PSU Repair: A Case Study

Auxiliary Output Results

Based on the weak and jagged main output ramp on the previous page and my earlier hypothesis that the main converter IC is not receiving sufficient power to operate normally, I would say almost certainly yes. The nice thing about knowing how the circuit works and having an oscilloscope is that I can hook it up and see how badly.

The first slide shows the bottom capacitor voltage, auxiliary output voltage and PWM controller supply voltage while the main outputs are off. Something looks very wrong here. I was expecting to find out what the nominal voltage on the auxiliary supply was, but its waveform peaks at -132V and does not appear to be holding any voltage at all, dropping to about -156V half-way between pulses. This is so different from what I was expecting to see that I had doubts about somehow soldering my probe wire to the wrong trace--the long, relatively fat, unmistakable one that links the auxiliary output to the PWM section. On the PWM's side, the fast edges from the auxiliary supply ripples are causing glitches to pass through. However, the output is otherwise holding steady at about 8V, which is consistent with the 3844B's under-voltage lock-out, so the PWM's bypass capacitor might still be good. Perhaps one of the other operating states will shed some more light.

The second slide shows what happens to the auxiliary and PWM supply nodes when the main outputs are trying to turn on. The math channel shows the voltage across the PWM's supply bypass capacitor. There is nearly 20VPP of noise on the controller's supply and the DC value is barely 10V with some dips below its under-voltage threshold. Not many chips will operate well when their supply has nearly twice as much noise as it does DC, on top of input voltage being barely above the minimum operating voltage. The W12NK80 MOSFET used in this power supply needs at least 8V to switch 6A, but the PWM controller needs to have a supply a few volts higher than that to offset its own gate drive's voltage drop and the voltage drop across external gate drive components, meaning 10V is nowhere near enough for reliable operation.

The final slide shows what happens when the PWM manages to turn on. I forgot to enable the math channel on this one, but if you use a little imagination to extrapolate the voltage between the bottom capacitor voltage and aux/PWM supply voltage, you can see the controller IC's input voltage is still quite messy. If you have trouble imagining this, look at areas where the cyan line and the two others move in opposite directions. Those would be spikes, many large ones at that. With finer time resolution, we can see that the PWM supply voltage is tracking AUX voltage almost perfectly, indicating that the photocoupler has no trouble driving its pass transistor into saturation to feed the main switcher, so we can dismiss the on/off control photocoupler and its pass transistor as potential concerns. We can also see fairly well that the auxiliary voltage is about 20V above the bottom cap voltage on average.

The auxiliary cap definitively needs to be replaced, and given the way the PWM bypass capacitor tracks it, there is a high probability it needs replacement too. If the PWM's bypass capacitor still had even 1µF of its nominal 22µF left, I would expect to see that charge decay over a few microseconds between top-offs by auxiliary peaks instead of momentary glitches.

If this was a normal repair job, I would replace both on the same soldering iron run and ask questions later. But since I am taking a case study approach, I am fixing problems singularly so that I can observe, analyze and comment on causes and effects, one variable at a time.