Overview: Principles of Full Matrix Capture (FMC) and Total Focusing Method (TFM) in Ultrasonic Inspections

Fundamental principles of FMC and TFM
technologies
Training Session
June 2017
11© Zetec Inc. - Proprietary and Confidential

22
Agenda
• FMC / TFM Principles
• FMC Characteristics
• TFM Characteristics
• Codes, Applications, and Limitations
© Zetec Inc. - Proprietary and Confidential

Basic Principles

44
Standard Phased Array Ultrasonic Testing
(PA UT)
PA UT technology uses multiple independent UT transmitting-receiving
channels for acoustic beamforming
Only the summed and digitized A-Scan signal is transferred to the
computer for display and recording; individual raw signals are not saved

55
Standard PA UT
Process:
1. Delay laws are generated
2. Instrument pulses every relevant probe element, using the delay defined by
the focal law
3. Inside wedge and component, energy from each probe element is summed
together, creating constructive and destructive interference
4. Instrument digitizes signal received back on each relevant probe element
5. Instrument performs a summation of the signals from each element, using
the delay defined by focal law
6. End result is a summed and digitized A-Scan
7. Process is repeated for every focal law to generate a complete sweep,
displayed as e.g. Sector Scan

66
Standard PA UT
Standard PA UT – 40 to 70 SW
• Standard PA UT, sector scan with
preprogrammed focal laws (beams)
• “Live image” during inspection

77
Terminology
• Focal Law
– Set of delays that describe how the signal from an individual probe element must be
offset to create a constructive interference along a given path and focal point.
• Elementary A-Scan
– A-Scan signal from a single pulsed element received on a single receiving element.
• Summed A-Scan
– Traditional phased array processing where elementary signals from all elements are
summed together using a focal law to generate an A-Scan.
• Sweep
– Group of summed A-Scans displayed in an organized way, e.g. Sectorial, Linear or
Compound Sweep.
• Dynamic Depth Focusing (DDF)
– Variation of a focal law; a single focal point is still used for pulsing but during reception,
delay laws are adjusted dynamically to focus at multiple depths at once.

88
FMC (Full Matrix Capture)
Full Matrix Capture (FMC) consists of capturing and recording A-Scan
signals from every transmitter-receiver pair in the array
From raw A-scan signals stored on a drive, it is possible to generate UT
imaging for any given focal law / beam (aperture, angle, focus depth),
and for improved algorithms (e.g. TFM) through post processing

99
FMC + Data Reconstruction
Process
1. Delay laws are generated, but not used during data collection
2. FMC, for each element in the probe, the instrument:
– pulses the element
– digitizes the elementary A-Scan from each pulser/receiver combination
3. FMC data can be processed in the instrument or transferred to host
computer for processing there
4. Set of FMC data is read and processed
– Using standard focal law to reconstruct standard phased array beam sweep
– Using TFM focal law to reconstruct TFM frame
– Using other innovative algorithms …
– Same set of FMC data can be processed multiple times to produce different
results, using different reconstruction parameters

1010
Focal Law Reconstruction
Software Summation – 40 to 70 SWSoftware Summation – -10 to 80 LW
• FMC elementary A-scan data
• Post-processing of focal laws
• Generate signal and view equivalent to standard
Phased Array

1111
TFM Reconstruction
TFM – Frame LW
• FMC elementary A-scan data
• TFM post-processing
• Generate signal and view different than standard
Phased Array

Total Focusing Method (TFM)
TFM – Frame LW
• TFM: A delay law is created
for every pixel in the TFM
frame
• Summation of constituting
FMC elementary A-Scans is
done for every pixel, using
local focal law
• Results: Every pixel has a
dedicated focal law perfectly
focused to its location

1313
Terminology
• Full Matrix Capture (FMC)
– Data collection technology in which every element is pulsed in sequence, and where
elementary A-Scan data is collected for each combination of pulsing and receiving
elements
• Frame
– Typical output of a TFM method; a rectangular area (“frame”) where each point in the grid
is displayed as a pixel
• Total Focusing Method (TFM)
– Processing method using FMC collected data to generate a frame of pixels where each
pixel is computed using a dedicated focused focal law
• Sectorial Total Focusing (STF)
– Also named “Almost-Total Focusing Method” (A-TFM)
– Processing method using FMC collected data to generate a sectorial sweep of A-scans,
where each sample of every A-scan is computed using a dedicated focused focal law

1414
Terminology
• Linear Total Focusing (LTF)
– Processing method using FMC collected data to generate a linear sweep of A-scans, where
each sample of every A-scan is computed using a dedicated focused focal law
• Compound Total Focusing (CTF)
– Processing method using FMC collected data to generate a compound sweep of A-scans,
where each sample of every A-scan is computed using a dedicated focused focal law
• Adaptive TFM (A-TFM again…)
– Variation of TFM where the surface of the component is not known prior to examination; FMC
data is used in a first phase to deduct the surface shape, generate a set of delay laws and in
a second phase re-process the same set of FMC data to obtain the final TFM frame.
• Adaptive TFM (A-TFM and again…)
– Same name as above, but a different concept. Normalization of names should be undertaken
by competent authority
– Variation of TFM where an amplitude normalization algorithm is applied on every pixel to
improve signal quality and reflector/indication shape.

FMC Characteristics

1616
FMC – Signal Characteristics
• Elementary A-Scan signals gathered by FMC require specific
characteristics, so they can be used for reconstruction later on
– RF signals (unrectified), because phase information is required to
compute constructive and destructive interference
– 100 MHz digitization (= 10 ns between samples); this resolution is crucial
for proper delay law forming in post processing
– High amplitude fidelity: digitization in 12 bits or better, along with very
low noise is required, since very weak signals are processed
– No compression – must keep the best possible resolution
– No smoothing because unrectified signals are used

1717
FMC – Signal Characteristics (cont’d)
• Capability is also often limited by current era electronics:
– These limitations may gradually disappear, as electronic capabilities
improve
– Limited or no gain control : gain is usually identical for each elementary
A-Scan generated from a single FMC data collection
• No way to calibrate probe elements electronically
– No Filtering: filtering is usually done in FIR on traditional phased array,
and is not possible due to the scale of parallel data acquisition acquired
simultaneously
• Analog filtering is currently possible, but takes space on acquisition boards
and is often discarded by manufacturer
– No Averaging: this would require a large number of multiple firing to
occur per probe, and would require a huge amount of memory to process
(see FMC scale factor later)

• Data example, 16 elements probe
• NOT to be used by human…
18
FMC – Data example
• Element 1 firing, receiving on 1
through 16• Element 2 firing, receiving on 1
through 16
• Element 16 firing, receiving on 1
through 16
• Main Bang of probe on wedge
• Signal inside wedge
• Interface signal between wedge
and component• Signal inside component
• Actual data used by TFM for image
construction

1919
FMC – Data Quantity
Function of Element Count
• Assuming a probe with n elements
• Probe must be pulsed n times (once per element)
• This implies that the acoustic travel time of an ultrasound pulse
occurs n times
• For each pulse, an n elementary A-Scan is digitalized, one per
element
• Implies the total number of elementary A-scans gathered after the
full firing cycle = n2

2020
Example
• For a 16-element probe
– Probe will be pulsed 16 times
– 256 elementary A-Scans are gathered

2121
File Size and Data Throughput
• Typically, FMC is performed with a 64-element probe, to enhance
aperture size, beam forming capability, and focusing power
• Typical elementary A-Scan lengths can be around 80 µs. Each
elementary A-scan must have a sufficient time base length so that the
maximum applied delay law + depth of coverage is fully contained
– Usually translates into an A-scan with 8192 samples, at 2 bytes per
sample (12 to 16-bits amplitude)
• The amount of storage require for FMC data recording at a single
probe location is 64 MBytes
– 4096 * 8192 * 2 (elementary A-scan * sample count * byte count)
• And this holds true for every probe location along a scan line ...

2222
File Size and Data Throughput
• Large amount of storage required for FMC (up to 64 MB for a 64-
element probe) has further implications :
• To achieve sufficient acquisition speed, the system data
throughput must be very high:
– To achieve 30 Hz, a data throughput of 1.8 GBytes/s is required,
i.e. 64 MBytes * 30Hz
• To scan a 12” pipe with a typical resolution, a huge data file will be
produced
– (12” * 3.1416 / 0.039” scan resolution) * 64 MB = 60 GB

FMC – Signal Equivalency
• Using FMC, every element is used to pulse and to receive; the
various element combinations can be represented in a firing matrix.
– Rows usually represent firing elements
– Columns usually represent receiving elements
• In the firing matrix, there are pairs of combinations that have
equivalent ultrasound paths and characteristics, resulting in
equivalent A-scans
– A21 is equivalent in path and signal to A12
– A41 is equivalent in path and signal to A14
– etc…
Probe
Pulsing 19
Receiving 1
Probe
Pulsing 1
Receiving 19 Equivalent to

2424
FMC – HMC
• Using the properties of equivalent elementary
A-Scans, alternative (more compact) methods
for performing FMC have been proposed
• Half Matrix Capture (HMC) is one example
• When using HMC, for each equivalent pair, only
one combination of elements is kept
FMC HMC
Firing element n n
Elementary A-Scan n2 (n2+ n) / 2
(Nearly half of n2)

2525
FMC – Data Processing Mode
Due to the large amount of data produced per second and in total, there are typically two ways
of handling FMC data today:
• Use and discard
– Some instruments are using the FMC data for a single probe location within the instrument, using
dedicated high-speed hardware.
– The FMC is converted on the fly into a TFM frame
– TFM frame is sent to the software for viewing and for storage
– Then FMC data is discarded to preserve usable space and speed
– Best for inspection speed, but no data validation or reuse capability is possible
• Transfer, keep, and reuse
– Some instruments also support a storage capability for FMC
– Requires a high-speed data link between the instrument and the computer. Highest Ethernet link is
still too slow, which forces the acquisition speed constraint
– Amount of data stored on hard disk of host computer is high
– But since the FMC is stored, it can be reused in analysis at will, with different sets of reconstruction
parameters.
– Offers absolute flexibility of use, at the cost of speed

TFM Characteristics

2727
TFM – Signal characteristics
• Requires elementary A-Scans for proper reconstruction
• Assuming elementary A-scans are available, reconstruction can
be performed “on the fly” during data collection, or post-processed
during analysis.
• Can be performed on multiple probe configurations:
– Pulse Echo
– Pitch & Catch
– Tandem and Self-Tandem
– etc…

2828
TFM – Frame Parameters
• A single set of FMC data (elementary A-Scans) can be re-used multiple times
in TFM to construct different images
• Each application of TFM can have different parameters
– Frame location
– Frame size
– Even the path reconstruction mode !
• It is possible to reconstruct direct paths, but also indirect and mode-
converted paths, in any combination
• Different TFM frames can be “merged” together to enhance detection and
sizing capability
TT TTL

2929
TFM – Total Focusing… or close …
• Every pixel in a TFM frame is a created using a dedicated focal law.
– This results in 1 focal point per pixel
– In theory, every pixel is perfectly focused
• But the laws of physics still apply
– Probe aperture size determines the near-field depth
– No probe is able to focus past its near-field depth
– Past the near-field, pixels will still have content, but it will be essentially similar to regular
phased array
– Note: Near-field depth depends on wedge thickness!
– Examples:
• LM-5MHz, Wedge, 64 elements, 38.4 mm aperture : near-field depth at 55 degrees = 147 mm
• LM-5MHz , Wedge, 32 elements, 19.2 mm aperture: near-field depth at 55 degrees = 28 mm
• LM-2.25MHz , Wedge, 32 elements, 19.2 mm aperture: near-field depth at 55 degrees = 6.3 mm

3030
TFM – Total coverage… or close …
• Every pixel in a TFM frame is a created using a dedicated focal law
– This result in complete rectangular area with ultrasound data
– In theory, every pixel is then perfectly positioned
• In practice, law of physics still apply
– Ultrasonic beams have a limit to the amount of steering that can be applied
– Maximum beam steering is a function of the beam spread of a single element aperture
– Past the maximum steering angle, focal law can still be computed, but beam does not really cover the expected area
– Leads to situations where pixel contains data related to a different location
• Example: LM 5MHz, Wedge SW55, Direct path
Data properly
located
For Pixel in
center of weld
For Pixel in top
HAZ
Data
improperly
located

3131
TFM – Amplitude
• Several phenomena can impact the precision of the amplitude
reading on a given TFM pixel within a frame
– Effect of frame resolution versus probe frequency
– Effect on interface signal

• The Nyquist Theorem, also known as the sampling theorem, is a principle
that engineers follow in the digitization of analog signals. For analog-to-digital
conversion (ADC) to result in a faithful reproduction of the signal, slices,
called samples, of the analog waveform must be taken frequently.
• Signal frequency can be expressed as period or wavelength
– 5 MHz signal = 200 ns period or 1.2 mm Wavelength (LW in steel)
• Waveform cannot be preserved with less than 2 samples per period
• Even when waveform is preserved, signal peak amplitude can be missed,
resulting in amplitude error in digitized signal
– Higher digitizing frequency reduce the amplitude error
• In Ultrasound methodology, an effective digitizing frequency of 5 times the
probe frequency is required to limit this amplitude error
– 5 MHz signal = 25 MHz minimum digitizing frequency (1 sample every 40 ns)
32
Digitizing Frequency of Signal

33
Digitizing Frequency of Signal
Digitizing at
20 X signal
Frequency
Digitizing at
5 X signal
Frequency
5 MHZ Signal
Digitized at 100 MHz
Digitized at 25 MHz
Sample positions are variable

• Low digitizing frequency is subject to variable error level in
amplitude
34
Amplitude Error related to Digitizing Frequency
Peak Value = 100 %
Digitized Value = 96 %

• At same digitizing frequency, exact error is dependent on signal
timing
– If the sampling is synchronized with the peak signal = no amplitude error
– If the sampling is exactly halfway before and after the peak = maximum
amplitude error
Synchronized signal, no amplitude error Worst case, maximum amplitude error

• Maximum error occurs at random occurrence, depending on signal generator location (flaw location)
• Maximum error can be computed
– Phase spread = number of degrees between two consecutive samples
– Phase error = number of degrees between a sample and peak location
– Maximum Phase error = Phase spread / 2
• Occur when samples are at same distance before and after the peak
– Amplitude error ratio in % = cos(Maximum Phase Error)
– Amplitude error in dB = 20 X log (cos(Maximum Phase Error))
Synchronized signal, no amplitude error Worst case, maximum amplitude error

• Sample Error level
– 5 MHz signal, digitized at 25 MHz = 5 points per period
– Phase spread = 360 / 5 = 72 degrees
– Maximum Phase error = 72 / 2 = 36 degrees
– Amplitude Error ratio in % = cos(36) = 80.9%
– Amplitude Error (AE) in dB = 20 X log (cos(36)) = -1.85 dB
• Other examples
– 20 samples per period: AE = -0.1 dB
– 2 samples per period: AE = infinite loss

3838
TFM – Amplitude subject to resolution
• Every pixel in a TFM frame is created using a dedicated focal law
– Corollary: each pixel has ONLY one focal law to cover it
– Focal law is tuned for center of pixel
• Depending on instrument and software capability, issues may be present:
– Size ratio of a pixel versus wavelength of the ultrasonic beam is critical for
appropriate imaging
– A large pixel will statistically miss the peak amplitude of a signal
– The Nyquist Theorem must be respected for a peak signal to be detected
– Systems with low and fixed frame resolutions (i.e.: 256 * 256) are more
susceptible to this issue
– Systems with higher and flexible frame resolutions (x * y) can overcome this
issue and generate high quality images

3939
TFM – Nyquist in TFM
• Given a pixel size, the acoustic wave may peak
right at its center, or elsewhere
• When the signal is at its peak elsewhere, the
amplitude used in TFM construction is lower
than expected
• When the size discrepancy between pixel and
wavelength is large, a large error in amplitude
is statistically expected.
– 20% (-2 dB) drop is measured when the
pixel size is smaller than 1/5 the wavelength
of the ultrasound beam
– 100% of the amplitude can be lost when the
pixel size is 1/2 the wavelength of the
ultrasound beam
Probe
0 LW
Example of signal at 2 different depth. Amplitude will
vary depending on location and Nyquist theorem

4040
TFM – Nyquist Example
Example A
• V Weld, Carbon steel (LW)
• Thickness 25mm
• Frame depth with coverage of first
and second leg require 50mm
• 50mm / 256 pixel = 0.2 mm per pixel
• 5MHz probe in steel = 1.2mm
• Factor: 1.2 / 0.2 = 6
• Maximum phase error: 360°/ 6 / 2 =
30
• Amplitude error:
– 20X log(cos(30)) = -1.2 dB
Example B
• Steel Bolt (LW)
• Thickness 100mm
• Frame depth with coverage of first leg
require 100mm
• 100mm / 256 pixel = 0.4 mm per pixel
• 5MHz probe in steel (LW) = 1.2mm
• Factor: 1.2 / 0.4 = 3
• Maximum phase error: 360° / 3 / 2 =
60
• Amplitude error:
– 20X log(cos(60)) = -6.0 dB

4141
6 Pixels per wavelength
• Maximum Amplitude: 97.3%
• Width (at -6 dB): 3.3mm
• Echo Dynamics Curve
– Regular amplitude distribution
2 Pixels per wavelength
• Maximum Amplitude: 69.4%
• Width (at -6 dB): 2.1 mm
• Echo Dynamics Curve
– Jagged amplitude distribution
Tilted Notch
• Single FMC data
collection
• TFM processed twice
• Only parameter modified
is pixel size in TFM
frame
• Every other parameter is
kept constant
• Same indication
visualized!

4242
6 Pixels per wavelength 2 Pixels per wavelength
Tilted Notch
• Same notch
• Same parameters
• Probe position is moved 0.5 mm
• Amplitude pattern changes
drastically when under sampling

4343
TFM – Amplitude and Interface/Dead zone
• As opposed to STF, where all beams transit through a small area and
cause saturation interference, TFM has a relatively smooth interface
signal
• In TFM, even when the frame is located below the probe, the
contribution to the top row of pixels is performed in such way that the
energy is distributed, and usually prevents saturation
• In fact, top rows of a frame dedicated to corrosion mapping will have an
interface signal that is less disturbing than with an equivalent phased
array approach
• This effectively reduces the dead-zone in the top section of the frame
compared to regular phased array

Codes and Applications

4545
Codes
• Currently, this new methodology is not supported by Codes
• It is not permitted to use this technique for “Code compliant”
inspections
• A working group on FMC/TFM was founded within ASME Section
V
– Schedule is to have a mandatory appendix for publication by
December 2019
– Draft not redacted yet
– Issues to resolve
– Zetec and other manufacturers are participating

• Currently, no system supports TFM calibration; this creates
issues for amplitude-based inspection techniques
• Draft proposal for amplitude calibration is “in course of
preparation”, scheduled mid-2017; No proposal yet for wedge or
velocity calibration
• Subject to change as technology evolves …
46
Calibration

4747
Application – Weld examination
Strengths
• Each pixel of the image is a focal point (as long as near-field rule is satisfied)
• Definition of a focal law is easier in calculator (no focalization parameters, no beam steering parameters)
• Multiple frame reconstructions can be overlaid to cover multiple mode conversions and may improve detection
Weaknesses
• Number of acoustic paths to perform the acquisition is dependent on the number of elements, not the number of laws; a
Typical 40 to 70 SW sweep with 1 degree resolution requires 31 travel times with standard PA, and 64 with a TFM frame for
a 64-element probe; this results in a slower inspection speed
• For small TFM frames (256 x 256), size and resolution can limit the weld thickness that can be examined due to the Nyquist
rule for a maximum amplitude drop of 2 dB (x5)
– for SW in steel, Wavelength is 0.65 mm at 5 MHz; so pixel size must be 0.13 mm and maximum frame dimension is 33 mm x 33
mm, therefore insufficient to cover weld thicknesses greater than 16.5 mm with full skip coverage.
– for Longitudinal wave in Steel, with proportional effect on wavelength, pixel size and frame size, maximum frame dimension is 60
mm x 60 mm; . Coverage of weld thicknesses up to 30 mm thick can be achieved with full skip coverage.
• Inaccurate pixel location at very high angles
• Live TFM does not store FMC data, which prevents data re-use and adds additional risk to qualification effort
• Post-processing TFM has FMC data, but is typically too slow for a production-environment inspection
• TFM does not provide A-Scan data, which may impact flaw characterization and qualification effort

4848
Application – Corrosion examination
Strengths
• Each pixel of the image is a focal point (as long as near-field rule is satisfied)
• Definition of focal law is easier in calculator (no focalization parameters, no beam steering parameters)
• Number of acoustic paths to perform the acquisition is dependent on the number of elements, not number of laws. The net
effect is the opposite of the weld. A typical 0LW, 64-element probe requires 58 ( 64 – aperture) travel times with standard PA
or 116 travel times when using improved resolution. A typical TFM frame with a 64-element probe will require 64 travel times,
independent of the final resolution. This may result in a faster inspection speed depending on TFM processing speed
• Coverage is performed at 0 degrees in standard PA, but TFM frames cover a wide range of angles, which may improve
detection of corrosion and image quality
• Averaging effect helps produce a crisp image
• Even with small TFM frame (256 x 256), size and resolution is within a typical inspection range and is not an issue
– Using Longitudinal Wave in Steel, with proportional effect on wavelength, pixel size and frame size, maximum frame dimension
of 60mm X 60mm is attained (for a 256 X 256 frame). Able to cover component up to 60 mm thick with direct path.
Weaknesses
• With typical frame resolution, lateral resolution is better than standard PA, but depth resolution is typically worse
• Live TFM does not store FMC data, which prevents data re-use
• Post-processing TFM has FMC data, but is typically too slow for a production-environment inspection
• TFM does not provide A-Scan data, which may impact thickness measurement accuracy

49
Plate Weld Inspection
Carbon steel, T = 19 mm, with realistic welding defects & SDH
Linear array LM 5 MHz (64 elements) on 55SW wedge

50
LOF, Incomplete Penetration, Toe-Crack, Porosity
Std PA UT, Merged data from Sector 40 to 70SW, focusing HP 50 mm

51
Reconstructed FMC data, Merge from Sector STF 40 to 70SW

52
Reconstructed FMC data, Merge TFM Frames SW

53
Reconstructed FMC data, Merge from TFM Frames SW
Rebounds included

54
Thick Vessel Weld
Carbon steel, T = 120 mm, Narrow Gap Weld

55
Thick Vessel Weld
Realistic Welding Defects:
F: Crack
NS5: Cluster of shrinkages

56
Thick Vessel Weld
Flaws F & NS5
Standard PA, Merge Sector 40 to 70SW, focusing HP 50 mm

57
Thick Vessel Weld
Flaws F & NS5
Reconstructed FMC data, Merge STF Sector 40 to 70SW
(STF = Sectorial Total Focusing)

58
Thick Vessel Weld
Flaws F & NS5
Reconstructed FMC data, Merge TFM Frame SW
(TFM = Total Focusing Method)

Closing words

6060
Expected improvement
• Many of the limitations of current-era FMC/TFM can be reduced as
technology matures.
– Calibration will be introduced
– Frame size and Pixel size can be improved
– Scanning speeds can be improved
– Live TFM could store FMC data
– …
• But the underlying laws of physics will still apply!

Overview: Principles of Full Matrix Capture (FMC) and Total Focusing Method (TFM) in Ultrasonic Inspections

More Related Content

What's hot (20)

Similar to Overview: Principles of Full Matrix Capture (FMC) and Total Focusing Method (TFM) in Ultrasonic Inspections (20)

More from Zetec Inc. (11)

Recently uploaded (20)

Overview: Principles of Full Matrix Capture (FMC) and Total Focusing Method (TFM) in Ultrasonic Inspections

Editor's Notes