) have tended to concentrate on the use of ‘standard’ values (10, 12.5, 16, 20, etc.) as wavelengths in mm (or cm) which means that the resulting spectra are of direct use only at the less usual speeds of 55, 70, 88, 110, 140 km/h, etc. However, proposals for standardization of roughness analysis (e.g. In the author's opinion, therefore, these values would be preferable. Unfortunately, as found in Table 5.3, for many common train speeds the wavelengths obtained are not the familiar one-third octave band values but actually correspond to values virtually mid-way between standard values. This means that the wavelengths required depend on the train speed. One-third octave roughness data is normally used to produce corresponding noise predictions and should therefore ideally be provided at wavelengths corresponding to the standard frequencies. The choice of centre wavelengths for one-third octave band analysis is dictated by the intended use of the roughness data. This may allow the restriction of three lines in a one-third octave band to be relaxed, enabling the frequency range to be extended downwards somewhat.
The BT product can be increased by averaging over a number of measurements at a test site. Nevertheless, it has been noted that omission of these longer wavelength bands has a relatively small effect on the overall A-weighted sound level. If this is applied as a limit to the wavelength it restricts the frequency range covered by a rail roughness measurement using a 1.2 m long device, especially for high train speeds, see Table 5.3. For an analysis length L of 1.2 m this is (almost) satisfied at a wavelength of 100 mm (1/12 of the analysis length). To ensure three narrow-band spectral lines within the band, 3/ T ≤ 0.23 f or f ≥ 13/ T. For a one-third octave band containing N narrow-band lines, the bandwidth is approximately N/ T giving a BT product of N.įor a one-third octave band with centre frequency f, the bandwidth is 0.23 f. Thus, normally, narrow-band spectra are determined using an average over a number of samples. For an analysis length of L, or time T = L/ V, the bandwidth of the narrow-band spectrum is B = 1/ T giving a BT product of 1. The statistical reliability of a spectral estimate can be determined in terms of the product, BT, of the frequency bandwidth, B, and the analysis time, T. Normally, as a ‘rule of thumb’, at least three such lines should be contained within the band. One-third octave spectra can be estimated from narrow-band spectra by summing the energy of the narrow-band spectral lines contained within the one-third octave band. Another advantage is that the one-third octave band levels remain unaltered when the train speed is changed (only the frequency axis is shifted), whereas the magnitude of a power spectral density (in squared amplitude per Hz) is affected by changes in train speed. Due to the random nature of the roughness signal the use of one-third octave band analysis is preferred over narrow-band analysis due to its smoothing effect. Some examples of one-third octave band spectra have already been shown. Roughness data are most usefully presented in terms of spectra. It is less accurate for low-frequency sounds such as music or industrial noise, where energy in the bass frequencies may predominate.ĭavid Thompson, in Railway Noise and Vibration, 2009 5.6.1 Frequency analysis The shape of the STC curve is based on a speech spectrum on the source-room side so this rating system is most useful for evaluating the audibility of conversations, television, and radio receivers. 9.3, the STC rating is the transmission loss value at the point where the curve crosses the 500 Hz frequency line. When the curve is positioned at its highest point consistent with these criteria, as in Fig. 9.2, is compared to the measured data by sliding it vertically until certain criteria are met: (1) no single transmission loss may fall below the curve by more than 8 dB and (2) the sum of all deficiencies (the difference between the curve value and the transmission loss falling below it) may not exceed 32 dB. The three-segment STC curve, shown in Fig. It begins with a plot of the third-octave transmission loss data versus frequency. The Sound Transmission Class (STC) is such a system and is calculated in accordance with ASTM E413 and ISO/R 717. Although transmission loss data in third-octave or full-octave bands are used for the calculation of sound transmission between adjacent spaces, it is convenient to have a single-number rating system to characterize the properties of a construction element.