gdt high frequency ring

[balazs]

Hey everyone!
I build kaizer sstc1. The output is present in a high frequency ring. When i test 30v  works, but when it is 230v then the Q1 mosfet is immediately destroyed. One of my questions would be: it’s written here: http://www.richieburnett.co.uk/temp/gdt/gdt2.htmlthat one damping resistor I should add, but where and how much. My other question would be, could this high-frequency ring cause the instantaneous death of the Q1 mosfet?



Thank you in advance for the answers, and sorry form my English

[John123]

Is that across the gate and source of one of the fets? If so then yes it will cause it to fail, put more series resistance in series with the 4.7 ohm gate resistor.

Can you show a real life picture of your circuit and gate drive transformer?

[davekni]

Looks like an issue of parasitic inductance in half-bridge interconnect.  Yes, pictures of construction will help.  Here's a link to my example of low-inductance half-bridge construction (overlapping parallel planes):
https://highvoltageforum.net/index.php?topic=1324.msg9795#msg9795

[balazs]

The signal is measured parallel to the primary coil. I exchanged resistances for 10, but the signal remained the same. The signal generator which interrupts the ucc chips looks like this because I'm using it from a signal generator, but I tried it with the original connection, but the signal was just as ugly. I tried to shorten the cables, but that didn't bring any improvement in the signal either. This structure is by no means definitive since the half-bridge can be converted to kaizer sstc 2.

[davekni]

Yes, definitely issues with wiring inductance.  Half-bridge has the highest current, so needs lowest inductance.  After fixing the half-bridge, you may still find issues with gate drive wiring inductance using plug-board construction.  Another recent thread shows gate drive oscillation due to plug-board construction wiring inductance:
https://highvoltageforum.net/index.php?topic=1499.msg11544;topicseen#new

Circuitry before the gate driver chip may be OK on plug-board.  Even the GDT chip might work if a bypass capacitor is added directly over the chip with leads cut to minimum length and plugged in adjacent the driver chip power and ground pins.

[balazs]

Thanks to everyone for the help!
 I rebuilt half-bridge and the problem was fixed immediately and 230 V is already working

[davekni]

Glad to see your good news.  Have fun with your coil!

[balazs]

I'd like to ask for some more help. True, the signal is much nicer, but it still has a little ringing. One of my questions would be how to improve the present.
The other would be that the interrupt signal can only be transmitted over an optical cable, or a coaxial cable is already good for this. Can I connect this coaxial cable so that both the negative and protective ground go on the outside?

[John123]

I'd like to ask for some more help. True, the signal is much nicer, but it still has a little ringing. One of my questions would be how to improve the present.

Weird how the ringing is asymmetrical, firstly I would put anti-series 18v zener/tvs diodes between each gate to source to clip anything from exceeding 20v and damaging the fets (small glass zener diodes are fine as long as they don't normally conduct: 18v zeners in anti series protecting a 20v max gate with 15v nominal gate drive), then add a pull down resistor between gate and source (4.7k-10k 1/4th watt resistors are fine). You can use odd sized zeners in series if you don't have ones with suitable voltages.

Downside to this is it will incur a slight switching speed penalty but in your case the pros far outweigh the cons (I include them anyway on all of my circuits like this for peace of mind).

I've noticed there's still a lot of flying wires in places particularly for the gate drive transformer, the wires coming out of the gate drive transformer should also be twisted in their respective pairs all the way to the fets legs/back to the gate driver chips. Alligator clip jumper wires are not good for this.

In other words the only place the gate drive windings should cease being a twisted pair is right at their place of termination, every untwisted inch is bad news.

Like this:


For example in your pics I can see a white gate drive wire terminating on the right hand side of your bridge circuit right next to a freewheeling diode, this increases the loop area, adds unnecessary inductance and capacitance which fuels things like the ringing you're seeing.

It may also force the gate drive current to share a current path with the tesla coils primary coil (bad since it further amplifies everything). Unsolder it, twist in a pair and resolder it right at the legs of the MOSFET (think of it like a delicate signal line).

Trying putting a low ohms resistor in series with the primary coil (2.2 ohm is an ok starting point).

Do you have plenty of decoupling capacitors at all the IC supply pins? You atleast a small (like 100nF ceramic) for each IC, the gate drive IC's like a larger bulk capacitor (low ESR electrolytic) in parallel with the smaller ceramic capacitor in my experience with short leads (every cm counts).

If it still won't calm down then there's always the brute force RC snubber approach across the primary of the gate drive transformer.

[balazs]

I twisted the cables and soldered them, the signal got a little better, but still not good. Salynos There are currently no zener diodes or potentiometers.
I tried the scroll a bit today. It works until I pick up the breaker frequency, maximum, because then it works for a few seconds and then the two mosfets burn.

[John123]

Much better but still appears to be ringing more in one half of the cycle, I think we need to find out why first. The other half of the switching waveform looks much better but still touches 20v (gate breakdown voltage).

It could still be the gate breaking down making the mosfets fail as it exceeds 20v.

[balazs]

I measured the primary coil in parallel, so there is 20v because I can't measure 230v with the oscilloscope.
The weird thing is that the mosfet broke because I tried several times and cut off the FI relay once. Other times, when the antenna was farther away, it worked. But the antenna was still not that close to 17-25 cm.

[John123]

I measured the primary coil in parallel, so there is 20v because I can't measure 230v with the oscilloscope.
The weird thing is that the mosfet broke because I tried several times and cut off the FI relay once. Other times, when the antenna was farther away, it worked. But the antenna was still not that close to 17-25 cm.

Oh, I was thinking you were probing across the gate and drain of one of the mosfets. Are you connecting your scopes ground clip to the circuit without isolation when it fails? It could be the cause of the fet failure as the ground connection is referenced to neutral where the electricity enters your property.

[balazs]

I use a battery-powered oscilloscope so its exterior is not grounded. But when the mosfets break, the oscilloscope is not connected because my probe only measures 300V. I have a 30V lab power supply that I can test on before 230V. I could find out what causes the asymmetry: I have a probe that can be set to 1X or 10X division, but with which you can set the probe not to be distorted, it only goes 10x. So the signal is already symmetrical, true, much uglier. I tried to connect the panel to the power supply on shorter wires, which helped a bit. I don't know if he can't ring from him because the DC cable is still too long.

[John123]

Ah gotcha now. Well the ringing you see across the primary is normal as long as it doesn't exceed the breakdown voltage of the fets, I was thinking that was the gate which would of been bad.

I think to get the full probe bandwidth it usually needs setting to x10, x1 will also load your circuit more so maybe an RC snubber could be useful here if the ringing changes with just the probe loading it.

Just out of interest what does the waveform look like on the gate?

[balazs]

I measured this signal on the gdt output. It is certain that the mosfets receive too much tension. But it’s weird because when I use 30v, the 230v shell doesn’t ruin the mosfet. And when it breaks, it explodes, which is more like opening the 2 mosfets together.
It would help to reduce the number of gdt revolutions from 15 to 12?

[John123]

That certainly looks more than 20v at the gate, it might well be inducing the failure especially when it's switching a real load. I'd include gate protection before proceeding.


When one MOSFET blows it take out the other because the remaining working FET shorts out the supply. Remove that mains filtering capacitor and use currenting limiting for now (or a big resistor), that way a fault has less chance of taking out more expensive MOSFETS.

Are you sure your gate drive chips are being supplied with a regulated 12v and you haven't put in another regulator by mistake?

To reduce the voltage supplied to the gates remove a few turns from the gate drive transformers secondary windings only, it's a transformer afterall so you can manipulate the turns ratio to get the voltage you want. I'd aim for 12v.

[davekni]

Zener clamps gate-to-source are a good idea.  It will also help some if the GDT output connections solder to FET source lead rather than to the copper planes.

I recommend setting your scope vertical mode to DC-coupled.  That makes it easier to see asymmetry accurately.

The gate itself will have less ring than GDT output.  That is what the series resistor is for.

Can you share a picture of how the scope is connected when measuring gate voltage or GDT output?  Ringing on the half-bridge output can couple into the loop between scope probe and scope ground clip.  Another suggestion:  probe GDT output and/or gate voltage with 0Vbus across the bridge.  That avoids any coupling from half-bridge output.  (It does change gate capacitance too, so isn't a perfect solution.  Still helpful for comparison.)

[John123]

I suspected his scope might be mistakenly adding some offset compared to the center of the screen, it's the asymmetrical ringing amplitude what's got me puzzled.

I second what Dave said about where to put the probes too, put it right at the pins of the MOSFET gate to source and use the short ground spring rather than the long ground wire.

[balazs]

I made a new gdt because I wasn't sure if the cables were the same. I also soldered the gdt output to the mosfets foot.
I switched the oscilloscope to DC mode and measured the output of gdt. The probe is placed on one leg of the mosfet and the part of the resistor where the gdt is connected with a cable. What does it mean if there is 0 Vbus on the bridge?
I added an offset because without it only the beginning of the signal is visible. The first image is offset.