Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 12:06am 13 Jan 2021
Copy link to clipboard
Print this post
Still not sure why the other inverters were failing though. I'm working on getting at least 2 going today and put a load on them..... Will write a detailed report in a few hours.
I know the 1n4007 won't cut it due to current handling. It's just what I have on hand. What's a good diode to use for this purpose?
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 01:07am 13 Jan 2021
Copy link to clipboard
Print this post
Ok I applied rectified mains, but again the medium inverter IGBTs popped.
So there is still something else going on. Oh well, one problem is at least fixed....
I'm all out of IGBTs so it will be a few days before I can try anything else.
It was working on 140v DC so my spirits are high
I don't have a variable power supply, so maybe I'll invest in a Variac.
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 01:12am 13 Jan 2021
Copy link to clipboard
Print this post
Could the toroid cores have become permanently magnetised?
Here was the working trace with 2 inverters, no load, 140vdc. I can see an offset from zero. Is that a clue?...
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 01:34am 13 Jan 2021
Copy link to clipboard
Print this post
In the last test, I still had 5 ohms in series with each of the two transformers. I would have thought that would be enough to save the IGBTs...
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 03:28am 13 Jan 2021
Copy link to clipboard
Print this post
Ah yes, that inverse diode is extremely important.
That last waveform looks o/k. Don't worry about the ragged looking steps, they always look like that until you have all four inverters going.
The reason being, that the transformers are trying to switch a current that is stepping up and down. Inductive components don't like instantly changing current. When all four inverters are running, the current in all of the transformers will be a nice smoothly changing sine wave.
All four transformers will be very happy, and all the waveforms look a lot cleaner and much nicer. Its an amazing transformation.Cheers, Tony.
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 03:32am 13 Jan 2021
Copy link to clipboard
Print this post
Without the inverse diode, there may have been a very destructive back emf which may have been high enough to break down your IGBT. The inverse diode clamps the flyback voltages so they can never go below ground, or above the dc supply.
Without that diode the IGBTs don't stand much of a chance, even with a low dc supply voltage. A series resistor is not going to help, as its an overvoltage spike.Cheers, Tony.
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 05:12am 13 Jan 2021
Copy link to clipboard
Print this post
You misunderstand; I added the diodes to the largest inverter, but the medium, small and tiny inverters use the ikfw90 IGBTs with built in diodes.
There is still another issue that I have to track down as the medium inverter blew up running on rectified mains.
Not sure what that issue is and it will have to wait till I get more IGBTs delivered.
So AFTER the added diodes, it WAS working on 2 inverters at 140V, but the medium inverter still blew up at 320v.
I plan to buy a Variac and slowly increase the voltage till I hear the transformer buzzing, and will see what is going on there.
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 05:18am 14 Jan 2021
Copy link to clipboard
Print this post
More blown IGBTs:
Ok so today I bought a Variac from Jaycar, with a view to slowly turning up the voltage while monitoring the inverter characteristics. Previously I could only select discrete large jumps, and things were either working at 140v or blowing up on 320v.
I only had one functioning inverter H-bridge left after yesterday's fireworks, so I was taking everything nice and slow today.
First I connected the largest warpverter output to the remaining H-bridge, and slowly increased the voltage to maximum 350vdc.
No problems at all.
I was monitoring the current to the transformer via 10 ohm resistor, and with no load, and no saturation, the current stayed very low as expected.
Next I plugged in the medium warpverter output into the H-bridge and again slowly turned up the DC voltage to 350v, while monitoring the output trace and current consumption of the transformer.
There was no sign of saturation, negligible current draw from the transformer, and the output voltage trace was correct.
Things were working well for a few minutes but again, without notice, one side of the H-bridge decided to vaporise.
Since both the large and medium inverters work well at lower voltages, but since only the medium inverter is blowing up at high voltages, my suspicion is that the culprit may lie in the fast transition from bottom peak to top peak of the rapidly changing voltage illustrated here:
I haven't ruled out the dead time as I don't have a 2 channel scope, but what might be more plausible is that a snubber circuit may be needed to keep this large voltage transition under control.
Any thoughts on this theory?
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 06:17am 14 Jan 2021
Copy link to clipboard
Print this post
Each side of the H bridge switches completely independently, and is clamped between ground and dc supply by the inverse diodes.
As you have noted, the largest inverter only ever switches one side at a time, whereas the other half bridges sometimes switch simultaneously in opposite directions to go from full peak voltage one way, to full peak voltage the other way across the primary.
I cannot see how that could be harmful, but obviously "something" is still not quite right somewhere. Klaus is still having mysterious random blowups as well, and we have not figured that out either. He is running mosfets at 48v, so his situation is a little different.
First thing to try might be to stress test your IGBTs. Just connect one up to a variable high voltage dc supply with the gate grounded, and see if it will withstand the advertised rated voltage. If it does, try a few others. If the game is fair, that should not be a problem. Its not unknown for Chinese components to fail at well below the published ratings.
When you have a blow up, is the gate driver chip destroyed as well ? I am wondering if its breaking down collector to emitter, or collector to gate ??
You might also like to try placing a protective zener directly between gate and emitter. It might be the gates that are being destroyed somehow from induced voltage spikes.
Its all a bit of a long shot, but by trying various things, we may get lucky and stumble upon a clue.Cheers, Tony.
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 07:05am 14 Jan 2021
Copy link to clipboard
Print this post
All great ideas warpspeed.
All of these IGBTs and opto-drivers are top shelf components sourced from digikey.
I have blown up more IGBTs than I care to remember now. They generally short from collector to emitter, but a few have shorted all 3 terminals.
The gate driver optocouplers generally survive. I have only blown one hi-link power supply module.
My hunch based on my last post is that something bad is happening during the peak to peak swing. In my case, that's a 700v swing!
The IGBTs are rated at 600v.
I could try a few things:
1. Use 1200v IGBTs. 2. Add ferrite beads to gates 3. Increase dead time- say 20nF capacitors across opto leds. 4. 15v zener on gate 5. Some sort of RC filter between collector and emitter?
What do you think is the best order of the above list? Has Klaus ever blown up the largest inverter, or are his failures only on the smaller 3?
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 07:21am 14 Jan 2021
Copy link to clipboard
Print this post
The IGBTs are never going to see more than the dc supply voltage, assuming that the power rail is not spiking up due to inadequate bypassing.
The 10nF that you have in conjunction with 180 ohm resistors should give a dead time of very close to 2.5uS which is pretty slow. Longer will not hurt, as our switching rates are in the low audio range, not tens or hundreds of Khz !
Fifteen volt zener on the gate might be worth a go. The voltages being switched here are pretty high and switching dV/dT fairly fast, and it might not take a lot of stray capacitance to couple into the gate circuit.
If a snubber fixes it, it will be due to slowing down the switching speeds, and masking the real cause, but anything you try that has some effect is useful knowledge and gets us closer to nailing the cause.
If your IGBTs are punching through collector to emitter without destroying the gate or its driver, it can only be breaking down from straight over voltage. Either the voltage really is excessive, or the IGBTs are failing from quality control problem and not meeting specification.
All Klause's blow ups as far as I know have been in the lower powered inverters, and not due to overload or doing something silly. Just really annoying random failures, cause unknown.
One theory we had is that Klause is using a system of piggy back boards for his gate drivers, and even a momentary poor connection could create a floating gate situation. It may pay to look seriously at your own layout and play devils advocate with any plugs, sockets, or connectors. Anything at all that might cause a noisy or intermittent contact in the gate circuits. Edited 2021-01-14 17:43 by WarpspeedCheers, Tony.
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 07:53am 14 Jan 2021
Copy link to clipboard
Print this post
Ah, if all Klaus's failures have been in the lower stages then we might have something in common...
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 08:01am 14 Jan 2021
Copy link to clipboard
Print this post
It can't be a quality control problem. I've blown up ikfw75's, ikfw90's and today my first two ixgn320n60's. Ouch! All from Digi-Key.
The ixgn320n60's don't have a built in diode. I was using some paltry 1N4007 diodes. But with no load connected, just the transformer, I thought that would be enough? Was that my fatal mistake?
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 08:16am 14 Jan 2021
Copy link to clipboard
Print this post
1N4007 are good for an amp, but they are very slow devices.
I really cannot offer anything other than to try a few different things and see what the effects are.
One thought just came to mind. The only really lethal fault I have found is when the crystal oscillator stops or fails to start up. That can leave two diagonal IGBTs turned on continuously.
Klaus as had that problem, caused by plugging the oscillator module into an IC socket. These oscillator modules MUST be soldered directly onto the board, that fixed one of his problems. I had a similar experience myself with an early Warpverter design, where the metal can of the crystal shorted to the pads directly underneath. It was intermittent too, which caused me a lot of grief until I discovered the cause.
Oscillator reliability is critical, and its why I went to using the more expensive oscillator modules instead of a simple discrete crystal oscillator. Edited 2021-01-14 18:34 by WarpspeedCheers, Tony.
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 10:23am 14 Jan 2021
Copy link to clipboard
Print this post
I'm going to try a higher value gate resistor and a longer dead time. Say 20nF and 30 ohm gate resistor.
I'd imagine the large old IGBTs you are using have a much larger gate capacitance than my little IGBTs, so I should be able to afford a slower switch on/off time which should decrease spikes caused by inductance in the cables.
Maybe some metal film capacitors right across the rails close to the IGBTs too.
I have some long wires going from the warpverter board to my power board. That's a concern that could be easily addressed. However the opto drivers are just millimetres away from the IGBTs, with their own separate "Kelvin?" non current carrying tracks straight to the IGBT gates. Not much more can be done there.
I'll decrease the rise/fall time and go from there...
I've been holding off on getting an oscilloscope for decades. Maybe now is the time to get one
Murphy's friend Guru Joined: 04/10/2019 Location: AustraliaPosts: 648
Posted: 10:53am 14 Jan 2021
Copy link to clipboard
Print this post
I doubt it. The #1 (big) & #2 drives are 5 parallel (20 mosfets) & 3 parallel (12 mosfets) mosfet drives. The #3 and #4 drives have just 4 mosfets per bridge.
Latest plan is to make the physical heatsink arrangement for drive #3 & 4 similar to that of #1 &2 heatsinks. Its very different now. I use mica insulators under each mosfet BTW. Then fit 2 parallel mosfets for #3 and a single mosfet for #4.
Doing that I can compare apples to apples so to speak and rule out layout problems.
That plan would increase the total mosfet count to 44 - expensive when things go bang. I dare say using IGBT's and a higher voltage (less current) makes PCB arrangements easier but I see now that other problems can sneak out of the bag.
Good luck, so far I have learned that this is an engineer's inverter. Myself, being just a retired electronics technician (mostly repairing things in working life)this project makes me wonder if I had bitten off more than I can chew. Anyway, it was fun building it, maybe a while yet to get it going reliably.
Haxby Guru Joined: 07/07/2008 Location: AustraliaPosts: 423
Posted: 11:31am 14 Jan 2021
Copy link to clipboard
Print this post
I'm not a power electronics guy either. One thing I'm learning is that it is a very unforgiving field. One little seemingly mild issue, and it all blows up. Either immediately, or worse- after a short period in service.
I felt I had great success while bench testing at 140v some weeks ago. It pretty much worked on day 1 of putting my board together, and I declared victory. Little did I know the battle had only started. But I like the challenge, and I'm learning a lot, so it's all good!
I bet the engineers at victron, goodwe, etc blow up stuff every day.
I tend to gravitate towards unforgiving hobbies. My other interests are in model aeroplane autopilots. One wrong parameter makes them fall out of the sky, or fly away out of sight never to be seen again!
CaptainBoing Guru Joined: 07/09/2016 Location: United KingdomPosts: 2075
Posted: 01:06pm 14 Jan 2021
Copy link to clipboard
Print this post
Long time lurker on this thread - what a great forum this is!
I think your idea of a EV with "auxiliary" output is fab - all EVs could do this, especially those aimed at more utilitarian tasks - at least offer as an option.
what a sense of de-ja-vu I am experiencing!!!
Murphy's friend Guru Joined: 04/10/2019 Location: AustraliaPosts: 648
Posted: 01:15pm 14 Jan 2021
Copy link to clipboard
Print this post
Welcome to the blowing up bits club, this is par of any inverter builders experience on this forum (engineers excepted).
Mosfets can go spectacularly flash bang but I found that by placing an alu clamping bar (20x3mm) on top of them only the source leg evaporates, limiting flying housing chips.
One thing I learned with IGBT's, I use them to turn on the selected panels for my solar grid tie back charging, is that they do not like to be switched off under load with 300 Volt DC solar or so, at least the ones I played with. Fixed that by removing the AC load of the grid tie inverter first.
You need to be careful with your 300+ voltages about PCB track spacing and the electric limits of some components. I test every opto driver chip after each 'event' but I think only one failed so far, 100+ mosfets were not so lucky .
Interesting hobby you have, at least with my drone it returns automatically when the signal is lost - assuming enough juice is left in the battery.
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 08:35pm 14 Jan 2021
Copy link to clipboard
Print this post
The big power block IGBTs are specified as 10nF gate capacitance, and I use ten ohm gate resistors.
This is why I think anyone attempting a project like this absolutely must own an oscilloscope, either PWM or Warpverter, an oscilloscope is absolutely vital. If there are high voltage spikes on the dc rail, you need to know and be able to actually see them.
The control board is designed with push pull antiphase outputs with 180 ohm series resistors. That will drive a twisted pair transmission line of any practical length perfectly. At the receiving end, there are back to back opto isolators providing a fully floating balanced input. That whole system provides incredibly high noise immunity, the twisted pair lines could be ten metres long if you wanted, and it would still work perfectly. The opto isolators also provide symmetrical dead time and absolutely fool proof protection from cross conduction.
Up to the opto isolators there cannot really be any problems, except from oscillator failure or a dead or sick ROM chip. Whatever your recurring problems are (both of you) will be either in the gate driver part, between the gate driver and mosfet/IGBT, or be some kind of layout problem, or faulty components.
Certainly worth a try, gate resistors of a hundred ohms or more should work, and snubbers across the IGBTs are also worth a try, but without an osilloscope you will be groping around in the dark.
Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.
Print this post
.