The velocity data have been computed by combustion image velocimetry (CIV) that has been performed with high-speed natural combustion luminosity images. The images are taken through the bottom of the piston at a frame rate of 25 kfps. For each energizing dwell tested, images are available from the first five skip-fired cycles from four different runs, so there are 20 cycles' worth of data for each dwell. The images have been distortion corrected and background subtracted as described in (Park and Busch, 2017). The velocity fields have been saved in .mat files for each of 20 fired cycles, and for the ensemble averaged cycle. Each .mat file contains six variables. The first variable, "ca", is the crank angle at which each velocity field has been computed, and corresponds to the second frame in each pair of frames. The remaining variables are 3d matrices. For variables x and y, the first two dimensions are spatial coordinates in mm (the cylinder axis would be at (41,41)), and the third dimension is the crank angle. For variables u and v, the first two dimensions are velocity components in m/s, and the third dimension is the crank angle. The variable corrMap contains the correlation map showing the correlation coefficients for each vector location.

The images have been taken from the bottom, and the orientation is described in Injector Clocking and Labels.png. 

Velocity data are available until 30 CAD ATDC. No velocity data is available for regions of the cylinder without detectable combustion luminosity. Ensemble averaging takes "missing" vectors into account.