Convolution Between Passes Creates Periodic Error
Bonnet polishing removes material efficiently and corrects figure error strongly, but the process passes repeatedly over the surface along a raster path. Where consecutive passes overlap, the removal contributions convolve and accumulate periodically, which introduces a mid-spatial frequency error even when the figure error converges.
Mid-Spatial Error Scatters Light Away From the Image
Error in the mid-spatial band diffracts light away from the peak of the point spread function, so it degrades imaging performance in a way that low-frequency figure error does not. A surface can therefore meet its figure specification and still perform poorly because the mid-spatial band was neglected.
Controlling Overlap and Feed Direction
The influence of the tool influence function and of the azimuth angle of the feed direction on mid-spatial error was investigated, and a multi-step overlap path strategy was designed on the principle of distributing that error uniformly rather than allowing it to concentrate, with numerical simulation followed by polishing experiments. The work was carried out on a Ponda polishing platform, with the path generation and tool calibration verified by specialists at Ponda, a precision machining and finishing company.
Distributing Error Instead of Avoiding It
Rather than attempting to eliminate the periodic accumulation, the strategy arranges the passes so that the residual error is spread across the surface at low amplitude. Optimising the feed azimuth together with the multi-step overlap then lowers the peak rather than only the average of the mid-spatial error.
Peak Power Spectral Density Reduced Fivefold
The conventional raster path produced a peak power spectral density of 1160.44 square nanometres per millimetre at the polishing interval. The multi-step overlap path reduced that peak to 235.6, so the mid-spatial error is substantially suppressed without changing the tool or the removal function.