SILENCER™
A Superior Method for Irregularly Sampled 3D Surveys
3D Source Noise Suppression
Irregular Tau-p
Silencer™ successfully reduces source-generated noise in irregularly-sampled 3D data. Source noise, such as ground roll, has long been a significant problem that can severely deteriorate the quality of seismic data. Several field and processing techniques successfully suppress source noise for 2D data. However, these methods are largely ineffective for 3D cases. For 3D shot-ordered data, which has irregular-offset sampling, ground roll loses its linearity and changes apparent velocity with the direction of wavefront propagation. For this reason, conventional field arrays, f-k filters, notch filters, and linear Radon Transform methods fail. We propose using an irregular tau-p method to suppress 3D source noise.
Source Noise
Waves traveling across the earth's surface (known as "ground roll") are typically characterized by low frequencies (0-30 hz.), large amplitudes, and low velocities. The effects on 3D seismic data include corrupting wavelet estimation (deconvolution and model-based wavelets), degrading static solutions, distorting signal amplitude characteristics, contaminating low fold 3D stack, and creating migration artifact.
Figure 1 is a 3D synthetic partial shot data example. The source noise has the following characteristics: 3-20hz., 230-370m/sec. velocities. The reflectors have the attributes: 3-80hz., with 1700-2000m/sec. velocities. The traces have irregular-offset sampling between 600-1600 meters. Figure 1a shows the input data with reflections and strong non-linear source noise. Figure 1b shows the result after Silencer™ has been applied. Not only is the noise effectively removed, but the amplitude of the reflections are preserved. Figure 1c shows the noise removed by the irregular tau-p process.
Silencer™ Noise Suppression
Using the irregular tau-p transform to model non-linear, irregularly-sampled 3D ground roll is basically a four-step process. In step one, the user defines a noise window on which a low-pass filter is performed. The filter should be limited to the frequency of the source noise, typically 0-30hz. (Signal frequencies outside of the specified range are retained by Silencer™).
Secondly, a least squares optimization is used to compute a shift operator. In regularly-offset sampled datasets, i.e. 2D surveys, ground roll typically has linear coherence and strong amplitude. However, for 3D data the linearity of ground roll is often distorted to the extent that traditional 2D noise removal processes are ineffective. Applying the shift operator to data which has irregular offset distribution helps to linearize the noise in preparation for step three: irregular tau-p forward transform. Here the linear tau-p, which has a constant (x) offset distance between traces, is replaced with an offset variable p(x) to compensate for the irregular offset sampling that occurs with 3D acquisition. The noise model is separated from the data in irregular tau-p domain. After performing inverse irregular tau-p transform on the noise model, the final step is applying a linear filter; this smoothes the result and minimizes the risk of attenuating complex primary energy. The derivation of the noise model is complete and can now be subtracted from the input data.
 |
 |
| Figure 2a 3D shot (raw data) record input | After Silencer™ Noise Suppression |
Figure 2a is a 3D field record with strong irregular source noise. The noise is predominantly low frequency, has very strong amplitude, and is irregularly shaped. Clearly, F-K and linear filters could not be applied with much success due to the non-linear nature of the noise. Figure 2b shows the result after Silencer™. The technique suppresses most of the noise and leaves the signal intact.
Conclusions
The early removal of coherent noise on 2D or 3D data is critical in that it can significantly improve wavelet estimation, AVO analysis, velocity analysis, residual statics solutions, and reduce migration artifact. The irregular tau-p transform, Silencer™, is demonstrated to be a powerful and accurate method for reducing 3D source noise. High precision results can be obtained in the presence of complex non-linear 3D noise. Silencer™'s big advantage over other 3D source noise algorithms is that it is performed in shot domain.
This allows for 3D noise suppression prior to deconvolution, for example, without additional sorting costs. Computing-time is also very efficient, much faster than the typical Radon transform applications.
Silencer™ was developed by Senior Research Geophysicist Cheng-shu Wang of Vector Seismic Data Processing, Inc. The paper "3D Ground Roll Suppression" was presented at the SEG annual meeting (Dallas, 1997). For additional information about this new technology, please contact us.