Or, think of it this way:
In a spur gear pair, there's a click or shock as the load transfers from one tooth pair to the next. The shock may or may not be smaller as the lash decreases or the number of teeth increase.
Imagine that you've split a pair of spur gears on their major planes, indexed them half a tooth, and pinned them back together. You've doubled the number of shocks per revolution, and maybe reduced their size.
Now keep doing it, splitting the gears again and again, indexing them slightly and rejoining, say in a spiral pattern. At some point you've got enough gear pairs that the teeth overlap, i.e. gears at one end of the stack are engaging just as gears at the other end are disengaging, so there are (nearly) always two gear pairs carrying the load. Now make the number of gear pairs infinite and reduce the width of each pair to almost zero. You've got an infinity of slightly staggered gear pairs, at least two of which are always carrying the load: helical gears.
The tooth action at any given cross section is no different from a spur gear, because the profiles are the same, but the teamwork of an infinity of infinitesimal gear pairs keeps the load always 'supported', greatly reducing the noise generated.
Of course it doesn't work if the width and helix angle are such that the gears are 'short', i.e. you can see something other than a perfect cylinder in axial projection, or if the gears are excessively compliant in torsion, but you get the idea.
Mike Halloran
Pembroke Pines, FL, USA