These 2 R-C combinations might be replaceable by simple diodes (cathode at 12 V, anode at the end of the coil) - you can try to put the SBLs there.
I'd expect this to work like a charm.

That won't work at all.
One MOSFET drain will be at 24V whilst the other is switched on, simply by the transformer action of the two parts of the primary winding.
In a less-than-perfect transformer there will be some overshoot.

 

That won't work at all.
One MOSFET drain will be at 24V whilst the other is switched on, simply by the transformer action of the two parts of the primary winding.
In a less-than-perfect transformer there will be some overshoot.

That much about my "might". Back to 2 * R-C

 

I think I'd be looking for the problem in the transformer itself.

 

I think I'd be looking for the problem in the transformer itself.

Seems like the diode idea was not that silly: from some solar inverter provider (sorry - German. But I think it's understandable):

 

Seems like the diode idea was not that silly: from some solar inverter provider (sorry - German. But I think it's understandable):
View attachment 237680

But you have R2 in series with the diode. There will be 12V less the diode drop across R2, which will dissipate 6.5W, so don't touch R1/R2 or you'll burn your finger!

 

But you have R2 in series with the diode. There will be 12V less the diode drop across R2, which will dissipate 6.5W, so don't touch R1/R2 or you'll burn your finger!

Touching power resistors during circuit operation has never been a reasonable idea

 

Touching power resistors during circuit operation has never been a reasonable idea

But it's such a handy way of checking to see whether things are getting too hot! I usually remember to ask myself "is that part live?" before going in.

 

How do you get it to calculate B?
Note: The energy recovery circuit only works for a non-isolated output.

Usually it is pumped back into the DC buss not the output and this is mandatory in a DC to AC converter.

 

Check out my solution. The energy stored in the leakage inductors is given to the load.View attachment 237675

Try pumping it back into the DC buss, that's the way it is usually done. For DC to AC converters (synth sine output) that's the only way. Still, snubbers are also used as they are the fastest being tightly connected to the transistors collector emitter or drain source terminals

 

This is a interesting shot. Yellow channel=gate, Blue channel=drain

So how did you wind the transformer?
How much power is it meant to deliver?
Are there enough turns for 16kHz? - which, as @Papabravo said, is quite a low frequency for this type of design.
Are the two primaries wound bifilar? That would be the worst case for capacitance between them.
It doesn't look like a very neat transformer - it looks as though there are more turns at one end than the other which would increase the leakage inductance.

 

Hi,

There are at least two connections to the MOSFET source terminal.
1. The heavy ground lead that carries the main power current which gets turned on and off.
2. The light gauge signal ground lead (from the driver circuit),
Now if the main power current (that gets turned on and off) flows though any part of the ground signal lead it modulates the gate to source voltage which when the MOSFET swtiches it could caues it to turn on and off several times before it settles to just one state (on or off). If the signal ground lead is separate, it goes to the source alone so none of the high current can flow though any part of it.
If the signal ground lead is not separate going to the source then that means is must have connected to the power ground lead somewhere before the MOSFET (not directly at the terminal) so when high power current flows it is flowing through the same wire that the signal current is flowing so the signal sees an increase (and later decrease) in current, which of course means that both the resistance of the shared wire and the inductance of the shared wire cause voltage drops. The resistance causes a set amount of voltage drop that starts when the MOSFET switches on and ends when the MOSFET turns off. The inductance of the shared lead causes a voltage drop spike that starts when the MOSFET turns on and decreases some short time later, then goes negative when the MOSET turns off then goes back to zero some short time later. The inductance is the most problematic and causes oscillation but the resistance doesnt help any either.
Keeping the leads separate avoids both of these effects that cause oscillation during part of the switching period which causes ringing which causes more power dissipation in both MOSFET and diode.
So the MOSFET source gets a heavy gauge lead direction from the input capacitors right to the source, and the driver circuit gets a separate lead from the input capacitors to the driver circuit, and then a separate lead directly to the MOSFET source. So there ends up being two separate leads connected directly to the source terminal of each MOSFET.
"Directly" here means as directly as possible. Sometimes a PC card is mounted right on the MOSFET heatsink so there may be only one short lead to the MOSFET, source, but then the separate lead connects to the PC card. If there is a PC card you can still keep them separate though just be running two separate leads from the PC card to the MOSFET source terminal.
So this is quite simple really, see the attachment. The big cap is the input capacitor bank.

View attachment 237669

Crystal clear. Many thanks. And thanks for the diagram. I've tried to apply some of your advices in my PCB originally. Had separated copper pours for GND, one for the chip + totem poles and other for mosfets and the rest in a try for any kind of ground separation. I presume ground loops is a escalating (and ringing) problem. I nearly got a fire and smoke on all other board's capacitors trying the big film capacitor in between Vds like I said (while trying for more output power). It was probably the exageration and abuse of the situations you've explained.

So how did you wind the transformer?
How much power is it meant to deliver?
Are there enough turns for 16kHz? - which, as @Papabravo said, is quite a low frequency for this type of design.
Are the two primaries wound bifilar? That would be the worst case for capacitance between them.
It doesn't look like a very neat transformer - it looks as though there are more turns at one end than the other which would increase the leakage inductance.

Yes totally right. Transformer is far from neat. Papabravo is partially right about: I'm somewhere in between making something seriously (with knowledge of the cause) and threwing a bunch of parts up against the wall in something that "kinda works". I know it looks like crap but it's a chance for me, from my point of view. Issues potentiation is also knowing these issues (for the first time) and opportunity to practice how to mitigate (or from where they come from).
So I dont have a "engineered" planned output power. More interested in testing how far can I go with the transformer (and others to be tested), the mosfet types (and others to be tested), the actual PCB (with/without further adjustments) and so on. More of a learning project (and it's doing!) than pretention.

Actual life's first transformer is a no gapped one that I have winded 4+4 primary turns, then 63 on top of the primary coils. Core is 1 square cm.
I just used some online calculator for this. This is dumb but I'm not even sure what frequency to inform to the calculator. If I have 20KHz for each side of the primary coils, transformer frequency is 20 or it's 40? considering PWM for each side is interleaved. Anyway after getting my transformer like this I tested changing the frequency (10~80KHz range) for the point of max output voltage with a fixed load and this point is 16KHz.

Been reading a couple of really nice texts yesterday about transformers and the relations between leakage inductance and core gaps. High frequency transformers is a mind blowing subject (and dificult) but really interesting so far. Trying to absorb/understand the basic (baby steps).

https://www.powerelectronics.com/content/article/21861299/why-have-an-air-gap
https://www.electronics-tutorials.ws/electromagnetism/magnetic-hysteresis.html
https://authors.library.caltech.edu/98808/1/07415565.pdf
https://www.ti.com/seclit/ug/slyu036/slyu036.pdf
https://www.edaboard.com/threads/wh...-leakage-inductance-and-core-gap-size.350032/

 

Just for a test, attached a small transformer from ATX smps that I didn't touch. Probably the fact I'm using it in reverse is not neat as well. The spikes are
Channels are gate and drain.

20KHz - 112V
40KHz - 106V
80KHz - 94V
150KHz - 94V

 

Seems like the diode idea was not that silly: from some solar inverter provider (sorry - German. But I think it's understandable):
View attachment 237680

Tried it. It did nothing. If I get rid of R2 and R1 spikes are there same way and my lamp dimmed low, but not considerable heat from schottky pair.
This scheme looks some any other animal than I have here (push-pull), some auto resonant circuit perhaps? From traductor rechteckige welle=square wave, then I can't decide from where it's coming from or going, from 12Vdc maybe?

 

Crystal clear. Many thanks. And thanks for the diagram. I've tried to apply some of your advices in my PCB originally. Had separated copper pours for GND, one for the chip + totem poles and other for mosfets and the rest in a try for any kind of ground separation. I presume ground loops is a escalating (and ringing) problem. I nearly got a fire and smoke on all other board's capacitors trying the big film capacitor in between Vds like I said (while trying for more output power). It was probably the exageration and abuse of the situations you've explained.



Yes totally right. Transformer is far from neat. Papabravo is partially right about: I'm somewhere in between making something seriously (with knowledge of the cause) and threwing a bunch of parts up against the wall in something that "kinda works". I know it looks like crap but it's a chance for me, from my point of view. Issues potentiation is also knowing these issues (for the first time) and opportunity to practice how to mitigate (or from where they come from).
So I dont have a "engineered" planned output power. More interested in testing how far can I go with the transformer (and others to be tested), the mosfet types (and others to be tested), the actual PCB (with/without further adjustments) and so on. More of a learning project (and it's doing!) than pretention.

Actual life's first transformer is a no gapped one that I have winded 4+4 primary turns, then 63 on top of the primary coils. Core is 1 square cm.
I just used some online calculator for this. This is dumb but I'm not even sure what frequency to inform to the calculator. If I have 20KHz for each side of the primary coils, transformer frequency is 20 or it's 40? considering PWM for each side is interleaved. Anyway after getting my transformer like this I tested changing the frequency (10~80KHz range) for the point of max output voltage with a fixed load and this point is 16KHz.

Been reading a couple of really nice texts yesterday about transformers and the relations between leakage inductance and core gaps. High frequency transformers is a mind blowing subject (and dificult) but really interesting so far. Trying to absorb/understand the basic (baby steps).

https://www.powerelectronics.com/content/article/21861299/why-have-an-air-gap
https://www.electronics-tutorials.ws/electromagnetism/magnetic-hysteresis.html
https://authors.library.caltech.edu/98808/1/07415565.pdf
https://www.ti.com/seclit/ug/slyu036/slyu036.pdf
https://www.edaboard.com/threads/wh...-leakage-inductance-and-core-gap-size.350032/

Hello again,

Yes the addition of an air gap is an interesting thing that has benefits and disadvantages.
Looking at the attached drawing, the original BH curve looks like a stretched out "S". Adding a gap stretches that out even more into a wider ramp (which would still be limited top and bottom). That means it takes more current to saturate. Since the permeability is the slope of the BH curve and the slope obviously decreases when the air gap is introduced, the permeability decreases which makes it necessary to add more turns to achieve a given inductance.

 

Tried it. It did nothing. If I get rid of R2 and R1 spikes are there same way and my lamp dimmed low, but not considerable heat from schottky pair.
This scheme looks some any other animal than I have here (push-pull), some auto resonant circuit perhaps? From traductor rechteckige welle=square wave, then I can't decide from where it's coming from or going, from 12Vdc maybe?

OK - you didn't recognize the self-oscillating multivibrator in the schematic I posted:
in that circuit the transistors drive the transformer, at the same time creating their control signals. Not very "modern" (bipolar), but very simple to implement.