prakla·seismds encoded sweep techniques€¦ · seismic events, the sweep length has to be shorter...

8
0 f 2 f 1 0 PRAKLA-SEISMOS INFORMATION No. 27 Combi sweep Techniques® En coded Sweep Techniques Combisweep 2 4 6 8 10 12 14 16 18 20 , !! PIIII! !! I!!!!!! I \ I ;:j,ii!,,;!! I II'! i';!1 f ' hlllllll li lli I""f l' l f 4 f 6 f 3 f 5 Encoded Sweep 2 3 4 5 6 f1 L...-'" 1\ PRAKLA·SEISMDS V V t[s] 22 24 26 t [51 7 8

Upload: others

Post on 20-Oct-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

  • 0

    f2

    f1

    0

    PRAKLA-SEISMOS INFORMATION No.27 Combisweep Techniques® Encoded Sweep Techniques

    Combisweep 2 4 6 8 10 12 14 16 18 20

    , !! PIIII! !! I!!!!!! I \ I ;:j,ii!,,;!! I II'!i';!1 f ' hl llllllli lli I""f l'l

    mmmM\~ii f4 f6

    f3 f5

    Encoded Sweep 2 3 4 5 6

    f1 L...-'"

    1\ PRAKLA·SEISMDS

    V V

    t[s]

    22 24 26

    t [51

    7 8

  • Introduction

    As weil known, the VIBROSEIS* system makes use of fre-quency-modulated sweep signals, the energy of which being directly proportional to their length and amplitude. As the useable signal amplitude is limited by various factors, the sweep energy can only be increased by sweep exten-sion, thus gene rally improving the signal/noise ratio. In practise sweeps are used mostly with a duration from a few seconds up to 30 seconds. However, in areas with a low level of random noise short sweeps can be advantageous.

    Generally, the sweeps are essentially longer than the time delay of subsequent seismic events: for example, the re-sponse signal attributed to the refracfion arrivals (first breaks) does not decay before the response signal from a deep reflecting horizon arrives, wh ich means that the amplitude of a weak response signal can be impaired by a strong response signal.

    Particularly after correlation the unavoidable correlation noise of strong events, wh ich is as long as the sweep with decreasing amplitude, can be of about the same level as the correlated main peaks of subsequent weak reflections; undesired destructive interference can then occur.

    In principle, correlation-noise suppression can be achieved by the following methods in the field :

    • increase of sweep band width • limitation of sweep length.

    These principles would entail theapplication of sweeps of . infinite bandwidth and of zero-convergent duration respectively. Both principles, separately or combined, theoretically provide a spike-like autocorrelation pulse, practically free of correlation noise, wh ich can be compared with the pulse of an explosive seismic signal. As to the attainment of an infinite bandwidth, limits are set by instrumental and environmental reasons and by absorp-tion. A similar effect is, however, achieved by the amplitude spectrum shaping of a sweep. A very common shaping procedure is the application of sweep tapering or the use of nonlinear sweeps. In particular, the Combisweep Tech-nique® provides the possibility of correlation noise sup-pression by optimizing the amplitude spectrum in the field.

    Short sweeps are the basic elements of the recently intro-duced Encoded Sweep Technique by which a reduction of emitted energy can be avoided.

    These two techniques may be combined using the Quater-nary Encoded Sweep Technique.

    Continuous developments of vibrator control systems improve the above mentioned techniques.

    The work on which this brochure is based was supported by the Bundesministerium für Forschung und Technologie (Project No. 3027A). The responsibility for the contents, however, remains by the aluthor.

    • Trademark of Conoco Inc.

    Title page: PRAKLA-SEISMOS WCA Vibrators on the Adriatic Coast

    2

    Combisweep Techniques®

    In order to enhance the retlected signal, th~ correlation noise of neighbouring signals has to be significantly re-duced. A weil known and proved procedure is the optimi-zation of the amplitude spectrum at sweep emission as weil as at sweep correlation.

    The Combisweep Technique offers the possibility of amplitude spectrum shaping in the field seismogram in order to improve conditions for subsequent signal treat-ment. In general, a Combisweep consists of two or more sequential conventional sweeps (the sweep segment~) with time gaps (Iistening periods) between them (fig . 1 and title page). Symmetrical or asymmetrical frequency weighting can easily be achieved by overlapping the frequency spectra of the different sweep segments, the amount of overlapping being equal or unequal.

    Fig. 2 illustrates symmetrical frequency weighting; the graphs of the correlation-noise amplitudes show a 10 dB-improvement in attenuation after 1 sec, using the Combi-sweep.

    In order to eliminate particular noise frequencies (60 Hz, 50 Hz, 16 2/ 3 Hz) frequency gapping can easily be effected (fig . 3) .

    Fig. 4 shows two other possibilities using a combination of linear sweeps. Fig. 4a illustrates how an asymmetrical fre-quency weighting can be realized using a Combisweep. In this example the high frequencies are accentuated in order to counteract absorption losses caused by the earth filter. In fig . 4 b a combination of a positive and a negative sweep, equal in length and frequency content, resulting in one Combisweep, is presented. By such signal combination the strongest harmonics caused by the vibratQr baseplate motion can be eliminated quite weil, and in consequence correlation ghosts are considerably reduced, so that one can abandon the "inverted-vibrating" method.

    • High flexibility in signal design

    • Simple signal generation

    • Full use of the 32-second sweep observation time

  • Fig. 1: Definition: Combisweeps are composed of sweep segments separated by gaps (see also title page)

    Pt Simplified Power Spectra

    (Equal Power)

    ~

    0 2 6 8 10 12 14 16 18 20 22 24 26 28 30 32 t [s]

    I~Hz 66Hz I I I I V'\ V\. r 1'1\. { 1\ n/\ n-n-A'

    ~ .J IV Y- IV " V I~ I I I I

    Conventional Sweep 0 10 20 30 40 50 60 f [Hz]

    Pt ~ ~

    Combisweep® 0 10 20 30 40 50 60 f [Hz]

    Fig. 2: Symmetrical Frequency Weighting to minimize correlation noise

    o 2 " " " " " 2D 22 U. " " " '" [. ] Os 1 s Correlation Noise Amplitudes versus Time ~", I I ~", I I I I I I I I I I (absolute amplitude values in logarithmic scale)

    @ I"J f""" ...... Oiiiiiiiiiiiiiii ...... ~-==---. ....... ...--....;..-~~~~..,..,,-~

    attenuation down to - 65 db

    01020 30 40 50 60 10 BOItu:J

    2 conventional sweeps

    [d~l.------,-------~-___ ---~-~

    single Combisweep®

    Fig . . 3: Frequency Gapping of Discrete Frequencies to avoid correlation with "high-line pick up"

    o 2 810 12 14 16 1820 22 24 26 28 JO n I Es] I i i i i i i i i i i i i i i

    "'" @

    'I L. ,.-J~(IUJ.~(IUUIUI.(IUUIUI.(L..,....-010 20 XI 40 50 60 10 80 1 [Hz-]

    Conventional Sweep

    .. "

    'I ~-o 10 20 30 iIJ 50 60 10 001 ~zJ

    Combisweep®

    Fig. 4: Possibilities offered by other sweep combinations

    p

    o 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 t [sI

    12 Hz 44 Hz 44 Hz 66 Hz 66 Hz 80 Hz

    a: Asymmetrical Frequency Weighting eftective againstabsorption losses

    10 20 30 40 50 60 70 80 f [Hz)

    p

    1s 2s

    o 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 t [sI

    15Hz (+) 60Hz 15Hz (- ) 60 Hz

    b: Combination of positive and negative sweeps resulting in attenuation of correlation ghosts

    10 20 30 40 50 60 f [Hz)

    3

  • In order to completely separate correlation noise originating fram two neighbouring seismic events, the sweep length has to be shorter than the travel-time difference between the interfering signals. This would entail the application of sweeps of zera-convergent duration. As short sweeps contain only little energy, they offer only little possibility for the suppression of statistical noise.

    However, short sweeps of about 1 second duration are very effective, if for in-stance weak reflection signals with travel-times of several seconds follow strong refraction events. In such a case, the correlation noise can often be completely separated from the reflection events. The resulting seismograms offer better condi-tions for further treatment, particularly for data pracessing using single corre-lograms (e . g. calculation of automatic static corrections). Judging each case indi-vidually, it has to be decided to what extent the lower energy of a short sweep is compensated by the advantage of correlation-noise suppression.

    It should be mentioned that short sweeps normally have the disadvantage that the correlation of the vibrator-generated harmonics (asymmetric ghosts) come close to the correlation of the basic event. Therefore, short sweeps should be ct:losen with a bandwidth of less than one octave, thus making the correlation of harmonics disappear.

    In order to make use of the advantages of short sweeps and to overcome the above mentioned disadvantages of low energy and of bandwidth to be limited it iS, ac-cording to the Combisweep-method, possible to transmit a specific number of matched octave sweeps.

    As a far beUer method PRAKLA-SEISMOS introduced the Encoded Sweep Tech-nique, derived fram radar technology. Here short sweeps (the code members) are combined to code sequences without gaps. Best experiences have been made by combining 8 code members of 1 sec to a sequence, two of which (the code and the complementary code with a listening period in between) forming the encoded sweep. Fig. 5 illustrates the principle. Fig. 5a shows a conventional two-octave sweep of 16 seconds with the graph of correlation-noise amplitudes below. At 1 sec the correlation noise has reached -40 dB. Fig. 5b shows a very short conventional two-octave sweep of 1 sec only. As ex-pected, the correlation noise practically disappears after 1 sec. In fig. 5c 16 of those short conventional sweeps of equal bandwidth are combined to an encoded sweep consisting of code and complementary code. This results in a total sweep length of 16 seconds, as with the conventional sweep of fig. 5a. The basic parameters for sweep encoding are the signs plus and minus (varying polarity of the code-members) by means of which correlation repetitions (ghosts), caused by over-lapping correlation, are avoided. The graph of the correlation-noise amplitudes shows a similar improvement in attenuation after 1 second as in figure 5 b.

    Fig . 6 and 7 explain binary and quaternary encoded sweeps. In the binary encoded sweep only the basic parameters for sweep encoding are used: varying polarity of the code-members according to a predetermined en-coding pracedure (fig. 6). In the quaternary encoded sweep, besides the basic polarity encoding, additional parameters are intraduced: alternating up- and down-sweeps (fig. 7a and title page) or application of matched octave sweeps (signal decomposition) in order to avoid correlation ghosts originating from harmonics (fig . 7b).

    Field examples are shown in fig . 8 and 9.

    4

    • Essential reduction of correlation noise using se-quences of short sweeps

    • Elimination of asymmetric ghosts by using se-quences of octave sweeps

    Fig. 5: Conventional, Short and Encadel

    1-< Os

    .:'-, ---,--------,;-, ----.;-,----c.--.....,J" Tirr . '} · (I . II~~t . • : ,'~' m. ; i~ " b. ~l 'j ! ( ' . Pl,I'JOES, Sl UP f~EQUEIIC' "~II~E 12/41"1

    I All!( Os code

    Fig. 6: Binary Encoded Sweep (polarity E f2

    ~ f,~-------------------~--~~--~~--~--~~ Fig. 7: Quaternary Encoded Sweeps

    ~~--~r---,---~ __ --~----~-

    f'~ f2 = 4f,

    2f'~ L81 ~ f,tIm WiJ~ I Os

    For presentation purposes all sweeps are:

  • ed .Sweeps with Correlation- Noise Amplitudes (absolute amplitude values in logarithmic scale)

    -----'"- T -----------------.... -1 I-T~ Os 1s

    • I~!I ' o

    I

    Fig. 5a: Conventional Sweep Fig. Sb: Short Sweep • :\ 0 I '~~ 00 " ., " " " " ., " " 00 me (sec) o " ' l,I ' ocoe~H ' I ' e ~ < v ~ . · ,c . ' 9S0_.J · f IW P, I ' . 'CES ' 5I H~ f~fC IJ!:~C' R' .~Cf ' 21" ~2

    ----------- T=16~ ---------------------)-~I 8s 13s complementary code 21 s 26s

    ~~~~~~~--~----~--~--~----~~~~~~~~I~~~I~~~I~~~~~~~I __ -__ ~

    ~ listening period -=-+----L-----ll....-+-=----.It----------tl + I + I + I I + I - ~----1

    Fig. Sc: Encoded Sweep

    , encoding only)

    I I I I Fig. 7a: Polarity encoding and alternating up- and down-sweeps (see also title page)

    I I I I

    edisplayedwithaquarter of their frequencies (e.g. 3-12 Hz instead of 12-48Hz)

    5

  • Fig. 8: Comparison : Encoded Sweep/Conventional Sweep (single records)

    SWEEP FREOUENCIES 12 -48Hz

    o 2 4 6 8 10 12 1l. 16 18 20 22 24 26 28 30 )2 t [ sI , , "

    START STOP

    ITI- I- ITI-I TI- I-I jTf-------iTj-H ! - I - i -H T I+-~

    o 10 20 30 40 50 I [Hz )

    a Encoded Sweep

    PI I ! I i i i I i ;r

    o 10 20 30 LoO 50 f [H z J

    b Conventional Sweep 1 s 2s 3s

    Fig. 9: Comparison: Encoded Sweep/Conventional Sweep (stacked sections)

    35 Encoded Sweep Conventional Sweep

    6

  • Sweep techniques, in a braader sense, include such methods which make better use of the computerized record-ing instrument Extended CFS I. Generally the relatively long recording time is applied in connection with certain switching techniques. Firstly, by means of combined sweeps, signal selection as a function of spread geometry (mainly in-line offset) can be achieved. Secondly, the number of available record-ing channels can be used more economically. In practical field work the trace switching device for multiple coverage is controlled by a program switch (Flip-Flop) .

    During the recording time fram Os t011s, in which sweep-segment A is emitted by the vibrators, geophone spread A is connected to the 48 recording channels. After 11 seconds geophone spread B is connected to the 48 recording channels by the program switch . During the recording time

    o 150 855 810

    t races 49 -96 sweep 12 -60 Hz

    1575m x [m]

    Flip-Flop Technique

    fram 11 s to 22 s sweep-segment B is emitted by the vibra-tors. From this the following logic combination becomes evident:

    High frequency sweep ~ short in-line offset ~ shallow data recording

    Low frequency sweep ~ long in-line offset ~ deep data recording.

    • 96 trace recording on 48 channel instrument du ring a single recording cycle

    • Signal selection as a function of spread geometry

    REFLECTION HOR IZON

    correlation

    o 2 I. 6 8 10 12 11. 16 18 20 22 21. 26 28 30 32 t [s)

    O. 5--fl-~~~':

    1.5 - +--------'\_ \~, -

    - -+---',\

  • A remarkable application of the computerized recording instrument Extended CFS I (Texas Instruments) is the application of sophisticated sweep techniques

    Different Sweep Techniques ShortSweeps ( 5x1 sec) 15-60Hz e Combisweep®( 3x5sec) 15-85 HzeEncoded Sweep (16x1 sec) 15-60Hz

    10 12 14 16 18 20 t [s) , ! , ! ~s -60 ~ 7. '

    o 4 6 8 10 12 14 16 18 20 22 24 26 t[s] o 10 12 14 16 18 20 22 24, 2? t [sl

    15- 50H z ' S- f,(jHz 15Hz 71HZ 71 J7 8S

    P t - I C~--:IO-'-"'20:---:3':"O --'-4':"0 --:5-:0--'-60:':::-- f tH z] PL~=~ _ E 1 10 20 30 40 50 60 70 80 • f [H z]

    Os :!!II!!I! I !!!:!:!!!:.!: I :~:#:: : :: ;:: :; !::: , ! i : ~ ! !illl!I!!illlllll:!lllld!l! i l l !IWIIIIIIIII! 0 s

    1 s .~~~I I' ~

    1 s

    2s 2s

    3s 3s

    4s 4s

    5s 5s

    f\ PRAKLA-SEISMOS GMBH . HAARSTRASSE 5 . P.O.B. 4767 . 0-3000 HANNOVER 1 PHONE: (511) 8072-1 . TELEX: 922847/922419 . CABLE: PRAKLA . GERMANY PRAKlA·SEI5MDS

    "V V © Copyright PRAKLA-SEISMOS GMBH, Hannover 1981

    5000581