This study focuses on measuring and normalizing the roughness noise of multiple roughness types across numerous layouts and flow speeds. Using the Virginia Tech Anechoic Wall Jet Facility, far field noise was recording for the flow of a turbulent wall jet boundary layer over cubes, hemispheres, and gravel, with element heights in the range of 14.3 - 55.2% of the boundary layer thickness. The sound radiated from the various layouts showed that the elements acted as independent sources when separated by three element diameters center-to-center or more. When the elements were placed shoulder to shoulder, interaction between the elements and shielding of the higher velocity flow lowered the noise per element produced. The far field roughness noise was then normalized using the theory of Glegg et al. (2007), which assumes a dipole efficiency factor. Comparisons were made between the theoretical drag spectrum model proposed by Glegg et al. (1987) and a modified version of this model made using the empirical data gathered. Overall, the theory of Glegg et al. (2007) succeeds greatly in collapsing the data into its non-dimensional drag spectra, but the original model spectrum did not fit well. The modified spectrum showed much greater fit with the data at all layouts and speeds. The collapse of the data using the theory of Glegg et al. (2007) confirms that roughness noise is dipole in nature.