Problem with SMPS design for high power ultrasonic

I'm trying to design a SMPS do power an ultrasonic transducer from the mains. I'm going to have rectified mains converted to around 200Vac @ 25kHz. I've wound a transformer 200 turn CT on primary (guessed) and tried to make a push pull supply. Positive going to centre tap and 800V FET on each side to ground. I made a 555 oscillator and driving FETs with an IR2184 driver, this pulls each one high and low, gives some dead time and has collision ovoidance etc.

Note attachment no load - it's ok - I'm not sure why the huge leading edge spike.

When I turn it on (I'm using low voltage to start with) it goes all swimmingly drawing about 30mA with transducer load, although I'll have to try and keep that big voltage spike in check, - note initial attachment. Then I notice the current start to climb and the wave form changes (the wave form is transformer output with piezo attached) it holds low for a while then begins to climb quite quickly, in 10 seconds or so it's pulling 1 amp and notice the output wave form drop and change. - Just as well I didn't plug it in the mains!!!

I'm assuming my oscillator output isn't giving identical times for each FET and/or I've misscounted a winding or two when I made my transformer so it may be magnetically out of balance. Is it likely what I'm seeing is the core of my Xfmr saturating? If so - how do I fix it? How come the big spike is on one side when I turn the circuit on, then migrates to the other side?

Are there better ways to accomplish what I'm wanting to do? And how do I snub that nasty spike?

I've been working on this for days and it's really annoying me so any help would be very much appreciated. My biggest issue is that I'm a mechanical engineer so my knowledge base on this sort of thing is quite limited.

Many thanks in advance for any help/suggestions

By the way, the supply current to the transformer with no load is around 4mA, I've just noticed it climbs too but not as fast, have attached an output wave form after the input has climbed to approx 500mA.

It may be bad ground, no decoupling capacitor on power lines.

If you use a E-I transformenr,a little gap may help avoid saturation. Put a thin paper in between.

HTH,
Ioannis

Thx - I'll try gapping the ferrite core a little - one thing I cant get my brain around tho is: How come a slight imbalance in drive can cause the core to saturate when push pulling, but it's ok to just switch DC pulses through a transformer and only drive it in one direction without it saturating?

I've found this in the data sheets - my core is N97 operating at 25kHz tho the data sheet says:

"Large E/ETD, RM/PM cores made of N27, N97, N87 (with large core cross sections (Ae ≥250 mm2), on account of eddy-current losses N87 must be used even where f <100 kHz)"

Is this likely to be my problem? I cannot find an N87 core in my transformer size. Is it worth me getting a different size bobbin and starting from scratch?

Gapping and decoupling didnt seem to help much, is the topology with half bridge suppling one side of transformer and a capacitive divider on the other (as attached) likely to be easier for me to accomplish successfully?

Do you have any means to compare the two transistors used in the circuit?

The idea of two capacitors as voltage divider in my opinion is very bad.

In general I avoid this kind of circuits, as the parts should be identical (as much as possible).

Sorry I cannot help any more.

Ioannis