Phase-estimate with Cross-correlation surprisingly dependant on the given true-phase

Dear All,

Thanks for a great site. I've build a wee program to estimate the phase difference between two sinusoidal signals simply using the built-in matlab cross-correlation function cross-corr as this http://dl.dropbox.com/u/12840152/CrossC_v4.m the programme uses a help function to draw a vertical line as http://dl.dropbox.com/u/12840152/vline.m

In the programme I simply give a true phase and then use cross-corr to estimate this phase. Surprisingly, the phase estimate is best at true-phase=0degrees. The estimate then gets worse as I take the true phase from 0 towards 90degrees. and then the estimate periodically gets better again when true phase is taken from 90 towards 180degrees. This puzzles me.

Can anyone tell me why this is so? That would be greatly appreciated

Sincerely, Kenneth