Description
Charge Exchange Recombination Spectroscopy (CXRS) provides 3D-localized measurements of ion temperature, rotation velocity, and impurity density profiles [1]. At the TCV tokamak, five CXRS systems monitor both the core (low and high field sides) and edge plasma regions in toroidal and poloidal directions. It uses charge-exchange interactions between, often highly charged, plasma ions and injected neutral atoms from Diagnostic and Heating Beams.
This work details upgrades of the TCV CXRS diagnostic to obtain real-time (RT) capabilities, a requirement for long-pulse devices like ITER. The RT analysis will include metadata of the production state and quality, compliant with ITER data handling requirements. These tags allow the plasma control system to track the data reliability, and provide alternative signals during low signal-to-noise ratio periods.
To ensure temporal determinism, the hardware architecture features a high-speed acquisition chain with a Camera-Link interface and dedicated PCIe frame grabber together with consistent time-stamping. This employs Direct Memory Access, reducing operating system bottlenecks providing raw spectral data with minimal latency.
RT operation necessitates efficient spectral software analysis. Currently developed within the MATLAB-Simulink environment, a Gaussian regression routine can process the signal from 20 optical spectra in less than 10 ms. To minimize execution time, the algorithm was optimized by benchmarking different control structures and comparing the algorithm’s performance within for and while loops. This work also prompted an investigation of methods for passive signal subtraction that remain compatible with the requirements of the RT analysis.
Beyond direct feedback on the plasma discharge, this upgraded CXRS can be integrated into state-observer frameworks, such as RAPTOR, to improve high-fidelity predictive transport simulations [2].
We present a performance evaluation comparing the currently-in-use Levenberg-Marquardt method against a Gauss-Newton algorithm, and include a timing analysis and benchmarking against offline analysis. This confirmed that this approach already provides sufficient robustness for direct feedback on the Neutral-Beam injection power, paving the way for the feedback control of the plasma torque.
[1] F. Bagnato et al 2023 Nucl. Fusion
[2] S. Van Mulders et al 2026 Nucl. Fusion