To filter a square wave output of the Tx into a sine wave - ideally you need a 2 stage filter - the Fo of these filters will generally be below 400Hz - the first stage by quite a bit i.e. 1 / (2. pi . sqrt (LC)) << 400Hz
the L in the 1st stage needs to be sized to limit the ripple current in the filter for the 220V pulses that are hitting it e.g. 200V for 1.25mS across 1mH = 250A
( V/L = di/dt ) so 1mH is too small ...!
To get down to +/- 1A delta I, we need L = 0.25 Henries, ( 250mH ) - i.e. quite large, and 1A on the HV side is 1 x 200/12 = 16.67A on the LV side
So your fets will need some pretty good heatsinking just to run at no load ...!
Say we make the filter Fo ( 1st stage ) = 33Hz ( well below 400Hz ) we need Co = 93uF at 275Vac rated ( and at least 1A rated )
So you need big L's & C's at 400Hz even to filter low power ( say 200 Watt ) even in the 1st stage of a filter
also if your power source cannot supply 12V @ 20A you will have real trouble with circuit stability ....
if your ferrite transformer is not rated for 200V 400Hz on the HV side and 12V 20A on the LV side - you will have more issues
dB/dt = V / ( N. Ae ), So allowing Bmax = 0.3T, dB = 2x = 600mT = dt. 12V / ( N. Ae ) [ Ae = core effective area in m^2]
if Ae = 1.5 cm diameter for example ( 176mm^2) , then N can be as low as 142 turns ( HV = 2400 turns )
I don't think your Tx is made this way ...(?) to get to 10 turns on the 12V side ( at 400Hz ) you need a core area of 2500 mm^2 ( and 180 turns on the 220V side )
or 50 x 50 mm ( think E65 core - but 5 of them side by side ), at the moment the transformer is shorting out the LV side as the core goes to saturation at fairly low currents ( i.e. the Bmax tries to go way above 300mT but plateaus and the pri current spikes way up high ... )
happy designing