ULTRASONIC TIME - OF -

FLIGHT DIFFRACTION (TOFD)

COURSE

i of 51
 CONTENTS

CONTENTS

DEFINITIONS
Figure 1 Co-ordinate System

1. INTRODUCTION
1.1 Scope of the Course
1.2 Requirements
1.3 Examination

2. BACKGROUND AND CONVENTIONAL ULTRASONIC TESTING

2.1 Background to Non Destructive Testing
2.2 Ultrasonic Non-Destructive Testing
2.3 Pulse-echo detection of flaws
2.4 Flaw Sizing with the Pulse-echo Technique
Figure 2.1 Flaw size in relation to beam spread
Figure 2.2 Determination of Flaw Size by 6dB Drop Sizing
2.3 Comparison of Flaw Sizing Accuracy for different Techniques

3. THE TIME-OF-FLIGHT DIFFRACTION TECHNIQUE

3.1 Background to the Importance of the TOFD Technique

Figure 3.1 Illustration of importance of accurate sizing measurements ...

3.2 History of TOFD Development

3.3 Principle of TOFD

3.3.1 Diffraction process

Figure 3.2 Representation of the Diffraction Process from Crack Tips
Figure 3.3 Illustration of Diffraction
3.3.2 Angular Variation of Diffraction Signals
Figure 3.4 Calculated Variation of Amplitude of Diffracted Waves with
Angle
3.3.3 Basic TOFD Set-up
Figure 3.5 Cross Section of Typical Probe
Figure 3.6 Wave Paths for the TOFD Arrangement
Figure 3.7 A-scan with no Defect Present
Figure 3.8 A-scan with Defect Present
Figure 3.9 Basic TOFD parameters
Figure 3.10 Consistent measurement of time of various signals

ii of 51
 Figure 3.11 Surface opening crack in a non-parallel scan
Figure 3.12 Surface opening cracks on the back wall for a non-parallel
scan

3.4 Types of TOFD Scan

Figure 3.13 Non-Parallel or Longitudinal Scan
Figure 3.14 Parallel or Transverse Scan
Figure 3.15 Difference between a non-parallel and a parallel scan
Figure 3.16 Uncertainty in lateral position in a non-parallel scan

3.5 Summary of Advantages and Disadvantages of TOFD

4. DATA AQUISITION SYSTEM

4.1 Basic Hardware

Figure 4.1 Typical Digital System for collecting TOFD Data

4.2 Advantage of Digital Recording

4.3 Digitisation of the Analogue Ultrasonic Signals

4.3.1 Amplitude
Figure 4.2 Digitisation of an A-scan
4.3.2 Sampling Rate
Figure 4.3 Digitisation with 5 samples per period
Figure 4.4 Sampling a 10 MHz sin wave with a 20 MHz digitiser
Figure 4.5 Sampling a 10 MHz sin wave with a 15 MHz digitiser

4.4 Selection of frequencies for filtering

4.5 Amount of data collected during an inspection

4.6 Grey scale imaging and B-scans

Figure 4.6 Grey scale representation of the amplitude
Figure 4.7 Appearance of a signal in a B-scan which occurs in a
succession of A-scans

4.7 Signal Averaging

Figure 4.8 Principle of Signal Averaging

4.8 Pulse Width Control

Figure 4.9 Effect of Drive Pulse on Transmitter Crystal Element
Figure 4.10 Illustration of Effect of altering Driver Pulse Width

iii of 51
 4.9 Software
4.9.1 On-line
4.9.2 Analysis and Of-line facilities

iv of 51
DEFINITIONS

Figure 1 Co-ordinate System

The following symbols are used in the text,
c velocity of sound
d depth of reflector below scanning surface
D thickness of sample being scanned
Parallel:scan in which direction of beam parallel to direction of scan
Non-parallel: scan in which direction of beam normal to direction of
scan
PCS probe centre separation
s half of probe centre separation 2s
t time of flight of signal from a reflector
2to time taken for sound to pass through two probe shoes, i.e. probe
delay
tl time of flight of lateral wave
tb time of flight of back wall signal
θ angle of beam centre with respect to normal to metal surface

v of 51
1. INTRODUCTION

1.1 Scope of the Course

The purpose of this course is to make the reader and student aware of the
Time-of-Flight Diffraction (TOFD) technique and its ability to perform flaw
detection, location and sizing. Although the course will be based around the
digital ultrasonic data collection system currently available a number of
similar systems are available and have essentially the same major features
and facilities.
The course notes are intended to be used as background and reference
material for both Level 1 and Level II qualification standard. Where there is a
clear distinction in the notes this will be indicated by typing the specific Level
II material in italics. Normal duration of the Level 1 and Level II courses are
40 hours.

Level 1 technicians should be capable of

i) setting up the TOFD equipment
ii) collecting TOFD inspection data
iii)recording and classifying the results according to written
instructions
iv) reporting the results.
Level II technicians should be capable of
i) setting up and calibrating the equipment
ii) performing and supervising the inspection
iii) interpreting and evaluating the results according to the
applicable standards
iv) define limitations of TOFD
v) write practical testing instructions and procedures
vi) supervise and train Level 1 technicians
vii) organise and report the results of a TOFD inspection.
Considerable time is spent during the course on the practical collection and
analysis of TOFD data since experience in the viewing and analysis of TOFD
data is very important. It is also very important to have in mind the different
nature of TOFD as compared to the more traditional Pulse-echo technique.
The course will hopefully extend the participants knowledge to be able to
recognise the limitations of both techniques and to be able to choose the
appropriate technique or both for the inspection in mind.
In addition to the current course notes the British TOFD Standard , BS 7706
and the European (Pre) Standard, CEN/TC 138 WG 2 N 143, should be
studied.

6 of 51
1.2 Requirements

It is assumed that a reasonable level of ultrasonic understanding has already

been achieved equivalent to at least a EN473 Ultrasonic Level 1 for the current
Level 1 course and a EN473 Level II for the current Level II course. All
candidates should have proof of a satisfactory vision test.

1.3 Examination

An examination will be set at the end of the course which will consist of,

Level 1 practical test - 2 hours

20 questions - 30 minutes

Level II evaluation of TOFD data - 1 hour

20 questions - 40 minutes
British Standard open book questions - 30 minutes

The object of the examination is to demonstrate that ultrasonic practitioners

are capable of applying the TOFD technique to defective components and
detecting, locating and sizing defects to an accuracy commensurate with the
intrinsic limitations of the technique.

7 of 51
2. BACKGROUND AND CONVENTIONAL ULTRASONIC
TESTING

2.1 Background to Non Destructive Testing

The detection of cracks, particularly in welds, before failure is very important

since the consequences may be very serious. There are various non-
destructive techniques available for testing metal components and they are
generally divided into two classes of detection, i.e.

surface opening cracks by magnetic particle

dye penetrant
eddy current
magnetic techniques
ultrasonics
and internal cracks by radiography
ultrasonics.
Surface opening cracks are usually considered the most important since they
are most likely to grow and cause failure. However, they have the advantage
that they can usually be repaired more easily than internal cracks. Dye
Penetrant and Magnetic Particle inspections have no depth sizing capability
and the magnetic techniques are also generally used for detection only. The
Eddy Current technique has a sizing capability but, depending on the
material, only within a few millimetres of the surface. Ultrasonics has both a
detection and sizing capability.
It is very important to detect and size internal cracks because repairs may
prove very difficult and expensive and if it can be proved that the crack size is
below some critical level the plant may be safely left running.
For most internal cracks there are only two inspection techniques available,
radiography and ultrasonics. Radiography is best suited to detecting
volumetric type defects and planar cracks may only be efficiently detected if
the beam of X-rays is directed down the crack. Also radiography needs access
to both sides of the sample and has safety implications. Radiography is very
widely used, however, and is a very reliable technique for detecting many
types of weld errors and will detect changes in density equivalent to 1% of the
wall thickness. It also gives a permanent record of the inspection which can
be stored away for future reference if required.
However, the most reliable technique for detecting and sizing planar cracks is
ultrasonics, but its sensitivity may only be equivalent to 2% or 3% of wall
thickness. Ultrasonics also has the advantage that it only needs access to one
side of the sample and there are no safety implications.

8 of 51
2.2 Ultrasonic Non-Destructive Testing

If an electric potential is applied to a piezoelectric type material it oscillates

and if it is of the right thickness will produce waves of ultrasound of the
frequencies most useful for inspecting metal components. This material is the
basis of ultrasonic probes which produce longitudinal waves, generally called
compression waves. If the longitudinal waves enter metal at an angle then
they refract in the metal and produce both longitudinal and shear waves, the
angles of the two types of waves depending on the velocity of shear and
longitudinal waves in the metal and the velocity of the longitudinal waves in
the probe shoe material. Shear waves are transmitted by a periodic shear
force and can only exist in materials like metals which possess shear elasticity.
Liquids cannot sustain a shear force.
For normal ultrasonic inspection of metals ultrasonic frequencies of between
2 and 5 MHz are used. The corresponding wavelength of the waves are
found from the formula,
velocity (c) = wavelength (λ) * frequency (f)
Velocity is usually defined in units of m/s and typical values in steel are 5950
m/s for longitudinal waves and 3230 m/s for shear waves. Since the probe
frequency is in units of MHz (and we shall later that time is defined in
microseconds in the TOFD technique) it is more convenient to define the
velocity units as mm/μs (i.e. 5.95 mm/μs and 3.23 mm/μs respectively). In
these units the wavelength in the above equation is given in millimetres. Thus
for the above frequencies the wavelength of longitudinal waves is in the range
1 to 3 mm and for shear waves from 0.6 to1.6 mm. For reflectors of size less
than half a wavelength interference can take place in the reflected waves and
hence the minimum size of cracks that can reliably be detected is equivalent
to one half the wavelength. To detect small cracks in thin materials higher
frequencies are used but in thick material the increasing attenuation with
increase in frequency generally prevents the use of much higher frequencies.
For conventional pulse-echo ultrasonic inspections angled shear waves are
very important since at a given frequency they have a wavelength half that of
longitudinal waves, allowing for the resolution of smaller defects. Also, as
will be seen in a later chapter, for a given size of crystal diameter and
frequency shear waves produce a smaller beam spread and a consequently
higher beam intensity and more accurate sizing ability than longitudinal
waves.

9 of 51
2.3 Pulse-echo detection of flaws

An ultrasonic inspection of a sample is carried by scanning the metal with a

beam of ultrasound. Any reflectors in the metal are only detected if the sound
is reflected back from the discontinuity and returns to the crystal element of
the probe, where it vibrates the crystal and is converted into electrical signals.
In order to reflect the waves back the beam must ideally be at right angles to
the reflector surface. This is the so called “specular” reflection. If the surface
is tilted with respect to the direction of the beam of ultrasound then the
reflected waves may miss the probe crystal altogether and the discontinuity
will remain undetected. The proportion of the sound beam getting back to the
crystal falls off rapidly with increasing angles of tilt and skew from this ideal
position. A tilt of only 5 degrees can cause the amplitude to fall by a factor of
2 (6 dB) and 10 degrees or more may result in loss of detection.
Thus for normal pulse-echo inspections with angled shear waves it is
important to scan the metal with a range of probe angles chosen to give the
best chance of detecting the expected orientations of the cracks of main
interest. In general a practical compromise has to be made. Vertical cracks are
difficult to detect because of the difficulty of getting an ultrasonic beam at
right angles to the crack surface and a pair of probes may be needed to detect
the cracks (tandem technique).
The decision to report the presence of a reflector in pulse-echo inspections is
based initially on the amplitude of the signal compared to some threshold
level. This is a relatively simple concept which works well practically. The
threshold level is usually determined from the amplitude of the signal
reflected back from a standard reflector in a calibration block. However, since
the roughness of the reflector surface, as well as its orientation, directly effect
the amplitude of the returned signal an awareness of the likely probability of
detection is very important. Ultrasonic inspection is not perfect and the skill
and experience of the operator is very important, hence the need for training.

2.4 Flaw Sizing with the Pulse-echo Technique

The method of flaw sizing used in the pulse-echo technique depends on the size of the
flaw with respect to the beam spread of the probe (see figure 2.1).

10 of 51
 Figure 2.1 Flaw size in relation to beam spread

The basic assumption in determining the size of flaws that are smaller than the beam
spread is that the amplitude of the reflected signal will be proportional to the area of
the flaw in the beam of ultrasound. In this situation the estimate of size is based on
the relative size of the amplitude with respect to that obtained from a standard
reflector at a similar range. Standard reflectors are provided in various calibration
blocks and consist of flat bottomed holes, side drilled holes or notches. Thus the
accuracy very much depends on the degree of roughness of the flaw surface and its
exact orientation with respect to the ultrasonic beam.

For flaw sizes larger than the beam spread increasing the size of the flaw will have no
effect on the amplitude of the signal since the area outside the beam will not
contribute to the reflected signal. The flaw size has to be obtained therefore by
scanning the probe over the flaw and noting the position where a certain change in
amplitude takes place. This is illustrated in figure 2.2 for an angled shear wave probe
sizing a lack of fusion

11 of 51
 Figure 2.2 Determination of Flaw Size by 6dB Drop Sizing

flaw. When the probe is moved towards the weld the flaw starts to appear in the
ultrasonic beam and the amplitude of the flaw signal rises. Once the flaw area fills the
beam the amplitude stays constant until the beam starts to pass the other end of the
flaw , when the amplitude starts to fall. It is assumed for this explanation that a
Distance Amplitude Correction has been applied so that there is no variation of
amplitude with range. The maximum amplitude trace across the flaw is called an
echodynamic trace and is shown in the bottom half of figure 2.2.
At the level where the signal amplitude is half that of the maximum signal it is
assumed that only half the flaw area is in the beam of ultrasound and that the centre
of the probe is opposite the edge of the flaw. Thus if the positions of the probe are noted
where the amplitude has dropped by 6 dB the size of the flaw can be measured and
hence the term 6 dB drop sizing. If the distance between the probe positions is x mm
then the width, w, of the flaw is given by w = xcos(θ), where θ is the angle of the
beam centre with respect to the normal to the surface of the metal on which the probe
sits. The through wall height of the flaw (the critical measurement), h, is then
h = xcos(θ)sin(θ).

To determine the length of the flaw along the weld the probe must be positioned so as
to obtain the maximum amplitude signal and then moved parallel to the weld to
determine the 6 dB drop positions. The length is the distance between these positions.

12 of 51
Again the main problem with the 6 dB drop technique is the variation in amplitude
due to the possible roughness of scattering surface and the fact that the flaw surface is
unlikely to be normal to the ultrasonic beam.
Some typical smallest sizes of flaws which can be detected and typical sizing errors
are listed below from document IIS/IIW - 580 - 86 from Commission V of the
International Institute of Welding for pulse-echo ultrasonics.

Defect positions Through-thickness Length

Smallest Accuracy, Smallest Accuracy,
size, mm + or - mm length, + or - mm
mm
Near surface, 3 3 4 5
0-5 mm depth
Mid-wall, mm
5-25 3 3 4 4
25-75 3 3 7 7
75-125 5 5 10 10
Back surface wall
thickness, mm
10-25 4 4 4 4
25-75 4 4 7 7
75-125 5 5 10 10

2.5 Comparison of Flaw Sizing Accuracy for different

Techniques

Other techniques are available for sizing flaws. For surface opening flaws, if access is
available to the surface the alternating current potential drop (ACPD) is an accurate
technique. The technique essentially measures the drop in voltage between two surface
contacts when placed on either side of the flaw, the additional voltage drop being due
to current having to flow around the flaw surface. Specialised eddy current type
techniques (ACFM) can also measure the size of near surface flaws. One of the most
accurate and useful techniques for measuring flaw height is the TOFD technique
which is the subject of this course. The importance of the TOFD technique is seen
from the comparison of the accuracy of different crack sizing techniques summarised
below,

Manual pulse-echo 4 mm
ACPD (surface opening only) 1 mm
TOFD 1 mm
TOFD when monitoring growth 0.3 mm

13 of 51
3. THE TIME-OF-FLIGHT DIFFRACTION TECHNIQUE

3.1 Background to the Importance of the TOFD Technique

The most serious types of defects in welds and metal components are planar
cracks since they are the most likely to grow and cause failure of the
component and hence the importance of ultrasonic inspection since
ultrasonics is the most suitable technique for determining the position and
sizing such defects. The importance of developing more accurate sizing
techniques than afforded by conventional pulse-echo inspections became
apparent in the 1960’s, especially in the nuclear and chemical plant industries.
When defects where found in these plants repair of the damaged components
was often very difficult or impossible, one of the problems being the difficulty
of carrying out the necessary heat treatment after welding. In the nuclear
industry there was also the problem of access and often high radiation levels.
To shut down and replace such plant is a very expensive exercise.
Thus the science of Fracture Mechanics developed to enable the growth rate
of defects to be predicted and to calculate the critical size of defect below
which it was possible to carry on safely running the plant Often, because of
the difficulties in determining all the properties necessary for the calculations
(e.g. fracture toughness) very conservative estimates had to be made and the
consequent lifetime predictions for the safe operation of the plant became
very short.
If it could be proved by successive ultrasonic examinations of the defects that
they are not growing or they are growing at a much slower rate than
predicted then this would be very important to the plant operator. If the
defects are stable and below the critical size then the plant can stay in
operation. If they are growing at a slower rate then the lifetime of the plant
can be extended. Also if the growth rate can be accurately measured then
suitable plans can be drawn for a programmed repair or replacement of the
plant, again saving the plant operator a lot of expense. Unexpected
shutdowns with the consequent loss of output and the need for unplanned
repairs are the plant operators nightmare.
The need, in order to measure the growth rate of cracks, is accurate through
wall sizing. The inadequacy of conventional pulse-echo techniques is
illustrated in figure 3.1. The two

14 of 51
 Figure 3.1 Illustration of importance of accurate sizing
measurements

plots show the predicted lifetime curve for a particular defect and it is
predicted to reach the critical size after about 1.5 years. In the top illustration
is shown the results of pulse-echo measurements of the crack height at the
yearly plant shutdowns. Because of the large error on the measurements
indicated by the error bar the measurements can not influence the
conservative predicted lifetime. The bottom plot illustrates the results of
measurements with TOFD. Since the errors are now much smaller it is
apparent that the actual growth is less than the predicted growth and the
actual plant lifetime can be extended.
The ability to size flaws more accurately should also help to limit the number of false
calls. If the requirement is that a very high probability of detection is required for
flaws above a certain size then the threshold for pulse-echo detection has to be set
considerably lower than this size because of the large error with the pulse-echo
technique. This means that a considerable number of flaws, which are actually below
this size, are reported as being above this size because they appear with the pulse-echo
technique to be larger. Thus while a very high probability of detection may be
obtained for flaws above the size of interest there will be a large false call rate. This is
made worse by the fact that the distribution curve of flaw size against number of
flaws usually rises towards the smaller sizes.

15 of 51
Thus in principle the detection threshold for the more accurate TOFD technique can
be set much closer to the size of interest and thus greatly reduce the false call rate.

3.2 History of TOFD Development

In the last section the importance of accurate sizing of cracks was described,
especially in the Nuclear Industry. For this reason the National NDT Centre at
Harwell (then part of the United Kingdom Atomic Energy Authority -
UKAEA) asked Dr Maurice Silk to try and develop an ultrasonic sizing
technique more accurate than the conventional pulse-echo method. In the
early 1970’s Dr Silk developed the technique known as Time-of-Flight
Diffraction (TOFD).

The two most important differences from conventional pulse-echo are its
more accurate sizing accuracy (typically + or - 1 mm and in a monitoring
situation + or - 0.3 mm) and the almost independence of angle of the flaw
for detection. The sizing is based on the transit time of the diffraction
signals and does not depend on the amplitude.
It will be seen that with TOFD a simple amplitude threshold cannot be
applied for rejecting or reporting flaws since the amplitude of the
diffraction signals does not depend on the size of the flaw and all the data
must be analysed before any flaws can be rejected. Hence with the TOFD
technique training and experience are essential.

For a number of years it remained largely a laboratory tool, but the

realisation of its importance and the proposed Public enquiry for a PWR
Reactor in the UK lead to a number
of major trials in the early 1980’s to evaluate the best possible ultrasonic
inspection technique for the reactor pressure vessel and other major
components. These trials were known as the Defect Detection Trials (DDT).
The trials were very important in view of the international PISC exercise in
the late 1970’s, which was aimed at establishing the capability of the ASME
Code ultrasonic procedures, and which obtained poor results for the
reliability and accuracy of conventional pulse-echo inspection. As will be
mentioned later, TOFD proved to be a very capable and accurate technique.
Many other trials and validations have been carried out comparing different
techniques and in all these tests TOFD has always proved to be virtually the
most reliable and accurate technique.
Because the digital systems available up to 1982 were relatively large it was
difficult to carry out inspections in the field. Thus the National NDT Centre
developed a more portable system for collecting and analysing the data. The
system was known as ZIPSCAN and was licensed by SGS Sonomatic in 1983
and sold throughout the world. Sonomatic is know part of AEA Technology.
Nowadays there are a number of commercial digital ultrasonic systems
available for carrying out TOFD inspections.

16 of 51
Initially TOFD was only developed and used as a sizing tool. It was expected
that flaws would be detected by the conventional techniques and then TOFD
used for more accurate sizing. A particular application was in the monitoring
the crack growth of cracks detected in major plant components, e.g. pressure
vessels.
However, has TOFD became more accepted and used during the later half of
the 1980’s, especially by the oil and gas industry both offshore and onshore,
its economic advantages for detection as well sizing became apparent and
these days TOFD is used in many applications for both detection and sizing.
Very often one scan with a pair of TOFD probes along a weld will reveal all
the defects in the weld volume. The ability to collect an image of the
inspection data (B- or D-scans) also affords advantages in difficult geometrical
situations or for weld roots since it is often easier to recognise new patterns in
an image than to characterise signals in a single A-scan. A good example of
this is major use of TOFD in the offshore oil industry for detecting weld root
erosion.
The TOFD technique has continued to be developed by the major research
establishments (e.g. modelling applications for helping with the analysis of
the data and optimising the collection of the data in complex geometry) and
by the inspection companies for different applications.

3.3 Principle of TOFD

3.3.1 Diffraction process

When an ultrasonic wave interacts with a long crack-like flaw it results in the
production of diffracted waves from the crack tips, in addition to any
ultrasonic waves reflected from the surface of the crack. The diffracted waves
are much weaker than specularly reflected waves used for conventional
ultrasonic inspection, but they radiate from the tips in all directions along the
same plane as the incident ultrasonic waves as indicated in figure 3.2.

17 of 51
 Figure 3.2 Representation of the Diffraction Process from
Crack Tips

The phenomena of diffraction is nothing new and occurs with all types of waves, e.g.
light and water waves. It is very well known in light especially when light is passed
through a slit or past an edge and contributes to the resolving power of telescopes and
other optical instruments. To explain diffraction when waves pass through a slot
Huygens’ proposed, three centuries ago, that “each point on a wave front may be
regarded as a new source of waves”. Hence, for illustration purposes, we can think of
reflection of a wave from a surface occurring as outlined in figure 3.3.

18 of 51
 Figure 3.3 Illustration of Diffraction

Each point on the surface ( on a scale smaller than the wavelength) acts as a reflection
point and produces waves. Then as the separate waves cross each other interference
takes place and as explained by Young in 1802 for light “the resultant displacement
is the sum of the displacements due to each separate wave” and in this case the result
is a reflected plane wave. However, at the edge of the surface the waves from the tip
are left and produce so called diffraction waves. Nowadays diffraction is more
accurately described by Kirchhoff’s theory.
The phenomena of tip diffraction is just another technique in the repertoire of
general ultrasonics. The use of tip diffraction signals as long been used in
pulse-echo sizing estimations since it often enhances the signal when the
probe is opposite the end of the defect and used in this way it is sometimes
called the maximum amplitude technique or backward scatter tip diffraction.

3.3.2 Angular Variation of Diffraction Signals

The angular variation of the amplitude of the diffracted waves with is
illustrated in Figure 3.4. The figure is not an exact reproduction and the
reader should refer to Charlesworth and Temple (1989) for an exact variation.
The variation of the top and bottom signals from a vertical crack located mid-
way between two TOFD transducers is shown as a function of of the beam
angle with respect to the normal.

19 of 51
 Figure 3.4 Calculated Variation of Amplitude of Diffracted
Waves with Angle

The amplitude has a maximum at around 70 degrees in steel, and has a slightly
greater amplitude from the bottom of the crack than the top, but the amplitudes on the
whole are very similar. Between about 45 and 80 degrees the variation in amplitude is
less than 6 dB. For the bottom of the crack there is a large reduction in amplitude at
about 38 degrees with the amplitude rising again around 20 degrees. Typical angles
in an inspection are 45, 60 and 70 degrees.
For shear waves the optimum angles in steel are between 50 and 55 degrees for the top
and bottom of a vertical crack.
For defects which do not lie in a plane perpendicular to the line joining the two
transducers the calculation is more complex and is described by Charlesworth and
Temple (1989), but relatively large angles of skew have little effect on the amplitude.
Thus one of the main advantages of TOFD, unlike pulse-echo, is that the
diffraction signal is largely independent of angle and thus almost
independent of defect orientation.

20 of 51
3.3.3 Basic TOFD Set-up

3.3.3.1 Probe type and wave type

The TOFD technique is a crack sizing technique based on locating the crack
tips from the ultrasonic energy re-emitted by diffraction. The basic
arrangement of the TOFD technique consists of two probes in the pitch-and-
catch configuration, with one probe acting as the transmitter and the second
probe the receiver. Such an arrangement provides a large volume for
inspection and provides for unambiguous location of the position and depth
of the reflectors.
A typical probe is shown in Figure 3.5 and consists of a compression
transducer mounted

Figure 3.5 Cross Section of Typical Probe

on a wedge of perspex or similar material. The compression probe should be

chosen to have a short pulse length in order to give good depth resolution.
The wedge angle is cut to give compression waves in the metal at typical
angles of 45, 60 or 70 degrees. The transducers often have screw threads so
that they are easily interchangeable with the wedges. Couplant must be
present between the transducer and wedge in order for ultrasound to be
transmitted. This design has the drawback that the couplant eventually dries
out and must be replaced.

21 of 51
The reason for using compression waves in the metal is that they travel at
about twice the velocity of shear waves and hence arrive first at the receiver.
In order to work out depth a knowledge of the velocity is required and it is
much easier to unambiguously to work out depth if one can assume that the
signals have a compression velocity. When either type of wave is reflected or
refracted it can partially mode convert into the other type of wave. Thus if
shear waves were used and they diffracted from a defect tip compression
waves might also be produced and these would then arrive at the receiver
first. If in this case it was assumed that the shear velocity was the correct
velocity the depth would be calculated incorrectly.
After the compression or longitudinal wave has travelled through the shoe,
however, it splits up into the desired compression wave at the required angle
and into a shear wave in the metal at approximately half the angle of the
compression wave. Thus shear waves are present in the metal but the
resultant signal generally occur after the compression signals. Thus the
resultant TOFD inspection data can contain waves that have travelled,
all the way as compression
all the way as shear
part of the path as compression and part shear.

3.3.3.2 Signals observed

The TOFD arrangement is shown in Figure 3.6. The A-scan appearance with
no defect present is shown in Figure 3.7 and with a defect present in Figure
3.8. The main types of waves seen are as follows.

Figure 3.6 Wave Paths for the TOFD Arrangement

22 of 51
Lateral wave
In general a weak lateral wave running between the two probes with the
compression velocity just below the surface of the metal is observed first. It
obeys Fermat’s principle in that a wave between two points takes the
minimum time. As we shall see later, for a curved surface it will travel
straight across the metal between the two probes. On material with a surface
cladding layer the velocity of sound in the cladding may be such that the
lateral wave travels most of the way in the material beneath the cladding.
The wave can be very weak for large probe separations and may not even be
recognisable.
Because of the basic pitch-catch probe arrangement the signals from the near
surface region are very compressed in time and these signals may be hidden
beneath the lateral wave.
Back wall signal
A much larger signal reflected/diffracted from the back wall is observed after
the lateral wave because of the greater distance travelled. If the probe beams
are only directed at the upper part of the metal or there is no suitable back
wall there may be no back wall signal.
Defect signals
If a planar type crack is present in the metal beneath the probes diffraction
signals from the top and bottom tips are seen between the lateral wave and
the back wall. These signals are generally much weaker than the backwall
signal but stronger than the lateral wave. If the defect has little height then the
signals from the top and bottom may run into each other. Thus the
importance of a minimum number of cycles in the signals in order to improve
the resolution of the signals from the top and bottom of small defects.
Because the diffraction signals are so weak they cannot often be easily seen on
a single A-scan and it is only by displaying the successive A-scans from a scan
in B-scan form that the pattern of the diffraction signals becomes clear. Signal
averaging is very important in these situations because it improves the signal-
to-noise ratio.
Again this is why TOFD is very difficult with an analogue flaw detector
where only a single A-scan display is available.
Shear or mode converted shear signals
After the compression back wall signal a much large signal generally appears
and is a back wall shear reflected signal and it is often mistaken for the
compression back wall signal. Between these signals other signals are
generally observed due to mode conversion at a defect into shear waves
which then takes a longer time for the signal to arrive at the receiver.
It is often very useful to collect signals in this region since genuine defect
signals are repeated at longer times and near surface defect signals may much
clearer since they are spread out in time more for the shear waves.

23 of 51
Ray Paths
The ray paths depicted in Figure 3.6 are simply those that happen to connect
the probes and crack tips. They do not represent special angles at which
diffraction only occurs. Diffraction can occur at all angles and those rays in
the beam spread which happen to hit the tips will be diffracted and hit the
receiver.

3.3.3.3 Phase Relationships

An A-scan is reproduced in Figure 3.7 and contains the lateral wave and back
wall signals.

Figure 3.7 A-scan with no Defect Present

When a wave in a medium with a higher acoustic impedance is reflected at

the interface to a lower acoustic impedance there is a phase change of 180
degrees (e. g. steel to water or steel to air) This means that if the waveform
starts with a positive cycle before it hits the wall it will start with a negative
cycle after reflection from the wall.
When a defect is present the situation shown in figure 3.8 occurs. The signal
from the top

24 of 51
 Figure 3.8 A-scan with Defect Present

of the defect acts as if it had undergone a reflection from a backwall and has a
phase change of 180 degrees, i.e. the phase is like a backwall and starts with a
negative cycle. The bottom of a defect, however, acts as if the wave runs
round the bottom without a phase change and the phase of the signal is like
the lateral wave, i.e. it starts with a positive cycle. Theory shows that if two
diffraction signals have opposite phase they must have a continuous crack
between them. Theory also shows that in a few cases the top and bottom
diffraction signals may not have a phase change of 180 degrees, but in general
they will.
Thus the recognition of phase change is very important for characterising
signals and for making the most accurate defect sizing. For example two
signals may be present which are from two slag lines rather than a single
crack. In this case there will not be a phase change. Slag lines and pores are
generally to thin to produce separate top and bottom signals.
Because the number of observed cycles in a signal very much depends on the
amplitude of the signals it is often difficult to recognise the phase. This is
especially true for the backwall which is generally saturated. In these
situations it is important to sit the probes on the sample being examined or a
calibration block and turn the gain down so that the backwall or any other
difficult signal has the same screen height as the defect signals and then
increase the gain making a note of how the signal grows with respect to the
order of the phases. It is sometimes easiest to concentrate on the two or three
most predominant cycles.

25 of 51
It is because the phase information is important that it is necessary to collect
the TOFD unrectified signals digitally.

3.3.3.4 Depth Calculation

The depth of a reflector is calculated from the arrival time of the pulse by
simple trigonometry. There is no recourse to measurement of amplitude. The
positioning of the tip signals then provides information on the actual flaw
sizes and height estimates below the scanned surface and through wall size.
For a reflector situated symmetrically beneath the probes, as shown in Figure
3.9 the

Figure 3.9 Basic TOFD parameters

distance travelled in millimetres between the probes is,
distance = 2(s2 + d2)1/2
where s = half the probe centre separation (PCS) 2s, mm
d = depth of reflector, mm
( )1/2 means square root of the number in the brackets.
Hence the time, t, taken in microseconds is
t = 2(s2 + d2)1/2/c
where c = velocity of sound in units of mm/μs.
Thus by rearrangement of the above equation the depth can be expressed as a
function of the measured time,
d = [(ct/2)2 - s2]1/2
Very often the reflector will not be symmetrically positioned beneath the
probes and there will a slight error in the calculation of depth. This is
discussed in more detail in the Chapter on Errors.
The typical size of the error for depth measurements is + or - 1 mm (see
Chapter on Errors), but when used in repeat inspections for crack growth
monitoring using the same probe array and equipment the error is reduced to
typically + or - 0.3 mm.

26 of 51
Because of the pitch-catch probe arrangement the relationship between depth
and time is not linear and has can be seen from the above equation it has a
quadratic form. Software should be available to convert the collected B- or D-
scan into a linear depth scale by a process known as linearisation. This then
provides a B- or D-scan linear with depth and this may be more useful to plot
for a report. Very often when analysing the data it is preferable to work from
the raw collected data with the data plotted on the time axis. Because the
paths to reflectors near to the surface are almost horizontal a small change in
time represents a large change in depth. Thus, conversion to a linear depth
scale tends to spread out the near surface signals and the lateral wave appears
out of proportion. An alternative method for depth measurements is to use a
cursor on the raw data and read out depth at the position of the cursor.
The other effect of this non-linear depth scale, has will be shown in the
Chapter on Errors, is that the error on the depth measurement increases
rapidly near the top surface. Thus because of the presence of the lateral
wave and the increasing depth error TOFD is not so useful for inspecting
the near surface region, typically the first 10 mm. By using higher frequency
probes ( ~ 15 MHz) and a small PCS, however, thin samples can be inspected
to within a millimetre or so of the surface.

3.3.3.5 Measurement of Time and initial Probe-

Centre- Separation
Depth Calibration
In practise the depth calculation needs to take into account the extra delay in
the measured time due to the passage of the sound through the probe shoes.
This delay is known as the probe delay, 2to microseconds. Thus the total
transit time measured, t, is in practise given by
t = 2(s2 + d2)1/2/c + 2to
and the depth is given by
d = [(c/2)2(t-2to)2 - s2]1/2
Thus it would appear that to calculate the depth for a reflector with a transit
time of t μs that a knowledge of the ultrasonic velocity, probe delay and probe
centre separation is required. However, it will be shown in the Chapter on
Errors that a more accurate calculation of the depth can be made if the
velocity and probe delay are found from the measurement of the positions of
the lateral wave and back wall signals. This procedure helps to reduce any
systematic errors such as errors in the PCS.
The lateral wave signal occurs at a time, tl μs, and is given by
tl = 2s/c + 2to
and the back wall at a time, tb μs, and is given by
tb = 2(s2 + D2)1/2/c + 2to
where D thickness of the sample.

27 of 51
By rearranging the two equation the probe delay and velocity can be found if
the PCS = 2s, and the thickness D is known, i.e.
c = 2(s2 + D2)1/2 - 2s
(tb - tl)

and 2to = tb - 2(s2 + D2)1/2/c

Thus the recommended method of measuring depth is to measure the PCS

and the thickness of the sample and enter the values into the file header
before starting a scan. Then to measure a depth in a B- or D-scan, if the
relevant software is available, first measure the position of the lateral and
back wall signals and let the computer automatically calculate the probe delay
and velocity. Then at each cursor position in the image the depth can be
calculated. Obviously if there is no lateral or back wall signal then only one
signal can be used for calibration and either the velocity or probe delay must
also be input to the program.
The PCS is measured between the index points of the two probes.

Measurement of time of arrival of various signal

Because of the phase changes in the various signals care must be taken in
when choosing the position for measuring the time of arrival of the various
signals in order to obtain the most accurate depth estimates. The prime
requirement is to be consistent in the various measurements. One could
choose to measure to the peak signal, but since the back wall signal is often
saturated this is more difficult. One suggestion is to measure the time at a
crossover point when the cycle changes from positive to negative. This is
quite easy if there is a readout of time and amplitude as a cursor is moved
about the B- or D-scan, since the reading should change at the cross over point
from negative to positive or vice versa. The sample nearest point nearest to an
amplitude of zero should be chosen.
A possible set of positions are shown in Figure 3.10.

28 of 51
 Figure 3.10 Consistent measurement of time of various
signals

If the lateral wave starts with say a positive cycle take the reading at the start
of the cycle. The equivalent time at the back wall signal will then be, because
of the phase change, the start of the first negative cycle. However, in the
example shown the start of the second negative cycle has been chosen because
the amplitude is much greater and there are more cycles. The second negative
cycle is deemed in this case to be the equivalent cycle to that used for the
lateral wave. For the tip diffraction signals the start of the first negative signal
is chosen for the top of the crack and the first positive signal for the bottom of
the crack.

Initial choice of PCS for an Inspection

The optimum choice of the PCS will be described in a later Chapter, but for a
new inspection the best choice of PCS is such that the ultrasonic beam centre
is aimed at two thirds of the depth of the sample. This will then in general
give coverage of the majority of the weld for example. If the beam centre
angle is θ, then
tanθ = depth/s and the PCS, 2s = (4/3)Dtanθ
where D is the thickness of the sample.

29 of 51
Check collecting correct part of A-scan
Since the lateral wave is often very weak and there is a very strong shear
wave back wall signal beyond the compression back wall signal it is easy to
take the later for the lateral wave and the shear wave signal for the back wall
signal. Always check that the observed signals correspond to the calculated
times of arrival of the lateral wave and back wall signal, i.e.
lateral wave tl = 2s/c + 2to
and back wall tb = 2(s2 + D2)1/2/c + 2to

3.3.3.6 Patterns for Surface breaking Defects

For surface breaking cracks the expected TOFD B- or D-scan will be modified.
Thus if the defect breaks the upper surface the lateral wave will normally be
eliminated ( as shown in Figure 3.11) or very much reduced in amplitude. If
the defect is not very long then the lateral wave may partially get round the
side of the defect.

Figure 3.11 Surface opening crack in a non-parallel scan

For cracks open to the back wall the resultant D-scan is shown in Figure 3.12.
The effect on the back wall depends on the height of the crack and the volume
of coverage of the probe set up.
For cracks with small penetration into the metal the diffraction signal should
be seen just above the back wall with normally no change in the appearance
of the back wall signal. This is because the majority of the ultrasonic beam
still gets past the crack. If the crack penetrates in somewhat further the back
wall signal may lose some amplitude and have a dip due to the longer path
for sound at the edge of the beam still being reflected from the back wall and
getting to the receiver. Finally, if the crack penetrates sufficiently far the back
wall signal will be cut off.

30 of 51
 Figure 3.12 Surface opening cracks on the back wall for a
non-parallel scan

3.4 Types of TOFD Scan

There are two types of scan. The initial scan generally used for detection is
shown in Figure 3.13 and is called the non-parallel or longitudinal scan
because the direction of scan is at right angles to the direction of the ultrasonic
beam. The resultant scan is known as a D-scan since it produces a cross
section along the weld. This type of scan is usually set up with as wide a
beam spread as possible to inspect as large a volume as possible with one
scan. Since the probes straddle the weld the weld cap does not interfere with
the scan. It is a very economical inspection and is the scan most frequently
carried out and often the only one.

31 of 51
 Figure 3.13 Non-Parallel or Longitudinal Scan

The second type of scan is shown in Figure 3.14 and is called the transverse or
parallel scan. The direction of scan is parallel to the ultrasonic beam direction.
The scan produced is called a B-scan since it produces a cross section across
the weld. In this case if there is a wide weld cap it is difficult to carry out the
scan and only a limited movement may be possible. This type of scan gives
the most accurate measurement of depth and should be optimised for this
purpose.

Figure 3.14 Parallel or Transverse Scan

The importance of the non-parallel scan is illustrated in Figure 3.15. In the

figure a lack-of-fusion defect with constant height and length in the plane of
the weld is assumed. The non-parallel scan will give a parallel set of
diffraction signal extendings over the length of the defect with characteristic
shaped arcs at the ends where the path length is longer as the probes
approach and leave the defect. However, there is no knowledge to be gained
from the D-scan of the location of the reflectors in cross sectional plane across

32 of 51
the weld. The defects could be from any in the beam coverage of the probes.
This is illustrated in Figure 3.16.
In Figure 3.16 a signal with a certain transit time could be anywhere on the
locus of an ellipse with the two probe positions as the focal points This means
that the reflectors may not be symmetrically situated beneath the probes and
hence the depth calculation will not be the most accurate. The error is in
general very small because due to the limitations of beam spread the variation
in depth is over the almost horizontal part of the ellipse.

Figure 3.15 Difference between a non-parallel and a parallel

scan

33 of 51
 Figure 3.16 Uncertainty in lateral position in a non-parallel
scan

With the parallel scan, provided a complete scan is obtained, the probes
traverse over the defect and at some point the reflectors are symmetrically
positioned beneath the probes, thus giving the most accurate depth. This is
shown in the figure by the characteristic arcs produced by this type of scan.
As the probes approach the reflectors the signal starts and rises as the path
length shortens until the minimum is reached when the reflector is
symmetrically beneath the two probes. The peak position corresponds to this
minimum time.
This peak position also gives the location of the reflector in the weld cross
section and the relative position of a crack top and bottom signal gives an
indication of the crack orientation. If the scan is carried out with an encoder
and the start position is known with respect to the weld centre line, then it
should be possible to locate the position in the weld cross section to within
one or two millimetres. This is very important since the location of the defect
will help to provide a clue for the characterisation of the defect. To obtain
good results, however it may be necessary to use a smaller PCS and narrower
beam width in order to obtain a sharp peak.
In many instances, because of the need to complete the inspection quickly or
because of financial constraints, inspections are only carried with the non-
parallel scan. However, in order to make a reasonable guess as to the type of
defect and to get the best sizing accuracy parallel scans should be carried out
over all the signals of interest found in the non-parallel scans. If the defects
are long then parallel scans should be carried out at different positions along
its length.

34 of 51
3.5 Summary of Advantages and Disadvantages of TOFD
Advantages
The two most important differences of TOFD from conventional pulse-echo
are,
i) the almost independence of angle of the defect for detection of the
diffraction signals
ii) the sizing is not dependent on the amplitude of the signals and the
corresponding errors
Thus the main advantages of TOFD are,
a) TOFD has a through wall sizing accuracy of + or - 1 mm and a crack
growth monitoring capability of + or- 0.3 mm
b) efficient detection of defects of all orientations
c) permanent digital record of the inspection data with cross-section
type views through the metal.

Disadvantages
The amplitude of the diffraction signals does not depend on the size of the
defects and a simple amplitude threshold cannot be applied for selecting the
important reportable defects, unlike pulse-echo inspections. TOFD easily
detects pores, slag lines, inclusions etc.
The main disadvantages of TOFD are,
a) no simple amplitude threshold for selecting reportable defects
b) all the TOFD inspection data has to be visually analysed in order to
select the reportable defects
c) not so suitable for defects near to the inspection surface since they
may be hidden by the lateral wave and the sizing accuracy
deteriorates rapidly near to the inspection surface.

Since a considerable human effort is involved in the analysis of the TOFD

data experience and training are essential.
TOFD should be regarded as just another tool in the field of ultrasonic
inspection and sometimes it is more appropriate than pulse-echo and
sometimes not. Very often a combination of the two techniques is the best
solution since the additional information is often vital for defect
characterisation.

35 of 51
4. DATA AQUISITION SYSTEM

4.1 Basic Hardware

Most ultrasonic inspections are carried out by manually scanning with angled shear
wave or compression probes using an analogue flaw detector. There is only generally
facilities for one probe to be used at a time. With an analogue flaw detector the
operator views the A-scan display and notes in a report sheet any indications seen,
there location and size. There is no record kept of the observed ultrasonic signals and
no permanent evidence of the coverage undertaken.
Since about 1990 these analogue systems have tended to become computer controlled,
with facilities like storage of DAC curves and set-up parameters, which make the flaw
detector easier to use. However, a small percentage of inspections have had to be
mechanised and handle arrays of probes for reasons of speed (steel mill), safety
(nuclear plant) or environment (under the sea). For these inspections very
sophisticated digital systems have been developed for handling arrays of probes,
controlling mechanised scanners and recording all the inspection data for subsequent
analysis and providing a record. They generally handle arrays of probes by rapidly
multiplexing around the probes and thus in practise only amplify and record the
signals from one probe at a time. For inspections of large volumes of metal at high
speed these systems tend to record just the range and peak height of any signals which
exceed a defined threshold level within a range gate. However, for TOFD inspections
these systems need to record the full unrectified A-scan at each inspection position or
sufficient information to be able to reconstruct the A-scan. In this mode the firing
rate (pulse repetition frequency) of the probes is much less than when just recording
peak height since a lot more data has to be stored on the hard disk.

A typical system for collecting TOFD data is shown below.

36 of 51
 Figure 4.1 Typical Digital System for collecting TOFD Data

The digital systems consist of , like the analogue flaw detectors, electrical circuits for
transmitting a firing pulse to the transmitting probe and amplifying and filtering the
received signal. The main difference from the flaw detectors is that there are a number
of output transmitter/receiver channels which are selected under computer control in
sequence by the multiplexer. The computer normally multiplexes at each inspection
point around those channels which have been defined in the inspection set-up. In the
TOFD configuration separate transmitter and receiver channels are defined for each
pair of TOFD probes used in the inspection.

Because the computer environment is often electrically noisy and there may be
external electrical noise it is usual to restrict the frequency bandpass of the amplifier
by means of filters to the spectrum of frequencies appropriate to the ultrasonic probes
being used. While this may reduce the overall amplitude it should improve the signal-
to-noise ratio of the amplified signals. Filtering is especially useful for TOFD signals
which are normally very weak and may require gains of about 70 dB.
If the TOFD signals are very weak or the probes are situated some distance from the
recording system (e.g. 100 m) then it is a good idea to use a separate head amplifier in
the receiving line, as near to the receiver probe as possible. Numerous battery
operated head amplifiers are available with gains of the order of 30 or 40 dB. The
battery operated amplifiers have the advantage that they are not connected to the same
mains supply as the main collection system and are therefor less subject to electrical
noise.
The resultant amplified and filtered analogue signals are then converted into a string
of digital numbers and stored in the computer memory by means of the digitiser.

37 of 51
The computer has attached to it the normal range of peripherals such as keyboard,
monitor, mouse, memory, floppy disk, etc. Additional devices such as printers and
optical disks can be connected to the normal communication ports. The systems
generally have provision for a motor control/encoder card for automatically
controlling a mechanised inspection system or for at least measuring probe position
by means of the encoder input facility. Several different axes of motion can be catered
for.

4.2 Advantage of Digital Recording

There are many advantages for the use of computer based systems for the digital
recording of ultrasonic data. The TOFD technique requires the full unrectified A-
scans to be stored at each inspection point along a scan. Some of the advantages of
digital storage are,
i) A permanent digital record of all collected data, inspection parameters
used, and calibration runs - i.e. complete record of the inspection.
ii) Various signal processing operations can be carried out on the collected
data allowing signal enhancement, etc. to be carried out (off-line
processing).
iii) Inspection results can be compared in detail in subsequent inspections,
(e.g. growth monitoring).
iv) Re-analysis of the raw inspection data can be carried out at any time.
v) A variety of visualisation displays can be used for data analysis and
presentation of results.
vi) Allows integration of the Data Acquisition System and the manipulator
controller.
vii) Allows multi channel pulse-echo and TOFD DATA acquisition and direct
comparison or overlay of the results.

For quality control the complete record of the inspection offered by digital systems is
a major advantage over the old analogue flaw detectors.

4.3 Digitisation of the Analogue Ultrasonic Signals

4.3.1 Amplitude

The pulsing of the probes and the subsequent amplification and filtering are carried
out carried out in analogue form. The great advantage of analogue systems is the
speed with operations can be carried out, i.e. firing rate of an array of probes and
subsequent peak-height detection. However, digitisation of the analogue signals, in
spite of its consequent reduction in speed due to the need to store all the data in the
computer memory, does as explained in the last section have a number of major
advantages and is essential for the subsequent analysis of TOFD data.
The conversion of the analogue A-scan into a string of numbers is known as
digitisation and consists of sampling the analogue wave form at a regular sampling
interval, as shown in the figure,

38 of 51
 Figure 4.2 Digitisation of an A-scan

The amplitude of each stored sample along with position and phase information (sign)
is passed to the computer for storage. Therefor the information stored in the computer
is a string of numbers representing the A-scan which can be recalled to reconstruct the
A-scan

for an A-scan display or by stacking together consecutive A-scans a B-scan can be

reconstructed.
The basic storage location of a computer is one bit which can be in one of two states
and can be used to represent the numbers 0 or 1. In ultrasonic systems a collection of
8 bits (one byte) has frequently been used to represent the amplitude of one sample,
however, some systems are now using 10 or 12 bits to represent the amplitude. In the
table below the corresponding range of numbers represented by 8,10 or 12 bits is
given for rectified and unrectified A-scan signals. For unrectified signals (TOFD) the
sign of the amplitude needs to be preserved and thus the number representing the
maximum amplitude is effectively half that for rectified data. Also in the table is
given the maximum range in decibels (i.e. dB = 20log10(A1/A2), where A is
amplitude) represented by one half cycle of the unrectified signals.

bits number range of range of range in dB for

representation numbers for numbers for unrectified data
rectified data unrectified
data
8 28 0 to 255 -128 to +127 20log10(127/1) = 42
10 210 0 to 1023 -512 to +511 20log10(511/1) = 54
12 212 0 to 4095 -2048 to +2047 20log10(2047/1) = 66

The digitiser operates between two specific voltage levels for the input signals (e.g. 0
to 1 volt or -0.5 to +0.5 volts). Thus any unrectified analogue signal outside these
limits will be saturated and stored as + or - 100% full screen height (FSH). For
example for the 8 bit digitiser these saturated values will be stored as 127 or - 128.
For the unrectified data the range in dB represented by the maximum amplitude of

39 of 51
one half cycle is only 42 dB for the 8 bit digitiser (i.e. from 100% FSH down to a
level of 1unit). If the noise level is only 2% the effective range for the 8 bit digitiser
for a linear amplifier is reduced to 34 dB (20log10(127/2.54).
Since the gain of the main amplifier is typically of the order of 80 dB then the
digitiser can only represent a limited fraction of the total gain range available. This is
not normally a problem for TOFD data since, because of the limits imposed by beam
spread, only a limited volume of metal is viewed by a pair of TOFD probes. The
more bits used to represent amplitude by the digitiser the less this is a problem. If
there is a problem then there are usually a number of options, i.e.
i) define a second pair of channels connected to the same probes but with a
different overall gain and collect data from both sets of channels
ii) use a DAC curve
iii) use a logarithmic amplifier

4.3.2 Sampling Rate

Most of the digital ultrasonic systems can store up to 4096 or 8192 samples per A-
scan, which is quite adequate for TOFD data. If the digitisation rate is F MHz then
the interval between digitisation samples is S microseconds (μs),
where S = 1/F.
Hence the number of samples in an A-scan length of t μs is,
number of samples = t/S or t*F.
For a typical TOFD A-scan the time between the lateral wave and backwall is 10 μs
and with a digitisation frequency of typically 50 MHz the number of samples is 500.
This is well within the total available. Consider a typical probe frequency of 5 MHz
and a digitisation rate of 25 MHz. The A-scan waveform is sampled every 1/25 =
0.04μs and within one period of a reflected signal (1/5 = 0.2μs) there will be 5
samples. This situation is illustrated in figure 4.3.

40 of 51
 Figure 4.3 Digitisation with 5 samples per period

The digitiser is working at the selected sampling rate and the A-scan comes along and
is sampled. The position of any peak in an ultrasonic signal is thus arbitrary with
respect to the sampling points. Thus for a sample to occur exactly at the peak position
of a positive or negative half cycle is a matter of chance and on average the peak
amplitude will not be measured. The greater the digitisation frequency the greater the
chance that one of the samples will occur at or very near the peak. However, the
higher the sampling rate the greater amount of data which must be stored in the
computer to represent the A-scan. Thus a compromise has to be reached. To obtain a
reasonable estimate (within 10% on average) of the peak amplitude at least 5 samples
per period of the signal are required. The measurement of peak amplitude is not of
great importance for TOFD data since the depth of a signal is derived from the time of
arrival of the signal and not on its amplitude. However, for accurate depth
measurements it is necessary to measure time accurately which means having
sufficient number of sampling points.
Thus for typical TOFD probe frequencies of 2, 5 10 and 15 MHz digitisation rates of
at least 10, 25, 50 and 75 MHz should be used. Most digital ultrasonic systems
nowadays have maximum digitisation rates in excess of 60 MHz and a choice of
several subdivisions of the maximum value.
For normal pulse-echo signals, since any sizing estimates depend on the
measurement of the maximum signal amplitude the choice of correct digitisation rate
is very important. However, the situation is eased for pulse-echo since rectification,
in effect, halves the frequency and the use of smoothing circuits means that the
effective frequency content of the signals is lower than the normal probe frequency.
So far we have considered the maximum digitisation frequency required, but we must
also consider the minimum which can be used. In order to represent the frequency
content of the digitised signals in the computer the analogue wave forms must be
digitised with a digitising frequency of at twice the maximum frequency in the wave

41 of 51
form, i.e. there must be at least one sampling point in each half cycle of the wave.
This is known as the Nyquist limit. This limitation is illustrated in the figures below.
The upper part of figure 4.4 illustrates the sampling of a 10 MHz sin wave with a 20
MHz digitisation frequency. There is one sample point per half cycle of the
waveform. The bottom part of the figure is the representation of the signal in the
computer and with this digitisation rate there is sufficient information for the
computer to recognise that the frequency of the signal is still 10 MHz. On average
the peak amplitude representation would be very poor.

Figure 4.4 Sampling a 10 MHz sin wave with a 20 MHz digitiser

The next figure (4.5) illustrates what happens when the 10 MHz sin wave is sampled
with a digitisation frequency of less than twice the 10 MHz frequency. The
digitisation frequency is 15 MHz and there is insufficient sample points to have a
sample in all the half cycles (see upper part of figure). The bottom part of the figure
shows the effect when the resultant sample points are reconstructed. The apparent
frequency of the waveform is now less than 10 MHz and the correct frequency cannot
be reconstructed. In addition if any further digital signal processing (e.g. filtering)
was carried on the data stored in the computer incorrect results would be obtained.

42 of 51
 Figure 4.5 Sampling a 10 MHz sin wave with a 15 MHz digitiser

Thus when digitising unrectified ultrasonic signals a digitisation rate of at least twice
the maximum frequency content of the signals should be used and ideally the
digitisation rate should be at least 5 times the maximum frequency.

4.4 Selection of frequencies for filtering

In the description of the basic hardware at the beginning of this chapter it was
explained that the filtering frequencies should be chosen to allow a pass band for the
spectrum of frequencies, appropriate to the ultrasonic probes being used, in order to
provide the best signal-to-noise ratio. In some circumstances, such as dirty metal or in
a noisy environment it may be necessary to be even more restrictive. This is best
accomplished by observing a signal from a known reflector in the noise and choosing
which filters give the best results.
There is usually a choice of high and low pass filter frequencies. The high pass filter
passes all frequencies above the chosen value and the low pass filter passes all
frequencies below its chosen value. For example a 5MHz probe may have a
frequency spectrum from about 2MHz to 6.5MHz and it would be appropriate to
choose a high pass filter of about 1.5MHz and a low pass filter of about 7.5MHz.
When working with high frequency probes it may be necessary to restrict the choice
of the low pass filter in order to meet the Nyquist sampling limit described in the last
section. For example, assume that 15MHz probes with a frequency spectrum between
10MHz and 18MHz are being used for an inspection and the maximum digitisation
frequency available is 32MHz. In order to prevent any signals with frequencies
greater than half the digitisation rate being sampled the low pass filter should be set at
16MHz.

43 of 51
 4.5 Amount of data collected during an inspection

When collecting TOFD data it is very easy to collect large amounts of data which
may fill the hard disk and provide a problem of storing the data. Fortunately the
advent of optical disks has largely overcome the latter problem. Backups of the
inspection data should be taken at the end of each day and where data is required to
be kept for some time two copies should be made and the copies stored in separate
locations.
The following example shows how to work out the storage requirements for a system
which uses an 8 bit digitiser and stores the full unrectified A-scans. For systems
which use a different number of bits to store the amplitudes or uses some compression
technique to represent the A-scan data the method and format of packing the data
must be known before an equivalent exercise can be carried out.
Consider an example where for a TOFD probe pair it is required to digitise a length of
A-scan of 10μs and an A-scan is collected every millimetre along a weld and the total
length of weld is 10m. If the digitisation frequency is 50MHz then each A-scan will
be represented by 10*50 = 500 samples and it will require 500 bytes of storage space.
Thus over the total length of weld there will be a need for 500*10,000 = 5,000,000
bytes or 5Mbytes of storage for a TOFD probe pair.
Each individual file containing the A-scans over a certain inspection length is slightly
longer than indicated above since it also has a file header containing information on
the various parameters used for the inspection. It is often useful if each separate scan
file does not exceed 1.44Mbytes since then it can be easily transferred by floppy disk.

4.6 Grey scale imaging and B-scans

The A-scan displays on the monitor screen are produced from the digital sample
points. Thus a single A-scan is plotted as a series of points representing the digitised
samples. If the A-scan is plotted horizontally across the screen then the vertical scale
will represent amplitude (+ and - 100% full screen height, i.e. the maximum voltage
range which can be input into the digitiser) and the horizontal will represent time
from the start of the firing or transmitter pulse. By joining the individual plotted
points (e.g. with straight lines) the appearance of a normal analogue A-scan display is
given. Cursors are generally available for taking amplitude and time measurements
from the displayed A-scan.
With TOFD data the succession of A-scans collected along a scan are displayed as a
B- or D- scan. B-scan generally means a scan across the weld and D-scan a scan
along the weld. Both have the same appearance and are made up of a succession of A-
scans and for the following description the general term B-scan will be used. Because
the TOFD diffraction signals are very weak they may not be easily seen on an
individual A-scan but in the B-scan display they are generally easily recognisable
since the eye is very efficient at picking out patterns in the data.
Since the B-scan may contain several hundred separate A-scans and the monitor
screen only contains, for example, some 640 by 480 individual plotting points (known
as pixels) it is impossible to plot the individual A-scans in the manner described at
the start of this section. Instead each individual A-scan is allocated a single column of
pixels down the screen (i.e. the time scale) with each pixel representing an individual
digitised sample and the amplitude scale is represented by the grey scale shading

44 of 51
given to each individual pixel. Colour could be used for the amplitude representation
but it is generally found that it is much easier for the brain to automatically
understand and interpret grey scale coding rather than a colour scale.
In grey scale coding the range of amplitudes is generally represented by a scale which
goes from pure white for a +100% FSH signal through mid grey at 0% to pure black
at -100% FSH (see figure 4.6).

+100% 0% -100%

Figure 4.6 Grey scale representation of the amplitude

Different digital systems may have a reverse scale going from black to white but the
choice is not important. The total number of shades available will vary with different
digital systems but typically the number may be 64 grey levels. For an 8-bit digitiser
+100% is represented by the number 127 and -100% by the number -128.
Thus a typical ultrasonic signal from a reflector will consist of series of bands of
shading alternatively of light grey and dark grey colouring. The number of bands
represent the number of half cycles present in the signal and their width represents the
number of sample points per half cycle. The appearance of a signal in a succession of
A-scans displayed as a B-scan is shown in figure 4.7.

Figure 4.7 Appearance of a signal in a B-scan which occurs in a

succession of A-scans

Contrast enhancement is normally available to enhance the low amplitude signals.

There are various algorithms for doing this but the simplest is to consider a linear
amplitude scale from -100% to +100% FSH. Contrast enhancement is obtained by

45 of 51
stretching the grey scale going from black to white over a smaller range of the
amplitude scale. For example if the full grey scale is applied to -50% to +50% FSH
then weaker signals will be more easily seen. Every sample with amplitude above
+50% FSH will be shown as pure white and every one below -50% FSH will be
shown as pure black.
The number of samples in an A-scan or the number of A-scans in the file being
plotted on the monitor may be larger than the number of pixels available. In order to
plot the complete image some of the A-scans may have to be omitted (e.g. every other
one) or some of the samples omitted ( in this situation the maximum amplitude
should be plotted regardless of any omitted samples). In general in these
circumstances the software will allow small areas to be selected and replotted with all
the data present or alternatively facilities may be available to scroll through the data
file without any loss of data for the area displayed. If the amount of data to be
plotted is less than the number of pixels available then the data can be expanded to fill
the screen by repeating the plot of a sample on successive pixels.

4.7 Signal Averaging

The TOFD diffracted signals from crack tips are very weak and need high levels of
amplification and hence are often difficult to see because of the noise in the signals.
Much of the noise is often due to random electrical signals which are picked up by the
system and hence this can be reduced by signal averaging. If N successive A-scans
are added together and the resultant sum divided by N then the signal-to-noise ratio of
a genuine signal over random noise is increased by the square root of N. This is
illustrated in figure 4.8 which shows in 5 successive A-scans a common signal and a
number of random noise signals. The effect of taking the average is shown at the
bottom of the figure. The consistent signal stays the same but the random noise
signals are reduced.

46 of 51
 Figure 4.8 Principle of Signal Averaging

Signal averaging can be accomplished most efficiently in the digitiser unit although it
is often done by software in the computer. Systems will generally allow averaging in
factors of 2 up to 256. A typical value would be N=16 which would give a signal to
noise improvement of 4. If the signals are very noisy a factor of 256 may be necessary
but because of the square root effect the improvement is only 16. Thus increasing the
number of averages does not give a simple linear increase in the signal-to-noise.
The rate of pulsing or firing the transmitting probe is called the pulse-repetition-
frequency (prf) and it is generally set by the user for the inspection. If there are more
than one transmitting channel then because of multiplexing in turn around all the
transmitters the effective prf for a pair of probes is the specified prf divided by the
number of transmitting channels. Signal averaging also uses N of these firings to
obtain an averaged waveform. Thus for a single pair of TOFD probes if the number of
averages is N then the effective prf for the inspection is the set prf/N.
TOFD data is usually collected as the probes are moved but the movement during the
collection of an averaged waveform is small. For example if the prf is set at1000 Hz
and the number of averages is 16 it will take 0.016 seconds to collect one averaged
waveform. If the scanning speed is, say, 50 mm/sec then the probes will travel in this
time 50*0.016 = 0.8 mm.
Signal averaging is very important for the collection of good quality diffraction
signals and is one of the reasons why adequate signals cannot usually be obtained
with a normal analogue flaw detector. However, signal averaging does not help when
the noise is coherent rather than random. This is the case for materials with large

47 of 51
grain structures such as austenitic steel when the additional scattering from the grain
boundaries is coherent and is not removed by simple time averaging.

4.8 Pulse Width Control

An ultrasonic probe consists of a piezo-electric material which when set into vibration
with a voltage pulse produces a burst of ultrasound. The voltage pulse is produced by
a transducer driver which produces a rectangular pulse of up to 400 volts amplitude
and with the width adjustable (usually between about 25 and 500 nanoseconds. The
use of different voltages ranging from 100 to 400 volts is generally dependent on the
probe frequency and the type of crystal element.
The pulse width is important when using TOFD in that it helps to optimise the shape
of the received signal. The first edge of the rectangular pulse sets the crystal element
into oscillation, as shown in figure 4.9. The second edge of the rectangular pulse also
sets the crystal element into oscillation again but the phase of the burst of ultrasound
is 180 degrees out of phase (i.e. the order of the + and - half cycles) with the first set
of oscillations.

Figure 4.9 Effect of Drive Pulse on Transmitter Crystal

Element

The two bursts of ultrasound generally overlap and interfere with one another.
Altering the pulse width causes cancellation or reinforcement of parts of the various
cycles. If the two bursts of ultrasound where composed of a single frequency and the
pulse width is set to one half of the period of this frequency (100 nanoseconds for 5
MHz) then as can be seen in figure 4.10 reinforcement takes place and a larger signal
is obtained, but it will ring for a longer time.

48 of 51
 Figure 4.10 Illustration of Effect of altering Driver Pulse Width

If the pulse width is set to 1 period of the wave frequency (200 ns for 5 MHz) then
after one period the two signals will be out of phase and a smaller amplitude signal
will be obtained, but the ringing will be reduced. This is a desirable feature for TOFD
applications since in order to resolve signals from the top and bottom tips of small
defects the signal should only last for a minimum number of cycles.
In practise a probe produces a range of frequencies and the optimum pulse width must
be found by trial. To do this obtain a signal from say the backwall at about 60% FSH
and adjust the pulse width starting at one period of the probe centre frequency.
Sometimes, because of the range of frequencies present altering the pulse width as
little effect on the overall shape.

4.9 Software

4.9.1 On-line
The basic software for most commercial digital systems which can perform a TOFD
inspection are very similar. Generally they operate via a menu system with a main
menu first appearing and selection of one of the options leading to a further set of
more specific set of menus and so on. All the normal operations for handling files
(e.g. copying , deleting, renaming, etc.) should be available. The main on-line
facilities which should be present are
i) capture and display of the unrectified waveform
ii) facilities to set-up and alter all the ultrasonic and collection parameters
iii) enter all data pertinent to the inspection (e.g. weld number, clients name,
etc.) for automatic entry into the file header information with the inspection
parameters
iv) save all the relevant inspection parameters for retrieval at a later date
v) motor control/encoder interface
vi) collection of inspection scan

49 of 51
 vii) display of the B- or D-scans and extraction and display of individual A-
scans
viii) display facilities such as scrolling, magnification of selected areas,
contrast enhancement and cursor facilities for sizing selected indications.

4.9.2 Analysis and Of-line facilities

The facilities available will vary between the various digital ultrasonic systems
available. Below is given a fairly extensive list of facilities which have proved to be
useful. These are being added to all the time.
The analysis facilities are,
i) cursor representation of a point reflector for the determination of depth and
lateral extent
ii) linearisation (conversion of the inspection data to a depth scale)
iii) lateral or backwall straightening (to enable more accurate depth
measurements and before lateral wave removal)
iv) lateral wave removal (to reveal any underlying indications)
v) Synthetic Aperture Focusing Technique or SAFT (used to clean up the
images and provide a more accurate method of measuring the lateral extent
of defects). In order to measure lateral extent 6 dB drop sizing facilities are
also required.
vi) digital filtering to remove noise or enhance signal-to-noise
vii) split spectrum processing (for improvement of signal-to-noise for
inspections carried out in materials with large grains)
viii) selection and storage of defect profiles and comparison (monitoring for
defect growth, display of root erosion defects)
ix) joining scans (e.g. when the inspection of a weld is broken down into a
number of separate scans)
x) addition of text and preparation of images for reports

There are two more categories of programs available although they are often written
more by individual companies or research establishments.
The first category is modelling which we may loosely divide into three sections
prediction - highly theoretical calculations of response from
reflectors
- calculation of beam path and effect in large grained
materials, e.g. austenitic
- probability of detection
complex geometry - calculation of optimum probe arrangement and
coverage for complex geometry situations, e. g.
nozzles
aids to analysis - either calculation of what the B-scan will look like for
a defined geometry or calculation of where a
specific signal could have come from
- calculation of phase of
signals

50 of 51
The second category of programs are generally fairly small specific programs which
we may term as useful aids. In this category some examples are,
calculation and plot of sizing errors as a function of depth
plot of beam spreads in simple geometry
calculation of location of mode converted and shear wave signals
corrections to be made when on convex or concave surfaces
signal shapes from reflectors with various orientations.

**********

51 of 51