Filtering Low-Frequency Pickup in LLAMA Diagnostic

In this contribution, we present a machine-learning model to infer the current induced in the internal components of the Lyman-Alpha Measurement Apparatus (LLAMA) [1]. LLAMA measures the intensity of the Lyman-alpha spectral line (121.6nm) emitted by atomic neutral deuterium during plasma operations at DIII-D to calculate the fueling of the plasma. The currents in DIII-Ds poloidal field coils induce a current in the internal components of LLAMA, producing signals comparable to the real measurements. Previous efforts to mitigate the high-frequency pickup signal produced by the coil control systems have proven successful [2] but not effective in solving the low-frequency pickup. We trained the machine learning model with LLAMA measurements with the shutter closed during standard operation (which reduces the measured signal to only the pickup signal) and during magnetic calibration (when there is only current being supplied to one coil), and the current supplied to the F-coils during operation. This model will filter the pickup signal from the real one, increasing the confidence in LLAMA measurements in DIII-Ds operation regimes with low signal levels like in wide pedestal QH mode scenarios.