Optimizing a pipelined MIMO sphere detector for energy efficiency

While transistor density continues to increase according to Moore's law, Dennard scaling ruling the downsizing of transistor power has broken down due to the increasing rise of leakage occurring in submicron technologies. With every process generation the speed at which transistors can be switched is increasingly exceeding the capacity to dissipate the generated heat, fact that demands highly energy-efficient systems. This is becoming especially problematic in the context of wireless communications, where battery-powered devices call for low energy consumption approaches. High performance must be simultaneously guaranteed, in order to satisfy the high data rates defined in modern communications standards, such as IEEE 802.11 WLAN or 3GPP LTE/-Advanced. In this work, the pipelined architecture of a MIMO sphere detector is analyzed and subsequently optimized by means of automated pipelining and retiming. As a result, a 40% reduction of the energy cost is achieved, without degrading the system performance nor incurring any additional area cost.