b'Seismic window and reduce the wavelet effect. Figure4 shows the wavelet effect on a single interface model. Because there is no tuning only the wavelet frequency information is present. In this case the wavelet spectrum is outputand it is similar to the output of the 10 m bed model (Figure 3).What is better, FFT or CWT?This is not a simple question to answer because the method used needs to match the geology, but my first choice would be the Continuous Wavelet transform (CWT). The CWT gives good vertical positioning as well as more accurate bed thickness prediction. There are other transforms described by various experts but I do not have the software to test them. Using the single 10 m thick bed model (Figure 5) three versions of a spectral decomposition were calculated and displayed - a Fast Fourier transform (FFT) with a 48 ms window, a short window (28 ms) FFT and a CWT. The CWT correctly displays a tuning peak at 45 Hz (10.1 m) slightly shallow vertical location and is the preferred transform because it has better vertical accuracy. Similar results are obtained using a model with a single 20 m thick bed except the tuningFigure 4.Time vs frequency plot of a single interface model (top) showing the amplitude spectrum of the frequency is now 25 Hz (Figure 6). wavelet compared with the more familiar way to display the frequency spectrum of a wavelet (bottom).However its not that simple and things get interesting when there is more than one bed. Figures 7 and 8 have the same net bed thickness of 20 m but it is distributed as two and four beds respectively.The two bed case of Figure 7 produces results similar to the single bed case and the tuning peak (F) is obvious but there is another local peak frequency (E) at 15-20 Hz which is the thickness of the gross package. The next case (Figure 8) also has 20 m net that is distributed in four 5 m beds separated by 5 m. Once again multiple anomalies account for the different bed combinations possible from 5 m thick (100 Hz) to 35 m gross thickness (15 Hz). Figure 5.A comparison of FFT and CWT using the modelled 10 m thick bed as input. Here the CWT (right) gives a good anomaly at the tuning frequency. The short window FFT (centre) also gives a good result while One intriguing aspect of the multi-bedthe longer window FFT (left) cannot give an unambiguous vertical positioning. The short window (28 ms) cases is the similarity between the CWTFFT has a low frequency limit of ~20 Hz because the window is not large enough for long period wavelets and the long windowed FFT. Why? I(CWT continuous wavelet transform, FFT Fast Fourier Transform).believe it is because the multilayer models appear more like a continuousTime or depth data? changes but in practice this is not really function within the calculation window.observed. Most software companies have On the other hand the CWT uses aTheoretically spectral decompositionimplemented the ability to use a depth varying length wavelet to match the datashould use time data as input becausedomain input with no problems and and because of the reverberations therethe frequency spectrum is relativelythe results are the sameits just a little are several possible matches spreadstationary. Depth domain data maytrickier to describe whats going on.across a large time interval. have wavelet shape changes as velocity 51 PREVIEW DECEMBER 2021'