OpenDPD

OpenDPD is an end-to-end learning framework built in PyTorch for modeling power amplifiers (PA) and digital pre-distortion. Developed by the Lab of Efficient Machine Intelligence @ Delft University of Technology, OpenDPD now ships as both a pip-installable package and a full research codebase.

• pip install opendpd exposes high-level Python APIs (train_pa, train_dpd, create_dataset, etc.) and an opendpd-cli entry point for quick experiments.
• The repository provides automation scripts, dataset tooling, quantization utilities, and reproducible baselines for RF researchers.
• Documentation includes rich examples (Python scripts + Jupyter notebooks) and detailed setup instructions for advanced users.

We invite you to contribute new backbone networks, pre-trained models, or measured PA datasets. This repository contains the complete training pipeline for OpenDPD using baseband signals from digital transmitters.

Where to start

• Install

  ---

  pip install opendpd, or set up the research codebase from source.

  Installation

• Learn the workflow

  ---

  PA modeling, DPD learning, quantization-aware training and validation, step by step.

  End-to-end training

• Bring your data

  ---

  The built-in measured datasets, their format, and how to add a PA of your own.

  Datasets

• Compare models

  ---

  The PA modeling and DPD benchmark with its methodology, evidence and reproduction script.

  Benchmark

• Script it

  ---

  train_pa, train_dpd, run_dpd, plot_dpd and the dataset helpers from Python.

  API reference

• Cite the work

  ---

  The papers behind OpenDPD, the measurement setup and the people who built it.

  About and citation

Quick start

Start with the Google Colab tutorial by clicking the badge below for a zero-install guided walkthrough of OpenDPD’s workflow.

Demo video

Watch our Demo Video to see a real OpenDPDv2-trained neural DPD model in action achieving close to -60 dBc ACPR on a 200 MHz 256-QAM 5G NR signal with a -41.5 dBm average output power GaN PA.

Versions

OpenDPDv2 embeds a new temporal residual (TRes)-DeltaGRU NN DPD algorithm and a new TM3.1a 5-channel x 40 MHz (200 MHz) test signal dataset, measured from a 3.5 GHz Ampleon GaN PA at 41.5 dBm average output power, named APA_200MHz, enabling fast prototyping of accurate and power-efficient NN-based DPD algorithms by streamlining learning and optimization through DPD quantization and temporal sparsity exploitation.

Version	Related Papers	Dataset	Supported Backbones	Performance on APA_200MHz
OpenDPDv1	Algorithms Hardware	Collected from a 40nm CMOS DTX @ 2.4 GHz DPA_100MHz, DPA_160MHz, DPA_200MHz	GRU, LSTM, GMP, RVTDCNN, VDLSTM, DGRU, TCN	DGRU with 1041 params: ACPR of -58.4 dBc, EVM of -39.1 dB
OpenDPDv2		Collected from a GaN Doherty @ 3.5 GHz APA_200MHz, APA_200MHz_b	PGJANET, DVRJANET, TRes-DeltaGRU	TRes-DeltaGRU with 996 params: ACPR of -59.4 dBc, EVM of -42.1 dB
Experiment code		Controlling, I/Q data upload/download MATLAB code	Rohde & Schwarz SMW200A, Keysight N9042B,	Matlab/.m