Comparison of edge turbulence velocity analysis techniques using Gas Puff Imaging data on Alcator C-Mod

Abstract

In the past, two methods for analyzing data from the Gas Puff Imaging diagnostic on Alcator C-Mod have been used. One uses temporal and spatial Fourier analysis to obtain wavenumber-frequency spectra, from which a phase velocity is computed [1, 2]. The other is based on time-delay cross-correlation of successive images used to track the motion of discrete emission structures [3, 4]. Several Gas-Puff-Imaging experiments were conducted to obtain data taken using the GPI Phantom Camera. The analysis of and results from these data are discussed in [3]. The results showed that the tracking time-delay-estimation technique found poloidal velocity magnitudes in the 0.1-1.4 km/sec range. However, independent examination of these data using the Fourier analysis yielded magnitudes up to a factor of 10 larger for the same data, and sometimes even disagreed with the direction of motion found. To understand the reasons for these discrepancies, we designed and generated synthetic data that mimics the real data. The user inputs the velocities, sizes, intensities, and distributions of the synthetic emission structures. We have used the synthetic data to test each code rigorously for strengths, weaknesses, and weighting. We have found that the Fourier analysis perfectly returns the correct poloidal velocity when there is no radial velocity component present. We have found that the tracking TDE analysis weights low frequency, low wavenumber features most heavily since they are typically the most intense, but systematically returns a smaller velocity than expected due to issues associated with averaging. After adjusting for these issues, the tracking TDE code now returns the correct value of the poloidal and radial velocities to within 10% for synthetic data as long as there is only one velocity present in the synthetic simulation. We applied these corrections to the analysis of the real data, and found that the measurements changed little in most cases. We then examined, in detail, the Fourier-analysis-derived "conditional" spectra for each shot, and determined that the likely causes for the discrepancies are due either to multiple velocities with emission structures moving in opposite directions in the same field of view or to non-zero "dispersion" in which lower-frequency/lower-wavenumber features are moving with one phase velocity and higher-frequency/higher-wavenumber features are moving with a different phase velocity. In a couple of cases, there may be a radial component in the actual images that may affect the poloidal velocity measurement for the Fourier analysis. Accounting for these explanations, we believe that we have resolved the discrepancies in many cases, and can explain it in the others.