 0%DC(this may extend slightly above 0% depending on the delay-times set for each side of the bridge but is still a very tiny margin) when both outputs are switching exactly in phase with each other, in which case it's irrelevant. Once the phase begins to shift, and the DC of the main transformer increases(in which the on-time is greater than either side's delay), at least one of the switches is on at all times(which switch depends on which part of the switching cycle you're looking at), and for the majority of the dead-time the main transformer sees, two switches are on connecting both legs of the primary to the same rail which gives zero volts across the primary and zero current being forced into the primary. So my question stands as, would this negatively effect, or prevent the use of, using a phase-shift controller for a strictly hard-switched full-bridge, given that the main transformer's dead time is essentially created by shorting both legs of the primary to the same rail? This doesn't pose a problem for the resonant bridge because there is a capacitor(along with the resonant inductor) connected in series with the primary and one side of the bridge.
 

kees54

kees57
@KX36

Thanks for your clear explaning about simulation of smps, I agree it is difficult to simulate such electronics but with help I
get there, thanks again.

I have make a schematic just only voltage feedback with the UC3875 chip, it do very well keep voltage om 16 volts, for 20 amps
pr 160 amps, but for welding these feedbacks are sensitive because need a fast response, therefore using a separate feedback
system with opamps or comparators, so it is also possible to use current and voltage feedback by put one of the pots max.

These feedback are nice for a high current supply or amp supply. different setups, serial RC parallel RC , serial 220k 1n integrator setup
give nice line, but RC time is maybe to small, give stability trouble, see first picture the wobble in the line, big RC time is the cause, needed
in welders.

Yes building is te real test but simulating is very helpfull, I use comparator setups, because most welders do better stability.

For the question of using a fase shift bridge without resonant I can not tel for shure, the dead time what is being programmed is used with the soft switch funtion on this ic, you have very fast low loss igbts I think just try, in my opinion I think you can also make a hard switch bridge, programming the two resistors for the deadtime properly, you can even program echt leg separate, but you need snubbers for protection of igbt.

ScreenHunter_598 Jun. 18 22.13.jpg

ScreenHunter_599 Jun. 18 22.16.jpg

ScreenHunter_601 Jun. 18 22.21.jpg

View attachment phase shift-UC3879.zip



see plots.

regards kees
 
Last edited:

KX36

New member
Blasphemy, resonant controllers tend to be completely different in the way they work to traditional PWM controllers. With zero current detectors, one shot generators, different timing on the outputs for PSFB to FB etc.

hard switching full bridge is not really difficult. The GDT and main transformer are driven in different directions on alternating cycles, so you don't have as much of a worry about balancing and saturation a you might think, especially if you use current mode control which is good on a full bridge where it would cause problems on half bridge. If you use a controller with 2 outputs you shouldn't need series capacitors on either of these transformers to balance them. I like the UCC38083 although I noticed on shopping for it recently it's getting low in stock and up in price in the major distributors. I don't know if that means it's discontinued.

Resonant converters running too low load for zvs isn't really a big problem on phase shifted full bridge and LLC converters as other resonant topologies, these are probably the most popular resonant topologies now. Even when it's not quite zvs it's still far lower dissipation than hard switching. I'd use controllers designed for resonant converters for those and ones for hard switching for those.
 

kees54

kees57
There are different aproaches, for example the quasi resonant who works indeed with a special chip UC3861/*

this uses a current driven vco, the resonance is max on lowest amperage and min on max aperage, it is also done
with a TL494 be using the freq setup as a adjuster with a pot, or the deadtime is adjusted keeping freq constant,
but here I can not use it because of mig, but with a modern chip who use a vco works. the output transformer sees
a very nice sinus not a square when in resonance.

Inverter popis has te be translated with google of pdf translator to get idea, this is a mig/tig/mma full bridge with
hardswitching.

Here some ideas.

ScreenHunter_602 Jun. 19 12.46.jpgScreenHunter_603 Jun. 19 12.46.jpg
View attachment LVI assisted soft welder.pdf
View attachment fasewelder.pdf
View attachment invertor_popis.pdf
 
Last edited:

kees54

kees57
All simulators beyond the most simple throw lots of errors unfortunately. There are things you can do with spice for switching converters and those are to add "uic" on the end of the execution line which skips dc operating point solution. Spice struggles to find the operating point in these non-linear models. You can set initial currents in inductors and voltages on capacitors and wires when necessary.

Also it can help to put a high value resistor in parallel with capacitors when the capacitor would have meant no dc connection from a wire to ground although LTSPICE is not so bad for this.

Also set a maximum time step of well under the switching period. I usually do about 1% of the switching period to avoid the sampling causing an oscillation which isn't there.

I did try IUC, but when I do this see result, without IUC it do simulate, the problems are because of U6 the Switching controller fail when using some kinds of opamps in feedback.

strange behavior, I ask myself are there better suited simulators for switching purpose, I now multisim do not, did try that, use it for audio amps.


View attachment 5201
 

blasphemy000

New member
Blasphemy, resonant controllers tend to be completely different in the way they work to traditional PWM controllers. With zero current detectors, one shot generators, different timing on the outputs for PSFB to FB etc.

hard switching full bridge is not really difficult. The GDT and main transformer are driven in different directions on alternating cycles, so you don't have as much of a worry about balancing and saturation a you might think, especially if you use current mode control which is good on a full bridge where it would cause problems on half bridge. If you use a controller with 2 outputs you shouldn't need series capacitors on either of these transformers to balance them. I like the UCC38083 although I noticed on shopping for it recently it's getting low in stock and up in price in the major distributors. I don't know if that means it's discontinued.

Resonant converters running too low load for zvs isn't really a big problem on phase shifted full bridge and LLC converters as other resonant topologies, these are probably the most popular resonant topologies now. Even when it's not quite zvs it's still far lower dissipation than hard switching. I'd use controllers designed for resonant converters for those and ones for hard switching for those.

Thanks for the explanation. I was just curious as to whether something like that would work and was interested in how the GDT and phase-shift controlled the duty-cycle of the bridge itself. I know the balance or saturation isn't that big of an issue with a properly designed and controlled full-bridge and I intend to use a standard CMC controller for this welder design. I wasn't very clear about that. Thanks for the suggestion of the UCC38083, I will check into it.

@kees54: I can't see your attachment, but it is "uic" (without the quotes), and it goes on the end of the simulation command. I haven't tried that in LTSpice yet, but it did help in some instances within MultiSim when it comes to SMPS stuff. And I do agree, MultiSim does have a particularly hard time simulating a SMPS for some reason. Not even talking controllers here, MultiSim doesn't really have any specific controller models that I've found useful, but even simulating a manually setup, free-running bridge is sometimes very difficult in MultiSim.
 

kees54

kees57
The problems with simulating is I thing more model problems, sub circuits not properly made.

I have tryed uic without quotes but then it stops and give singularity errors.

Also when I put this small feedback schematic in the circuit it is totaly not simulatable, nothing works.

So strange stuff, it give error about the controller comparator.

I read some stuff, it seems that a resonant bridge of the ZCS type better is then ZVS because of the tail current of igbt,s
I go try that out to see what happens I had somewhere a pdf about it, very nice stuff.

here sim of a zvs in that case if seen good..

soft switching;-).jpg

feedback, this time current feedback.

ScreenHunter_605 Jun. 19 15.39.jpgScreenHunter_606 Jun. 19 16.05.jpg

learnstuff.

View attachment Aerospace_Electronic_systems.pdf

regards
 

Attachments

  • ScreenHunter_608 Jun. 19 23.02.jpg
    ScreenHunter_608 Jun. 19 23.02.jpg
    17.2 KB · Views: 19
Last edited:

kees54

kees57
ZCS tryout 60Khz igbt.

If I am right afcourse but I see current is zero when gatedrive get on, both legs, 200 amps out.

it still has zcs with 30 amps

ScreenHunter_611 Jun. 20 00.31.jpg

LTspice has strange behavior, sometimes it seems to crash and can not stop simulation anymore,
menus do not respons and after that I get this of the same schematic als first.

So strange, I go try Hspice, or micro cap 11 what that does, multisim does no good at all, a pity
because it has such nice environment.

ScreenHunter_609 Jun. 20 00.17.jpg

regards

kees
 

kees54

kees57
Wel I did test a feedback with comparators, this because I think it is much more stable when welding.

for normal dutys like a supply voltage feedback van made easy with the chip own error amplifier.

I did also find this a welding inverter with fase shift uc 3875 controller, maybe you can do get ideas from
it.

here are the sims with the feedback, I did use a oscillator with make a saw teeth to make a pwm control
voltage tot the chip error amp, and it do well, also I have in meantime the igbt,s and such getting with mail.

You can see max amperage and min amperage get pretty wide.

ScreenHunter_616 Jun. 24 21.26.jpg ScreenHunter_617 Jun. 24 21.45.jpg ScreenHunter_619 Jun. 24 21.50.jpg
ScreenHunter_620 Jun. 24 21.58.jpg

regards

kees
 
Last edited:

twenglish1

New member
I don't have any real advice to add at the moment, but i have been off and on working on a very similar project, switched mode plasma cutter, i am still in early design stages and haven't worked too much on it recently, mostly due to lack of knowledge, it will be producing 400Vdc open circuit and as of now i will be starting with 20amps, and then increase max current from there, at my location i have a 50 amp 240v circuit available, which should be sufficient to power this machine. i have been following this thread and i wanted to ask, what exactly is the PFC for?
 

kees54

kees57
I don't have any real advice to add at the moment, but i have been off and on working on a very similar project, switched mode plasma cutter, i am still in early design stages and haven't worked too much on it recently, mostly due to lack of knowledge, it will be producing 400Vdc open circuit and as of now i will be starting with 20amps, and then increase max current from there, at my location i have a 50 amp 240v circuit available, which should be sufficient to power this machine. i have been following this thread and i wanted to ask, what exactly is the PFC for?

If you do not now what A PFC is and want to play with a very dangerous grid high power supply I gess learn first what a PFC is.

I will tell you this is a power factor controller, needed to push input voltage up to 400 V DC wo there is more buffer for heavy demands.

https://en.wikipedia.org/wiki/Power_factor

Plasma cutting

https://en.wikipedia.org/wiki/Plasma_cutting

regards kees
 
Last edited:

twenglish1

New member
Did some reading and I think I understand the powerfactor, I know I still have alot to learn before I start playing with this kind of stuff, as of now I have been trying to focus on the theory and design aspects. Looking at schematics of commercial machines welders and plasma cutters) I've noticed that they seem to use the two switch forward converters, but I was confused because from what I have read, a full bridge topology is recommended for multiple kilowatt supplies, could someone explain this? I hope I'm not hijacking this thread, figured this would fit in since about building a welder
 

kees54

kees57
You have to now, for me it is also new, I do study and learn about it, first this and then practice with a test, these things are so sensitive about pcb layout, dead time, tail current etc.

But it is nice stuff to play , and hee, be very carefull, these voltages can kill easely and without someone around your gone.

Power factor is not always needed, I do not now if it helps, just a pwm coil who keep a buffer elco on high voltage, need a big elco if you want to have profit.

I was planned a full bridge with two forward bridges and two primairy coils, but struggle with the fases in input of igbt,s, these double forward full bridge seems to work
very good with zvs also, soft switching is good for your plasma cutter also, but when beginning I do not now if it is interesting, this comply also for me wanted to much and to fast.

regards
 

twenglish1

New member
My understanding is that the PFC creates a constant current load on the mains, which in turn prevents the current spikes seen when main filter capacitor voltage falls below mains voltage and it charges back up, is this correct? i have always had an interest in power electronics, and i have been wanting to learn more about designing and building switching supplies, i probably should be starting with something less than an 8kw supply that produces 400vdc... i probably will decide to start with something simpler and of lesser power, just something about the high pitched whine of these high powered switching supplies, i love it!
 

kees54

kees57
Well I have find out why it did not work, for LTspice, watch out the labels, I had three of the same name, this did let it not sim right.

Always it are the small things who make the differents, as people say the simplest things are the the most difficult.

I was right about the fases of transformers who drive the igbt, both bridges has one fase shifted output connected.

Now it go easy to 300 amps, yes I need half.

ScreenHunter_663 Jul. 12 23.26.jpg
 

kees54

kees57
Yes, it do keep the cap charged on plus minus 400 volts, keep a more constant voltage.

Learn first build a smps working on low voltage for get familiair with it, I do also.

I read also a lot of papers, google on fase shifted weld inverter.

regards
 
Top