I'm working on a Nixie tube clock and I'm having issues with the power supply. It's a flyback converter based on the LT3757 and the Coilcraft DA2032-AL flyback transformer. I designed to work with an input of 19 V and an output of 230 V at 38 mA of maximum load. The problem is that at its initial intended switching frequency of 100 kHz, it outputs 230 V under no laod and as soon as the tubes turn on, the voltage drops to about 140 V and the chip seems to enter some sort of protection mode, outputting a very low duty cycle gate drive signal. This happens at a load current of less than 10 mA.
I tried to change anything that could cause this: loop stability compensation capacitor, softstart capacitor, decreased current sense resistor, changed diode, tested with and without snubber, nothing helps.
I increased the frequency to 275 kHz and this caused the transformer to heat up to 80 degrees Celsius and the diode to fail after about 2 minutes. So I found a sweet spot at 185 kHz that provides enough power without excessive heat, but I'm still not happy with the final result because I did not find the root cause.
I simulated the circuit in LTSpice at 100 kHz and it can provide up to 70 mA, so there is a huge discrepancy between simulation results and practical results.
I think the limitation is due to the transformer somehow, but I'm not sure. So my question is: what could cause the circuit to fail to provide enough power at 100 kHz?
Here are the schematic and the layout:
Here are some scope captures at 100 kHz under load, with snubber populated, during the described behavior:
CH1 - Q2 Gate
CH2 - Q2 Source
CH3 - Q2 Drain