Abstract: In order to offer a new theoretical foundation for constructing huge capacity communication system without increasing system bandwidth, it has been verified that under additive white Gaussian noise environment, if the transfer function H(f) of the channel has a limited bandwidth B (single side), the duration T of the impulse response function h(t) goes to infinite, the space between independent samples are 1/2B and 0 respectively in time and frequency domain. It is not sufficient to only employ independent signaling samples separated by 1/2B Sec in time domain like Shannon, it should be also to employ independent signaling samples separated by B/K Hz in frequency domain. In this way, the channel capacity is proportional to K. Since the space between independent samples is 0 in frequency domain, K has no limited. The Shannon capacity is only a special case of ignoring the independent signaling samples in frequency domain (K=1). Similarly if the impose response function h(t) of the channel has a limited time duration T, the bandwidth B of the transfer function H(f) goes to infinite, the space between independent samples are 0 and 1/T respectively in time and frequency domain. It is not sufficient to only employ independent signaling samples separated by 1/T Hz in frequency domain like orthogonal frequency division multiplexing, it should be also to employ independent signaling samples separated by T/K Sec in time domain. Similarly, the channel capacity is proportional to K. Since the space between independent samples is 0 in time domain, K has no limited. And the essential of overlapped multiplexing principle is to construct independent parallel Gaussian channels.

Key words: overlapped multiplexing principle, independent parallel Gaussian channels, overlapped time domain multiplexing, overlapped frequency domain multiplexing, overlapped X domain multiplexing, channel capacity