Building my first High Current SMPS...

[GeorgesWelding]

Hi everyone. This is my first post here. My background is actually in computer programming but I'm a quick study and have been reading and learning a lot about building electronics the last few months. I've built a couple small DC-DC converter SMPS the largest running off of a 25VDC input with a regulated 5V/2A output. I am currently prototyping and developing my own SMPS that I will be using for TIG Welding. TIG inverters are VERY expensive, so I've decided on trying to build my own just for kicks.

I'm using a TL494 controller to handle generating the PWM signals and the current control. In push-pull mode, the max duty-cycle is 48% giving a 2% dead-time between alternating pulses. I think the 2%DT is plenty for the IGBTs I'm using. I'm using IR2110s to drive my H-Bridge IGBTs which are IRG4PC40UDPbF. The TL494 and IR2110s are powered with a simple regulated 15V supply and the H-Bridge will be powered from a rectified mains voltage of +320VDC to +340VDC. Now, I built the main part of the controller board and built the H-Bridge and powered it up with a low-voltage feeding the H-Bridge(25VDC) and everything worked fine w/o the transformer attached(No load on the H-Bridge's output). Nothing got hot or anything of the sort. Here is where things went wrong. I decided to try the bridge at 1/2 of it's normal operating voltage(~170V) it made a funny noise at power-up and before I could switch it off the noise stopped and the breaker in my bench's power strip popped. Upon examination I found that 1 of my IGBTs had become a dead short between Collector & Emitter and there was only 10Ohms between the Gate and either the Collector or Emitter. It also fried at least one of my IR2110 driver ICs, although I replaced both of them to be safe. Upon these findings I made some changes to my controller card and tested 1/2 of the bridge at 12V driving a small electric motor and all ran fine(I tested it for 10 minutes of on-time with no ill effects). Attached are 2 PDF files of my circuit and I'll list the changes that I made after I blew part of my bridge.

Notes:
IGBTs are: IRG4PC40UDPbF
D1 & D2 are: FR107 (1000V, 500ns)

Changes:
I changed R2 & R3 from 2.2KΩ to 100Ω.
I added resistors R8, R9, R10, & R11.
Hoping these two changes will speed up the Turn-Off time.

Now since I blew one of my High-Side IGBTs I can only test 1/2 of my bridge until my new IGBTs get here, then I can test the full bridge. As I said earlier, after making these changes, I powered the bridge with 12V and powered a small DC motor from 1/2 of the bridge and all ran fine. The motor was running at 1/2 speed due to the 48% duty-cycle. I just wanted somebody to examine my circuit before I tested it at a higher voltage and blew it up again.

Attached are 2 PDFs of my circuitry.
View attachment Control Card Circuits(1).pdf
View attachment Control Card Circuits(2).pdf

- - - Updated - - -

I edited the second PDF file as I made a mistake in the power source for the H-Bridge.

 

[GeorgesWelding]

Just tried the circuit at 65V with a resistor in series with the supply to limit the current to just 100mA in case of problems. Circuit seems to switch erratically at less than the intended frequency. I'm unsure of where the problem is. I was to understand that the bootstrap capacitors(C2 & C3) only need to be rated for the VCC voltage, is this correct or do they need to be rated for the full voltage of the bridge?

 

[Orson Cart]

First off D1, D2 need to be 600V fast or ultrafast types, e.g. 600V, 1A, 50nS, or better, e.g. BYV26C or better, the FR107's are too slow and may overheat and die with time, better check em.

OK, most likely as you wind up the voltage, the turn on transient (dV/dt) provided by your igbt's is getting into your control circuit and upsetting things to the point where two igbt's are on long enough at the same time to short the bus and kill each other. Good idea to put the smallest fuse that will allow low power testing in line with your bus supply to hopefully prevent more blow ups.
Also put a 275Vac rated 1uF across each half bridge, short leads, to soak up some energy when the igbt's turn off.
Also you likely need a DC blocking cap in series with your transformer (a few uF, 275Vac rated for now for safety), if the PWM is a bit asymmetric due to interference the flux in your transformer can staircase and overcurrent your devices - hence the DC blocking cap.
To sort the dV/dt, you need to up your gate drive resistors to say 1k for starters, R1,5,6,7, put a reverse schottky on them to still give a quick turn off, this should get you up to 325VDC bus for no load on your transformer. Change R8,9,10,11 to 10k, or to 16V zeners, 1W.
When you get to the stage of running load (start at low bus volts again and wind up gradually) if the igbt's turn on too slowly and get too hot, you will want to trim down the 1k's to get a balance of heat and interference to your control.

When you are running load you will want some serious C on the 325 bus, e.g. 100 - 470uF or more depending on your full load rating.

Oh yes, put some 1uF MLC caps right across Vcc to gnd on all your IC's, short leads, and a 10uF too, amazing what a stiff supply to the control IC's will do.

Let us all know how it goes....! good luck.

 

[GeorgesWelding]

Thanks so much for the reply. I do have a 0.1A quick-blow fuse on my supply line as well as a current-limiting resistor now for testing purposes until I get this sorted. I also have filter caps across the VCC/GND of all the ICs to stiffen up their supply lines, I just drew up the schematics I posted in haste and forgot to include them in the drawings.

I didn't know that the 500ns of the FR107 would be too slow, but that would explain some weirdness in some of the readings I was taking yesterday.
I'll make the changes to the gate drive resistors and add the reverse schottky diodes and add the 1uF cap across each 1/2 of the bridge and see how it works then.

As for running load, I'm redesigning my transformer right now. I'm waiting for the new E80 cores to come in the mail. Output will be approximately 100A in this "test" version of the welder, with that said I know I will need a significant amount of capacitance on the rectified mains voltage feeding the bridge.

Thank you so much for the reply. I'm going to start working on this again and I'll post back with my findings.

- - - Updated - - -

Would these caps be proper for placing across the bridge as you mentioned? If not, what cap from alliedelec(I have an account there) would be a good choice for what you mentioned.

https://www.alliedelec.com/search/productdetail.aspx?SKU=70102582

Also...When you say to place them across each 1/2 of the bridge, do you mean between the positive and negative rails supplying the bridge? Or am I misunderstanding you?

 

[GeorgesWelding]

I tried the changes to the gate drive resistors you mentioned to no avail. Also, thinking maybe the IGBTs were partially damaged I built 1/2 a bridge with 2 MOSFETs that I had been using in another circuit and still got the same odd results when the bridge voltage is above about 25V. Just seems like I'm doing something obviously wrong that I can't figure out. I'm gonna breadboard some test circuits and see if I can figure it out.

 

[DeepOne]

Hi.
I think the main reason of blasting is absence of correcting chains between third and second pins of TL494. And is reasonable to do experiments with more cheap key transistors, like IRF740.

 

[GeorgesWelding]

I'm currently doing testing with a couple of MOSFETs I pulled from some PC power supplies that I was using in another project, so if they explode, no big deal. The results are still the same as first described though.

Not sure what you mean about a correcting chain between the feedback and negative pin of the TL494's #1 ErrAmp. I know I can place a resistor in between pins 2 and 3 to adjust the gain of the ErrAmp in the TL494 but I currently have both ErrAmps disabled as is described in the 494's datasheet. No matter what the bridge voltage is the signals between the TL494 and the IR2110 are fine, it's after the IR2110 that the signals are getting all screwy when the bridge voltage gets above a certain point. I don't have a variable voltage supply so I can't tell the exact voltage where things get weird, but it runs fine on my 22V and 12V supplies but on my 50V supply it don't. I've built other power supplies using the TL494 and always had the FEEDBACK(Pin 3) grounded and used the two ErrAmps for voltage and current feedback with out any issues. Only difference in those supplies was that I used a Gate Drive Transformer and a Gate-Shaping circuit to control the switches whereas I'm using the IR2110s for this supply. I'm just not sure what I'm doing wrong.

 

[Orson Cart]

Also...When you say to place them across each 1/2 of the bridge, do you mean between the positive and negative rails supplying the bridge? Or am I misunderstanding you?

Yes, no you are not mis-understanding.

make sure all the grounds are connected between all your control circuits, and the power ground, and that all your wiring to the power circuits is correct, post a photo?

My company has designed and built a fair number of high power igbt and mosfet converters, and for newbies the problem is almost always layout and too aggressive gate drive, leading to dv/dt & di/dt induced interference to the control.

Try running first with no transformer - then a dummy load - e.g. 4k7 or 10k (10W or higher as appropriate), and then with the tranny with DC blocking cap.

If you can't get operation with a 10k dummy load (325 bus) - things are seriously wrong.

Keep an eye on the lower devices gate drive when you are running up - it tells you things.

It may pay to buffer the 2110's with an BC327/337 emitter follower pair, this gives current gain (good for turn off) and some isolation for blow ups - we did this with all our old 2110 ckts, as the 2110 isn't really rated for 45kHz, 15V. Also putting a 2k2 on the bases to gnd guarantees the power devices are off when they are supposed to be.

Keep going and tell us the news....

 

[DeepOne]

I currently have both ErrAmps disabled

Yes, you right. I mistaken about this.

 

[GeorgesWelding]

Ever since the original failure I've been running with a 150Ω dummy resistor as the load on the bridge. I haven't tried any testing above 50V as my 50V supply has a 500mA limiting setting on it to which it instantly pegs regardless of the value of the dummy load(I've tried as high as 1MΩ). I wasn't aware that the IR2110 wasn't designed for those speeds, I guess I should have done more reading, luckily they weren't too expensive. I did try slowing down the freq of the TL494 all the way down to 1.16KHz and I still get the exact same results. I also tried grounding the LIN pin of the IR2110, locking the low-side FET in the off-state and somehow they still are both managing to get turned on and creating a short across the bridge. Guess I will have to wait until I get my scope repaired before I can pinpoint this issue. I'm at a loss at the moment.

 

[Orson Cart]

Hello Georges, the 2110's could have low side damage giving you the results you see, when your scope is fixed, you will be able to see if they are running correctly (no bus power), we use an old tektronix 730 handheld scope for a lot of run up work it has 600v isolation between the two inputs allowing us to look at the high and low side gate drive (or any other waveforms that don't share a common gnd) simultaneously, these can be picked up for <US1000 these days and if you replace the battery they are great...and extremely useful in power electronics.

 

[GeorgesWelding]

Thanks for the tips Orson. When my bridge originally got damaged I replaced the 2110s with new ones just in case they were damaged when the bridge died. I'm back to chasing the problem today but if I can't figure it out today then I'm just going to have to wait until I get my scope back. Thanks everyone for the advise. I'm looking forward to getting this problem solved and completing this project and sharing it with all of you. This seems like a really great electronics community you all have here.

-Brad

 

[GeorgesWelding]

You were right about the low-side damage to the 2110. The LO pin was shorted to VCC, locking the low-side MOSFET in the on-state. I replaced the 2110 with the last good one I had on hand and was much more careful about my testing today(everything survived). I'm still getting sub-par results. Both the high and low-side MOSFETs on my test rig are hanging on about 22%(I estimated this from the average DC voltage coming from the bridge's output with one of the two MOSFETs disabled and either pulling the output up or down with a 47Ω 5W resistor.) after they should have shut off. I have the G-S resistors soldered directly to the FETs so the leads can't possibly get any shorter on those.

I did just try driving my low-side MOSFET(With the High-Side disabled) directly from the TL494 using the same technique of pulling up the output with the 47Ω resistor and letting the MOSFET drive it low, and the average DC output from the bridge was almost exactly 48% of the bus voltage. Which is perfect, as the max duty-cycle of the TL494 in push-pull mode is 48%. So somehow, somewhere, the IR2110 is lagging behind my turn-off signal by 22% and this remains constant regardless of the freq or the dead-time I set on the TL494.

Also, I thought this problem was nonexistent with the bridge bus at only 12V and that it only arose when the voltage was ramped up. I was wrong, it is just much less noticeable at the lower voltage due to my 12V supply's very low amperage output(about 100-200mA).

 

[Orson Cart]

Buffering the 2110 o/p's with an emitter follower pair may well solve a lot of your problems, you really need your scope to ensure the original gate drive waveforms from the TL494 are clean with no VHF switching at the transitions, and that they propagate thru properly and there is definite dead time on the drive pins to the 2110's.
Once you have this and a slowed turn on for the igbt/mosfets then you will begin to see proper results.

p.s. check all your power devices too....!

 

[GeorgesWelding]

Thanks for the tip. That was my next step was to take your advice about boosting the 2110's drive with a BC327/337 pair. If that doesn't solve the issue completely then I'm just going to give up until I get my scope back in my hands and I'll let everyone know my results.

I am very appreciative to all of the help you have given me so far. Thank you so much!

 

[GeorgesWelding]

I boosted the 2110's outputs each with a bc327/337 pair of transistors with the same results as before. One thing that made the biggest difference was replacing my 15V regulator(It was making 16.8V) that was feeding the TL494, and the VDD/VCC of the IR2110 with a 12V regulator(That makes 12.05V). This change reduced the time the FETs were hanging on after they should have shut off to only a few percent(Again based on average Vout readings from the bridge). For now I'm just going to wait until I get my scope back on July bench and can take some precise measurements.

 

[Orson Cart]

Don't forget the de-coupling, esp across Vcc & Com on the Lo, 100nF (or better 470nF) on the 2110 low o/p's

It appears from the above the output of the 494 is holding up a tiny bit too long, you should be able to suss this fairly quickly when the scope returns...

possibly there is some VHF switching going on tht isn't immed obvious...

 

[GeorgesWelding]

The output of the TL494 is triggering properly. If I drive my MOSFETs with my gate drive transformer and gate signal shaping circuit everything functions fine up to the full bus voltage of 340VDC. When I get home today I have an idea for something I'm going to try. I'll let everyone know the results.

 

[Orson Cart]

Good to hear, often if a single tranny is used to drive the 4 fets, the dv/dt causes little spikes on the gates of the fets that are trying to stay off, there are a number of papers out there (and the ray ridley site) that suggest using 2 gate drive trannys, one for each diagonal pair of mosfet, rather than one for each totem pole pair (which gives the dv/dt gate induced turn on) - hope this may be useful to you...

 

[GeorgesWelding]

So what you mean when using a GDT is to use two of them driving the FETs diagonally so that each transformer is turning on/off both it's FETs at the same time instead of each transformer driving it's two FETs in opposite directions. That makes a lot of sense. Thanks for the tip. If I can't get this problem with the 2110s figured out, I have no problem using Gate Drive Transformers along with my Gate Signal Shaping circuits to drive the FETs as that is what I have used in all of my other power supplies with no issues at all. Only difference is my other power supplies I either used the Two Switch Forward Convertor or the Half-Bridge Push-Pull designs. This is the first power supply that I've attempted to use a Full-Bridge topology as I figured it would allow for more power output than the other designs I had made before.