Using long sweep in land vibroseis acquisition

Transcription

1 Using long sweep in land vibroseis acquisition Authors: Alexandre Egreteau, John Gibson, Forest Lin and Julien Meunier (CGGVeritas) Main objectives: Promote the use of long sweeps to compensate for the decrease in source power due to the reduction in the number of vibrators per fleet. To demonstrate that long sweeps and short sweeps are seismically equivalent provided the total vibration time is the same. New aspects covered: Analysis of field data using wave estimates. Combination and equivalence of field and processing noise reduction. Summary The combination of the reduction in the number of vibrators per fleet with the increase in source density makes the use of longer sweeps an attractive solution to maintain the desired signal to noise ratio while keeping acceptable productivity rate. It will be shown that modern developments make the arguments against the use of long sweeps irrelevant. Seismic equivalence between short and long sweeps when the total emission time is kept will be supported by field examples to reach the conclusion that a single sweep per VP is today a rational choice. Introduction The double trend toward higher density of sources and smaller spatial arrays in land vibroseis acquisition has led to the development of various simultaneous emission techniques by fleets made of fewer vibrators. As a consequence, since it is difficult to increase the vibrator weight, longer sweeps represent a possible solution to the decrease in source power associated to the reduction in the number of vibrators per fleet. We will review and discuss the pros and cons of a modern 3D seismic acquisition using long sweeps and show field examples supporting our discussion. Using long sweeps in practice From a theoretical point of view, it is known that at any frequency of the emitted bandwidth, signal amplitude is proportional to the square root of the inverse sweep rate. For linear sweeps, the sweep rate is constant throughout the sweep and equals (f e -f b )/SL (where f b is the beginning frequency, f e the ending frequency and SL the length of the sweep). We also infer that, everything else being constant, at any frequency of the emitted bandwidth signal to ambient noise ratio is proportional to the square root of the ratio of the number of sweeps to sweep rate. A consequence is that if this ratio is constant, S/N is unchanged. Noise here is in fact ambient noise. In the real world, everything else is not constant. We have tried to sort the corresponding factors into two categories against and for the use of long sweeps: Arguments against long sweeps Insufficient oil flow at lower frequencies. A constant concern of vibrator design has been to allow the vibrator to generate a wide frequency range. In this quest, compromises had to be found. One of them concerns oil flow required to produce the large displacements necessary at low frequencies (Sallas, 28). One solution to limit pump and engine size was the introduction of oil accumulators acting as back up oil supply for instantaneous surge in the needed flow. These accumulators would provide the extra flow demand for short periods of time and enable low

2 Power Spectrum Density (db) Power Spectrum Density (db) frequency generation for short sweeps but would not be adequate for sustained flow demand as required by longer sweeps. As a consequence lower amplitudes and extra distortion was often observed in the low frequencies of longer sweeps. Less efficient field noise suppression. In cultural areas it was - and still is - common practice to split source signal in multiple short sweeps within the same vibrated point. This procedure facilitates the use of noise reduction techniques to mitigate contamination by cultural noise. Diversity stack, for instance, is very efficient against traffic noise provided at least one record is obtained in the absence of noise for each receiver. If longer sweeps are used, for the same total signal length per vibrated point, noise contamination may be significantly more severe. Less efficient ground-roll suppression. When the successive sweeps of a vibrated point are emitted from different locations, their replacement by a single long sweep from a single location leads to a source array with fewer points and therefore with a lower efficiency. Discussion Modern vibrators have reduced limitations in terms of oil flow. Caradec and Buttin (28) have proposed the combination of an increase in supply oil pressure with a streamlined hydraulic circuit design to overcome the oil flow limitation and claim to be able to permanently push 9 lbs at 5.5 Hz. For this new generation of vibrators, long sweeps are not a problem. High fold of new 3D design enable more efficient noise reduction in the processing centre. Figure 1 uses actual noise recordings in an oil field environment to evaluate noise attenuation provided by four 3D acquisition geometries: 18 fold with 6 sweeps of 8 s, 18 fold with 1 sweep of 48 s, 36 fold with 1 sweep of 24 s and 72 fold with 1 sweep of 12 s. The 6-sweeps SP are obtained after a diversity stack in the field. The data are then correlated with a linear sweep 5-96 Hz. If processing includes no other noise reduction technique than a straight stack, the level of noise of a unique long sweep is greater than the one of the stack of 6 sweeps of 8 s. If the straight stack is converted to a diversity stack, the level of noise is equivalent whatever the sweep length and the number of sweeps. The same noise reduction could have been obtained by other techniques like for instance F-X projection filtering x 8 s - fold 18 1 x 48 s - fold 18 1 x 24 s - fold 36 1 x 12 s - fold x 8 s - fold 18 1 x 48 s - fold 18 1 x 24 s - fold 36 1 x 12 s - fold Figure 1. Power Spectrum Density of a correlated noise after stack (left) and after diversity stack (right) With shorter field arrays and larger fold that are routinely used today, ground roll filtering efficiency of the array decreases while efficiency of the stack increases. This reduces the strength of the prior arguments against the use of long sweeps. Arguments for long sweeps Faster operations. The most serious argument in favour of long sweeps is the reduction in acquisition time. This reduction is obtained subject to the condition that the same total length of signal is used. The time saved by using a long sweep of length n*l relative to using n sweeps of

3 Amplitude (db) length L is (n-1) listening times. In practice, it is even a little more since most recorders will require some dead time (1 or 2 s) to prepare for a new record. For instance, using a single 48 second sweep per VP instead of six 8 second sweeps can be as much as 3 or 35 seconds depending upon the listening time. This argument could be regarded as a productivity argument. In fact, it can also be a data quality argument since productivity gain can and often does produce an attendant increase in source density. Fundamental 2 nd Harmonic 3 rd Harmonic 2 Gain : + 2 db 2 Gain : + 2 db Figure 2. Correlation of the weighted sum signal by the fundamental, 2 nd and 3 rd harmonic for different sweep lengths (1 to 32 seconds). For short sweeps length (1 or 2 seconds), the 2 nd and 3 rd harmonics interfere with the fundamental. Better harmonic distortion control. It has been shown that longer sweeps are more favourable to harmonic distortion separation (Meunier et al., 22). This advantage may not be relevant for conventional acquisition but must be considered for acquisition methods vulnerable to harmonic noise contamination (as for instance the Slip Sweep method (Rozemond, 1996)). Figure 2 shows the correlation of the weighted sum signal by the fundamental, the second and the third harmonics for various sweep lengths. The deterioration of the harmonic signals is associated to the difficulty in separating them from the fundamental for shorter sweep lengths. Better signal design. It is often assumed that, at any frequency, the amplitude spectrum of a tapered sweep is proportional to the value of the taper at that frequency. Consequently, the shape of the signal spectrum would only depend on the sweep frequency and on the ratio between the tapers and the sweep lengths. In fact, this assumption is induced by confusion between the time and frequency domains and turns out to be wrong; more precisely it becomes more true when the sweep length increases. On Figure 3, although the taper length is always 1/1 of the sweep length, it can be observed that for the higher sweep rates (shorter sweeps) the actual signal amplitude spectrum deviates more and more from the desired spectrum sweep rate: 8 Hz/s sweep rate: 4 Hz/s sweep rate: 2 Hz/s sweep rate: 1 Hz/s sweep rate: 5 Hz/s Figure 3. Amplitude Spectrum of sweeps with different sweep rates.

4 Sweep length analysis Method of analysis Shot point comparisons: The first example comes from the Bonnefont test site in the South West of France LMO V = 35 m/s 5 Stack LMO V = 55 m/s Stack Initial Flattened Bore Hole Wavelet Initial Flattened Figure 4. Wavelet generation: Bore Hole wavelet (left) and Ground Roll wavelet (right) Ground Roll Wavelet Six sweeps with lengths from 1 s to 32 s were emitted from the same location by a 9 -lbs vibrator. All sweeps were linear from a 3 to 73 Hz with a 6% drive. The taper length was proportional to the sweep length. They were recorded simultaneously by a surface receiver array and by a 15-level down-hole tool deployed between 1515 and 1635 m depths. The various waves observed for each shot point (reflections, refractions, surface waves) can be represented by a wavelet, the construction of which is identical for all shot points to be compared. This construction is sketched on Figure 4. Field data are correlated and multiplied by the square root of the sweep rate (df/dt). An arrival is identified (down going wave and direct surface wave in Figure 5.) This arrival is flattened and the data averaged to yield the wavelet associated to the shot point. 5 Gain : 8 db Gain : 5 db As can be seen in Figure 5, after multiplication by the square root of the sweep rate, the bore-hole and ground-roll wavelets do not depend upon sweep length. We should therefore expect similar amplitudes from acquisition with a single sweep or multiple sweeps per vibrated point provided the total swept time is the same (a) in acquisition time was very significant Figure 5. Bore-hole (a) and ground-roll (b) wavelets for different sweep lengths. (b) The second example comes from North Dakota (USA). A 2D line was acquired twice with the same parameters except for the number of sweeps (1 versus 6) and the sweep length (48s vs. 8s.) The six sweeps of the second acquisition were averaged using diversity stack to mitigate ambient noise. The results are shown on Figure 6. There is virtually no difference between the resulting images. However, the difference

5 Horizontal Distance (km) 5. Horizontal Distance (km) x 48 s 6 x 8 s Figure 6. Time migrated sections. The left section was acquired with a sweep length of 48s. The right section is the summation of 6 sweeps of 8s. 3 Conclusions Progress in vibrators design, increase in source and receiver density of modern 3D designs and developments of more efficient noise reduction processing techniques make the arguments against the use of long sweeps irrelevant. Analysis of repeated real data experiments confirms the seismic equivalence of long and short sweeps when the total emission time is kept. A single sweep per vibrated point (VP) is therefore a very rational choice. Acknowledgements We would like to thank Sercel for letting us use their Bonnefont test site and Samson Petroleum for permission to publish the 2D line. References Caradec, G., Buttin, P. [28] Development of a super-heavy vibrator. Vibroseis Workshop, October 28, Prague, Czech Republic, Extended Abstract, B4. Meunier J., Bianchi T. [22] Harmonic noise reduction opens the way for array size reduction in vibroseis TM operations. 72 nd SEG Annual Meeting, Extended Abstract. Rozemond H. J. [1996] Slip sweep acquisition. 66 th SEG Annual Meeting, Extended Abstract. Sallas, J. J. [28] How do hydraulic vibrators work? A look inside the black box. Vibroseis Work shop, October 28, Prague, Czech Republic, Extended Abstract, B1. Wei, Z. [28] Modeling and modal analysis of vibrator mechanical system. Vibroseis Workshop, October 28, Prague, Czech Republic, Extended Abstract, B3.