This comment carries a lot more weight than thought, although sometimes too much weight. There is a thinking amongst some that disturbances on the output of a switched mode power supply are more attributed to surges and spikes on the incoming Neutral rather than the Live. Let's be fair, the Live swings about by ±325V relative to Ground while the Neutral stays fairly close to 0V making it fairly logical the power supply will be more susceptible to fluctuations on the Neutral. This is not the case. If the output is to be affected by any form of surge, transient, or spike on the incoming mains, it will be regardless of whether it came in on the Live, Neutral, or both.
The mechanics of this are usually not clear at first. When viewing a switch mode power supply circuit diagram the incoming live and the positive high voltage of the smoothing capacitor are seen as being synonymous with one another. This mode of thinking is prompted further by diagrams having a tag showing the expected voltage e.g. +350V shown relative to the negative side of the capacitor tagged as 0V, which is the voltage associated with the Neutral especially when it's remembered that Neutral and Ground are one. Get the idea? The thinking that the 0V of the high voltage section of the PSU and the 0V expected on the Neutral are one and the same voltage, can lead to errors in the emphasis given to surges on the incoming Neutral and in filter design and spike suppression.
Attention needs to be turned to the rectifier and to both it's obvious and hidden characteristics. The obvious characteristic is to convert AC to DC by having the more positive of the Live and Neutral fed towards the positive side of the smoothing cap, and conversely the more negative towards the negative side of the cap. This function is taken in turns by each pair of diodes with each half cycle. Ok, sorry to bore you with the obvious, we all know how rectifiers work! But what about diode switches?
This more unknown characteristic of a diode is the fact that while it is conducting it is a very low impedance path. If you doubt this please investigate the use of diodes as 'solid state antenna switches' in radio systems. What is often unknown is that the diode is fed with only a few mA yet the RF current passed through it is in the amp range, this done with very little degradation to the RF signal.
This same effect exists within a normal rectifier set such that on one half of the cycle the Live must be seen to be directly connected (switched) to the positive with the Neutral to the negative, and the opposite during the other half of the cycle. To add a further complication there is a rather large cap (no less that 100µF) across the positive and negative and therefore, during the conduction phase, is effectively connecting Live and Neutral to each other. In this mode any spike on the Live will be found back on the Neutral and vice versa.
Furthermore, while continuing with this mode of thinking, spikes that affect the power supply output can therefore arrive either through the Neutral, or Live, or both. With the low impedance of the smoothing capacitor it becomes relatively unimportant whether the spike originates on the Live or Neutral (except for power quality engineers busy tracing potential faults for predictive maintenance purposes), as well the level of spike required to cause effect on the output is equal on both Live and Neutral.
Adding insult to injury is the RFI filters on the input to the power supply. As can be seen in the diagram there is a total of over 0.3µF between Live and Neutral. Smoothing capacitors are known to not handle high frequency components very well, but this is negated with the smaller sub-µF caps between the two incoming lines. It must be noted that these filters are never designed with a view to protecting the power supply from transients. Their one and only purpose is to block any very high frequency components generated by the switching reaching the mains wiring and being radiated.
One fortuitous fact is the filters are typically passive LC types and, although designed for the function laid out above, do attenuate to some extent any high frequency pulse coming in. What is unwise is to place too much emphasis on this attenuating characteristic believing the filter will eliminate all harm from impulses. Transients can quite comfortably make their way through these filters.
In brief: The effects of incoming spikes on the PSU output can be attributed to two main causes. The first is the coupling (mainly capacitive) between the primary and secondary windings on the transformer. There are often static shields between the windings but if not securely grounded within the power supply, spikes will make their way through to the secondary. The second reason is again coupling (and again, mainly capacitive) between the input side components and their surrounds making them vulnerable to 'sudden shifts' in voltage (as a result of a spike) which is reflected in the output as a disturbance. Again, these spikes, if having the required characteristics to affect the output, can appear on either the incoming Live or Neutral without distinguishing between the two.
When designing power supply transient protection give equal attention to both the Live and Neutral and, more importantly, the Ground. The latter is often forgotten as being the means by which transient currents are safely dispersed. Without a solid Ground no amount of transient protection or filtering will stop the effects of incoming spikes on the output of switch mode power supplies. Another forgotten aspect, drain the energy before it gets into the power supply. The high magnetic fields generated by grounding systems within the power supply during a spike may well find their way through to the sensitive electronics and thus influence the output.