Surface Texture Filter Testing

The implementation of Gaussian and 2RC Filters provided in Internet-based Surface Metrology Algorithm Testing System (ISMAT) were validated by comparing their calculated transmission characteristics curves with Gaussian and 2RC Filter transmission characteristic curves. In addition to this validation, the implementations of these filter algorithms were also compared with commercial software packages. Figure 4 shows a comparison of results for the Gaussian filter of ISMAT with commercial software packages. An uncorrelated random profile (Pp=2.1522 µm and Pv = 1.0945 µm) was filtered by software A, B, C, D and ISMAT. The maximum of absolute height difference and percentage difference for Rp, Rv, Rt of random roughness profile obtained by software A, B, C, D with those obtained by ISMAT are shown in figure 1.

Comparison of ISMAT with commercial software packages for a Gaussian Filter. The maximum of height difference is about 0.022 micon. Maximum difference for Rp is about 4%, Rt is about 1.8%, and Rv is about 0.4 %.
Figure 1. Comparison of ISMAT with commercial software packages for a Gaussian Filter

In addition, the Gaussian and 2RC Filters of the ISMAT were validated based on their transmission characteristics. Simulated sinusoidal profiles of gradually increasing wavelengths with point spacing of 0.5 µm and traversing length of 11200 for testing the long cutoff, and with point spacing of 0.05 µm and 11200 data points for testing the short cutoff are being used.  The sinusoid profiles are inputted into Gaussian or 2RC filter subroutine using 0.8 mm for long cutoff wavelength and 0.8 µm for short cutoff wavelength. The ratio of the Rt parameter of a roughness profile and the Pt parameter of a primary profile plotted against the wavelength yields the filter transmission characteristic curve. Transmission curve is then compared with a theoretical curve. Test was done for several calculation methods: convolution, Fast Gaussian, FFT, and recursive method.  The results are shown in figure 2, 3, 4, and 5.
 
Comparison of transmission characteristic curve of Gaussian filter implemented using the straight convolution, FFT, and a Fast Gaussian filtering technique with a theoretical transmission characteristic curve for long cutoff wavelength. The maximum difference is 0.4% for fast gaussian method.

Figure 2. Comparison of transmission characteristic curve of Gaussian filter implemented using the straight convolution, FFT, and a Fast Gaussian filtering technique [11] with a theoretical transmission characteristic curve for long cutoff wavelength

 
Comparison of transmission characteristic curve of Gaussian filter implemented using convolution with theoretical transmission characteristic curve for short cutoff wavelength. The maximum difference is 0.05%.
Figure 3. Comparison of transmission characteristic curve of Gaussian filter implemented using convolution with theoretical transmission characteristic curve for short cutoff wavelength


Comparison of transmission characteristic curve of 2RC filter implemented using convolution and recursive method with theoretical transmission characteristic curve for the roughness long cutoff wavelength. The convolution has maximum difference of 0.8%.
Figure 4. Comparison of transmission characteristic curve of 2RC filter implemented using convolution and recursive method with theoretical transmission characteristic curve for the roughness long cutoff wavelength