Neural-Network DPD via Backpropagation through a Neural-Network Model of the PA

Design and Implementation of a Neural Network Based Predistorter for Enhanced Mobile Broadband

New Version of my DPD Matlab Library Available

I released a new verison of my DPD library for Matlab yesterday. You can find the source code here. This software can enable you to easily add DPD to any of your Matlab based wireless prototyping. This new version now supports the Generalized Memory Polynomial (GMP). This is a more expressive model than the standard memory polynomial. The extra memory effects captured in this model are especially helpful for signals with larger bandwidth such as what we see in 5G NR.

Two Papers Accepted for Asilomar 2019

The two papers I was involved with and submitted to Asilomar this year were accepted. The first paper is “Neural Network DPD via Backpropagation through a Neural Network Model of the PA.” This work is the foundation of the idea which I expanded on for my SIPS submission. In this paper, we propose performing digital predistortion by modeling the PA as a neural network (NN). We can then backpropagate through the PA NN model to train a NN for performing the DPD.

New Preprint Available. Neural Networks are taking over DPD!

I have a new preprint available for my submission to the 2019 IEEE International Workshop on Signal Processing Systems in Nanjing, China. The paper is titled “Design and Implementation of a Neural Network Based Predistorter for Enhanced Mobile Broadband” and is available here. In this paper, I use a neural network (NN) to implement digital predistortion (DPD) to correct for power amplifier (PA) nonlinearities. The main contributions are: A novel training method where we learn the NN DPD by first modeling the PA with a NN and backpropagating through the PA NN model to update the DPD NN weights.

2019 ECE Affiliates Day

The ECE Corporate Affiliates Day was recently on March 29th. I was involved in two research posters. 1st was the neural network based DPD that I am working on with our VIP team. The undergraduates won the “Best Undergraduate Research Award.” Their poster is available here. This is unpublished work so far that is promising. I have worked on DPD for my entire time at Rice, and it is still an interesting problem to me.

DPD shoutout at FOSDEM 2019

One of my papers from 2017 was given a shoutout by Derek Kozel in February 2019 at the FOSDEM conference. The talk is shown is an overview of DPD and is on YouTube here. In the talk, the speaker refers to my work as “a really interesting find.” The result shown is from my paper titled “Multi Component Carrier, Sub-band DPD and GNURadio Implementation” which was published at ISCAS. In this paper, we feature a sub-band approach to DPD which only targets spurious emissions that are in violation of some emission mask to save on the complexity.

Digital predistortion with low-precision ADCs

Digital Predistortion (DPD) is a popular technique for linearizing a power amplifier (PA) to help reduce the spurious emissions and spectral regrowth. DPD requires the learning of the inverse PA nonlinearities by training on the output of the PA. In …

Multi Component Carrier, Sub-band DPD and GNURadio Implementation

Digital predistortion (DPD) is an effective way of mitigating spurious emission violations without the need of a significant backoff in the transmitter, thus providing better power efficiency and network coverage. In this paper, the IM3 subband DPD, …

“Parallel Digital Predistortion Design on Mobile GPU and Embedded Multicore CPU for Mobile Transmitters”

Digital predistortion (DPD) is a widely adopted baseband processing technique in current radio transmitters. While DPD can effectively suppress unwanted spurious spectrum emissions stemming from imperfections of analog RF and baseband electronics, it …