We present a simple classical (random) signal model reproducing Born's rule. The crucial point of our approach is that the presence of detector's threshold and calibration procedure have to be treated not as simply experimental technicalities, but as the basic counterparts of the theoretical model. We call this approach threshold signal detection model (TSD). The experiment on coincidence detection which was done by Grangier in 1986 [22] played a crucial role in rejection of (semi-)classical field models in favour of quantum mechanics (QM): impossibility to resolve the wave-particle duality in favour of a purely wavemodel. QM predicts that the relative probability of coincidence detection, the coefficient g((2)) (0); is zero (for one photon states), but in (semi-) classicalmodels g((2)) (0) >= 1 : In TSD the coefficient g((2)) (0) decreases as 1/epsilon(2)(d); where epsilon(d) > 0 is the detection threshold. Hence, by increasing this threshold an experimenter can make the coefficient g((2)) (0) essentially less than 1. The TSD-prediction can be tested experimentally in new Grangier type experiments presenting a detailed monitoring of dependence of the coefficient g((2)) (0) on the detection threshold. Structurally our model has some similarity with the prequantum model of Grossing et al. Subquantum stochasticity is composed of the two counterparts: a stationary process in the space of internal degrees of freedom and the random walk type motion describing the temporal dynamics.