One of the more interesting and frustrating issues is that of power supply sequencing. Power supplies generally are analyzed for startup issues, but less often for other sequences of events. This can lead to interesting failures.
Even though supplies are analyzed for turn on conditions, even this is sometimes not done correctly. For turn on issues, the most common is that of inrush current. Much work is done under steady state conditions, but the startup can lead to interesting issues. The inrush current from an AC line can destroy the input diodes. This happens when the AC line is at a high voltage, and the input capacitance of the supply is not charged. This leads to a large amount of current to flow in order to rapidly charge the input capacitance. This “snap on” can damage the input diodes as a result.
Inrush currents can also affect transformers — the remanence flux of a transformer is not known at startup. If the AC line is the correct polarity, it will saturate the transformer on the first few cycles until the hysteresis loop reaches its steady state.
For other cases, the inrush current leads to excessive energy being stored in line inductance. This leads to high voltage spikes on the inputs to the end equipment. This isn’t desirable in most cases, and can damage equipment.
Motors can also lead to a heavy load at startup. Motors require a large amount of power to move from stopped to steady state. This large current can adversely affect control loops, resulting in a lockout condition. In this case, the current is excessive, resulting in a reduction in a severe reduction in voltage. This prevents the motor from starting, but the stopped motor can still draw a lot of current.
This condition might even be stable, meaning the supply will never reach rated voltage and the motor will never start. This is annoying because, had the motor begun spinning, the supply would have been able to achieve rated voltage and keep the motor spinning!
Perhaps the most interesting and overlooked case is that of turn-off-turn-on events. This type of use case is rare in the field, but is common in factory and verification tests. In this case, the power supply is disconnected from its input. This causes the supplies to sag, depleting the capacitors/inductors.
Analog circuits begin to unbias. This might take 5-10 seconds. Impatient technicians might reconnect the power at just this time. At this time, some circuits are biased, some are not. This can lead to very complex and potentially destructive behavior. It’s interesting because, depending on the product, the supply might never be toggled in this manner in the field. But in a lab or production test, it probably will at least once.
This problem is very nuanced. There are usually dozens of possible sequences. If the loading can change on each supply, then each might decay at a different rate. The amount of time can be crucial too. Further, most supplies are designed to “ride though” a long disconnection in the input for as long as possible — to improve reliability.
This isn’t an entire list of everything that can go wrong. The intent of this article is to suggest a bit more thought about getting from on to steady-state, as well as from steady-state to off. Such failures can be difficult to analyze if you don’t know what you’re looking for.