Noise in Mermin's n-particle Bell inequality

We calculate the signal-to-noise ratio in Mermin's n-particle Bell inequality [Phys. Rev. Lett. 65, 1838 (1990)] based on source noise, data error, and statistical uncertainty. We find that the signal for violation grows exponentially faster than the exponentially growing noise, as long as the noise per detector or per particle is less than about 14%. Alternately, in the absence of noise a loophole-free test could be made with detector efficiencies as low as 71%.