where expertise comes together - since 1996

Web's Largest Portal of Nondestructive Testing (NDT)
Open Access Database (Conference Proceedings, Articles, News), Exhibition, Forum, Network

All Forum Boards
Technical Discussions >
SSP does it work?
Career Discussions
Job Offers
Job Seeks
Classified Ads
About NDT.net
Articles & News

VisiConsult Systems & Solutions GmbH
VisiConsult X-ray Systems & Solutions – the leading specialist for customized X-ray systems, portable DR equipment and automation solutions.

633 views
06:46 Nov-24-2004
Pim van Andel
SSP does it work?

The objective of (C)SSP is to reduce scattering noise for improved flaw detection. This could be a valuable tool for course grained materials, composites, etc. Perhaps you would use it if:

- the signal to noise ratio improves by at least 10dB
- if it is reliable
- if it can be cost effectively implemented in ultrasonic instrumentation such that SSP filtering can be done in real time on a A-scan with a PRF of e.g. 25Hz or higher.

I am not yet convinced, though there are many publications on the subject.

I have been testing the SSP modeling software that was made available on ndt.net by Philippe Rubbers (thanks Philippe).
The diffraction model is easy to understand. I have more problems with the CSSP pulse propagation model:
- what kind of noise is present in this model?
- can we make conclusions on signal to noise ratios with this model?

Anyway, I highly appreciate the initiative of making software like this available.



 
09:35 Nov-25-2004

Philippe Rubbers

Engineering
SCM,
South Africa,
Joined Nov 1998
21
Re: SSP does it work? Dear Sir

Concerning CSSP, and processing of ultrasonic signals:
CSSP does work for improving signal to noise ratios however there are a few things that a user needs to be aware of, which are not immediately evident:
1) The probe bandwidth must be high. For <80% bandwidth probe, the improvement is marginal, with only the edge wave being affected (fewer mode conversion signals in complex geometries). <80% bandwidth will therefore not make a visible improvement in coarse grained materials. Between 80% and 110% bandwidth the edge wave is progressively removed, and the main wave front is narrowed, improving resolving power and slightly improving angular resolution. At bandwidths >110% the signal to noise improvement becomes quite evident, and I have managed to find flaws which were 6dB below the noise floor before processing (>12dB improvement in S/N).
2) The choice of filters is crucial. I do not believe in the huge number of filters used in most literature on SSP, and find excellent results with just 6 well selected filters.
3) The data needs to be oversampled to be filtered and processed correctly. I have not yet been able to create analog filters for CSSP, and all my filtering is done digitally, requiring huge processing power.

Concerning the performance, yes there is a huge processing requirement, and my 5 year old system (hardware has not changed in 5 years) can process an A-Scan (16384 data points) in 15ms. This is a PRF of 66.... not too bad for manual UT, but pathetic for phased array or mechanised UT....

Concerning the simulation software, this was actually written (to explain to myself) why the S/N improvement was so dependant on bandwidth (which the software confirmed) and to see what CSSP actually does to a UT signal in fine grained material. Following the interest in the simulation software, I have shared it on NDT.net, and I would like to thank NDT.net for the exposure they have given to this tool (which is still a work in progress). Note that the software performs CSSP not SSP.

Concerning the cost effectiveness of CSSP.... well this is similar to phased array, as hardware improved and volumes increase, the cost effectiveness increases.
Time will tell.

The simulation software does not add any simulated grain noise to the signal, however I believe that the edge wave is an undesirable effect (causing a second wave front), and imperfections in the probes crystal will also create signals between the 1st and 2nd wave front, which are also undesirable. I have called these undesirable signals ‘noise’ for the purpose of the software, however some academics have objected to this terminology, which is criticism I can live with.

We cannot make direct conclusions on how CSSP will improve Signal to Noise ratios in coarse grained materials from the simulation software. We can however see how it improves depth and angular resolving power for fine-grained materials. As for all validation exercises, first you need to know what you can’t do before you look for what you can do, otherwise you may find yourself going down the wrong path.

Thanks for the interest in CSSP and in my simulation software.

Best regards


----------- Start Original Message -----------
: The objective of (C)SSP is to reduce scattering noise for improved flaw detection. This could be a valuable tool for course grained materials, composites, etc. Perhaps you would use it if:
: - the signal to noise ratio improves by at least 10dB
: - if it is reliable
: - if it can be cost effectively implemented in ultrasonic instrumentation such that SSP filtering can be done in real time on a A-scan with a PRF of e.g. 25Hz or higher.
: I am not yet convinced, though there are many publications on the subject.
: I have been testing the SSP modeling software that was made available on ndt.net by Philippe Rubbers (thanks Philippe).
: The diffraction model is easy to understand. I have more problems with the CSSP pulse propagation model:
: - what kind of noise is present in this model?
: - can we make conclusions on signal to noise ratios with this model?
: Anyway, I highly appreciate the initiative of making software like this available.
------------ End Original Message ------------




 
03:07 Nov-25-2004
Pim van Andel
Re: SSP does it work? Philippe,

Thanks for your explanation. I had not considered the spatial resolution effects.

Now suppose that I want to inspect coarse grained material and I want to select the best transducer. Without CSSP I would select a broadband transducer, because it has a short pulse length and that should improve S/N according to:
http://www.ndt-ed.org/EducationResources/CommunityCollege/Ultrasonics/Physics/signaltonoise.htm
Then when you add CSSP the S/N ratio would improve even further. Thats fine. Intuitively however you may take a narrow band transdurer to have a stronger signal. And for thick components you may not have enough electronic amplification.

You may also loose consideral amount of bandwidth due to the increase of attenuation with frequency.

I find 15ms for a CSSP filtering on a 16K A-scan quite encouraging.

Seems to me that it becomes important to thoroughly test and demonstrate the robustness and reliability of CSSP for various applications.

Regards,
Pim van Andel

----------- Start Original Message -----------
: Dear Sir
: Concerning CSSP, and processing of ultrasonic signals:
: CSSP does work for improving signal to noise ratios however there are a few things that a user needs to be aware of, which are not immediately evident:
: 1) The probe bandwidth must be high. For <80% bandwidth probe, the improvement is marginal, with only the edge wave being affected (fewer mode conversion signals in complex geometries). <80% bandwidth will therefore not make a visible improvement in coarse grained materials. Between 80% and 110% bandwidth the edge wave is progressively removed, and the main wave front is narrowed, improving resolving power and slightly improving angular resolution. At bandwidths >110% the signal to noise improvement becomes quite evident, and I have managed to find flaws which were 6dB below the noise floor before processing (>12dB improvement in S/N).
: 2) The choice of filters is crucial. I do not believe in the huge number of filters used in most literature on SSP, and find excellent results with just 6 well selected filters.
: 3) The data needs to be oversampled to be filtered and processed correctly. I have not yet been able to create analog filters for CSSP, and all my filtering is done digitally, requiring huge processing power.
: Concerning the performance, yes there is a huge processing requirement, and my 5 year old system (hardware has not changed in 5 years) can process an A-Scan (16384 data points) in 15ms. This is a PRF of 66.... not too bad for manual UT, but pathetic for phased array or mechanised UT....
: Concerning the simulation software, this was actually written (to explain to myself) why the S/N improvement was so dependant on bandwidth (which the software confirmed) and to see what CSSP actually does to a UT signal in fine grained material. Following the interest in the simulation software, I have shared it on NDT.net, and I would like to thank NDT.net for the exposure they have given to this tool (which is still a work in progress). Note that the software performs CSSP not SSP.
: Concerning the cost effectiveness of CSSP.... well this is similar to phased array, as hardware improved and volumes increase, the cost effectiveness increases.
: Time will tell.
: The simulation software does not add any simulated grain noise to the signal, however I believe that the edge wave is an undesirable effect (causing a second wave front), and imperfections in the probes crystal will also create signals between the 1st and 2nd wave front, which are also undesirable. I have called these undesirable signals ‘noise’ for the purpose of the software, however some academics have objected to this terminology, which is criticism I can live with.
: We cannot make direct conclusions on how CSSP will improve Signal to Noise ratios in coarse grained materials from the simulation software. We can however see how it improves depth and angular resolving power for fine-grained materials. As for all validation exercises, first you need to know what you can’t do before you look for what you can do, otherwise you may find yourself going down the wrong path.
: Thanks for the interest in CSSP and in my simulation software.
: Best regards
:
: : The objective of (C)SSP is to reduce scattering noise for improved flaw detection. This could be a valuable tool for course grained materials, composites, etc. Perhaps you would use it if:
: : - the signal to noise ratio improves by at least 10dB
: : - if it is reliable
: : - if it can be cost effectively implemented in ultrasonic instrumentation such that SSP filtering can be done in real time on a A-scan with a PRF of e.g. 25Hz or higher.
: : I am not yet convinced, though there are many publications on the subject.
: : I have been testing the SSP modeling software that was made available on ndt.net by Philippe Rubbers (thanks Philippe).
: : The diffraction model is easy to understand. I have more problems with the CSSP pulse propagation model:
: : - what kind of noise is present in this model?
: : - can we make conclusions on signal to noise ratios with this model?
: : Anyway, I highly appreciate the initiative of making software like this available.
------------ End Original Message ------------




 


© NDT.net - The Web's Largest Portal of Nondestructive Testing (NDT) ISSN 1435-4934

Open Access Database, |Conference Proceedings| |Articles| |News| |Exhibition| |Forum| |Professional Network|