On 15 Mar, 15:33, Vladimir Vassilevsky <nos...@nowhere.com> wrote:
> Rune Allnor wrote:
> > Again, I've messed with that sort of stuff in the past.
> > I am not tempted to reacquaint myself with these things.
>
> BTDT with electromagnetic problems :))))
>
> But, what is the point in solving the problem if it is soluable?

?????

My impression is that a lot of people think that because
one can express a chain of cause(s) + effect(s) that somehow
explain a certain scenario, one can also 'undo' a number
of destructive or unwanted effects.

My favourite example to show why this is erratic thinking,
is plain ol' Newtonian gravity: The formula is simple - it
explains the effects of gravity - but one can not use it
to remove gravity in the real world:

- Aerospace systems will have to deal with weight,
  irrespective of the gravity formula.
- Floating vessels will have to deal with stability
  issues irrespective of the gravity formula.
- Building foundations will have to be dimensioned to
  the loads they are supposed to support, irrespective
  of the gravity formula.

Don't get me wrong - the gravity formula is essential to
describe each individual scenario, and is key to be able
to dimension the systems correctly. The formula only
allows people to *handle* the gravity; not *alter* it.

But then - I'm merely an engineer who wants to make
systems that actually work.

"Making systems that actually work" is by no means the
same as "getting an income". It is my impression that
a lot of people go with these inverse methods because
it is impossible to prove thay they *can't* work. A
lacking result in this test can always be blamed on
some missing or poorly handled detail. So it is always
possible to obtain more funding. Prticularly if this is
a long-running project that has already soaked immense
amounts of $$$.

Oh well.

Rune

>
> It is well-known that seismic waves travelling through the
> earth experience frequency-dependent attenuation. It's the
> same effect people hear in thunderstorms or near shooting
> ranges: Thunders or shots that are set off close to the
> listener are percievet as sharp 'snappish' bangs, while
> thunders or shots that are set off at a distance is percieved
> as a low rumbling.

Yes - this is a good explanation of the physics.  I couldn't have 
written it better myself.

>
> The difference is that the higher frequency bands are
> attenuated faster than the lower frequency bands, thus
> distorting the transient as a function of travelled
> distance.
>
> This effect has a huge impact on matched-filter processing
> in seismics: What are the chances of detectning the pulse
> waveform if the filter is set up to match the original
> emitted pulse, while the recieved pulse has undergone
> significant frequency-dependent attenuation?
>
> This is the question the constant Q algorithms attempt to
> answer.
>
> The problem is - as always - that one needs some sort of
> analytic model of this frequency-dependent attenuation
> in order to incorporate it in the processing. While the
> analytic model might be simple, the reality is not.
> Just about anything you will find in the real world will
> impose some sort of frequency-dependent response:
>
> - Grain&  pore sizes in sedimentary rocks - frequency
>    dependent response
> - Rough surfaces on boundaries between rocks - frequency
>    dependent response
> - Micro-layered anisotropy in sedimentary rocks - frequency
>    dependent response
> - Micro-cracks in rocks - frequency dependent response
> - Volume inhomogeneities in rocks - frequency dependent
>    response
> - Fluid contents in pores - frequency-dependent response
> ---
> - Reflections off sequences of rock layers - frequency-
>    dependent response
> - Conversions between wave types - frequency-dependent
>    response
>
> So if one wants to start messing with these things, one
> has to be aware that *all* the factors influence *everything*
> *all* the time. As always, it's the one you don't see
> (or ignore) that will eventually take you down.

Seismology (or the related field of non-destructive testing, NDT) can 
sometimes get very complicated.  That's the reason why I do more 
research in NDT rather then seismology.  IMHO, the distances over which 
measurements are taken are smaller and the physics is simpler.  But even 
NDT poses a number of significant challenges.  Nothing is ever really 
simple.

>
>   All the items on the list above the '---' separator (and
> presumably a number of other items) are 'obvious' factors
> that affect the Q factor of the recieved pulse. Don't be
> surprised if you find a number of such factors discussed
> in the literature, with associated analytic or empirical
> models for the Q factor.
>
> As far as I am concerned, it is the latter two factors,
> reflections and conversions between waves, that are most
> likely to throw you off. These are not considered Q factors,
> but dominate the propagation path of the seismic energy.
> In other words, you will have to get those factors ~100%
> right before it makes sense to discuss the other, minor
> factors on the list.
>

Thanks Rune--

Nicholas


Rune Allnor wrote:

> Again, I've messed with that sort of stuff in the past.
> I am not tempted to reacquaint myself with these things.

BTDT with electromagnetic problems :))))

But, what is the point in solving the problem if it is soluable?

> 
> Rune

VLV

On 14 Mar, 20:54, Rune Allnor <all...@tele.ntnu.no> wrote:

> Just browsing the table of contents of the book,
>
> http://www.amazon.com/Seismic-Inverse-Filtering-Yanghua-Wang/dp/14051...
>
> sends cold shivers down my spine. The author's premise seems
> to be that there exists 'a' mathemathical model that influences
> the Q factor of the trace. Then he goes on to list almost a
> dozen different models for the effect (see the list of subchapters
> in chapter 3). That alone sets off all the alarms in my mind.
>
> I used to do those kinds of things ages ago and my experience
> is that the earth is a random medium in every sense of the
> word. Unless you sample it directly - drill a hole and
> dig stuff out of it - one does not know much. While any of
> the proposed models might be defended in one setting, the
> reality is that there are gazillion other situations in the
> same survey where the model is invalid.

OK, maybe I should elaborate a bit on these things and why
my spine goes cold.

It is well-known that seismic waves travelling through the
earth experience frequency-dependent attenuation. It's the
same effect people hear in thunderstorms or near shooting
ranges: Thunders or shots that are set off close to the
listener are percievet as sharp 'snappish' bangs, while
thunders or shots that are set off at a distance is percieved
as a low rumbling.

The difference is that the higher frequency bands are
attenuated faster than the lower frequency bands, thus
distorting the transient as a function of travelled
distance.

This effect has a huge impact on matched-filter processing
in seismics: What are the chances of detectning the pulse
waveform if the filter is set up to match the original
emitted pulse, while the recieved pulse has undergone
significant frequency-dependent attenuation?

This is the question the constant Q algorithms attempt to
answer.

The problem is - as always - that one needs some sort of
analytic model of this frequency-dependent attenuation
in order to incorporate it in the processing. While the
analytic model might be simple, the reality is not.
Just about anything you will find in the real world will
impose some sort of frequency-dependent response:

- Grain & pore sizes in sedimentary rocks - frequency
  dependent response
- Rough surfaces on boundaries between rocks - frequency
  dependent response
- Micro-layered anisotropy in sedimentary rocks - frequency
  dependent response
- Micro-cracks in rocks - frequency dependent response
- Volume inhomogeneities in rocks - frequency dependent
  response
- Fluid contents in pores - frequency-dependent response
---
- Reflections off sequences of rock layers - frequency-
  dependent response
- Conversions between wave types - frequency-dependent
  response

So if one wants to start messing with these things, one
has to be aware that *all* the factors influence *everything*
*all* the time. As always, it's the one you don't see
(or ignore) that will eventually take you down.

 All the items on the list above the '---' separator (and
presumably a number of other items) are 'obvious' factors
that affect the Q factor of the recieved pulse. Don't be
surprised if you find a number of such factors discussed
in the literature, with associated analytic or empirical
models for the Q factor.

As far as I am concerned, it is the latter two factors,
reflections and conversions between waves, that are most
likely to throw you off. These are not considered Q factors,
but dominate the propagation path of the seismic energy.
In other words, you will have to get those factors ~100%
right before it makes sense to discuss the other, minor
factors on the list.

Again, I've messed with that sort of stuff in the past.
I am not tempted to reacquaint myself with these things.

Rune

>
> Just browsing the table of contents of the book,
>
> http://www.amazon.com/Seismic-Inverse-Filtering-Yanghua-Wang/dp/1405185406/ref=sr_1_1?ie=UTF8&s=books&qid=1268595741&sr=8-1#reader_1405185406
>
> sends cold shivers down my spine. The author's premise seems
> to be that there exists 'a' mathemathical model that influences
> the Q factor of the trace. Then he goes on to list almost a
> dozen different models for the effect (see the list of subchapters
> in chapter 3). That alone sets off all the alarms in my mind.
>
> I used to do those kinds of things ages ago and my experience
> is that the earth is a random medium in every sense of the
> word. Unless you sample it directly - drill a hole and
> dig stuff out of it - one does not know much. While any of
> the proposed models might be defended in one setting, the
> reality is that there are gazillion other situations in the
> same survey where the model is invalid.

Okay, thank you for your post, Rune.

Nicholas

On 14 Mar, 19:10, Nicholas Kinar <n.ki...@usask.ca> wrote:
> Hello Rune--
>
> Thank you for your response.
>
>
>
> > Seismic traces are usually modeled using some Finite
> > Difference or Ray Tracing scheme. Either way, the
> > constant Q question is irrelevant, as all kinds of
> > physical effects mess the signal up.
>
> > Decide on a Ricker wavelet and use it. No need to
> > fuzz things up with Q values; you will have more
> > than enough other stuff to worry about.
>
> Yes - you are right about the physical effects, and this is what I would
> normally do if I wanted a "realistic" seismic trace: use
> finite-difference or ray-tracing algorithms.
>
> However, I am reading a book "Seismic Inverse Q Filtering" by Y. Wang,
> and the author describes how to build an extremely simple seismic trace
> with a known Q value to test an inverse seismic Q filtering algorithm.
>
> The idea is to generate a (non-realistic) simple seismic trace. &#4294967295;I've
> tried to implement this algorithm, but I find that some of the salient
> details of the implementation are tricky.
>
> Could I send you some further information via e-mail to read?

No.

>&#4294967295;Then
> perhaps we could summarize on the newsgroup what would be required to
> fully implement this synthetic algorithm.

Just browsing the table of contents of the book,

http://www.amazon.com/Seismic-Inverse-Filtering-Yanghua-Wang/dp/1405185406/ref=sr_1_1?ie=UTF8&s=books&qid=1268595741&sr=8-1#reader_1405185406

sends cold shivers down my spine. The author's premise seems
to be that there exists 'a' mathemathical model that influences
the Q factor of the trace. Then he goes on to list almost a
dozen different models for the effect (see the list of subchapters
in chapter 3). That alone sets off all the alarms in my mind.

I used to do those kinds of things ages ago and my experience
is that the earth is a random medium in every sense of the
word. Unless you sample it directly - drill a hole and
dig stuff out of it - one does not know much. While any of
the proposed models might be defended in one setting, the
reality is that there are gazillion other situations in the
same survey where the model is invalid.

Rune

Hello Rune--

Thank you for your response.

>
> Seismic traces are usually modeled using some Finite
> Difference or Ray Tracing scheme. Either way, the
> constant Q question is irrelevant, as all kinds of
> physical effects mess the signal up.
>
> Decide on a Ricker wavelet and use it. No need to
> fuzz things up with Q values; you will have more
> than enough other stuff to worry about.

Yes - you are right about the physical effects, and this is what I would 
normally do if I wanted a "realistic" seismic trace: use 
finite-difference or ray-tracing algorithms.

However, I am reading a book "Seismic Inverse Q Filtering" by Y. Wang, 
and the author describes how to build an extremely simple seismic trace 
with a known Q value to test an inverse seismic Q filtering algorithm.

The idea is to generate a (non-realistic) simple seismic trace.  I've 
tried to implement this algorithm, but I find that some of the salient 
details of the implementation are tricky.

Could I send you some further information via e-mail to read?  Then 
perhaps we could summarize on the newsgroup what would be required to 
fully implement this synthetic algorithm.

Thanks Rune.

On 14 Mar, 18:23, Nicholas Kinar <n.ki...@usask.ca> wrote:
> Hello--
>
> I would like to generate a synthetic seismic trace using the real-valued
> Ricker wavelet. &#4294967295;(Other wavelets could also be used.) &#4294967295;Normally I think
> that this should involve a convolution operation. &#4294967295;However, I would like
> to ensure that the synthetic trace has a constant (and known) Q-value.
>
> I would like to find a monograph/book/paper or some other type of
> procedure to efficiently generate this synthetic trace. &#4294967295;Any suggestions?

Seismic traces are usually modeled using some Finite
Difference or Ray Tracing scheme. Either way, the
constant Q question is irrelevant, as all kinds of
physical effects mess the signal up.

Decide on a Ricker wavelet and use it. No need to
fuzz things up with Q values; you will have more
than enough other stuff to worry about.

Rune