Mod12003 mod12003 is an open-source printed circuit board by PCBench, from the PCBench/PCBench repository on GitHub, published under the MIT licence. It is a 2-layer board of 90 × 90 mm with 452 parts on 100 nets, routed with 2,210 tracks and 85 vias.
It is built around the ATXMEGA32E5-M4U, LM358DGK and AOZ1282CI, with the MAX5215GUA+, MCP9701-E/TO and OPA2376AIDGK among its other parts. Its parts include 59 capacitors, 53 resistors, 12 ICs, 12 transistors and 8 connectors. It belongs with the AVR (ATmega, ATtiny) boards of this library.
PCBench: A Dataset for Printed Circuit Board Routing
PCBench is a dataset for PCB routing task, it includes a dataset consisting of 164 printed circuit boards (PCB), a data augmentation script to expand the dataset for supervised learning (SL) tasks, and a reinforcement learning (RL) environment.
Dataset
Folder Structure
All the PCB designs in our dataset are stored in the folder PCBs, where each PCB design is stored in a subfolder named {author/org}_{PCB_name}_{design_version(optional)}. Below is the folder structure of the dataset.
The raw.kicad_pcb is the raw PCB design file, process.kicad_pcb is the file after cleaning up, final.json stores only routing-related information used for ML tasks. metadata.json stores the meta information for each single PCB design. visual.png is a visualization of each single PCB design. master_meta.json contains metadata of all PCBs and global-level information.
PCB Routing Description Language (PCB-RDL)
We propose a JSON specification (final.json for each PCB design) called the PCB Routing Description Language (PCB-RDL) that expresses a PCB routing problem and its solutions intuitively using basic concepts in order to facilitate research into automated PCB routing using ML. Following figure shows is an example of PCB with marked components in PCB-RDL (Left) and it corrsponding PCB-RDL (Right).
Data Augmentation
Run
The augmentation data can be generated with the script Scripts/Data_augmentation/augmentation.py. To generate 3 samples by randomly extracting 50% nets from the PCB 1Bitsy_1bitsy, please run the following command
text
cd Scripts/Data_augmentation/
python augmentation.py --num_samples 3 --net_ratio 0.5 --pcb_name 1Bitsy_1bitsy
Output
All the generated PCBs will be stored in the folder named augmented_data under Scripts/Data_augmentation/. In augmented_data, all the generated samples will be stored in a subfolder with the name of selected PCB. Each generated sample has the name sample_index.json. For example, the above command will generate the following file under Scripts/Data_augmentation/
It measures 90 × 90 mm and has 2 copper layers, 1.6 mm thick.
What parts does the Mod12003 mod12003 use?
452 parts, built around the ATXMEGA32E5-M4U, LM358DGK, AOZ1282CI and MAX5215GUA+. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.
Where can I download the Mod12003 mod12003 KiCad files?
From the PCBench/PCBench repository on GitHub, at the commit this page was made from: the board is PCBs/mod12003_mod12003/processed.kicad_pcb. PCBKit opens the board in the browser, with nothing to install.
Can I use the Mod12003 mod12003 design in my own project?
Its repository publishes it under the MIT licence: you may reuse it on that licence's terms (keep the author's notice, and for a copyleft licence share your changes alike).