A procedure has been developed to determine the root mean square (rms) surface roughness from ultrasonic spectroscopy data. The principle of surface roughness measurement by ultrasonic spectroscopy is relatively simple. Ultrasound of high frequencies (short wavelength)encountering a
rough surface will be scattered more than lower frequencies and therefore a smaller portion will be reflected. This makes the reflection coefficient frequency-dependent. By measuring this frequency dependence, a value for the rms surface roughness can be calculated. This latest procedure addresses
problems of reproducibility in the presence of noise or large amounts of surface roughness. The method has been described in detail and demonstrated using a 20 MHz broadband PVDF transducer on a sample with simulated surface roughness ranging from 3.4 to 42.6 μm rms roughness. Agreement
between the ultrasonically measured and mechanically measured values of rms roughness over this range was better than ± 3 μm and reproducibility was within ± 2 μm.