One method you could use is look at the mitial function yalue {rst in your list}: then cast out points based on this salon 'iE-ie know that the signal is approximately sinusoidal and the frequency is at around 10 Hz. That means that there will be one zero crossing per 500 recorded points. Suppose that we saw that our initial recorded value was 1 . We could start by choosing as our brackets for zeros the l"1 501\": l00l\" etc. points in the recording, and iis might work, in fact it probably would work quite well in many cases for a well behaved function like a sinusoid. Let's go back to this original recorded value of 1. We don't know if this is small or large on the scale of the measurements. It might be the case that the 501\" yalue is +1: that the two points are themselves both near zeros and that the maximum Tralue are actually +f- 300. In this case the bisection method might work but it would be very ine fcient. Therefore, a test for maximum and minimmn y-alues, or a prior knowledge of the scale would help you know whether you were near a maximum f minimum or zero at your rst sampled point If you found your rst sampled point was Targe' then you might stay with it and cast out to the 501\