You can find detailed description here (the second article has an example):

http://office.microsoft.com/en-us/assistance/HP010322691033.aspx

http://www.beyondtechnology.com/tips016.shtml

Quote:

**STDEV** is used when the group of numbers being evaluated are only a partial sampling of the whole population. The denominator for dividing the sum of squared deviations is **N-1**, where **N** is the number of observations ( *a** count of items in the data set* ). Technically, subtracting the **1** is referred to as "*non-biased*."

**STDEVP** is used when the group of numbers being evaluated is complete - it's the entire population of values. In this case, the **1** is NOT subtracted and the denominator for dividing the sum of squared deviations is simply **N** itself, the number of observations ( *a** count of items in the data set* ). Technically, this is referred to as "*biased*." Remembering that the **P** in **STDEVP** stands for "*population*" may be helpful. Since the data set is not a mere sample, but constituted of ALL the actual values, this standard deviation function can return a more precise result.