Adaptive RF Front-end Design via Self-discovery: Using Real-time Data to Optimize Adaptation Control
Adaptive RF Front-end Design via Self-discovery: Using Real-time Data to Optimize Adaptation Control
复制标题
通过自我发现的自适应射频前端设计：使用实时数据优化自适应控制
DOI：
--
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
A. Chatterjee
中科院分区：
文献类型：
--
作者：
D. Banerjee;A. Banerjee;A. Chatterjee
在先前的研究中已经确定，通过动态地权衡各个射频模块的性能来改变信道条件下的功耗，可以节省大量的功率。还表明，重新配置射频前端的控制律必须考虑到选择射频设备的过程角，以便以最佳方式（在规定的数据吞吐量下最小能量/位）权衡性能和功率。由于在所有可能的信道和设备工艺条件下模拟射频前端的复杂性，这种最优控制律的设计实际上是难以处理的。因此，现有的控制算法基于通道处理空间的粗采样，存在建模不准确的问题，并且从性能与功率的角度来看，本质上是次优的。相比之下，在本文中，我们提出了一种射频前端控制方法，该方法在实时操作期间进行了优化，不需要在所有信道过程条件下进行预先模拟，并且不容易受到模拟不准确性的影响。与当前的实践相反，这将产生更加稳健/最优的控制。提出了一种基于模拟退火（SA）的工艺优化框架，并使用内置传感器监测性能和功率。仿真结果和硬件数据验证了该方法的可行性。
It has been established in prior research that significant power can be saved by dynamically trading off the performance of individual RF modules for power consumption across changing channel conditions. It has also been shown that the control law that reconfigures the RF front end must take into account the process corners from which the RF devices are selected in order to trade off performance for power in an optimal manner (minimum energy/bit at prescribed data throughput). Design of such an optimal control law is virtually intractable due to the complexity of simulating the RF front end across all possible channel and device process conditions. Hence, existing control algorithms are based on a coarse sampling of the channel-process space, suffer from modeling inaccuracies and are inherently sub-optimal from a performance vs. power perspective. In contrast, in this paper, we propose a RF front end control methodology that is optimized during real-time operation, does not require upfront simulation across all channel process conditions and is not susceptible to simulation inaccuracies. This results in far more robust/optimal control as opposed to current practice. A simulated annealing (SA) based framework for process optimization is proposed along with the use of built-in sensors for monitoring of performance and power. Simulation results and hardware data are presented to show the feasibility of the proposed approach.