Research

Nonlinear Digital Compensation

We often assume signal processing systems to be linear: if we put a band-limited signal in, the output will be confined to the same frequency band; if we put the sum of two signals in, we will get the sum of the corresponding output signals as the new output. Real systems are almost never linear, yet we make this assumption because (a) we have well-established tools to deal with linear systems and (b) nonlinear systems appear linear over small enough domains. It can be beneficial to extend the linear range of a system. For example, power amplifiers used in wireless transmitters are typically most efficient in their nonlinear regime, but the signal processing that prepares information for tranmission assumes a linear system. I have worked to develop digital compensation of nonlinear systems to make them appear as linear devices. My work has spanned pre-distortion and post-distortion as well as theoretical and practical.

Physical Layer Fingerprinting

Suppose my iPhone and your iPhone transmit the exact same signal (same modulation scheme, same meta-data, etc.) over the air to a base station. Can the base station tell our two devices apart? While the two phones are supposed to be sending the same information over the air, manufacturing differences can result in slight variations between the two signals. This line of work aims to develop practical methods for fingerprinting devices and explore the limits of physical layer device discrimination.