About the CD4049 there are contradictional info’s. My data book from 1981 states that pin 16 is “3 to 15 V but more than at pin 1 ” and it works fine when connecting them (voltage is equal). The infos in www.alldatasheet.com indicate that pin 16 is not connected. There must have been a change some time ago.
There are 6 inverters. The first two are connected to form an oscillator, relevant for the frequency is the 100k resistor, the 330p capacitor and the potentiometer. The 2 diodes on the trimmer makes the impulse/period-relation variable. The four parallel inverters are just to drive the driver transistors in a push-pull configuration.
These drive the power MOSFET, I used an old BUZ11, works well, but a more modern IRFZ44 would work at least equally well.
The transformer is the key component. The winding data are given in the schematic. I used Teflon tape for insulation. Remember that if you stretch it, it becomes thinner. You need an adequate insulation between primary and secondary, and a very good insulation between secondary and the “upper” heater winding. This is because there are several hundred volts over the capacitor.
The capacitor is needed to keep off DC content, because the lamp will try to generate a DC-content and pre-magnetize the core and bring it into an unwanted operational state. The lamp partly behaves like a diode. This is one of the odd features of fluorescents, in addition to having a nonlinear response to voltage and frequency.
The inverter is a forward converter. If we just switch a MOSFET on the primary side on and off, it is not yet clear if we use the inverter in forward mode or flyback mode. This is decided by the other components around. Here we have a demagnetization winding, wound bifilar in parallel to the main primary winding. The switch-off voltage spike is fed through the BYS- Schottky power diode back into the supply capacitors. We could also destroy it by using a snubber or RC-combination. But recovering is far better for the efficiency and heat dissipation.
The starter relay simply switches off this de-magnetisation winding for a short moment and the inverter generates the mentioned voltage spikes of a flyback converter. These appear on the secondary side and start the lamp immediately.
For this short moment, the built-in zener diode of the BUZ or IRF cuts the voltage off at ca 60V on the primary side. It would not be permitted to use it continuously in this mode, it would probably overheat.
After switching the device on, the relay becomes active for about 1/4 second and opens its normally closed contact. Then the capacitor is fully charged and the relay drops off. The 10k resistor just discharges the capacitor in reasonable time, so that you can restart again after having switched off.
So the starter is very simple, just three parts.
You need to take out the built-in normal starter, if there is one embedded in the socket, as with the PL-11 or Osram-Dulux 11W lamps.
Use a plastic screw to slightly bolt the pot core halves together. These are common for model airplanes. If you cannot get plastic, use hard wood, cut a thread on it and harden the thread with CA glue. Remember that there is an air gap and exerting too much force makes the core crack due to internal bending. Metal screws will generate eddy current losses and are not recommended.
Take great care that all wires come out of the pot core without touching the ferrite and, of course, without touching each other. Only the filament heating wires can be twisted, the enamel insulation will stand the ca 6V.
Alternatively to Teflon tape you can use model airplane tissue or even paper from the copy machine, secure with thread and dip it into a can of clear cellulose lacquer. Suitable is such one, which is called Banana Oil in Britain. Boiling in paraffin is best, but you cannot use the standard thermoplastic winding former, it melts away. This insulation technique needs special materials.
Try to get enameled copper wire from an armature winding workshop, they have the best quality. (The darker the better). It must withstand the heat of a soldering iron.
The inverter has a very high efficiency. I have no measurements, but as an indication, the estimated stationary temperatures at 20°C room temperature:
Pot core 35°C. MOSFET on a small heat sink 40°C. Secondary capacitors 40°C. Primary capacitor 470µF 50°C. Surprise, how come?
The loading with a pulsed current of 18 kHz generates losses. As an improvement you could try to enlarge the 1µF foil capacitor to ca 5µF and put some ferrite beads (or even a small choke) between the electrolytic capacitor and the foil capacitor.
But it works also as it is documented here, and has been in use now for 5 years.
The small PC-board is just for the impulse generator, the rest can be wired by hand. You may change it to fit to your trimmer (many versions) and capacitors (standing or horizontal). Check the printout scale before making the board, from pin 1 to pin8 should be 7 x 2.54 = 17.78mm distance. The jpg shows the component side.