The noise would be more like X_1 + 0.7 * X_2 + 0.7^2 * X_3, where X_1, X_2, X_3 are uniformly distributed on [-1,1], and with X_2 having a frequency double of that of X_1, X_3 one double of that of X_2 (that is: there are smaller-scale variations). Thus, a threshold of 0.55 is quite less that 50% c...