1) typically performance impact is negligible
2) we have no way of knowing whether or system fits into the definition of "typical"
3) the best way would be to do that on a test system and measure the impact
4) I would be concerned more about SYSAUX growth than performance impact -- you need to make sure that you won't run out of space because of x4 more frequent snapshots
You're going to be running it four times more often and using four times the space.
But would snapping every 15 minutes normally be a concern?
No but individual mileage will always vary depending on the profile of the system.
You must have a reason for wanting to snap more often - that reason would normally have to be more significant than any additional minor overhead.
What can you do to mitigate that concern?
On your test system, you could at least see what additional resources are consumed when running 15 minutes vs 1 hour, although the volume of data being snapped by AWR will be a factor.
But in the worst case scenario, you can change the interval right back if you do see an impact.