It was thought some years ago that using npn inputs i.e. +24v signal off & 0v signal on would reduce the chance of spikes usually positive on a grounded system.
This did appear to be the case on a system I once had the misfortune to find a problem.
The company had 3 identical machines, originally these were japanese origin with plc's using npn inputs.
One had been upgraded to a siemens 110 using pnp inputs, only the sensors had been changed to account for the new input type.
This machine would on a number of occassions do funny things, now the software was simple logic nothing fancy & was a direct reproduction of the original (they had even had symbols with the I/O numbers presumably to copy directly the code as they only had a dump of the code.
I could find no reason for the faults but going back to my TV days remembering that on the old 405 line system the video was inverted this reduced the white spot syndrome with noise in the transmission so I fitted some 6.8k resistors across all inputs to the ov, this cured the problem, now was this noise on the inputs that did not show up on npn or was it generated from the sensors however it occured it never happened again so who knows
By the way, before anybody replies with why not use a scope, I was on-site not expecting to have to use one as this was a plc control system with a "BUG".