I have a working fluorescent lamp inverter!
This time I built a H-bridge. I used bipolar transistors, which is going to add another 40mA or so to my current consumption; but at least it doesn’t need a higher voltage or P-channel power MOSFETs (which are about as rare as rocking-horse muck). Maybe I can do that for the Mark Two version — after I’ve built a little stereo amp with the remains of the TDA2030 jobbie. In the meantime, this bridge will do. And it’s inherently safe against both transistors on the same side turning on (barring a driver transistor failure).
The transformer is just a 12V / 1A one from my junk drawer (otherwise, I’d have chosen one with a centre tap and not had to faff about building bridges). This circuit runs at about 90Hz in order to try to squeeze a bit more power through without saturating the core. If you’ve ever opened up an eco-bulb, you’ll know that the supply gets rectified anyway; it’s really just a fancy switched-mode power supply.
One thing I didn’t do with the failed TDA2030 design was to ensure a 50% duty cycle. Of course it should be 50% anyway, barring any deviation from ideal op-amp behaviour — and indeed, was almost bang-on until I connected that pesky load (and I think you will agree that an extra couple of ohms output impedance is a deviation from ideal).
The easiest way to get a 50% duty cycle is to cheat a bit, and halve the frequency using a flip-flop. This also has the side benefit of offering two complementary outputs, which is just what my H-bridge needs. And, as luck would have it, there exists an IC with both a monostable/astable multivibrator and a flip-flop with complementary outputs: the HEF4047BE. It even helpfully provides the astable output from upstream of the flip-flop, in case you need it to do other things (like maybe run a step-up converter to turn on the top FETs in a H-bridge, for instance?)