+1 to bill's post. then perform a spectral analysis of the content in order to see what the peak to average difference is. we tend to set the "volume" at the average value, but in order not to compress anything, you need to allow headroom for the peaks. how much compression we can hear is debatable, but if you don't want any with a well recorded track, you might need 15-20db of headroom. most folks don't listen at much more than 100db at the listening position average, which is why the "rule of thumb" for an accurate system starts around 115-120db at the listening position.