In modern communication systems, a key measure of quality is the Error-Vector-Magnitude (EVM) of a system component such as an Amplifier, Frequency Converter or Transmitter.  Linear frequency response, non-linear distortion and noise all contribute to the symbol errors that impact EVM.  For designers of transmitter systems, the amplifier stage of the system is often the limiting factor in the system EVM.  In this presentation we will discuss the most precise way the characterize amplifier non-linearity that is critical to EVM using a novel frequency domain technique. The mathematical accuracy of this method will be examined and compared with traditional time-domain demodulation methods.  We show methods to remove the confounding factors of linear frequency response and noise so only the non-linear effects appear and allow one to optimize amplifier linearity absent these other effects.  A direct comparison to traditional methods is presented, that will show where traditional demodulation fails to properly report errors due to distortion.

Joel Dunsmore has worked for Keysight Technologies (formerly Agilent and Hewlett-Packard) at the Santa Rosa Site, since graduating from Oregon State University with an MSEE (1983). He received his Ph.D. from Leeds University in 2004. He is a Keysight R&D Fellow focused on component test.  He was a principle contributor to the HP 8753 and PNA family of network analyzers, with recent work in non-linear test, including differential devices, and mixer measurements, as well as modulated and spectrum measurements.  He has received 31 patents related to this work, and authored the “Handbook of Microwave Component Measurements (John Wiley, 2012)”