Algorithm-Architecture Co-Design for Digital Front-Ends in Mobile Receivers

Abstract: The methodology behind this work has been to use the concept of algorithm-hardware co-design to achieve efficient solutions related to the digital front-end in mobile receivers. It has been shown that, by looking at algorithms and hardware architectures together, more efficient solutions can be found; i.e., efficient with respect to some design measure. In this thesis the main focus have been placed on two such parameters; first reduced complexity algorithms to lower energy consumptions at limited performance degradation, secondly to handle the increasing number of wireless standards that preferably should run on the same hardware platform. To be able to perform this task it is crucial to understand both sides of the table, i.e., both algorithms and concepts for wireless communication as well as the implications arising on the hardware architecture. It is easier to handle the high complexity by separating those disciplines in a way of layered abstraction. However, this representation is imperfect, since many interconnected "details" belonging to different layers are lost in the attempt of handling the complexity. This results in poor implementations and the design of mobile terminals is no exception. Wireless communication standards are often designed based on mathematical algorithms with theoretical boundaries, with few considerations to actual implementation constraints such as, energy consumption, silicon area, etc. This thesis does not try to remove the layer abstraction model, given its undeniable advantages, but rather uses those cross-layer "details" that went missing during the abstraction. This is done in three manners: In the first part, the cross-layer optimization is carried out from the algorithm perspective. Important circuit design parameters, such as quantization are taken into consideration when designing the algorithm for OFDM symbol timing, CFO, and SNR estimation with a single bit, namely, the Sign-Bit. Proof-of-concept circuits were fabricated and showed high potential for low-end receivers. In the second part, the cross-layer optimization is accomplished from the opposite side, i.e., the hardware-architectural side. A SDR architecture is known for its flexibility and scalability over many applications. In this work a filtering application is mapped into software instructions in the SDR architecture in order to make filtering-specific modules redundant, and thus, save silicon area. In the third and last part, the optimization is done from an intermediate point within the algorithm-architecture spectrum. Here, a heterogeneous architecture with a combination of highly efficient and highly flexible modules is used to accomplish initial synchronization in at least two concurrent OFDM standards. A demonstrator was build capable of performing synchronization in any two standards, including LTE, WiFi, and DVB-H.