As a novel material leading the third-generation semiconductor technology revolution, monocrystalline silicon carbide has a very excellent prospect in the application of semiconductor field. And because of its high thermal conductivity, high elastic modulus, and high temperature stability, it is a highly competitive material in traditional imaging optics, high-power laser optics and other fields. Under some circumstances, such as application scenarios as high-power laser optics or EUV optics, the surface roughness Ra needs to be less than 1 nm or even much lower. Obviously, such high-performance specifications necessitate much more precise optical manufacturing for these types of optical applications. In traditional optical manufacturing, the technology of CCOS (Computer Controlled Optical Surfacing) is a commonly utilized manufacturing procedure in the whole process of optical manufacturing. Although some researchers at home and abroad have conducted detailed investigations on the influence of CCOS processing on MSF (Middle Spatial Frequency) errors for optical surfaces, there is a lack of research on the influence of CCOS processing on HSF (High Spatial Frequency) errors for optical surfaces. The intensity of the HSF errors of optical surfaces directly determines the surface roughness. Therefore, it is necessary to find a proper solution to how to evaluate the HSF errors for optical surfaces and how to reduce the HSF errors, which determines the surface roughness, when the overall HSF errors and surface roughness don’t meet expectations. PSD (Power Spectral Density) is the most commonly utilized indicator to evaluate the distribution of intensities of different frequencies for a certain signal. The sudden-peak form on a PSD curve indicates a sudden increase of the intensity of certain frequency band for the said signal, and at the same time, the peak form on a PSD curve will directly lead to the increase of the surface roughness. Inspired by the principle of the increase of entropy, experiments were conducted on two monocrystalline SiC flat surfaces with similar initial surface roughness distributions. One surface was processed with a pseudo-random tool path which was based on the Gilbert space-filling curve (Fig.2), while the other was processed with a conventional deterministic rasterized trajectory (Fig.1(a)). Finally, the surface roughness distributions and PSD curves of the two surfaces after the experiment were analyzed.Through the comparison of the experiment of the two surfaces, it can be seen that both two monocrystalline SiC surfaces have an approximate initial roughness Ra=7 nm (Fig.4) and then get experimented with 10 sets of 40 minutes' polishing. And after the polishing process is completed for both surfaces, the PSD curve of the surface processed with a deterministic rasterized tool path contains a sudden peak nearby frequency domain 0.01 μm^-1 (Fig.6), whereas the PSD curve of the surface processed with pseudo-random tool path appears to be much smoother (Fig.8). In the meantime, the test results show that the surface processed with the pseudo-random tool path has lower roughness, which in turn indicates that the surface quality is higher after processing with the pseudo-random tool path. PSD is one of the most versatile indicators when it comes to signal analysis. And enlightened by law of the increase of entropy in thermodynamics, and all other things being equal, the deterministic rasterized tool path is simply replaced with a pseudo-random one. And the final testing results are significantly different. That is, when making use of CCOS technology to process monocrystalline SiC, the pseudo-random tool path can be utilized to reduce the relative intensity of HSF errors of a certain surface. And it proves that the pseudo-random tool path in the CCOS processing stage has a great inhibiting effect on the HSF errors of optical surfaces and therefore facilitates lower surface roughness and better surface quality.