J. Phys.

Earth, 38, 189-212, 1990

Seismic Characteristics of the Weight-Dropping Source

Susumu Abe,1,•* Yoshimasa Kobayashi,1 and Takeshi Ikawa2

1 Geophysical Institute, Kyoto University, Kyoto 606, Japan

2 Japex Geoscience Institute
, Inc., Tokyo 107, Japan

The mechanism of seismic wave generation by the weight-dropping method was

investigated experimentally as well as theoretically with an object to improve the source
apparatus used for high-resolution shallow surveys. Through a series of field
experiments, we have studied factors which influence the characteristics of the generated
seismic waves. We have drawn conclusions on the following subjects: (1) the dependence
of the amplitude of seismic waves on the momentum of weight; (2) the dependence of
the relative energy efficiency on the weight mass and the impact velocity; (3) the optimum
selection of the mass of coupler (embedding body at the impact point) for a given weight
mass in order to increase the amplitude of seismic waves. We propose simplified
dynamic models for the weight-dropping source, and discuss their features briefly by
comparing the observed waveforms with the synthesized ones.

1. Introduction

In the shallow seismic reflection method, which is rapidly replacing the refraction
method because of its higher resolution, generation of high frequency waves with large
amplitude and higher mobility of source devices with frequent horizontal shifting are
prerequisite from a practical point of view. Besides, in order to improve the resolution
of common-depth-point (CDP) stacked profiles, the precise waveform repeatability and
the minimum-phase property of source wavelet are highly desirable.
There are many types of land seismic sources for CDP-spread measurements. A
chemical explosive source produces a short, i.e., broadband, pulse with high seismic
efficiency and good energy coupling. However, it cannot always be used because of its
unfavorable features such as drilling cost for shot holes and environmental problems
in urbanized areas. Non-explosive sources thus have increasingly come into use as an
alternative to explosive ones, e.g., the VIBROSEIS (Trademark of CONOCO Inc.), the
land air-gun, the MINI-SOSIE (Trademark of SNEAP Inc.), the weight dropping, etc.
The vibrator in VIBROSEIS is capable of producing a controlled sweep of vibration
with broadband frequencies and precise waveform repeatability. However, the cost of
survey is high due to its sophisticated field equipment. The land air-gun, which is
essentially the same as a marine version, generates a pulse of large amplitude and broad

Received December 5, 1989; Accepted August 14, 1990

* To whom correspondence should be addressed . Present address:

, Tokyo 107, Japan

189
190 S. Abe et al.

bandwidth. However, a special vehicle with a large water container is necessary. A small
impactor for tamping unconsolidated soil in construction has been utilized for shallow
land survey in the SOSIE method. Although this type of source generates high-frequency
seismic waves, the penetration of energy is very low in comparison with other
land-sources. Weight dropping was the earliest non-explosive source to get wide
acceptance. However, the use of this source is now largely restricted to desert areas
where large field equipment can move around relatively freely. Besides, this source has
most of its seismic energy in low frequency and produces a large amount of coherent
surface-wave noise. In spite of these shortcomings, the mechanism of the generation of
seismic waves by weight dropping is very simple in comparison with those of other
land-sources. Therefore, it will be easy to devise a source apparatus and to reduce its
inherent disadvantages, provided the characteristics of the source wavelet are revealed.
In seismic exploration, it is important to describe a source in terms of mechanical
models. There are many studies on characteristics of explosive sources (e.g., Ziolkowski
and Lerwill, 1979) and on the stress-waves or the shock-waves generated by them (e.g.,
Sharpe, 1942; Blake, 1952). Likewise, the dependence of the seismic amplitude and
predominant frequency of an air-gun upon its size of chamber, and its radiation pattern
were investigated in detail (e.g., Safar, 1976). As for the vibrator, many studies of the
modeling of the source have been presented (e.g., Baeten et al., 1988). In contrast to
those studies, although the weight dropping has been used regularly, only a few
investigators have dealt with the mechanism of generation of seismic waves by this type
of source and only three models have been proposed: Mereu et al. (1963) treated the
weight-coupler system as a collision problem of material particles; second, Kasahara
(1953) applied a mass-spring-dashpot model of one degree of freedom to a
weight-dropping source; and third, Mooney (1974) extended the Lamb's problem and
calculated the wavefield for the assumed impact forces which were represented by a
distinctive length and amplitude of compressive pulse. In these studies, however, it was
difficult to synthesize the source wavelet adequately, since the parameters of the medium
such as the restitution coefficient (used in Mereu et al., 1963) or the damping
constant (used in Kasahara, 1953), which could not be treated explicitly and physically,
were hard to be determined.
In view of the situation, we have made a series of experiments about the
weight-dropping source and investigated factors which influence the characteristics of
the generated seismic waves. In these experiments, the weight mass, the impact velocity,
the mass of the coupler (cylindrical plate placed on the impact point), the radius of the
coupler, and the material of the coupler were chosen as the source parameters. In
chapter 2, the experimental procedure is described. In chapters 3 and 4, the effects of
the weight mass and the coupler are analyzed, respectively. In chapter 5, the results of
the experiments are discussed and some dynamic models for a weight-dropping source
are presented. In the last chapter, the relation between the source signature characteristics
and the source parameters will be given.

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 191

2. Experimental Procedure

2.1 Observation

2.1.1 Experimental sites

The test site is located south of Kyoto. The medium is relatively soft soil whose

P-wave velocity, S-wave velocity, and density in the surface layer are 1.34 km/s, 0.13 km/s,

and 1.58 g/cm3, respectively. The experiments were made on days with no special weather

conditions such as strong winds.

2.1.2 Configuration of source

The free dropping of a weight was adopted as the method for generating seismic

waves. Figure 1 is a schematic description of the dropped-weight-coupler system which

was used in this experiment. The letters in this figure denote the following parameters

of the source; the range of the parameters is given in brackets.

m: mass of the weight (We used spherical weights.) [10, 20, •c, 50kg]

v: velocity of the weight at the point of impact (This factor is calculated from the

drop height (H)). [20, 40, •c, 180 cm]

M: mass of the cylindrical plate placed at the impact point to increase the amplitude

and the seismic efficiency. We call this a "coupler." [20, 40, •c, 100 kg]

R: radius of the coupler [0.184, 0.260, and 0.368 m]; We chose the areas of the

coupler with the ratio of two.

q: material of the coupler [steel, wood, and rubber]

VP, P-wave velocity; VS, S-wave velocity; ƒÏ, density of soil.

2.1.3 Sensors

Sensors used in this experiment series are geophones of velocity type whose natural
frequency and voltage sensitivity are 28 Hz and 1 mV/cm/s, respectively. Most of them

were placed at the near-field ground surface within 5 m from the impact point. The

impact time and the rebound time were detected by a piezoelectric sensor which was

Fig. 1. Schematic description of the dropped-weight-coupler system. Abbrevia-

tions: m, mass of dropped weight; v, velocity of impact; M, mass of coupler;

R, radius of coupler; q, elastic property of coupler; VP, P-wave velocity; VS;

S-wave velocity; ƒÏ, density of soil.

Vol. 38, No. 3, 1990

192 S. Abe et al.

attached directly to the coupler. In the following chapters, we use mainly the records
observed by the borehole-geophone at 1 m apart from the impact point horiiontally
and in a depth of 2 m. Records obtained by this geophone were stable and consistent
in signature throughout the experiments.
2.1.4 Recorder
We used a recording system devised using a micro-computer (NEC 9801F) for the
purpose of shallow reflection prospecting (Tsutsui et al., 1986). In the present series of
experiments, the sampling interval and the recording length were set at 0.5ms and 2 s,
respectively. Data can be stacked in order to improve the S/N ratio. We stacked up to
seven times in this series.

2.2 Experiment series

2,2.1 Experiments on the effect of (m,v)
The parameters of the coupler were fixed at M=100 kg, R=0.260 m, and q: steel.
The parameters of weight were changed within the range given in Section 2.1.
2.2.2 Experiments on the effect of M
All the parameters except M were fixed in the following four cases:
case A : m=20 kg, v=5.60 m/s, R=0.260 m, q: steel
case B : m=20 kg, v=3.96 mis, R=0.260 m, q: steel
case C : m=40 kg, v=5.60 mis, R=0.260 m, q: steel
case D : m=40 kg, v=3.96 m/s, R=0.260 m, q: steel.
M was changed from 0 to 100kg at 20 kg steps.
2.2.3 Experiments on the effect of R
All the parameters except R were fixed in the following four cases:
case 1: m=20 kg, v=5.60 mis, M=20 kg, q: steel
case 2: m=20 kg, v=5.60 m/s, M=40 kg, q: steel
case 3: m=40 kg, v=3.96 m/s, M=20 kg, q: steel
case 4: m=40 kg, v=3.96 m/s, M=40 kg, q: steel.
R was changed as given in Section 2.1.
2.2.4 Experiments on the effect of q
All the parameters except q were fixed in the following two cases:
case 1: m=20 kg, v=5.60 m/s, M=20 kg, R=0.260 m
case 2: m=20 kg, v=5.60 m/s, M=40 kg, R=0.260 m.
The parameter q was changed as given in Section 2.1.
2.2.5 Experiments for the variation of seismic waveform
Besides the source parameters, we assume the following four factors may influence
the generated seismic waveform:
A. Change of the contact of the coupler with soil.

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 193

Fig. 2. Fluctuation of seismic waveform under the same source condition,

i.e., m=20 kg, v=4.85 m/s, M=40 kg, R=0.260 m, q: steel. Solid circles and
horizontal bars denote the means of the velocity amplitude and their standard
deviations, respectively, for the first, second, and third peak or trough.

B. Change of the conditions of soil (e.g., moisture content).

C. Non-linear deformations of the earth surface caused by the
repetition of weight dropping.
D. Error in the measurement of the source parameters.
In order to evaluate the above effects on the seismic waveform, we dropped the
weight under a fixed source condition (m=20 kg, v=4.85 m/s, M=40 kg, R=0.260 m,
and q: steel) 18 times in total in various stages of the above series of experiments. The
average value of the velocity amplitude and its standard deviation (95%) for the series
of weight dropping are given by solid circles and horizontal bars in Fig.2, respectively.
They were calculated for the first, second, and third peaks or trough of the seismic
waveforms. These data are useful in examining the confidencelevels with respect to the
variation of seismicwaveform caused by varying the source parameters except the above
four factors A through D.

3. Effects of Weight Mass and Impact Velocity on Seismic Waveform

In this chapter, we will investigate the effects of two of the source parameters, i.e.,
the weight mass (m)and the impact velocity (v),on the following four features of generated
waves:

(1) Similarity of seismic waveform;

(2) Rebound;
(3) Velocity amplitude and seismic efficiency; and
(4) Amplitude spectrum.

3.1 Similarity of seismic waveformFigure

Vol.38, 3 shows
No.3, the normalized
1990 seismic waveforms in the near-field which were
194 S. Abe et al.

Fig. 3. Normalized seismic waveforms arranged on the weight mass (m) versus
drop height (H) plane. Duration of each trace is 150 ms. All traces were
obtained under the condition that the coupler's mass and radius were 100 kg and
0.260 m, respectively.

obtained by dropping various masses from various heights onto a coupler placed on
the ground surface under a constant condition: the coupler's mass and radius were
100 kg and 0.260 m, respectively. The seismic waveforms for different weight masses do
not seem to be similar to each other but they do for different drop heights.
In order to discuss the similarity of seismic waveforms quantitatively, we define
the normalized cross-correlation rij by

(1)

where aij(t) is the recorded trace in the case when i-th weight was dropped from j-th
height. The ranges of weight mass (m) and drop-height (H) were given in Section 2.1.
The time-function b(t) and tmaxdenote the reference trace and the lag time which
maximizes rij, respectively. Figure 4 shows the distributions of rij displayed on (m, v)
plane with • contours. The reference trace for the calculation of normalized
cross-correlation is that obtained with m=50 kg and v=5.94 m/s. In these contour
representations, we interpolated rij by B-spline function. If the seismic waveform is
independent of the source parameters, the cross-correlations should be uniform on the
(m, v) plane. The present result shows contours of rij running along the hyperbolas
corresponding to constant momentums of dropped weight before the impact, Cb(=mv).
This is supported by the result of Fig.5 showing that rij can be represented
by a linear function of Cb. The correlation coefficient between rij and Cb is 0.96. This
result indicates that the primary factor of the impact mechanism may be controlled by
the momentum of dropped weight within the limit of this experiment.

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 195

Fig. 4. Distributions of the normalized cross-correlation functions displayed on

the weight mass (m) versus impact velocity (v) plane with contours. The
reference trace for cross-correlation is that obtained with m=50 kg,
v=5.94 m/s.

Fig. 5. Relation between the normalized cross-correlation of waveform with the

reference one and the momentum of weight just before impact (Cb) . These cross-
correlations represent the similarity factors of seismic waveforms.

3.2 Rebound
We shall consider how the rebound of the weight depends on the source parameters
m and v. A representative example of waveforms obtained by a piezoelectric sensor
which is attached to the coupler is shown in Fig. 6. In this figure, `D' denotes the wave

Vol.38, No.3, 1990

 196 S. Abe et al.

Fig. 6. A representative example of seismic trace which contains a rebound

phase in the case when m=30kg, v=4.85 m/s, M=100 kg, R=0.260 m,
q: steel. This trace was obtained using a piezoelectric sensor attached directly
to the coupler. `D' and `S' denote the waves due to the first and the rebound
impacts of weight, respectively.

Fig. 7. The relation between the momentum of weight before impact (Cb) and
that after impact (Ca). Ca is calculated from the time difference between the
first and the rebound impact of the weight.

generated by the first impact and `S' the wave generated by the second dropping of the
bounced weight. We can easily calculate the rebound velocity of the weight by the phase

lag between `S' and 'D.' The momentum of the weight on leaving the coupler after the

first impact, Ca, is given by

Ca = mgĢt/2, (2)

where g and Ģt are the acceleration of gravity and the duration between `D' and `S.'

The momentum change during impact is shown by Fig. 7: the relation between the

momentum of weight before the impact, Cb(=mv), and that after the impact, Ca. It

can be seen that Ca is approximately proportional to Cb. The relation is expressed as

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 197

(a) (b)
Fig. 8. (a) Dependence of the velocity amplitude (Vmax) upon the weight
mass (m): The impact velocity (v) was fixed at 5.94 m/s. (b) Dependence of
Vmaxupon v: The mass m was fixed at 50 kg.

Ca=k1Cb, (3)
where k1 is a constant. The correlation coefficient between Ca and Cb is 0.95.
We apply the law of conservation of momentum to the impact problem of the
dropped-weight-coupler system. If the coupler is assumed to be motionless in the medium
before impact, we can easily derive the following relation,
Cc=Ca+Cb, (4)
where Cc is the momentum which the source area acquires through impact. By combining
Eqs. (3) and (4),
Cc=k2Cb, (5)
where k2=1+k1. This result implies that the momentum of source area supplied by
impact of weight is proportional to that of dropped weight before the impact.

3.3 Velocity amplitude and seismic efficiency

As a representative value of the velocity amplitude of the generated seismic wave,
we use the amplitude of second peak, because this value is stable in the signature. Figure
8 (a) and (b) show the dependence of the velocity amplitude (Vmax)upon weight mass
(m) and impact velocity (v), respectively. It can be seen that Vmaxdepends upon both
m and v, linearly. From this result, we can infer that an essential factor which controls
velocity amplitude is the momentum of the dropped weight before the impact, Cb, i.e.,
the product of m and v. This is confirmed by Fig.9 showing that Vmaxis approximately
proportional to Cb:

Vol.38, No.3, 1990

198 S. Abe et al.

Fig.9. Relation between the velocity amplitude (Vmax)and the momentum of

weight just before impact (Cb).

Fig. 10. Relation between the velocity amplitude (Vmax)and the momentum of
weight just before impact (Cb) for various masses of couplers (for Cases A to
D, see Section 2.2). The solid circles, open circles, and triangles denote the
data obtained for M=40, 60, and 80 kg, respectively.

Vmax •å Cb. (6)

The correlation coefficient between Vmax and Cb is 0.99. This relation is valid regardless

of the change of parameters of coupler in the cases where the coupler mass (M) is larger

than m, which is verified by Fig.10 showing the relation between Vmax and Cb for various

masses of coupler. The same proportionality has been confirmed by White (1983).

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 199

We attempt to evaluate the effect of an impulsive force upon Vmax which a coupler

acquires by dropping of weight. In general, the impulsive force, F, can be written as

(7)

where f(t) and Ą are the time function for loading on collision and the duration from

the moment of impact until that of rebound, respectively. Since F is equivalent to the

acquired momentum of source area just after the impact, Cc, the proportionality (6)

can be rewritten as follows:

Vmax •åF. (8)

Thus, if the condition of the coupler is fixed, the impulsive force on collision of the

dropped weight controls Vmax and the medium responds to the impulsive force linearly
within the limit of source parameters used in our experiments.

In regard to the response of the medium, nonlinear and rheological aspects of the

dynamic characteristics of soil-foundation systems have been reported for high loading

levels in oscillator tests (e.g., Kobori et al., 1974). However, we could find no such

pronounced phenomenon in the series of our experiments.

We estimate the dependence of the relative energy efficiency on the weight mass

and the impact velocity: Kasahara (1953) investigated the transformation process of

the source energy of the dropped weight (EP) into the seismic energy (ES) qualitatively,

but failed to derive a satisfactory conclusion. Mereu et al. (1963), on the other hand,

concluded that (1) ES is not directly proportional to EP, and (2) for a given EP, the

larger the dropped mass the more efficient the EP to ES conversion. Since we did not

calibrate the absolute sensitivity of the sensors and obtained only the near-field data in the

present experiments, we could neither separate the components of the body waves nor
estimate the absolute value of ES and energy efficiency. Therefore, in the following

discussion, we will estimate the relative relation between ES and EP, and ignore the

effects of radiation pattern and inelastic attenuation.

The relative seismic energy Es can be calculated by

(9)

where u(t) and T are the time-dependent ground velocity and length of data, respectively.

To investigate the relation between ES and EP, we consider the relative seismic

efficiency ā which is given by ES/EP. Figure 11 is a contour expression of ā displayed

on (m, v) plane with a 3 dB interval. We observe that the contour runs approximately

parallel to the v-axis. If ES is directly proportional to EP, the distribution of ā should

be uniform on the (m, v) plane. The present result suggests that the seismic efficiency is

constant for varying impact velocities but can be improved with the increase of weight

mass. This fact may be explained by the scaling law that the velocity amplitude of the

ground surface is directly proportional to the impulsive force on collision of a weight.

Thus, ā can be represented as

Vol.38, No.3, 1990

200 S. Abe et al.

Fig.11. Distributions of the relative seismic efficiency (ā) displayed on the weight

mass (m) versus impact velocity (v) plane in contour representation.

(10)

where k3 and k4 denote the proportionality constants. This relation fu lly explains the
experimental results and is consistent with Mereu's result.

3.4 Amplitudespectrum
We shall discussthe fluctuation of amplitude spectra caused by the change of the
weight mass and the impact velocity. Figure 12 (a) is a contour representation of
amplitude spectra with a 3 dB interval in the frequency-weight-mass(f-m)domain for
the impact velocityof 5.94m/s. Figure 12(b)shows a similar representation of amplitude
spectra in the frequency-impact-velocity(f-v) domain for the weight masses of 50 kg.
In these contour representations,wenormalizedthe amplitudespectra by the momentum
of dropped weightjust before impact (Cb) and interpolated them by B-splinefunction.
If the seismicwaveformsare similar to each other regardless of source parameters and
if these velocityamplitudes are directlyproportional to Cb, then the normalized spectral
amplitude distributions in the (f-m) and (f-v)domains are approximately uniform in
the ordinate direction.
At first glance, the amplitude spectra do not seem to be affected by change of the
weight mass or by the impact velocity.However,a closer examination shows that the
low frequency components tend to be dominant over the high-frequency ones in the
case that the weight mass is much smaller (say 10 to 20kg) than the coupler mass
(100kg). As the impact velocity increases,the high-frequencycomponents tend to be
slightly dominant over the low-frequencyones. Summarizing these results, we can
consider that the low-frequency components tend to predominate over the high-

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 201

(a) (b)
Fig. 12. (a) Contour representation of the spectral-amplitude distribution
on the frequency versus weight-mass (m) plane. The impact velocity (v) was
fixed at 5.94 m/s: (b) Contour representation of the spectral-amplitude
distribution on the frequency versus impact velocity (v) plane. The weight
mass (n1)was fixed at 50 kg.

frequency ones if the momentum of dropped weight is small.

4. Effects of the Mass and Radius of Coupler on Seismic Waves

The most important reason for us to place a coupler on the impact point of weight
is to suppress the conversion of the initial source energy into the energy of plastic
deformation which results in nonlinear effects. Further, in regard to the control of the
source characteristics, it is of principal importance to determine the optimum parameters
of the coupler for an efficient excitation of seismic waves. In this chapter, we will change
the parameters of the coupler [mass (M), radius (R), and material(q)] while fixing those
of the weight (m, v) to investigate the effect of a coupler upon the generated seismic
waves from two viewpoints, i.e., velocity amplitude and amplitude spectrum.

4.1 Velocity amplitude

Figure 13 shows the dependence of the velocity amplitude (Vmax)upon M in four
cases. The source parameters of these cases were given in Section 2.2. The values plotted
at M=0 are those obtained when we did not use any coupler and dropped each weight
onto the naked soil. We obtained these data only for Cases A and D. From this figure,
we can find that Vmax'shave an overall tendency to decrease with the increase of M.
Referring to the data for evaluation of errors shown in Fig.2, the fluctuation of Vmax
for M seems to exceed the confidence levels. A closer examination reveals that Vmax
seems to increase greatly as the mass of the weight (m) approached that of the coupler

Vol.38, No.3, 1990

202 S. Abe et al.

Fig. 13. Dependence of the velocity amplitude (Vmax)upon the coupler mass (M).
The circles with centered dots denote the data obtained under the condition
that M is approximately equal to m. Case A : m=20 kg, v=5.60 m/s,
R=0.260 m, q: steel. Case B: m=20 kg, v=3.96 m/s, R=0.260 m, q: steel. Case
C: m=40 kg, v=5.60 m/s, R=0.260 m, q: steel. Case D: m=40 kg,
v=3.96m/s, R=0.260m, q: steel.

Table 1. Variation of amplitude (Vmax) with coupler area.

(M). The circles with a centered dot in Fig. 13 show the results when M is approximately
equal to m. Further, comparing Vmaxat M=0 with that at M=20 kg in Case A or that
at M=40 kg in Case D, we can see that the effect of using a coupler is pronounced.
However, when an unsuitable coupler is used, i.e., when M is much larger or smaller
than m, the effect of the coupler is reduced..
We examine the relation between Vmaxand the radius of coupler (R). Table 1 shows
the dependence of Vmaxupon R. The source parameters except R are fixed in the four
cases given in Section 2.2. It can be seen that Vmaxtends to decrease with the increase
of R in Cases 1 and 2. In contrast to this, Vmaxincreases greatly in Case 3 when R is
changed from 0.260 to 0.368 m, which is supposed to be caused by a permanent flection
of the coupler itself. If M is much smaller than m, the thickness of the cylindrical steel
coupler is very small in comparison with the radius of the dropped weight. Therefore,
the coupler cannot behave as a rigid body and the seismic signals cannot be consistent
since they are affected by the coupler's permanent deformation even if the source

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 203

Table 2. Variation of amplitud e (Vmax

) with the material of coupler.

parameters are kept constant. As for the relation between Vmax and R, we can note that
the contact area of a coupler on soil can be regarded as a point in comparison with
the wavelength of generated seismic waves. Accordingly, it is possible that the fluctuation
of Vmax by varying R is smaller and unremarkable in the far-field even if Vmax of the
direct wave in the near-field is affected considerably by R.
We examine how Vmax depends upon the elastic property of a coupler. In the
experiments, three kinds of material were chosen: steel, wood, and hard rubber. Table
2 shows the dependence of Vmax upon M for different kinds of couplers. It can be seen
that Vmax's for steel couplers are about the same as those for wooden couplers. On the
other hand, Vmax's for hard rubber couplers seem to be much smaller than those for
others, which suggests that the transformation of the source energy to the seismic energy
is inefficient if the collision of a weight is made against a soft coupler.

4.2 Amplitude spectrum

In order to distinguish the fluctuation of the amplitude spectrum caused by changing
source parameters of a coupler, we calculate spectral ratios for various cases to a
reference instead of individual raw spectra. Figure 14 (a) and (b) show the variation of
spectral ratio with the change in M for the Cases A and D, respectively. Since it was
found that Vmaxis greatly increased when M approximately equals m, we chose the
amplitude spectrum obtained under such a source condition as the reference. These
reference spectra are shown in the lower panels of Fig.14. The source parameters except
M are the same as those for the Cases A and D given in Section 2.2, respectively. In
regard to the shapes of the spectral ratio, we could find no significant tendency with
the change of M, but noted a slight difference of the spectra between the cases for
M>m and M<m in Case D: larger low-frequency components in the range of 10 to
25 Hz for M<m. It is probable that the condition m=M is a border between the linear
and nonlinear behaviors of a weight-dropping source. The effect of M on the spectral
ratio may be summarized as follows:
(1) In the case that m is larger than M, some couplers do not behave as rigid
bodies due to the nonlinear behavior (e.g., the flection). It is possible that some couplers
cannot prevent the soil around the source area from being permanently deformed when
M<m.
(2) In the case that m is larger than M, the dropped mass cannot rebound but

Vol.38, No.3, 1990

204 S. Abe et al.

(a)
Fig. 14. Variation of the spectral ratio
panels show the spectral ratios and
respectively. The reference chosen
approximately equal to m. (a) Case
Section 2.2).
(b)
with the change in M. Upper and lower
the amplitude spectra of reference data,
is the spectrum obtained when M is
A. (b) Case D. (for Cases A and D, see

keeps on contacting with the coupler. In other words, the physical condition on collision
of the dropped weight with the coupler may be essentially different between the cases
M>m and M<m.
As for the effect of other source parameters (i.e., R, radius of coupler; q, material
of coupler), first, no pronounced variation of the character of the amplitude spectrum
with the change of the radius of the coupler could be detected. Second, in regard to
the effect of the elastic property of a coupler, the lower-frequency components seem to
increase appreciably only in the case when hard rubber was used.

5. Dynamic model for weight-dropping source

From the experimental results presented in the foregoing chapters, it seems that a
critical condition between elastic and inelastic phenomena is realized when the coupler
mass, M, is approximately equivalent to the weight mass, m. The seismic amplitude
becomes maximum when m_??_Mand a slight difference in predominant frequency can
be detected between the cases M>m and M<m with the parameters of weight being fixed.
On the experimental results in the case of M<m, we can infer that great loss of
seismic energy should occur in the region of soil below the coupler which cannot behave
as a rigid body any more. In such a case, in order to construct a model of weight-dropping
source, we must accumulate more data to derive the empirical laws which describe the

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 205

nonlinear behavior of the coupler. However, since our data from the experiments are
not sufficient to discuss these effects, we will confine our efforts for seismic modeling

in this study to the elastic case only.

The experimental results, which can be explained within the elastic and linear limit,

i.e., in the case M>m, may be summarized as follows:

(1) If the parameters of a coupler are kept constant, the amplitude of the seismic
wave is directly proportional to the momentum of the weight before impact.

(2) Conversely, if the parameters of a weight are kept constant, the amplitude of
seismic wave tends to decrease with the increase of M.

In the mathematical modeling of a weight-dropping source, desirable conditions

for a model are as follows: first, the synthesized source wavelet should be derived in

the full-wave form, or the model should contain the information of both the seismic

amplitude and frequency contents. In addition, no abstract parameters, which cannot

be treated explicitly and physically such as the restitution coefficient introduced by

Mereu et al. (1963) or the damping constant introduced by Kasahara (1953) between

the coupler and the soil, should be contained in the model. Further, the source model

should be represented by the least possible number of parameters.

Bearing these desirable conditions for modeling in mind, we review existing studies

about the model of a weight-dropping source briefly and give some comments on them.

5.1 Review of weight-dropping source models

Only a few attempts have been made to theoretically investigate the generation of

seismic waves caused by a weight-dropping source. First, Mereu et al. (1963) applied

the collision problem of material particles to the weight dropping and proposed that

the amplitude of the seismic wave (Vmax) is given by

Vmax = C(1 + e)mv/(m + M), (11)

where e, v, and C are the restitution coefficient, the velocity of the dropped weight, and

the coupling constant, respectively. And C is expressed as

C=kMP, (12)

where k and p are constants. The value of p was determined to be 2/3 in Mereu's data. •

This model is rather simple and easy to understand. However, it represents only

the amplitude of seismic disturbances and it is impossible to state the dependence of

frequency characteristics upon the source parameters in more detail. It can be seen that

the evaluation of the restitution coefficient, e, is difficult. However, if the collision is

confined only in weight-coupler system ignoring the effect of soil, e can be calculated by

e = ƒ¢t•Eg(1+m/M)/(2v)+m/M, (13)

where Ģt and g are the duration of rebound shown in Fig. 6 and the acceleration of

gravity, respectively. In this formula, e depends on all the source parameters.

Furthermore, the formulation (12) is nothing but an empirical law, since the coupling

constant, p, must be determined for each medium after some experimental studies.

In contrast to the model above, two approaches which are designed to derive the

wavelet form from a weight-dropping source were presented. Kasahara (1953) attempted

Vol.38, No.3, 1990

206 S. Abe et al.

to explain the mechanism of the source by a simple mass-spring-dashpot model of one

degree of freedom. The velocity amplitude of an oscillating coupler, u(t), was calculated
by solving the response equation as follows:

(14)
where
(15)

(16)
and

(17)
v, E, and q are the velocity of dropped weight, spring, and damping constants between
the coupler and soil, respectively. In this study, the response characteristics of a
coupler-soil system were constructed by two visco-elastic elements, i.e., a dashpot and
a spring. However, there are some difficulties in estimating these parameters, since we
cannot treat the elastic property of the coupler and that of soil separately. Furthermore,
the problem becomes more complex if we take the effective mass of soil which is
concerned with the vibration of a coupler into consideration.
Mooney (1974) applied the Lamb's problem to the modeling of weight dropping
and synthesized the wavefield generated by an arbitrary force function. In this model,
the seismic wave which is produced by an impact of weight mass on the semi-infinite
elastic solid can be evaluated. However, in the case when a coupler is settled on the
impact point, the seismic wave cannot be calculated similarly because we have to consider
the coupling factor between the coupler and soil.

5.2 Simplifieddynamicmodel
The primary aim of our study in this sectionis to make a proposal for modeling
the weight-droppingsource. Under various approximations, the successiveeffectsthat
are involved in the source waveletwithin a fixed time gate can be envisagedby means
of the convolutional equation - (source wavelet)=(compressive-forcefunction)*(cou-
pling factor between coupler and soil)*(absorption)*(instruments), which we write
mathematicallyas
w(t)=f(t)*c(t)*a(t)*i(t), (18)
where * denotes convolution.
In this model, i(t) can be obtained by the electric impulse response of the total
recording system.However,there are difficultiesin estimatingthe factor of absorption,
a(t).Though the primaryfactor whichaffectsthe absorption of sourcewaveletis supposed
to be the plastic region created in the impact point, we cannot determine the size of
such region and the effective contact area of the coupler with the surface of soil.
Therefore, for simplicity, we confine our discussion to only the elastic behavior by
dropping a(t) from this model. Thus the source waveletis defined as

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 207

w(t)=f(t)*c(t)*i(t). (19)

We calculate the coupling factor c(t) first by solving a problem of forced vibration

of a rigid body on the semi-infinite elastic space. We assume that a coupler which is

regarded as a rigid cylindrical body is placed on the semi-infinite space and is subjected

to a vertical harmonic force. The displacement under the coupler is specified and the

remainder of the surface is kept stress-free. This condition leads to a mixed-boundary-

value problem represented by a dual integral equation, which can be reduced to a

Fredholm integral equation using a standard method (e.g., Robertson, 1966). Following
Robertson's procedure, we solved this equation and calculated the dynamical ground

compliance (f1+if2), which represents the relation between the displacement of coupler

(W) and the harmonic loading [F•Eexp(iwt)] by

(20)

where ƒÊ, R, and ƒÖ are the rigidity of the medium, the radius of the coupler, and the

angular frequency, respectively. Using this dynamical compliance, we solved the equation

of motion for the coupler and derived the amplitude, A(ƒÖ), and phase responses of

coupler, ƒÓ(ƒÖ), as follows (Bycroft, 1956):

(21)
(22)
where

(23)
(24)
(25)
and

Fig. 15. Non-dimensional amplitude responses (A(ƒÖ)ƒÊR/F) of the coupler to a

vertical vibrational force applied to the semi-infinite medium, as a function of

the normalized frequency (ƒÖR/VS). (for parameters, see Section 5.2)

Vol.38, No.3, 1990

208 S. Abe et al.

q=ba02; (26)

b and a0 are the reduced mass and the angular frequency, respectively. Figure 15 shows
the amplitude responses when M is set to be 10 and 100 kg.

We consider next the shape of the compressive-force function, f(t), which the weight

exerts on the coupler. It is difficult to estimate f(t) theoretically, since the problem of

collision between the weight and the coupler is hard to be solved dynamically even if

both the weight and the coupler were approximately two elastic spheres in contact.

Therefore, we assumed the following three types of representative source-time functions:

Case 1 : f(t)=C1mv•EƒÂ(t),

Case 2 : f(t)=C2mv•EƒÂ(t), (27)

Case 3 : f(t)=C3mv•Esin2(ƒÎt/Td) for 0<t<Td,

=0 otherwise,

where C1, C2, and C3 are constants which are independent of source parameters. The

parameters m, v, and Td are the weight mass, the impact velocity, and the duration of
impact, respectively. These functions are almost the same as those proposed by Mooney

(1974), who considered the Lamb's problem in which a vertical time-dependent point
force acts on the uniform half-space. In Case 3, we assumed a causal function instead

of a cosine-type one as used by Mooney (1974), since the sine-type function well

approximates the shape of compressive force which acts between two elastic spheres in

contact. The problem of the compressive force between two elastic spheres can be solved

by applying the Hertz stress theory (Timoshenko and Goodier, 1951) if the collision

problem is treated statically. In regard to the effects of source parameters for a dropped
weight upon the seismic wave, we supposed that f (t) is proportional to the momentum

of the dropped weight on impact. In other words, the parameters of the weight are

regarded only as a scaling factor.

Finally, using the estimated coupling factor and assumed compressive-force

function, a synthesized velocity response of the coupler, u(t), can be calculated by

differentiating the source wavelet, w(t). Figure 16 shows the comparison of normalized

synthetic waveforms calculated by our model for each case of assumed forces with the

observed waveform at the near-field which is corrected for i(t). The source parameters

are given as m=50 kg, v=4.85 m/s, M=100 kg, R=0.260 m, and q: steel. In the following

discussions, we use the values of C1, C2, C3, and Td determined under the above source

condition. In regard to Td, we chose the representative value (25 ms) by comparing the

modeled waveforms with the observed ones for various couplers. In this figure, we see

that the Case 3 is similar to the observed waveform if the appropriate duration of

impact is chosen. On the other hand, in Cases 1 and 2, the predominant frequencies

of synthetics seem to be much higher than that of observed waveform.

We investigate the relation between the maximum velocity amplitude (Vmax) and

the mass of coupler (M) for the synthesized waveforms. Figure 17 shows the dependence

of Vmax upon M for each source-time function. We assumed that the radius of the

coupler is 0.260 m and its material is steel. In Cases 1 and 2, Vmax's tend to decrease with

J. Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 209

Fig. 16. Synthesized velocity waveforms for three

functions (Cases 1 to 3) and the observed waveform.
types of
The
source-time
source parameters
are m=50 kg, v=4.85 m/s, M=100 kg, R=0.260 m, q: steel.
Case 1: f(t)=C1mv•EƒÂ(t).

Case 2: f(t)=C2mv•EH(t).

Case 3: f(t)=C3mv•Esin2(ƒÎt/Td) for 0<t<Td,

=0 otherwise.
(for parameters, see Section 5.2)

Fig. 17. Dependence of the velocity amplitude (Vmax)of synthesized source wavelet
upon the coupler mass (M) for three types of source-time functions (Cases 1 to
3). The numbers attached to the solid lines are the types of source-time
function shown in Fig.16.Vol.38, No.3, 1990
 210 S. Abe et al.

the increase of M. On the other hand, in Case 3, Vmax's tend to increase slightly.
Comparing these results with the experimental ones shown in Fig.13, the tendency of
Vmax-M relation for synthetics in Cases 1 and 2 agrees with that of the observed seimic
waveforms qualitatively. Further, a closer examination shows that the rate of decrease
of Vmax with increasing M in Case 2 is approximately equal to that of the observed
seismic waves.
Summarizing the above results, in regard to the source waveform, the synthesized
waveform of Case 3 agrees with the observation if the duration, Td, is chosen
appropriately. On the other hand, in regard to the relation between the maximum
velocity amplitude, Vmax, and the coupler mass, M, the synthetics in Case 2 have the same
tendency as in the observation. In conclusion, the experimental results could not perfectly
be explained by a simplified dynamic model at the present stage. However, we can point
out two notable defects in these models in order to give some insight for devising a more
realistic model of weight-dropping source.
First, it does not seem feasible to calculate an appropriate synthetic waveform
using the given compressive-force functions. Therefore, it is more realistic to estimate
the source-time function by solving the inverse problem, i.e., by deriving the function
from the observed wavelet and the theoretical response of the coupler. Then we should
examine the dependence of the extracted force function upon the parameters of source
statistically.
Second, the coupling factor between the coupler and soil is influenced by the
strength of compressive force. Therefore, the coupling factor cannot be represented by
a linear function of compressive force. Moreover, the absorption is not negligible in
estimating the source wavelet. If we take both nonlinear effects into consideration, we
cannot reduce the model of source wavelet to the convolutional one anymore.

6. Conclusions

In order to investigate the factors that influence the characteristics of seismic

waveform by dropped-weight-coupler system, we have made some field experiments
and derived a method to synthesize the velocity response of the coupler. The following
conclusions are drawn:
(1) In the case when a weight bounces on a coupler, the momentum of the weight
after impact is directly proportional to that before impact. Accordingly, the momentum
which the source area acquires through impact depends linearly upon that of the dropped
weight before impact because the law of conservation of momentum holds.
(2) In the case when the parameters of the coupler are fixed, the velocity amplitude
of a seismic wave is directly proportional to the momentum that the source area acquires
from the dropped weight. This scaling law explains the experimental result that the
seismic efficiency does not vary with varied impact velocities but can be improved with
the increase of weight mass while the potential energy of the source is kept constant.
(3) In the case when the parameters of the weight are fixed, the velocity amplitude
of a seismic wave can be greatly increased as the weight mass, m, approaches that of
the coupler, M, and, for larger M than m, it tends to decrease with the increase of M.
(4) In regard to the shapes of seismic spectrum, we found some differences among

J, Phys. Earth
 Seismic Characteristics of the Weight-Dropping Source 211

cases depending on whether M>m or M<m. The low-frequency components seem to

be slightly dominant over the higher-frequencyones when the momentum of dropped
weight is small. The change of the amplitude spectrum with source parameters is
consistentwith the result that the maximum values of the normalized cross-correlations
(rij) between seismic traces can be represented by a positive linear function of the
momentum of dropped weight in the case of M>m.
(5) We reviewed some studies of the modeling of weight-dropping source and
proposed a simplifieddynamicmodel whichis composed of a compressiveforce, coupling
factor, and response of instruments. The synthetics by this model are not always in
good agreement with the observed waveforms with respect to the amplitude and the
frequency characteristics. In view of these disagreements, we infer that not only the
magnitude of compressiveforce but also its duration is controlled by parameters of the
source, and that the coupling factor between the coupler and soil is influenced by
inelasticity which can arise depending on the strength of the force. It will be necessary
to solve these difficult problems in order to construct a more realistic model for
weight-dropping source.

We would like to express our appreciation to Profs. Soji Yoshikawa, Kojiro Irikura, Drs.
Fumio Amaike and Masanori Horike for their encouragement during the course of this study
and critical comments on the manuscript. Thanks are also due to Mr. Toru Kuroda for valuable
discussions and helpful suggestions, and to Dr. Nobuyoshi Baba for critically reading the
manuscript. We are also indebted to Mr. Hideo Yukutake for providing us with his equipment,
to Mr. Tomoki Tsutsui for his aid in operation of data acquisition systems, to Mr. Shinzo Fujii
for his assistance in assembling hardware used in the experiments, and to Messrs. Takafumi Idei,
Takashi Imazumi, Takao Kagawa, Mitsuyuki Hoshiba, Eiichi Fukuyama, Tomotaka Iwata, and
other colleagues for their assistance in the experiments. All computations were run on the FACOM
M-382 at the Data Processing Center of Kyoto University.

REFERENCES

Baeten, G. J. M, J. T. Fokkema, and A. M. Ziolkowski, Seismic vibrator modelling, Geophys.

Prospect., 36, 22-65, 1988.
Blake, F. G, Spherical wave propagation in solid media, J. Acoust., Soc. Am., 24,211-215, 1952.
Bycroft, G. N., Forced vibrations of a rigid circular plate on a semi-infinite elastic space and on
an elastic stratum, Philos. Trans. R. Soc. London, Ser. A, 248, 327-368, 1956.
Kasahara, K., Experimental studies on the mechanism of generation of elastic waves IV, Bull.
Earthq. Res. Inst., Univ. Tokyo, 32, 67-77, 1953.
Kobori, T., R. Minai, T. Suzuki, and K. Kusakabe, Dynamic characteristics of soil-foundation
systems detected from oscillator tests. Part 1. Nonelastic dynamic characteristics, Ann. Disas.
Prey. Res. Inst., Kyoto Univ., 17B, 115-138, 1974 (in Japanese with English abstract and
captions).
Mereu, R. F., R. J. Uffen, and A. E. Beck, The use of a coupler in the conversion of impact
energy into seismic energy, Geophysics, 28, 531-546, 1963.
Mooney, H. M, Some numerical solutions for Lamb's problem, Bull. Seismol. Soc. Am., 64,
473-491, 1974.
Robertson, I. A., Forced vertical vibration of a rigid circular disc on a semi-infinite elastic solid,

Vol.38, No.3, 1990

212 S. Abe et al.

Proc. Cambridge Philos. Soc., 62, 547-553, 1966.

Safar, M. H., The radiation of acoustic waves from an air-gun, Geophys. Prospect., 24, 756-772,
1976.
Sharpe, J. A., The production of elastic waves by explosion pressures. I. Theory and empirical field
observations, Geophysics, 7, 144-154, 1942.
Timoshenko, S. P. and J. N. Goodier, Theory of Elasticity, McGraw-Hill, New York, 1951.
Tsutsui, T., S. Abe, and Y. Kobayashi, Data acquisition system for seismic prospecting using
personal computer system, Butsuri-Tansa, 39, 106-113, 1986 (in Japanese with English
abstract and captions).
White, J. E., Underground Sound, Elsevier, Amsterdam, 1983.
Ziolkowski, A. and W. E. Lerwill, A simple approach to high resolution seismic profiling for
coal, Geophys. Prospect., 27, 360-393, 1979.

J. Phys. Earth