I am trying to figure out how many samples to take from known population to be accurate within 95%.

For example to find average height and standard deviation of college student in Pennsylvania and let's say we know that there are 500,000 students.