摘要： Soaring demand for high data rate services entails high throughput satellite (HTS) systems with multi-beam architecture, and full frequency and time resources reuse. However, interference among simultaneously served users is the fundamental factor that is needed to be addressed before enacting HTS system with this architecture. Beamforming has been proposed as a potential technique to mitigate the interference in the literature. Different types of beamforming techniques proposed including beamforming at payload (on-board), beamforming at a gateway and hybrid beamforming. On-board beamforming prevails over other techniques due to its advantages—channel information at payload is more recent than gateway and sharing overhead of channel and symbols across multiple gateways is reduced in a multi-gateway architecture to name a few. Despite these advantages, beamforming at the gateway is usually preferred due to the heavy processing cost incurred in beamforming. Beamforming processing cost can be split into two factors: design cost and implementation cost. While design cost accounts for the cost involved in the design of beamformer, implementation cost accounts for multiplications and additions involved in applying calculated beamformer coefficients to data symbols. Through our study, we noticed that the major contributing factor to processing cost is the implementation cost which accumulates for every data symbol rather than design cost which is incurred only once per channel coherence time which usually relatively longer than many data symbols. Furthermore, the implementation cost is dominated by the multiplications involved. Hence, in this work, we address the issue of implementation cost from the perspective of on-board multiplications. We formulate the problem of minimizing on-board implementation cost (multiplications) of a beamformer as a second-order cone programming problem with the help of (cid:96)1 norm constraint on the beamforming matrix subjected to a minimum signal-to-interference-noise ratio of simultaneously served users and classical total power constraint. We show the efficacy of our algorithm over the traditional power minimization method through Monte-Carlo simulations.