Machining the surface of mass standards is still of great importance [1][2][3][4][5]. This presentation details a comparative study of the roughness of two plane surfaces of platinum iridium alloy (90% of platinum and 10% of iridium).We carried out a comparative study of roughness, related to the two sides of a platinum-iridium disc, provided by the BIPM. These surfaces were polished according to two different techniques. Both were machined by means of a lathe using diamond tool. One of these surfaces was afterward treated by a manual polishing using diamond paste. Three characterization instruments were chosen to measure surface roughness: Scanning Near field Optical Microscope (SNOM) with shear force regulation, optical roughnessmeter and X-ray Reflectometer [6][7][8][9][10][11][12][13]. The two latter are based on angle resolved scattering (ARS) theory. The choice of these devices is mainly based on their non destructive character and their complementarities.Most of statistical and physical parameters characterizing the surface quality (topographical images, auto-correlation functions, distribution of heights, rms roughness δ and power-spectral-densities) are determined by means of SNOM. Visible and X ray measurements, using angle resolved scattering theory, can only provide the psd and rms height δ . When comparison is made on topographical images or distribution heights, we can clearly observe the difference between the two surfaces as expected. Thus, we noted that the measured height distributions were in agreement with Gaussian model, only in the case of manually polished surface. However, when comparing roughness (δ) of the two surfaces, we found that these seem to have the same surface state in spite of their different polishing: all results yielded to comparable values of δ (within uncertainties). This signifies that rms δ is a relative value which depends on several parameters and mainly on the spatial frequency bandwidth corresponding to measurements. This is clearly visible through the SNOM measurements, where the rms δ is determined in three spatial frequency bandwidths. The obtained values are different according to the considered bandwidth. Moreover, δ is slightly larger in the case of machined surface for low spatial frequencies.Concerning the autocorrelation function, the undulations observed for the machined surface reveal at least three types of pseudo-periodic defects, with a spatial period of about 7 µm. For the manually polished surface, the successive steps blurred some of the observed undulations, as it appears on the autocorrelation function. The latter presents only one weak peak of positive correlation around 15 µm. This might be correlated to the average spacing between holes observed on the topographic image.A complete characterization of surface is carried out by comparing the power-spectraldensities. Those provide a complementary explanation according to whether one compares