Sometimes in the recording and digitizing process a wave will have it's zero crossing point offset - meaning there will be more of the wave to one side of the zero crossing point than the other. This will limit your ability attain maximum volume from the wave, since one side of the wave would reach 0db sooner than the other.
The "eliminate dc offset function" simply resets the zero crossing point by calculating the amount of offset present and then adding (or subtracting) that amount to the entire waveform.
Most of the files that I have worked with have very small offsets, but I tend to use this function as a matter of habit for all my work. I figure it can't hurt, and there may come an occasion when it will help alot.
The question teacher has raised, though, is whether additional dsp can introduce more offset - or is the offset pretty much established at the time the signal is digitized by the A/D converters?