Sure, it sounds clever but taking the second derivative of noisy data is poor statistics as far I'm concerned.