Marko's SSTC

Marko, Fri Jul 11 2008, 01:58PM

This is what I have been busy through most of last year. It is not really anything special to you guy's standards, but since it wasted me a great deal of time and since there's no chance I'll work on it anymore I really think I should have it posted.

Basically I started from this stupid idea of building my own heatsinks from aluminium sheets.
At first I wanted 3mm aluminium but only got 1.5mm. For some reason I thought it would be ''just fine'' (??)

I started cutting them up and building the structure I wanted, which wasted me a lot of time again. I wanted to have entire inverter and control electronics built into something 'solid'.

At first I wanted to install a stack of fans around the edges, but small 25mm fans were too expensive and I abandoned that idea (completely ignoring the fact that there will be nearly no natural convection from what I did).





I wanted to use 2 parallel bridges of IRFP450's. In fear of parasitic oscillations I separated them completely apart the power supply. Note the odd orientations of switches: I had 4 Al plates, and I wanted the thermal contact




I used one of these coils for a secondary. Resonant frequency was around 650kHz without and 400kHz with dedicated toroid.





How I built the H bridges. Using cables to provide gate drive for each device separately proved to be a mess and a problem. Later I found out how huge difficulty is to assemble this thing together.




This is how the first assembly looked. Despite I glued them sil pads kept falling off, and I only succeeded after several attempts. Thought to have succeeded, at least. Note the weird circuit boards that I put in, they didn't last for long.

The circuit consisted of nothing but gate drivers fed directly by clamped CT output. When CT is clamped with zeners, it will put out a very good approximation of square wave since it is effectively a current source. I believed this is just OK and a test with small half-bridge coil proved that it works. I for some reason believed that simple circuit will minimize delays and be simple to debug.

The circuit is started up by a short pulse given into one of gate driver inputs at rising edge of enable signal. (enable signal was just a simple switch at that time).






Of course, after I powered everything up for a basic test, it didn't work. UC3710 gate drivers I used 2+2 in parallel seemed to shut down, latch up or whatever, for no apparent reason.

Also, since I wanted this coil to run from full wave rectified mains, with no filtering (because I did not have any filter caps) or interrupter. The problem was that even with the small coil the circuit tended to be unstable with fullwave rectified mains, tending to quit oscillation at zero cross. Some smaller filter caps helped, but then other problem appeared - if enable signal was held to 1 circuit could spontaneously start without warning, despite it didn't receive start pulse. Transients caused by applying the +325V were enough to start it.

When I tested the circuit in the big inverter I just blew mosfet's for unexplained reason.

Because of this and the first unexplained problem I started to dislike the circuit and quickly abandoned it.


My next circuit used 74HC14 and mostly antenna feedback (CT never worked with it). I wanted to use NE555 to constantly provide enable signal, as well as I switched to only 2 UC3710 gate drivers - it is completely pointless to parallel drivers in a system where currents are anyway dictated by leakage inductance of the GDT.

I tested the circuit in a small experimental coil and it worked well.

To much of my frustration, circuit didn't work again when I put the board into final setup! The UC3710's were shutting down again!

I was very confused and unable to find what problem was.

Apart from that, circuit had pitfalls (I was basically trying to send DC through the GDT, weird waveform in case oscillation fails to start, and potentially damaging to mosfet's).
It also blew a mosfet in the test circuit once I tried to use it with CT instead of antenna.






Desperate, I ordered a lot of IXDD414's believing there is something wrong with UC3710 gate drivers.

The new circuit was probably the most complex thing I ever designed completely myself.

For start-up, it used a dedicated 50% duty cycle oscillator tuned close to soil's resonant frequency. The oscillator would drive the output first. As feedback signal appears, it disables the oscillator and gates itself to output. If feedback signal disappears for any reason, after 200us delay oscillator starts up again.

I used 74HC14 gates with RC networks to generate deadtime on the output.

To my huge surprise, the circuit seemed to work *excellently* in test setup.





I put the board into the big inverter, set it to oscillate at desired frequency (400kHz) and put some light bulbs onto bridge outputs.

Oddly enough, mosfets blew again.

I dismantled everything, removing the boards and putting in a single lash up air-wired H bridge. I powered it up, it seemed to work fine with a light bulb - then I switched to resonator - and witnessed some 30cm sparks! Only for few seconds, because mosfets blew again then.

And after inspecting everything very carefully, I realized only one thing can be the cause - opened the bridge, and after some inspection I found out that a sil pad is punctured by chunk of aluminium and was shorting mosfet drain to ground. I somehow failed to see it in continuity checks, and it seemed to be short or open at random.

I never changed the glued sil pads even though I've changed mosfets a lot, and god knows how long was the Al chunk in there causing problems.

It may have been there since the beginning and maybe it was the reason why the first circuit didn't work??

Anyway after assembling the coil back it worked sort of well, for a while.

I also got the second H bridge to run in parallel, for a short time, producing some 40cm sparks (not a too great improvement from a single H bridge!)

I reduced the primary turns for about sqrt2 to get the current in all mosfet's same as it was with 1 H bridge.

So here are some spark pics:








I think I never actually photographed the 40cm sparks I got from paralleled bridges. Sparks on the pictures are 30cm at most, still very hot.

The heatsink performed very poorly - sil pads added a huge amount of thermal resistance, and thin 1.5m aluminium conducted it poorly even if I used external convection. Run times were like 2 minutes a most!

But I didn't care, I was so excited about the sparks I got that I wanted to wrap the coil up.

I started designing various 'upgrades' like a tappable primary coil (intended to be ran with and without breakout point, for wireless power demonstrations), a remote control, proper power supply, etc.



The coil sat for few weeks, and after I was done, I powered the coil up -on my horror mosfets blew again after only few seconds of run time!

As I needed the coil for demonstration I started to panic - in debugging mania I took everything apart and put together many times, replacing sil pads, mosfets, mutilating H bridge boards, and blew like 20 mosfets trying to find what was wrong.

But I didn't.

MOSET's blew instantly, or blew fuse at some voltage level, even completely cold. Circuit was extremely hard to analyze because of it's crappy design.

Desparate, put back the lash-up bridge from beginning, which somehow worked (I have no clue why).

Replacing the mosfets was such hard work, and I could damage something every time I do it.

So the coil still sits with mock H bridge, I don't run it anymore as it can just blow any time(I never found the fault) and I'll never be bothered to replace mosfets again. It failed to meet any reasonable reliability standards.

So this is how project ended, I just messed up so hard that nothing can save it any more.

To summarize some of flaws:

-Completely silly heatsink design. I put such a hard work into it, and really failed to understand basic thermal management. The thin metal conducted heat poorly, and I had no forced air stream which is absolutely required for a heatsink in horizontal position.

The entire effort was absolutely not worth buying some CPU heatsinks or even large new heatsinks from ebay. Only explanation of why I did this was because I was an idiot.

-Sil pads are extremely poor conductors of heat probably no better than 3-4K/W and heatsink of any size is vain while they are used.

-Trying to assemble some odd sort of structure I put way too much pressure onto sil pads, which was one of reasons why they got damaged.

-mess of gate drive cables was just horrible, any of them could have easily got damaged or shorted to something.

- not enough space left for a power supply, I ended always using external one.

- extreme difficulty of changing mosfets. I was arrogant enough to think I won't do it often.

- paralleled H bridges are just unnecessary bother over using bigger mosfets. No need of complicating things, especially if mosfets are so hard to change as they were. etc.

So yes, it does spark, but after all these pitfalls, unreliability and final problem I couldn't solve I don't think I can really consider this project a success.
I hope I at least learned something for my future coiling.

After I collect some money I'll start anew, but probably after resting from ''three legged fuses'' for a while.

Some older pics of lighting bulbs wirelessly:





I added some more pics of current state of the project.
I have no more motivation of repairing it's full capability or doing anything more with it.

Hope you enjoy the pics. Thanks for reading!