This paper investigates a distributed cyclic delay diversity (CDD) transmission scheme for cyclic-prefixed single carrier systems in non-identically and identically distributed frequency selective fading channels. The distinguishable feature of the proposed scheme lies in providing a transmit diversity gain while reducing the burden of estimating the channel state information, which is a challenging task in distributed and cooperative systems. To effectively use the distributed CDD scheme at the transmitters, two sufficient conditions are derived to eliminate the intersymbol interference at the receiver and leveraged to convert the multi-input single-output channel into a single-input single-output channel. These conditions allow the system to achieve the maximum diversity for frequency selective fading channels at a full rate. To achieve this maximum diversity, a fixed number of CDD transmitters are selected based on the channel conditions, symbol block size, and maximum time dispersion of the channel, and a new two-stage transmission mode is proposed. Based on the distributed CDD and the proposed selection schemes, a new expression for the signal-to-noise ratio at the receiver is obtained with the aid of order statistics, and then closed-form expressions for the outage probability and average symbol error rate (ASER) are derived. As far as the identically distributed frequency selective fading channel model is concerned, the achievable maximum diversity gain is proved, with the aid of asymptotic analysis, to be equal to the product of the total number of transmitters in the system and the number of multipath components. Link-level simulations are also conducted to validate the analytical expressions for outage probability, ASER, and maximum achievable diversity gain.