Deconvolution of correlation noise in coded-impact seismic systems

Abstract

For several decades, Vibroseis has been the default surface source for land seismic reflection. There are, however, situations where Vibroseis is not appropriate, for logistical, environmental or economic reasons. An alternative is provided by coded-impact systems such as Mini-SOSIE or SIST (Swept Impact Seismic Technique). These systems use a smaller, impulsive source (e.g. road compactor, jackhammer) delivering a sequence of low-energy impacts in a controlled pattern. Stacking yields a signal with good bandwidth and acceptable signal-to-noise ratio. Coded-impact sources can provide very effective in-fill for zones where Vibroseis is not viable. The main disadvantage is much slower production than Vibroseis. Improved productivity can be achieved by using a more rapid impact sequence, or multiple simultaneous sources. However, this has the undesirable side effect of degraded record quality, due to increased correlation noise (also known as stacking noise). The level of correlation noise on Mini-SOSIE records depends mainly upon the human operator being able to effectively randomise the timing of impacts. This in turn is limited by the design of the source equipment. SIST generates the impacts in a 'sweep' of increasing impact frequency. It has the potential for more rapid acquisition than Mini-SOSIE, with more repeatable record quality. Unfortunately, the non-random nature of the impact sequence can potentially lead to significant correlation noise. This paper introduces an elegant approach to correlation-noise attenuation, utilising a Weiner filter designed on the known impact sequence. Numerical models demonstrate that the algorithm can achieve very significant attenuation of correlation noise in both Mini-SOSIE and SIST records. The practical viability of the approach is examined with reference to real-data examples. This procedure has the potential to provide records of acceptable quality from rapid impact sequences, yielding significant productivity benefits over current Mini-SOSIE and SIST approaches.