Hi Wanderson,

The answer depends on which is your use of the Ricker wavelet (RW); for example, you could use RW to simulate the time variation of a (point) seismic source or you could convolve the RW with a seismic section, made up of dirac spikes and representing subsurface reflections in order to generate a synthetic seismogram. If I recall correctly (I haven't read the paper ina while) it mainly investigates this second case (check the attachment).

In the section "conclusions" you can read that the reasons not to use this family of wavelet are (just made copy and paste):

1. They are not observed in nature, being based on the

wrong theory for seismie attenuation.

2. As 'standard' computable wavelets for rough and

ready synthetic seismograms, they suffer from a con-

fusion of definition-there

are several types and

several 'conventions' for specifying them.

3. Having only a single parameter

definition makes

them too inflexible for synthetic seismogram applica-

tion.

4. The Ricker types which show the most similarity

with wavelets encountered in synthetic seismogram

work may have two loops, three loops or be inter-

mediate fractional-derivative

wavelets.

5. Minimum-phase Ricker wave lets suffer from a delay

which depends on the choice of sample interval. Such

a delay can be subtracted out in an arbitrary way,

but again no standard convention exists for doing

this.

Hope this was useful.

S.

Abot point source generation: During a MASW survey, just for my curiosity, I decided to test if a slammerhand generated signal recorded 2 meters far from the source (where geophone was not saturated) was actually comparable to a RW. I found that it was possible to decompose the recorded signal with a superposition of 2/3 RW with different central frequencies.

Hi Samuel,

Thank you for your answer and for the paper. I will read this article as soon as possible. I wondered about the statement number 3 but the others are new to me.

Can I ask you if you know any free existing software where I can play around to simulate seismic acquisition and the influence of its parameters and basic processing (despite of SU)?

Your experiment with the source generator brings me another kind of question. What control the shape of the wavelet in Earth? For example, should I expect to find a sharper wavelet in a well consolidated sediment rather than in a poorly sedimented region?

Thank you again for your valuable information.

Best,

Wanderson

About synthetics:

You can generate very good quality synthetic seismograms using SPECFEM both in 2D or 3D approximatio. If you set correctly the program, the seismograms will look like a reflection seismic survey acquisition.

In this program the source can be set to be a Ricker wavelet, ( you can  set source properties such as amplitude onset time and central frequency)

You also can buid a more involved source time function using different sources with same location but different properties.

You can also take a look to HERRMANN's "computer programs in seismology".

About signal processing:

I guess "seismic unix" could be a good (FREE) candidate even if is not so user friendly.

Mhhh, I would expect, considering body waves, in homogeneus isotropic elastic media,

since Velocity = f * wavelength,

that (keeping fixed frequency), the faster the wave, the longer the wavelenght. I.e. It should be sharper in unconsolidated, with no change in shape.

For non-elastic soils, expecially if some frequency-dependent effect comes into play I would expect a change in shape.

Just a guess, though.

Good day

I attach to you the paper of Hosken (1988). Hopping it could help your answer.

Regards