Benjamin Penaflor

At the DIII-D tokamak, the Charge Exchange Recombination (CER) spectroscopy diagnostic routinely measures impurity temperature, rotation, and density. The ability to make these measurements in real time within DIII-D’s Plasma Control System (PCS) has recently been expanded and improved. High-speed network connections send raw data from 24 tangential CER views to the PCS, so spectral analysis can be performed. These views make measurements on the outboard midplane of DIII-D from the magnetic axis to the scrape off layer. Spectral analysis is completed quickly by a simplified least-squares fitting routine. Results are typically available within a millisecond of when the raw data become available and new raw data for all 24 views are typically sent to the PCS every 5 ms (for the most common CER sampling frequency). Fully ionized carbon is the most common impurity used for analysis, and analysis can also be performed for other highly charged impurities with relatively simple spectra such as boron, neon, and argon. The large majority of the real-time temperature and rotation measurements are within 20% of the most accurate automatic analysis available. As DIII-D discharges are typically separated by 10–15 min, the more rapid availability of the real-time analysis can be a major aid for control room analysis during an experiment. The real-time analysis can also be used in the PCS for control purposes (e.g., plasma profile feedback).

The DIII-D Plasma Control System (PCS) application has experienced significant growth, with new control algorithms, large-scale data processing, and additional real-time CPUs straining the current PTDATA archiving format. PTDATA's limitations—including a fixed header format, inability to store IEEE-754 floats, restricted signal name length, and lack of higher-dimension data storage necessitated a replacement to support future expansion. General Atomics (GA), in collaboration with Princeton Plasma Physics Laboratory (PPPL), and Massachusetts Institute of Technology Plasma Science and Fusion Center (PSFC) launched a multiyear project to migrate PCS data acquisition from PTDATA to MDSplus. Following stakeholder interviews with physicists, control experts, and software developers, the team selected MDSplus, a mature, well-supported database in the fusion community. The project aims to maintain data fidelity, add features, improve archiving performance, and ensure seamless integration with DIII-D's data acquisition and analysis software. In year two of implementation, DIII-D PCS archives data to both PTDATA and MDSplus servers, adding approximately 30 seconds, lengthening the plasma shot cycle by only ~3%. In 2025, DIII-D accumulated 849 GB in PTDATA format and 537 GB in MDSplus format. The archiving code was refactored using a layered architecture pattern, enabling future database replacements. New PCS gateway computers serving as satellite MDSplus servers were configured, and networking infrastructure is being upgraded to support larger data transfers. The 2026 project objectives include end-to-end PCS simulation using MDSplus, State Vector data archival and retrieval, completion of model tree population, early user engagement, and plasma simulation server demonstration. Beyond DIII-D, this work could benefit the broader fusion research community by sharing both the data format evaluation process and the lessons learned during the PCS overhaul. For example, efforts to expand universal access to a wide range of proprietary fusion data silos served as one of the driving forces behind the US DOE Fusion Data Platform project.

The MAST-U plasma control system platform had become increasingly difficult to maintain due to the presence of legacy devices lacking modern operating system support, resulting in a backlog of dependent software and execution framework upgrades. The original proposed approach to upgrading would have required multiple complex transitions to interim solutions that could not be easily reversed if unsuccessful, thus presenting a high risk to operations. To bypass these complexities, a complete “drop-in” platform replacement was developed, comprising a new control computer running the latest execution software framework on a new operating system with new hardware devices to interface with the same plant input and output connections. This “clean break” approach eliminated all the development bottlenecks caused by the chain of legacy dependencies and allowed us to skip over many years' worth of pending incremental upgrades. The strategy was fundamentally reliant on building a complete end-to-end replacement that could be installed and operated in parallel with and independently of the existing system, with pin-compatible pluggable input and output interfaces and retaining the control logic and user experience of the original system. By decoupling the installation from the commissioning phase, the team was able to perform exhaustive in-situ testing prior to the final switchover. This strategy transformed a high-risk system overhaul into a controlled, reversible event, resulting in a significantly shortened recommissioning window and a seamless transition to a modern, maintainable platform.