diy solar

diy solar

DIY Franken-Inverter

this thread is totally crazy and i love it. taking lots of notes and having fun learning about the innards of inverters by following along.

btw idk if this is helpful but this project i saw a long time ago popped into my head


edit: they used EconoDual IGBT
 
Do you have link to open inverter forum?
This is the page I'm referencing https://openinverter.org/docs/index.html?en_gate-driver,23.html

I found an interesting document about the test driver board that Infineon makes for this specific IGBT module. Attached it to this post.

It seems like there's some sort of internal resistance already (1.5Ohm) and for some reason RGon is not the same as RGoff. So in the driver board they use another resistor going back with a diode in place. At this point it's far beyond my rudimentary understanding of these things, but it looks like they use a similar driver system to the one my board is using, a PNP and NPN transistor to drive the gate. Except in this case they use 4 in parallel and each is able to drive 4.2A continuous, so they have a lot of headroom.

Back to the board I'm using,
I THINK the board is having transient issues due to the way they create the negative gate drive. They use 24v linear regulators fed filtered 30v from the transformer. From this they somehow get -5.8v and +18v and store it in two separate capacitors for each gate. These capacitors which are 100uF 35A electrolytics were getting quite warm. I believe around 50C. Now, I think this is because of the transients generated by the L7824cv regulators being pushed quite hard, as they are only 1.5A parts. I did note that the hottest components on the board (besides the resistors of course) were these regulators. I believe they got to upwards of 65-70C, which is within operating temperature but unsettles me quite a bit. I do have a couple adjustable LM2596 buck converter modules that are rated for 3A that I could sort of hack on in place.
 

Attachments

  • Infineon-Module_Adapter_Board_for_PrimePACK-ApplicationNotes-v01_02-EN (1).pdf
    1.1 MB · Views: 7
I sat down and looked at my driver board for a bit, analyzed the drive circuit with a multimeter and made a mental schematic of it. Really weird setup for the negative gate drive, but here we go.

1) 12v power comes in through the input. A UC3843A pwm driver IC drives a 80N75 mosfet, which pwms the 12v at 500khz into a ~0.2ohm winding in the transformer. The transformer then has 3 isolated ~0.4ohm windings, which are connected each to their own 35v 100uF electrolytic in series with a diode. We now have three isolated 30v sources. One is shared between the low side IGBT gates (which don't require isolation, one gets a 60Hz square wave and the other outermost one 23kHz swpm) and the other two go to each of the high side IGBT gates which require isolation, the outermost is driven by 23kHz down and the innermost gets the 60Hz square wave. So the two resistors that heat up are the outermost ones, the inner ones are perfectly cool.
IMG_20210418_235749__01.jpg
2. On the backside of the board, there are three L7824cv 24v (1.5A) max regulators that each take the 30v and turn it into 24v and pass it back to the front side of the board.
IMG_20210419_000916__01.jpg
3. Now back to the front, this is where it gets weird. The 24v is put into two 35V100uf caps in series. Now, in a normal scenario it would be split up almost perfectly 12v and 12v between the two. But there's what seems to be a 5.6v Zener diode in parallel with the left capacitor. So, the voltage is split around 5.6-5.7v on the left capacitor and 18.5v on the right.
IMG_20210418_235749__02.jpg
And then of course the NPN and PNP transistors switching with the point in between the two capacitors referenced as ground, so they switch -5.6v and +18v from the caps into the gate.

This is what worries me, because if the left capacitor (5.6v, used for negative driving) has more draw than the right at any moment (when turning gate to low, for example) all the voltage would shift to the right one, and possibly bring it above 20v?

So when the gate is first turned high, all is good, but when turned low, the right capacitor might charge above 20v, and can't discharge cause of the diode. And given that the maximum gate voltage is +-20v for the igbt, it could explain the excessive heat generated from the resistors, if it's trying to push more than 20v into it.

I'll probe it probably tomorrow and see if the voltage in the caps fluctuates significantly when switching. I have a plan if it has problems.




Also @Bob, the power draw of the driver board is exactly the same with and without any collector voltage, and remains constant from 0-33v, scoping the gate it doesn't have the peak anymore though.
 
Last edited:
So an update, pretty much on hold waiting for some parts to arrive.

I tested multiple configurations of gate resistors, and anything above 4.7ohm causes overshoot at higher collector voltages (at 30v, current draw goes up from the constant ~0.06A to around 0.1A), and I suspect that at even higher voltages this will turn into a very significant short.

So sticking with the 3.3ohm, I scoped around the -5.6v and +18v caps while gate driving and it's super stable. The caps at least seem to do a good enough job of suppressing any possible noise.

Also bought 5W resistors, which won't fit on the board so I'll have to jump the original mount holes and splice them in the wires.

The main concern I have right now are temps. I ran the board for maybe 20 minutes (ignoring the health of the resistors, as they were reaching over 130C), going around with a temp gun measuring everything. The second hottest component on the board were the 24v regulators. They stabilized around 80C. I think I'm drawing too much current from them. And the TIP41/42s were getting around 70C.

For the 24v regulators I purchased a couple of the same ones and plan to just put another in parallel with them in whatever way it fits.

I also purchased a bunch of little stick-on heatsinks like those used on motherboard VRMs.

Now on the crazier plan B I also purchased some massive TIP3055 and TIP2955 TO-247 transistors, which I could try to somehow use to replace the much weaker surface mount TIP41/42s. These are 15A parts and should provide a huge upgrade in headroom. I have very very little experience with basic NPN and PNP transistors to know if they would actually work or not, but they looked interesting.

I wish I could reverse engineer the schematic of this PCB and add my own upgrades to it, like beefier caps and whatnot. I'm planning on at least trying to in the upcoming weeks. It would be amazing to have a proper board that can pretty much handle whatever IGBT I throw at it without having to fiddle so much.
 
When you get a circuit that you can short the output with abandon then you'll have
something useful. If it keeps blowing up I can help.
 
Just wond'ring aloud... please tell me what's wrong with the reasoning. Just out of curiosity.
To make a true sine wave inverter... if one took an electric motor, connected an alternator to it with say, a belt, got the AC out of the alternator, the wave would be "pure", right? Would the efficiency be that bad?
 
When you get a circuit that you can short the output with abandon then you'll have
something useful. If it keeps blowing up I can help.
Unfortunately I haven't worked on this project and probably won't for quite a while as I have begun in-person college classes again and have been quite busy.

I was able to reverse-engineer a schematic of the driver board about 70% of the way, including the gate drivers and isolated power supply, and at the same time beefing up the components, for example the caps used for the + and - gate drive voltages were heating up a lot, I suppose b/c of ripple, and placed bigger ones more able to withstand it, instead of TIP41s and 42s used TIP3055 and 2955 (which were just really placeholders, I have no idea if they're even compatible frequency wise).

I never got around to drawing out how they worked the safety's. I know they have an STM32 and various AND gates, so the gate drivers only see the gate signals if the STM32 allows them to after checking if there's any error (as the gate drivers somehow have a way to see if the igbt is being overloaded and send a signal to the STM32 albeit that's way out of my scope of knowledge.

I could send a picture of the schematic I made if you wish, it's likely not the neatest thing you'll see but it's the first time I used any circuit design tool. Probably won't be able to share the file itself as I had to install like a million different footprints libraries and whatnot to be able to even get started.
 
Just wond'ring aloud... please tell me what's wrong with the reasoning. Just out of curiosity.
To make a true sine wave inverter... if one took an electric motor, connected an alternator to it with say, a belt, got the AC out of the alternator, the wave would be "pure", right? Would the efficiency be that bad?
Yes, the efficiency would be horrible. Not to mention the amount of failure points added to the system, bearings, belts, brushes for the motor, and of course, with the cost of a giant DC motor and giant alternator required to generate say 5kw continuously, you'll be so much better off just buying an inverter.
 
Not sure about the cost and failure points actually. The "giant" components aren't all that "giant", and can be had for less the $500 each.
Maybe 1000 if you really want to over-engineer it - and make failure points even less likely than electronics ones.
The weight and space parameters would obviously be on another level, as might be the noise ones... even though a 5KW inverter running continuously at half-power tends to be rather noisy too (the fans...)

Of course weight and space would be the obvious reasons why commercially it isn't done.
But... efficiency-wise, honestly, can it be that bad?
Because, for a DIY thing...
 
Carrier used on these boards may be hi for your IGBT, for this reason gate rezistors run hot, if you manage to find silicon carbide bricks things will look better, you must use lo esr capacitors for clamping dc bus, and snubbers.
A single electrolytic cap is not enough to clamp bus for switching spikes.
You may ensure enough dead time to avoid cross over, for these monsters 1500ns may be not enough, same rating silicon carbide run well with 500ns.
EGS002 knows tho types of modulation unipolar and bipolar, in unipolar one leg runs hot (pwm) and one runs cold, 60hz square(reference).
 

Attachments

  • 20160814_010524.jpg
    20160814_010524.jpg
    127.8 KB · Views: 3
  • 20160814_005358.jpg
    20160814_005358.jpg
    134.6 KB · Views: 3
Silicon carbide is a giant step foreward in inverters, those IGBT bricks are used at 1-5 khz carriers for motor inverters, but silicon carbide can run above 20khz even large bricks.
 
this is a great thread. i have also been playing with egs002 and big igbt bricks but im backing down and using some isotop mosfets instead.
i might try to use the igbt for a modified sine inverter. in fact its only my fridge and freezer and occasional air compressor that uses ac. almost everything else uses smps so runs just fine from my 220vdc string
 
Maybe more technical than appropriate here, but here is an app note regarding inverter output stage design and reliability:
(Also attached below)

Transformerless output inverters that have output device failures can start fires. Did a video on that at:
 

Attachments

  • H2PToday2201_design_RedHillLabs.pdf
    1 MB · Views: 11
  • H2PToday2201_design_RedHillLabs.pdf
    1 MB · Views: 5
Back
Top