High-Frequency Signal Generator Using Cascaded Frequency Multiplier

Abstract

This paper presents a method for increasing the operating frequency range of a sub-6 GHz transmitter to millimeter wave (mm-wave) using two cascaded frequency doublers. The usage of such a frequency up-conversion approach is constrained since applying the first frequency doubler’s generated harmonics to the second doubler for mm-wave signal creation results in severe distortions and bandwidth expansion of the transmitted signal, limiting the use of such frequency up-conversion method. We have proposed a combined gain and phase retarder and physically inspired cascaded feed-forward neural network (CFFNN) based digital predistortion to mitigate the bandwidth expansion and distortions. A comparative study of feed-forward neural networks (FFNN) and CFFNN has been carried out to establish CFFNN as a significant and less complex model for effective DPD of frequency quadrupler. For proof of concept, we have experimentally validated this method for Long-Term Evolution (LTE) 20 MHZ (64-QAM) signal at 24 GHz frequency band using commercially available frequency multipliers. The measurement results show the bandwidth expansion is reduced approximately 11–12 times, and the adjacent channel power ratio (ACPR) of around −45.1 dBc and error vector magnitude (EVM) of approximately 3.07% is achieved.

Keywords

• Bandwidth reduction
• cascade feed-forward neural networks (CFFNN)
• digital predistortion
• error vector magnitude (EVM)
• frequency quadrupler
• millimeter-wave