Description
Laser-based Thomson scattering is a standard diagnostic for the electron density and temperature in basic plasma physics and fusion research. While it is one of the most convenient diagnostics to measure these quantities spatially resolved and reliably for a wide temperature and density range, only a small fraction of the incident laser light is scattered and, consequently, signal levels are typically low. A common way to mitigate these low signal levels is the use of so-called polychromators, which integrate the spectrum of the scattered light over a number of different wavelength bins. The set of wavelength bins observing measureable signals is determined by the width of the spectrum (temperature) and its intensity (density). In combination with noise, this can lead to large measurement errors in temperature ranges where the spectrum broadens enough to just extend into a new wavelength bin. Large noise levels are, for example, found in large fusion experiments, where the high density leads to high plasma radiation levels, which together with the large plasma volume causes a substantial background signal in the polychromators. In this contribution, we document noise-induced features in the temperature dependence of the Thomson scattering measurement errors. Furthermore, we show that the impact of noise can be mitigated. Measurements of low temperatures are particularly affected by noise (since only few wavelength bins detect a signal). Making use of the symmetry of the TS spectrum at low temperatures, we show that the error in temperature and density can be reduced by introducing an additional filter on the “red side” (longer wavelengths than the incident light) of the spectrum. The additional filter does not mirror a filter from the blue side for redundancy, but adds additional sensitivity in a temperature range, where one filter starts to loose sensitivity, while the next filter still detects only a small signal. The additional red-side filter does not reduce the overall error (for all temperatures) substantially, but mainly avoids an uncertainty spike in that critical temperature range.