Processors have been playing important roles in both communication infrastructure systems and terminals.In this paper,both application specific and general purpose processors for communications are discussed including...Processors have been playing important roles in both communication infrastructure systems and terminals.In this paper,both application specific and general purpose processors for communications are discussed including the roles,the history,the current situations,and the trends.One trend is that ASIPs(Application Specific Instruction-set Processors) are taking over ASICs(Application Specific Integrated Circuits) because of the increasing needs both on performance and compatibility of multi-modes.The trend opened opportunities for researchers crossing the boundary between communications and computer architecture.Another trend is the serverlization,i.e.,more infrastructure equipments are replaced by servers.The trend opened opportunities for researchers working towards high performance computing for communication,such as research on communication algorithm kernels and real time programming methods on servers.展开更多
Baseband design and implementation for micro/pico base stations （mBS） in 5G ultra-dense network （UDN） is studied. Low cost is an essential requirement for mBS baseband in UDN. Digital baseband cost of ASIC/ASIP （...Baseband design and implementation for micro/pico base stations （mBS） in 5G ultra-dense network （UDN） is studied. Low cost is an essential requirement for mBS baseband in UDN. Digital baseband cost of ASIC/ASIP （Application Specific Integrated Circuit / Instruction-set processor） is of the most uncertainty in roBS system. However. the actual costs and hardware feasibility of the baseband are yet unknown to network deployers and researchers. In this paper, we studied the baseband hardware system design and implementation for low-cost roBS. We analyzed popular baseband algorithms and architectures for both full-digital and hybrid beamforming （BF） for UDN. We then proposed feasible chip-level solutions for the baseband with up to 128-antenna BS system, and estimated their implementation cost. Results show that among lull-digital BF algorithms, zero-forcing is a choice of high performance and low cost; for hybrid BF, 4×32 architecture （32 RF chains） provides good reduction in baseband cost with acceptable performance loss, thus it can be a preferable solution under low cost consider- ation. The proposed system planning method can also be used for the design of other related systems.展开更多
A new power estimation method is proposed for base station(BS) in this paper.Based on this method,a software platform for power estimation is developed.The proposed method models power consumption on different abstrac...A new power estimation method is proposed for base station(BS) in this paper.Based on this method,a software platform for power estimation is developed.The proposed method models power consumption on different abstraction levels by splitting a typical base station into several basic components at different levels in the view of embedded system design.In particular,our focus is on baseband IC(Integrate Circuit) due to it's the dominant power consumer in small cells.Baseband power model is based on arithmetic computing costs of selected algorithms.All computing and storage costs are calibrated using algorithm complexity,hardware architecture,activity ratio,silicon technology,and overheads on all hierarchies.Micro architecture and IC technology are considered.The model enables power comparison of different types of base stations configured with different baseband algorithms,micro architectures,and ICs.The model also supports cellular operators in power estimation of different deployment strategies and transmission schemes.The model is verified by comparing power consumption with a real LTE base station.By exposing more configuration freedoms,the platform can be used for power estimation of current and future base stations.展开更多
There are already several power models to estimate the power consumption of base stations at system level. However, there is so far no model that can predict power consumption of the future base station designs based ...There are already several power models to estimate the power consumption of base stations at system level. However, there is so far no model that can predict power consumption of the future base station designs based on algorithms and hardware selections with insufficient physical information. We present such an energy model for typical base stations. This model can help designers in estimating, evaluating and optimizing energy/power consumption of candidate designs in early design stages. The proposed model is verified by an LTE extreme scenario. The estimated results show that digital front-end, channel equalization and channel decoding are three major power greedy modules(consuming 39.4%, 16.3%, 13.4%) in a digital baseband subsystem. The power estimation error of the proposed power amplifier(PA) power model is 3.5%(macro cell). The major contribution of this paper is that the proposed models can rapidly estimate energy/power consumption of 4G and the future base stations(such as 5G) in early design stages with well acceptable precision, even without sufficient implementation information.展开更多
Baseband ASIP designs for handsets are discussed based on the author's R&D backgrounds. Algorithms for 4G, 3G, and WLAN are analyzed and selected for implementation based on the trade off of cost and performan...Baseband ASIP designs for handsets are discussed based on the author's R&D backgrounds. Algorithms for 4G, 3G, and WLAN are analyzed and selected for implementation based on the trade off of cost and performance with power consumption in mind. A SDR ASIP baseband system architecture is proposed for 4G and 3G mobile handsets. Function partitions for heterogeneous symbol processors are introduced to get higher performance over cost. Three structures for DFE, FFE, and Matrix symbol ASIP are proposed. The concept of bit parallel processor is introduced. Challenges of baseband processors for UDN of 5G were briefly introduced. Conclusions on ASIP architecture and system design are given for different baseband processors on different products.展开更多
基金The National High-Tech Research and Development Program of China(863 Program)2014AA01A705
文摘Processors have been playing important roles in both communication infrastructure systems and terminals.In this paper,both application specific and general purpose processors for communications are discussed including the roles,the history,the current situations,and the trends.One trend is that ASIPs(Application Specific Instruction-set Processors) are taking over ASICs(Application Specific Integrated Circuits) because of the increasing needs both on performance and compatibility of multi-modes.The trend opened opportunities for researchers crossing the boundary between communications and computer architecture.Another trend is the serverlization,i.e.,more infrastructure equipments are replaced by servers.The trend opened opportunities for researchers working towards high performance computing for communication,such as research on communication algorithm kernels and real time programming methods on servers.
基金supporting from National High Technical Research and Development Program of China(863 program)2014AA01A705 is sincerely acknowledged by authors
文摘Baseband design and implementation for micro/pico base stations （mBS） in 5G ultra-dense network （UDN） is studied. Low cost is an essential requirement for mBS baseband in UDN. Digital baseband cost of ASIC/ASIP （Application Specific Integrated Circuit / Instruction-set processor） is of the most uncertainty in roBS system. However. the actual costs and hardware feasibility of the baseband are yet unknown to network deployers and researchers. In this paper, we studied the baseband hardware system design and implementation for low-cost roBS. We analyzed popular baseband algorithms and architectures for both full-digital and hybrid beamforming （BF） for UDN. We then proposed feasible chip-level solutions for the baseband with up to 128-antenna BS system, and estimated their implementation cost. Results show that among lull-digital BF algorithms, zero-forcing is a choice of high performance and low cost; for hybrid BF, 4×32 architecture （32 RF chains） provides good reduction in baseband cost with acceptable performance loss, thus it can be a preferable solution under low cost consider- ation. The proposed system planning method can also be used for the design of other related systems.
基金The finance supporting from National High Technical Research and Development Program of China(863program)2014AA01A705
文摘A new power estimation method is proposed for base station(BS) in this paper.Based on this method,a software platform for power estimation is developed.The proposed method models power consumption on different abstraction levels by splitting a typical base station into several basic components at different levels in the view of embedded system design.In particular,our focus is on baseband IC(Integrate Circuit) due to it's the dominant power consumer in small cells.Baseband power model is based on arithmetic computing costs of selected algorithms.All computing and storage costs are calibrated using algorithm complexity,hardware architecture,activity ratio,silicon technology,and overheads on all hierarchies.Micro architecture and IC technology are considered.The model enables power comparison of different types of base stations configured with different baseband algorithms,micro architectures,and ICs.The model also supports cellular operators in power estimation of different deployment strategies and transmission schemes.The model is verified by comparing power consumption with a real LTE base station.By exposing more configuration freedoms,the platform can be used for power estimation of current and future base stations.
基金supporting from National High Technical Research and Development Program of China (863 program) 2014AA01A705
文摘There are already several power models to estimate the power consumption of base stations at system level. However, there is so far no model that can predict power consumption of the future base station designs based on algorithms and hardware selections with insufficient physical information. We present such an energy model for typical base stations. This model can help designers in estimating, evaluating and optimizing energy/power consumption of candidate designs in early design stages. The proposed model is verified by an LTE extreme scenario. The estimated results show that digital front-end, channel equalization and channel decoding are three major power greedy modules(consuming 39.4%, 16.3%, 13.4%) in a digital baseband subsystem. The power estimation error of the proposed power amplifier(PA) power model is 3.5%(macro cell). The major contribution of this paper is that the proposed models can rapidly estimate energy/power consumption of 4G and the future base stations(such as 5G) in early design stages with well acceptable precision, even without sufficient implementation information.
基金supported by the National HighTech Research and Development Program of China (863 Program) 2014AA01A705.
文摘Baseband ASIP designs for handsets are discussed based on the author's R&D backgrounds. Algorithms for 4G, 3G, and WLAN are analyzed and selected for implementation based on the trade off of cost and performance with power consumption in mind. A SDR ASIP baseband system architecture is proposed for 4G and 3G mobile handsets. Function partitions for heterogeneous symbol processors are introduced to get higher performance over cost. Three structures for DFE, FFE, and Matrix symbol ASIP are proposed. The concept of bit parallel processor is introduced. Challenges of baseband processors for UDN of 5G were briefly introduced. Conclusions on ASIP architecture and system design are given for different baseband processors on different products.
