Channel adaptive process resilient ultra low-power transmitter design with simulated-annealing based self-discovery
MetadataShow full item record
Modern day wireless communication systems are constantly facing increasing bandwidth demands due to a growing consumer base. To cope up with it, they are required to have a better power vs performance from the RF devices. The amount of data being exchanged over wireless links has tremendously increased and simultaneously, there is a need to switch to portable RF devices and this has in turn forced the issue of low-power RF system design. Therefore, what we need is an RF transceiver that operates at high data rates and over adverse channels with a low power consumption. A major portion of the power is utilized by the RF front end of the wireless system. Many methods like controlled positive feedback, re-utilizing bias current, etc have been employed to reduce the power consumption of the RF front end. The most modern wireless systems adapt to the channel quality by adjusting the data transmission rates and by adjusting the output power of the RF Power Amplifier. However, each of these methods concentrates on working for the worst case channel and giving the highest data rate. What needs to be known is that the channel conditions are not always worst. Even for a normal channel, the system is going to utilize a lot of power and give the highest possible data rate which may or may not be necessary. And thus, for the most part, the system is going to use up more power than necessary. What we need instead, is a system which works nominally for a normal channel and exhaustively for a harsh channel condition. This requires the system to adapt to the channel conditions. Also another major factor causing fluctuations in the performance is the process variations. This calls for a channel-dependent dynamic transceiver with adequate power management and tuning. In our work, we try to devise a method to dynamically minimize the power considering the varying channel conditions and process variations. We first use companding to reduce the dynamic range of the signal so that it can be used on facilities with smaller dynamic range. This brings down the transmitted power. We also create multiple instances of the Power Amplifier to simulate process variations. After finding the optimum tuning knob settings for one instance of the PA, we try to use it to obtain the optimum settings for another instance. This requires the use of some heuristics and in our work, we have supplemented it with Simulated Annealing. Using SA, we can dynamically tune the power of a system for changing channel conditions and existing process variations. Towards the end, we have also proved that the slower the cooling rate of the experiment, the more elaborate the search space is and the more accurate the result is.