Unexpected boost action in a push-pull converter?

[Artlav]

Greetings.

I'm trying to make a 12V to 320V DC-DC converter with TL494, and there is something weird going on.
Schematic: https://i.imgur.com/u4qGQbL.png
50Khz frequency, 5+5:168 transformer (XF), no isolation, 50W expected power.

The problem is the diode D1 between the input and all the control circuitry.
With it, i'm getting over 20V after it (at 12V input!), or more - causing all sorts of explosive problems.
Without it, i'm getting chaotic performance and all sorts of oscillations.
I've tried to avoid the voltage rise by putting a resistor or a zenner parallel to the diode - then it works perfectly fine, but either start heating a lot.

Surprising part is, it's the second time i'm using this exact schematic (originally the diode was put in to avoid battery voltage sags affecting the control part), and the first time around there was not a hint of either problem.
And i still have that first build, identical down to component values used - and there is no such problem there.

What is going on here?

 

[fethiyeli]

It may be flux imbalance. You may need to connect a capacitor between source of high side mosfet and transformer.

 

[FvM]

Sounds like a large cable inductance in battery connection. Either if flux imbalance or other problems are causing excessive curents in the push-pull stage or not, effective supply bypassing would be required anyway.

 

[Artlav]

It may be flux imbalance. You may need to connect a capacitor between source of high side mosfet and transformer.

Flux imbalance?
What is that?

Also, there is no high-side fet - i might have named the topology wrong, but it's a center-tapped primary winding, with the ends going to mosfet drains, and the centre is 12V.

effective supply bypassing would be required anyway.

Nice...
Adding a 3300uF capacitor next to the transistors was quite effective.
There is still a bit of a noise, but nothing is heating up any more.

So, just wire inductance and source impedance?
Now that i think about it without being distracted by the weirdness of the symptoms, it's quite obvious that lack of input capacitance is not exactly healthy for such a circuit.

Thank you for helping.