First, you mean most likely in general? Or first as in, first in the cascade of events?
Because failures are most often cascades. Consider:
1. Suppose an output diode fails shorted (due to overheating at high current and high temperature).
2. Now the primary side transistors are switching into a short. Then those transistors die (possibly explosively, as they discharge the big electrolytic capacitor in the process).
The diodes won't have failed visibly, but will be damaged.
3. Surge current is drawn from the mains, eventually blowing the fuse. This current (perhaps 2000A) is drawn through the input FWB or PFC rectifier.
4. That current possibly blows up one or more of those diodes (there are always at least two diodes).
5. If the unit has an active PFC stage (looking at pictures of the unit in question, it looks like it does), then those transistors will be trying to switch against the PFC diodes, which are now carrying fault current. So those might die, too.
Anything connected to a faulting transistor will probably also die:
6. In a conventional flyback supply, one transistor does the switching, driven directly by a controller IC (e.g., UC3842) on the primary side. During the fault, the transistor's input (gate) terminal surges to some ~100V, damaging the chip's gate driver section. The series gate resistor if present) may be blown like a fuse.
In a conventional forward converter, the transistor drive is transformer coupled, which is pretty safe. (An old ATX power supply is a typical design using bipolar transistors and transformer coupling.)
A newer style design, including two-switch forward or flyback types, half bridge and push-pull forward converters, and resonant converters, might use an HV gate driver IC (IR2110 as an example). This IC would be toasted. The controller IC connected to it may also be damaged (when the driver IC fails, it lets through some surge voltage as well, through its input or supply terminals).
Things that aren't likely to be destroyed:
- Low voltage secondary stuff. Surges don't couple through transformers well, so the secondary side usually rides it out through fault conditions like this.
- Most passives. Surges cause thermal stress in resistors and voltage stress in capacitors, but they're much more robust than semiconductors are. Needless to say, transformers and inductors are just balls of wire; a fuse is enough to protect them.
- Junction diodes. They're amazingly robust; together with SCRs, they are the only semiconductor types that can actually be protected by the right kind of fuse (you need a special ultrafast "semiconductor" type).
So, between these lists, you can see, if you have a damaged power supply, it's quite likely there are many components, buried deep within its bulk, that need to be tested and replaced. Many of them can't be tested in circuit (e.g., the diodes are wired across a 0-ohm transformer winding, good luck ohming that out!). So, you're better off just buying a replacement.
Things that are likely causes:
- Capacitor failure (electrolytics drying out; ESR rises; heat rises; eventual release of guts)
- Thermal overheating, subsequent diode or transistor breakdown
- Surges applied to the input or output (wiring faults in a system?, lightning induced mains surges)
- Poor design, operation, environment, etc. (overheating, clogged with dust, marginal design limits, poor current-limiting design, etc.)
As for more mundane things like open circuits, like cold solder joints, those are possible too, but don't usually result in catastrophic faults so much as errant behavior (output runs in pulses; voltage too high or too low; doesn't start up; etc.). You can trace the circuit (once you're familiar with the basic building blocks of a power supply, that is) to find where the miscommunication is taking place.
Tim