MC34063 based Nixie tube power supply oscillations

[ntoskrnl]

I built a Nixie tube clock with an MC34063 based switch-mode power supply for the tubes. The input voltage is 12V and the output is 180V. Here's the schematic:



The circuit was mostly inspired by this page: https://threeneurons.wordpress.com/nixie-power-supply/

The power supply works perfectly. However, it has a rather low efficiency (around 50–60%), and when I looked at some voltages with my oscilloscope, I noticed some oscillations that should not be there.

Vds of Q1:


(Vdiv: 50, tdiv: 10µ)

Pin 2 of U1:


(Vdiv: 2, tdiv: 5µ)

The circuit works by switching the FET approximately 2–3 times quickly in succession and then waiting a little longer before repeating. This is why the waveforms look a bit weird on the oscilloscope. However, my concern is the oscillations that happen after these 2–3 pulses. (Maybe the notches on the falling edges of the pulses are also related?) These turn the FET partially on and cause significant power loss.

The oscillations seem to be coming from the chip rather than from around Q2/D2/R1. What could possibly cause this? Coupling from the inductor maybe, but where?

[Mads Barnkob]

Parasitic inductances in the circuit that oscillates with the gate capacitance, add a gate resistor to the MOSFET

[ntoskrnl]

Hmm, the frequency of those oscillations seems too low to be caused by that – the frequency is only around 400 kHz and I would expect gate capacitance oscillations to be in the MHz range. Or maybe they can be that slow?

There is also almost no ringing whatsoever on the rising edge of the gate pulse:


(Vdiv: 2, tdiv: 50n)

I could try adding a gate resistor, but it's a bit of a hassle. I wish I had caught this issue before I made the PCB... I also don't have any resistors between 1 ohm and 1 kohm at hand right now. But I guess it's worth a try if you think it will help?

Do you have any idea if it could be caused because I used a BC558 as Q2 instead of an MPSA55 as specified?

Edit: Oh my god. Q2 was installed backwards. I turned it around and now the waveforms are super sharp. I'm surprised the circuit even worked with the transistor backwards.

This seems like a curse. I don't think I've ever managed to finish a project without installing a transistor backwards. Not even double-checking seems to be enough.

[ntoskrnl]

No, I spoke too soon... The gate waveforms and efficiency are better but there are still small oscillations which cause large oscillations in Vds:


(Vdiv: 50, tdiv: 10µ)

The frequency of those oscillations is still the same 400 kHz.

[T3sl4co1l]

That's the clamped oscillation of L1 with Q1 Coss.

Obligatory:
MC34063, really?

Tim

[ntoskrnl]

Yep, I figured it's probably something that should be there because I get those oscillations even when simulating this circuit. There's very little current flowing through the FET during that so I guess it doesn't waste much power. I get around 75% efficiency now which is pretty good for an MC34063.

I'm using an MC34063 because I've heard it's easy to use with its very simple feedback loop. The circuit worked perfectly even on a breadboard and I didn't have to apply any black magic to my circuit board design.

[Mads Barnkob]

Did you try to add a gate resistor? Just something really low like 1R

[ntoskrnl]

No, I didn't try that. But I tried it in the simulation and it didn't change anything. I will try it if I ever build a similar circuit on a breadboard again.