Topologies for the Digital Predistortion Techniques — Lesson 4

This lesson covers the fundamentals of digital predistortion techniques, focusing on the topology used in these techniques. It delves into the models used to represent the inverse or forward model for the power amplifier, such as the LUT model. The lesson also discusses the memory effect of power amplifiers and the different structures proposed by researchers to model this effect. It further explores the concept of memory polynomial models and neural network models, highlighting their application in fitting the output of a system. The lesson concludes by discussing the two factors in a model: topology selection and filtering algorithm selection.

Video Highlights

00:18 - Wiener structure and Hammerstein structure
02:55 - FIR filter and LUT
09:00 - Concept of normalized mean square error
11:38 - Adjacent channel error power ratio for measuring out of band performance
16:16 - Volterra model and the feed forward neural network model
20:40 - Memory polynomial model
23:40 - Introduction to the neural network models
24:53 - Topology selection and filtering algorithm selection

Key Takeaways

- Digital predistortion techniques use models to represent the inverse or forward model for the power amplifier.
- The memory effect of power amplifiers can be modeled using different structures, such as the Wiener and Hammerstein structures.
- Memory polynomial models and neural network models are used to fit the output of a system.
- The two factors in a model are topology selection, which involves choosing the nonlinearity order and memory depth, and filtering algorithm selection, which involves tuning the coefficients.