In this study, real time rectangular pulse detection problem is addressed when the channel noise distribution is not known in advance. Since there is no information about the noise parameters, sub-optimal detectors such as soft limiter, sign correlator and the proposed signed power are utilized for real time pulse detection problem using only received noisy samples. Noting that the unknown channel noise is not necessarily to be Gaussian, symmetric α-stable (SαS) distribution is given as non-Gaussian noise model. Since one of the main objectives is to detect the existent pulse within minimum observation interval, detector performances characterized by detection and false alarm probabilities are analyzed with respect to pulse length under both Gaussian and SαS noise. It is shown that not only the given sub-optimal detectors can exhibit performance close to optimal linear detector under Gaussian noise, but also they provide superior performance under SαS distribution. When the channel has strong impulsiveness, it is observed that the sign correlator and signed power detector introduced in this study exhibit better detection performance compared with soft limiter detector. Consequently, these detectors can be practically implemented to determine existence of pulse within a certain observation interval when there is no prior information about channel noise which is most likely non-Gaussian. Among the other sub-optimal detectors, the proposed signed-power detector is observed to exhibit more stable detection performance under channel noise having varying impulsiveness.