Chapter (4): Digital Aspects Of TOFD

CHAPTER (4): DIGITAL ASPECTS OF TOFD

We have already touched on the fact that aspects of TOFD (and NDT more generally) are based on
digital technology. In fact, there are so many aspects of NDT that rely on digitisation that a separate
coverage of the topic is merited.

Definitions of the concepts of analogue and digital are as follows:

Analogue: Measuring or representing dat4 by means of one or more physical properties that can
express any value along a continuous scale. For example, the position of the hands on a clock is an
analogue representation of time.

Digital: Relating to or being a device that can generate, record, process, receive, transmit, or display
information that is represented in discrete numerical form. For example, a digital clock indicates the
change in time by displaying the sequence of digits from 0-60 (seconds), without considering the
range between the individual units.

The difference between analogue and digital is that, analogue is using a continuous scale, whereas
digital uses discrete intervals.
Some aspects of UT are not practical or possible to convert to digital; e.g., input Power supply,
transmitted and received ultrasound are always analogue. Howevet, many input controls and some
outputs are feasible as digital signals.
Digitally controlled ultrasonic instruments have many advantages over the older analogue units:
. Accuracy (time or clock-based, instead of deflector plates)
o Repeatability (exact settings can be recalled)
r Storage of settings to memory (all parameters stored)
. Speed of setup (simply recall the stored parameters)
. Signal processing to improve assessments
. Display options, e.g., projection scans (tomographic presentation of data)
o Data recall for reporting

4.7 NDT INSTRUMENTS - OVERVIEW

Test setups may include both digital and analogue equipment. Where several analogue input or
output channels are used in a test system, they are often added to an analogue to digital converter
(ADC) or digital to analogue converter (DAC) through a multiplexer (MUX). A MUX is used to select
which of the analogue signals will be converted al any given time. Figure 4-1 shows a block diagram
of how a variety of analogue and digital inputs and outputs might be arranged on a computer.

69
 Chapter (4): Digitat Aspects Of TOFD

Computers used in data acquisitiory

come in a variety of sizes, formats and
price ranges. At one time, three groups
applied: microcomputers,
minicomputers and main{rames. These
were based roughly on size of
memory. Today, the differences are not
so clearly defined. Except for situations
where the amount of data to be
collected is large and must be
processed quickly, in addition to
performing many other functions
(multitasking), most inspection
systems can be automated with some Figure 4-l Multiplexed inputs and outputs via ADCs and
form of a personal computer. DACs

Integral to any automated data acquisition system, is the data acquisition software. Data acquisition
software is used to collect data, analyse the data and display the results. Without the ability to analyse
and display the results of data collectiory the millions of bits of data that can be collected would be
unintelligible to the average operator. Processed data can be output to monitors or printers in the form
of tables, graphs or even be made to duplicate strip chart or oscilloscope (A-scan) presentations.

Several varietiesof specialised software exist in NDT inspection systems. These are often used in
conjunction with specialised instruments. The data acquisition software collects all aspects of the
signals and records all instrument parameters. Some software also addresses motion control and
positional information.

4.2 ASYNCHRONOUS VERSUS SYNCHRONOUS SYSTEMS

Most ultrasonic technicians, having used a traditional En@derPu,seU
mono-element probe in pulse-echo mode, are familiar UIPRF W
urchi ru
with the concept of PRF (pulse repetition frequency). urch2 M
This is the rate that the ultrasonic instrument pulses wch3 ru
the probe. In a multi-channel system, where several urch4 M
probes may be used, or in a phased array system, PRF Contrclled

where the focal law repetition is controlled, PRF En@derputseU

controlled systems require high clock speeds to ensure UTPRF M
that all the channels are fired in the allowed sample fch 1
u
U
interval. When the encoder position pulses are UTCh 2

interlaced with the ultrasonic pulses, the ultrasonic

UTCh3

UTCh 4
U
U
PRF and position pulses are said to be asynchronous.
Firs On Position
This is illustrated in the upper portion of Figare 4-2. Figure 4-2 Multi element scanning system -
Asynchronous (upper), Synchronous (lower)

70
Chapter (4): Digital Aspects Of TOFD

Synchronising the UT pulsing with encoder position pulses ensures that a1l chalnels are fired in the
sample interval. The only limit using this system is computer through-put rate. Slmchronous, multi-
channel timing sequence is illustrated in the lower portion of Figure 4-2.

When discussing optical encoders we made reference to converting the photodetector voltage to a TTL
output. TTL is an electronic convention that stipulates the form of voltage to be achieved. Because it is
a voltage level, the TTL is actually an analogue signal. But it is used to provide digital responses (e.g.,
on or off conditions).

To be a true TTL signal, certain tolerances must be observed on the voltage. The rise and fall times are
described in Figure 4-3.

FallTime

Figure 4-3 TTL voltage rise and fall times

Tolerances for voltage are also to be observed. The output voltage and current specifications are as
follows:
. The off voltage level lies between 0V and +0.4V
. The on voltage output lies between +2.4V and +5.0V
. The region between 0.4V and 2.4Y is undefined
. The maximum rise/fall time is 50 ns between Vu- and Vrn
Most TOFD systems, where an encoder is used, are slmchronous. The encoder pulse triggers the
pulser and all the other data acquisition events follow in sequence prior to the next encoder pulse.
Data acquisition, by the asynchronous systems, often uses computer algorithms to select the
maximum, minimum or average values of the gated information received by the ultrasonic instrument
between encoder pulses. That value is then transferred to the computer for data display (and to
memory). Synchronous systems have only a single firing for each channel (unless averaging is used),
so the single gated value (per channel or focal law) is transferred to memory for display.

71.
Chapter (4): Digital Aspects Of TOFD

4.3 MEMORY AND DIGITISATION ASPECTS

4.3.1 INSTRUMENT OUTPUTS
Physical properties measured in NDT can include: temperature, pF{, pressure, distance, velocity, mass
or optical, acoustic and electrical energy. The sensors used convert these properties to an electrical
quantity: voltage, current or resistance. The sensors, as a result of changing the physical properfy to an
electrical quantity, are also termed transducers (transducing energy from one form to another).

Electrical data can be considered a signal or "waveform". This is usually a voltage vawing with time.
Since computers prefer digital signals the analogue form of the wave needs to be changed to a digital
form. These are discrete values in specified constant time intervals. If the digital signal amplitude
intervals are small and the time intervals are also small, the resultant digital waveform can closely
approximate the analogue waveform. The process of converting an analogue waveform to a digital
format is done in an electronic component called an analogue to digital convetter. The process is in
several steps:
. Step one in sampling the waveform is to choose a signal range. This is a range of voltage levels
large enough to cover all the variations, up and dowry that the signal makes.
. Step two is to divide the range into two equal halves. In the diagram this dividing line is shown
by the red line. We then determine if the point we are sampling is above this line. If it is above
the line, we assign a value '1', and if below the line we assign a value'0'. (e.g., the first digit in
the 3-digit value of the vertical scale).
. Step three divides the 'half-range' which the sample occupies into two equal halves. (Since in
this case it is the top half, this division is at the yellow line.). Once again, if it is above the line
we assign a value '1', and if below the line we assign a value'0'. (e.g., the second digit in the 3-
digit value of the vertical scale).
We can repeat step three of this process as many times as we like, each time gaining another blf of
information about the point whose level we are trying to determine. In Figure 4-4we have done this a
third time (the blue line). As a resulf we can say that the level is in the band of voltages which we can
represent by the number '101'. The more times we divide the range and determine which side of the
line the sample lies on the more bits of information we get. As the number of bits increases, our
approximation of the amplitude level becomes more precise. In practice, we can stop asking after a
while, because we will reach the level when random noise blurs the signal, making any further bits
meaningless.

By taking a series of samples at equal time intervals along the time-base, we can convert all the details
of the analogue waveform into a stream of binary numbers. In the example in Figure 4-4, tllte wave can
be represented by the series: 710,170,777, 1,1,1,, 101, 100, 011, etc. This is accomplished by taking three
bits per sample for the sampled points indicated by the circles.

72
Chapter (4): Digital Aspects Of TOFD

t1: I
1'r0 !

/,t
!r)1 !
I

t 00
I !I sry*.il
r:tn!'-
Yoh*ge
orr I

oro I
00! |
*ol
finr* ---+
Figure 4-4 Analogue to digital shaping{zs)

By decreasing the time interval between the samples and by increasing the number of bits used at each
sample we can generate a series of dots that very closely resembles the continuous line that represents
the analogue signal.

In order to maintain the "binary" code required for computer communication the vertical division of
thesignalsintobitsusespowersof two, e.g.,26,27,28,ze.Useof 256divisions(i.e.28)hasbeenthemost
common. Sampling along the horizontal axis is time based and a single sample is taken at each time
interval. The closest whole value that the interval corresponds to, in the given time interval, is the
value assigned to the amplitude point.
The process of changing an analogue signal to the computer friendly digital signal is called
digitisation. The electronic device that accomplishes this is called an analogue-to-digital converter
(ADC). The associated electronics to accomplish this conversion are usually incorporated on a printed
circuit board inserted as a separate card in the computer, so the hardware is often called, an A to D
card. The digital signal that results from this conversion is composed of digital values of a known
range termed the scale factor. These values are separated by a fixed time interval known as the
sampling interval.
The reverse process is also used, i.e. converting digital signals to analogue signals. The device that
accomplishes this is called a digital-to analogue converter (DAC). Most people are aware of the
reverse process in entertainment devices whereby music information on a CD or DVD diskette is
played back on a set of speakers (the sound wave ou@ut of the speakers is always analogue).

4.3.2 BIT-DEPTH AND SAMPLING RATES

In the description of analogue to digital conversiory it was noted that typically the vertical increment
scale uses an ADC of 28 or 256 levels of vertical (resolution). This is considered to be an 8 bit
digitisation. Separate from the vertical sampling is tlne sampling rate of the ADC. This will dictate the
time interval along the A-scan that is captured and digitised. Flash converter boards, A to D, are
available in a variety of speeds, typically 20 to 100 MHz, but slower and faster varieties are also
available. If 100 MHz ADC is used, sampling occurs every 0.01ps. This is determined by dividinglby

/J
 Chapter (4): Digital Aspects Of TOFD

the digitising rate (e.g., for 100 MHz it is 1/100,000,000). In pulse-echo, this provides a resolution in
steel of 0.016mm (shear) and 0.03mm (long). Temporal (time) resolution also dictates the quality of
signal reproduced from the analogue. Figure 4-5 shows an analogue signal from a 10 MHz probe.
Digitising at 100 MHz allows reasonable reproductioru with 10 points between each division. But, at
20 MHz, the original analogue trace is just barely recognizable, as a result of only 2 points between
vertical graticule (the dashed line of the analogue trace is provided as reference, only the dots will
appear on the scope).

f' 1i

10MHz analogue signal 10lV1Hz analogue signal 10fu1H2 analogue signal

0.1 ps per division 100MHzADC (10 samples perdiv) 20MF{zADC (2 samples perdiv)
0.'1 ;rs per divisron 0.1 ps per division

Figure 4-5 Digitising effects on oscilloscope waveform quality

For general work in NDT a minimum sampling rate, of four times the nominal frequency of the probe
used, is usually adequate. This will ensure the digitised amplitude will be within 3dB of the analogue
value. Five times the nominal probe frequency is required if the digitised sample is to be within less
than 1dB of the analogue signal amplitude. E.g., for a 10 MHz probe, an ADC rate of at least 50 MHz is
recommended for amplitude critical work. Some TOFD codes recommend a minimum of 6 times the
nominal probe frequenry be used, as the ADC rate. As well, it will be seen by operators that the
quality of the recorded signal at a higher ADC rate is much closer to the original (analogue), and
makes for improved signal characterisation.

Whereas, temporal or distance resolution is solely a function of ADC rate, amplitude resolution is a
function of both ADC rate and the number of levels of sampling e.g., number of bits. For UT data
acquisition systems, 8 bit sampling is presently the most common. .
An important aspect of digitised amplitude is the effect on dynamic range. As noted, the most
common bit sample has been 8 bit digitisation. Accuracy of amplitude assessment is based on the
number of divisions of sampling in the vertical direction. The term bit is derived from binary
treatment of data, whereby there are 8bits to a Byte in computer terminology. Here a bit is one of two
options, i.e. the values '0' or '7'. Computer-based ultrasonic systems have been increasing the bit-

74
Chapter (4): Digital Aspects Of TOFD

sampling used and the higher-end units are sometimes using 12 bit digitisation. The product of the
bit-size is the number of samples that the vertical (amplitude) range can be divided into, e.g.:

8 bit = 28 :256 intervals of vertical sampling

10 bit : 210 = 7024 intervals of vertical sampling
1.2bit:212 = 4096 intervals of vertical sampling

4.3,3 RF VERSUS RECTIFIED DIGITISATION

Ultrasonic signals begin with the A-scan display. In its raw form, the impulses are seen as voltage
variations on either side of a zero volt baseline. When the variations on either side of the zero volt
level are preserved as the display format, the format is called the radio-frequency or RF display. This
is the form used in TOFD. In pulse-echo flaw detection UT, the convention has been to rectify the
signal so only the positive portions of the signal are displayed. This can be accomplished by
eliminating the negative voltages displayed (positive rectification); or by processing the negative
portions, such that their sign is changed to positive and then flipped to also appear on the positive
side of the voltage base line (fullwave rectification).

This can be illustrated graphically. Figure 4-6 shows an RF waveform and a rectified waveform
presented on a graph with a colour code for amplitude on either side. The amplitude scale on the
TOFD is indicated from negative 128 to positive 127. With the added zero at the midpoint this
accounts for 256 levels. The offset is to ensure that all the negative voltages are equally positioned, as
in our first step of digitisation where we set signal range. This is the range of voltage levels large
enough to cover all the variations, up and down, that the signal makes. For the rectified signal, zero is
the lowest value, so all voltages displayed are above the zero line.

\ I \\
I \
\
I
I \
\
I
\ a\i \
\

I Bit F.F l-l')FDl

Signa t-1,)FDl E Sit ll-'ctllied Sign.:l lPirlse-echo;

Figure 4-6 Sbit digitising effects on dynamic range

Signal amplitudes are usually stated in dB and the concept of dB is simply a ratio from:

)D h'
aa = 20log,o(l,r) (41)

Where; hr and hz are the relative amplitudes of two signals.

75
 Chapter (4): Digital Aspects Of TOFD

For a rectified signal in an Sbit ADC unit, the voltage bias places the zero point at the bottom and
shifts all points positive, so the dynamic range is (l/256) or 20logro(a) = -4}d.n

The smallest 7" interval on the screen is (1/256) x 100 = 0.39o/o

For the RF signal, the same 8 bit ADC has no bias and signals are positive and negative. The dynamic
range is determined from the zero point to the maximum displacement (i.e.128).

Therefore the dynamic range will be (L/1.28)or 20logro *= -nrb,

NOTE: half the amplitude is -6dB so reducing the number of points by a factor of 2 reduces the dB
dynamic range by 6dB (i.e. 48 - 6 = 42) and the smallest vertical screen interval is 0.8% for the RF
display. \zVhen we use higher bit-sizes there is an interesting bi-product.
Large amplitude signals that are 1,00"/" or gteater, of course, cannot be assigned a real value. They are
simply considered "saturating". This applies to analogue or digital displays. When amplitude is an
important factor and its absolute value is required, this means that a re-scan is required to assess the
actual amplitude with respect to the reference level. In an 8 bit digitised rectified signal, once the
signal has reached the 256 level, it is saturated. A signal greater than 256 levels may be 101% or it may
be >500% with respect to the full scale display. There is no way to determine how much saturation has
occurred.

With a 10 bit digitisation rate, the vertical range of any signal is divided into 7024 equal intervals. This
allows for the collection of signals at a lower receiver gain; and after the data is collected,
electronically adding gain. With 1024 amplitude levels, the resolution is four times that of the 8 bit
systems. That means calibration at a reference level of 20% can be done (instead of the typical 80% on
an 8 bit system). Thus all the A-scans are collected at lower amplitude. Signals on the new 10 bit
display, reaching 25o/" screen heighf would be 100% on the 8 bit display. Therefore, the likelihood of
getting troublesome saturating signals, when using the 10 bit digitisatiory is greatly reduced. Using
the same assessment of dynamic range, as for the 8 bit system, the 10 bit system is seen to have a
dynamic range of 60dB for rectified signals or 54dB for RF signals.
There are two methods that this extra dynamic range can use. One uses the standard display, whereby
the 100% level on the display is the maximum displacement of the signal amplitude. Thery setting the
reference level to a lower point on the amplitude scale allows a direct reading off the scale for the
amplitude. Another method is shown in Figure 4-7, where the gates are used to measure the
amplitude and even though the display no longer shows an increase in signal level, the operator can
read the measured amplitude as a digital numeric output. In the example, the digitisation is 9 bit and
the gate output of amplitude shows that the signal on the extreme teft is indicated as having 200./"
amplitude and occurs at 10.00mm. The next signal has a separate gate (green) and the signal also
saturates the "display" , but r:rot the gate level. That signal is indicated as having an amplitude height
of 776.1% and occurring at 20.03mm.

76
Chapter (4): Digital Aspects Of TOFD

200.0 10.00 mm 176.1

^", o" 8"/"

"r.

Figure 4-7 gbit digitising gate display to increase dynamic range

4.3.4 LOGARITHMIC AMPLIFIERS

Since the earliest days of ultrasonic inspection the display format used on the A-scan, related
amplitude to voltage. This is a logical treatment since the piezo-element voltage response is linear
with respect to applied pressure. However, the instrumentation used in NDT preferred to use gain
adjustment knobs that were rated in dB instead of voltage. Since dB is a logarithmic ratio, it means
that in order to double the applied voltage, 6dB of gain are added. Such an increase means that the
amplification is linear with respect to voltage.
At some point in the past, some manufacturers of UT instruments decided to provide an amplification
that was logarithmic with respect to voltage. This means that, by doubling the dB setting on the
attenuator (or gain dial), the signal amplitude being displayed doubles. This type of instrument uses a
logarithmic amplifier. One can imagine that a response from a reference target being set to 45"/" screen
height on an instrument with a logarithmic amplifier has the gain setting at 40dB. When the operator
doubles the gain setting to 80d8, the signal increases to 90Y" of the screen height. The dynamic range
of logarithmic amplifiers is significantly greater than that of linear amplifiers.

Converting the concepts of linear amplifiers to logarithmic amplifiers involved a re-thinking of the
way operators dealt with ultrasonic signals. The conversion from the traditional linear amplifier units
to the logarithmic amplifier units was not popular and as a result not many are found in use today.
However, with the use of computer technology to simplify operator functions, there is a great
opportunity to take advantage of logarithmic amplifiers, once again. One of the manufacturers of
TOFD equipment that has taken advantage of logarithmic amplifiers in its system is Force
Technologies (Denmark). They have incorporated a logarithmic amplifier (log amp) that provides
120d8 of dynamic range. Since this is a fixed amplificatioru there is no need for a hardware range
selector (dB attenuator) or hardware implemented TVG-circuit.

Figure 4-8 shows the input signal displayed in logarithmic scale. In the logarithmic mode all the echo
details within an amplitude range of 120dB are recorded, whereas the linear recording can only cover

77
Chapter (4): Digital Aspects Of TOFD

a range of 48dBfor the rectified signal. The signal displayed in a logarithmic scale looks different from
the conventional linear echo signal. Since the log amp is a fixed hardware setting, virtually all signals
can be accommodated without saturation.

-tt-..,,,-. ....1:..:':.:::--'-::l- :: '....,...::'--. . ... :: -:

Figure 4-9 P-Scan linear display

78
Chapter (4): Digital Aspects Of TOFD

A software conversion routine is included, in the Force P-scan system, to optionally display a linear
echo signal on the screen. In Figure 4-9, the same signal is shown as it would appear in linear scale.
When the maximum signal is set to 80% screen height, one can see how signals, 40% screen height at
the midpoint of the horizontal trace of the log amp display, are not seen at all on the linear display.

In addition to the data collection by a log amp, the operator has the option to digitise data in 8 bit, 9 bit
or 10 bit format at the time of collection. In a logarithmic display the bit-value directly relates to the
amplitude resolution in dB, as follows:
. 8 bits: 1.00d8
. 9 bits: 0.50d8
o 10 bits: 0.25dB

The advantage of this acquisition technique can be seen by observing the display of a TOFD signal.
Figure 4-10 illustrates two views of a TOFD image using the log amp. On the left the display is
adjusted to produce an amplitude range of 80d8. This means that from the zero volt level, at the
middle of the A-scan, 80dB of range exists on either side. At this level, the lateral wave is barely visible
and the back-wall and mode-converted signals, at approximately 40mm and 70mm depths
respectively, have amplitudes of approximately 30% screen height. The image on the right has the
same 80dB amplitude range but the display is adjusted so the green bars, indicating the 80% screen
height level, are at 35dB. This provides a lateral wave that is approximately 30"/" screen height.

Figure 4-10 P-Scan log amp display of a TOFD scan (80d8 full screen left and 50dB full screen right)

The same scan portion is displayed in Figure 4-11. The image looks very similar to that using the log
amp, where the lateral wave is set to 30% screen height. The only significant difference is the lower
amplitude signals in the log amp display are not present in the linear display and the back-wall and
mode-converted signals saturate (i.e. exceed 1007o screen height.

79
Chapter (4): Dieital Aspects Of TOFD

5 20

Figure 4-1L P-Scan linear amp display of a TOFD scan (30% lateral wave using 60'beam)

The advantages of a log amp are not limited simply to increased dynamic range. The fact that
increased dynamic range can avoid unrecoverable saturating signals, allows for subsequent digital
processing of the entire waveform. This also means that no rescans are required where amplitude data
is required.

4.3,5 FILE SIZE CALCULATIONS

Because data is saved to computer-readable memory/ there is often concern for how much memory
will be required to store the data. Images we use in TOFD, to construct the TOFD B-scans, are
composed of the A-scan waveforms.

Digitising an A-scan is the first step in constructing a B-scan. But each sample must be saved to
computer memory, therefore, larger scan lengths and larger time of the gated period, require more
memory than small scans and gated times.
File size (f.! is given by the equation:
FS=txf,xBxlx.S (4.2)
\A4rere;
t: Time required for the sound wave to traverse the desired material distance and back
f : Digitising rate or sampling frequency
B: Number of bytes of amplitude information per sample point on each A-scan
l: Length of scan distance travelled
S: Number of A-scans collected per mm of travelled distance

80
Chapter (4): Digital Aspects Of TOFD

As an example, consider a simple thickness scan using a 5 MHz contact normal beam probe on a
25mm thick plate. Use a basic ADC with a bit depth of 8 bits per sample.

The goal is to gate the entire thickness for display, by using the ISO standard's recommended
minimum ADC rate, of six times the nominal frequency, so the ADC sample rate is 30 MHz. Consider,
also, that 50mm is the time equivalent for a longitudinal wave to cover just the 25mm thickness (i.e.
25mm to the far side and another 25mm to return to the probe). Since the velocity of a compression
wave is 5.9mm/prs, the time required for the wave to traverse 50mm is:

t: g#: B.45prs

At 30 MHz ADC, the digitisation rate or sampling frequency f,(samples per ps) is:

3o,ooo,ooo YSgq
sps
: 3oto*Pl"t

If a portion of the A-scan display of 8.45prs (f as calculated above) is gated, the number of samples
recorded for each A-scan is 30 x 8.45 : 253.5 samples.

At each point, the number of bytes of amplitude information, i.e. bytes per sample point on each A-
scan collected, is 1 byte (8 bits). This is the symbol B in the equation.

If the 0" B-scan is to be collected across a weld and includes heat affected zones, a sufficient scan
distance travelled is 50mm either side of the weld centreline (i.e. l:100mm). If an A-scan is collected at
every 0.5mm interval, the number of S scans per 1 mm is 2.

By substituting in the equatiory the size of data generated or file size is:
FS=txfrxBxlxS
=8.45x30x1x50x2
= 25350 bytes (-25k8)
To generate full volume scans, for depth encoded C-scans and D, E, and/or P scans, requires several
such scans be made in a single process. Even if a small square 100x100mm is scanned, with a 1mm
raster step with the above conditions, 2.S4Megabytes would be generated.

A TOFD scan on a relatively thin plate (e.g., 25mm thick), with a 20ps time window, gated for the A-
scans using a 10MHz probe with a standard 100MHz ADC, can be made on a weld 3 metres long. At
1mm per sample, the file size of even such a simple linear scan, parallel to the weld axis, could quickly
result in file sizes of several tens of Megabytes (MB).

4.3.6 DATA PROCESSING

An added feature of digital data acquisition systems is the ability to perform subsequent processing of
the stored signals. Several techniques have been derived to enhance the information collected. This
process is generally termed digital signal processing (DSP). DSP will be further discussed in next
chapters.

81
 Of TOFD

include:
Effectiveness of DSP relies on the quality of the captured signal. Factors determining quality
. How well the transducer and data acquisition system are matched
. SamPling period
. Signal quantisation level (bit depth)
o Calibration
o Material attenuation
Any unwanted disturbance, in the useful frequency band that is introduced to the signal, is
spurious
considered noise. Noise may have several sources: the transducer itself, instrumentation,
noise'
waves from scatter, geometry and mode conversions, as well as surrounding electricaf
entraPments for
Defects may originate in areas where geometric configurations form stress raisers or
may be corrupted
chemicals that can lead to corrosiory cracking or both. Defects occurring in this area
geometries may be misinterpreted
or completely masked by the surrounding conditions. Conversely,
the big
as defects. B-scans, C-scans or other imaging displays allow defect detection by illustrating
picture, where subtle trends are noticeable that might not be evident in the static A-scan display.
a result of
DSp is mostly used in the analysis process. There are many DSP options now available, as
computer processing. These analysis tools will be further examined in a later chapter.

A process option, that is used prior to analysis and at the time of acquisition, will be considered for
now. The process is signal averaging.
4,3,6.1 SIGNAI AVERAGING
A somewhat simpler form of processing is signal averaging. Signal averaging allows a flaw signal to
be drawn out of the background noise. This is the principle that a flaw signal is coherent,
but noise is
not. A coherent repetitive signal added to itself "n" times will increase by a factor of "n", whereas
noise added to itself 'n' times will increase by the square root of "r(' . Aflet "n" iterations, the signal to
noise ratio of the averaged waveform is improved by ",li.The effect of such simple averaging seen
is
in Figure 4-12.

I st Average

10th Aver;lge

l00th i\vcrage

Figure 4-12 Signal averaging {zr)

82
Chapter (4): Digital Aspects Of TOFD

There is a subtle difference between signal averaging and other one dimensional DSP. Signal
averaging is done as part of the data acquisition software and requires the pulse to be repeated several
times. Signal averaging is therefore not a "post" digital processing operation. \zVhen averaging is used,
the output of the instrument is displayed to provide the average of the repeated pulses received. It
would not be possible to take a single acquired signal and then average it against itself since every A-
scan would be just a repeat of the others and no "average" difference would result.

4,3.7 SCANNING SPEED

If large gated times are required for B-scans, the ability of the computer to process the information
may present a speed limitation. If the computer has not had enough time between sdmple intervals,
blank lines on the B-scan result, indicating missed data points. The throughput processing ability of
the computer will be one of the most important limiting factors to the scanning speed of any TOFD
system.

The other most significant factor inlluencing scanning speed will be how robust the mechanics of the
scanning system are. At some point, the speed of the scan may be such that the probe holder may not
be able to maintain good contact or the scanner risks shaking apart. Empirically established scanning
speeds may be found, or specification or code can stipulate maximum speeds that can be set, based on
probe or beam size and PRF.

Scanning speed is the maximum speed at which a system can operate, which in turn controls the
minimum time it will take to scan a part. A rule of thumb is often used in ultrasonic data acquisition
systems. It requires that at least three firings of the ultrasonic pulse occur, for each channel, over a
distance equal to the 6dB dimension of the beam. In fact some Codes or Standards actually state this
in an equation format:

tl-
W.
'
x PRF
v-a (4.3)
J
Where;
7: Scanning speed
Izll": Narrowest -6dB width at the appropriate operating distance of the transducer determined by
design requirements
PRF: Effective pulse repetition frequency for each transducer

This example requires three firings within the 6dB beam width.

In a system where many probes are sequenced via a multiplexer, the PRF is divided among the total
number of probes. Although many units have PRF's of 2kH4 when 20 probes are used in the system,
the effective PRF at each probe is only 100 Hz. If the -6dB beam width is 3mm and the effective PRF is

l[OHzthe maximum scanning speed should not exceed ry 3 =100 mm/second.

This equation addresses scanning speeds for asynchronous systems. The equivalent is maintained for
synchronous systems, by having three samples within the 6dB beam width. Therefore, for
"synchronous" or "fire on position" systems, 1 sample every 1mm will achieve this, if the beam width
is 3mm or greater.

83
In TOFD setups, the beam width is significantly greater than 3mm; however, in order to provide an
image that is adequately smooth (i.e. not pixelated) and to allow for length sizing estimates
typically
."qui."d by codes, the sample interval is kept reasonably small. Sample intervals of 0.5mm may be
used for thinner materials and a maximum of 2mm sample intervals are usually allowed
for materials
over about 50mm.
Many functions are occurring in the UT system, based on the initiating pulse from the encoder'
Typically systems use scan speeds around 100mm/sec. This means that the sequence of events
required to carry out all the steps in a single millimetre, are repeated 100 times each second.
There is still a limit to what a computer can do
in a short period of time and if the scanning
speed is too fast to complete all the functions
required in the sample interval, then all the
information for that step is lost. On a TOFD B-
scan this is seen as a constant grey line along
part or all of a sample interval on the display.
The effect is seery when the scanning speed is
too higlu as shown in Figure 4-13. Portions, of
the black to white cycles, are seen to be lost
Figure 4-13 Missing data Points
and replaced by grey lines.

4.3.8 ACQUISITION RATE

The ability to scan without missing data poi.nts can be assessed by the acquisition rate' This is
generally the number of A-scans per second that can be collected. When simple amplitude or time
data is collected the concept breaks down. A sample of the amplitude, in a gate at a single location,
requires only 2 bytes of memory (one for the position and one for the amplitude data). But in general
the maximum acquisition rate is determined by:

. Length of A-scan being collected

o Digitising frequency
o Data compression if used
o Instrument PRF setting
. Averaging if used
In general, the inspection speed, divided by the scan axis resolution, must be less than the acquisition
rate in A-scans per second.

Inspection speea ff) 1 Acquisittonrate (A - scans per second) (4.4)

Scan axts resolution (mm)

84
Chapter (5): TOFD Calculations

CHAPTER (5): TOFD CALCULATIONS

In earlier chapters we briefly touched on the loss of resolution that occurs in the lateral wave due to
the ring-time effects. This was one of the limitations of TOFD. Most users of the TOFD technique are
familiar with the loss of information that occurs immediately under the test surface. This is sometimes
referred to as the lateral wave dead zone. It results due to the pulse ring-time limiting the resolution of
flaws, immediately below the surface, for a time approximately equal to the depth eqgivalent to that
ring-time for the probe configuration used.
However, some users of TOFD seem to be unaware that the same effect occurs at every diffraction
interface below the surface and a similar effect is had from the back-wall reflection boundary.

These three "dead zones" are identified and approximated in the EN standard EN-583-6. No special
corrections are made for wedge angle variations with depth and the assumption is made that the
indication is at the midpoint of the Probe Centre Spacing (PCS). It is important to note that these are
approximations; however, they are generally adequate for most applications.
Estimations of resolution limits of TOFD are one of the most important considerations, when using
TOFD with fracture mechanics-based acceptance criteria. When using TOFD to estimate flaw size
(height), the smallest resolvable flaw is a function of the PCS, probe frequency and damping quality as
well as the depth of the flaw below the surface. Failure to understand this could lead to grossly
overestimating the capabilities of a system and impossible expectations of TOFD capability. 8.g., a
specification issued for a large corporation required extreme sizing capabilities on the inside surface of
a putatively critical component. Small surface breaking flaws less than 0.5mm vertical extent were
considered critical and TOFD was identified as the means to both "detect" and "size" any flaws on that
surface. This seems to have been the result of an assumptioru that the improved time resolution, as
one approaches the far-wall, would allow sizing and detection on this minute scale. Such assumptions
ignore the ring-time limits. Even when using a 10MHz probe with a 45" L wave, detection and sizing
capability of <0.5mm would be unlikely.

5.1 DEPTH AND RING-TIME CALCULATIONS

Calculations of the three dead zones are derived from the Pythagorean Theorem. This is the simple
trigonometric equation relating to the right-angled triangle. Figure 5-1 illustrates the basic layout and
associated A-scan. It is assumed that the ultrasonic energy enters and leaves the specimen at the index
points of probes. When a flaw is assumed to be midway between the two probes, depth "d" to the
flaw is given by:
Where c is the acoustic velocity, to the time in the wedge and S is half the probe spacing (i.e. half the
PCS) and t is the time from the initial pulse to the upper tip signal.

85
Chapter (5): TOFD Calculations

l';gnr. gcn.rrt.d rslnF ls]erm I onl

Figure 5-l Calculating depth to upper tip

Simply identifying the same parameters for the lower tip signal allows the flarv height to be
determined. This is seen in Figure 5-2'

+2xto (s.1)
f:

o: (r - zto)2 - s2 (5.2)
^l(;)'x

Figure 5-2 Flaw height from TOFD

86
Chapter (5): TOFD Calculations

Depth dz is calculated using the same formula, as used for the upper tip, so we need only subtract the
depth to determine the flaw height. This is then given as
h=dz-dt (5.3)

The same parameters can be used to identify the limits of resolution in TOFD. The latera.l wave is seen
as a pulse. This pulse requires a set amount of time based on its frequency. The start of the pulse may
be considered the entry point ot zero depth. Therefore, the time required to complete the ringirig time
defines the pulse duration. Typically a broadband TOFD transducer is designed to have a ring-time of
not more than 1.5 cycles. When seen on the A-scan the lateral wave has a depth, in the metal,
equivalent to the pulse duration. This is shown in Figure 5-3 and given the symbol ti,. Note that we
typically measure pulse duration from the point where the pulse first rises above the 10% screen
height level and finally drops below the 10% level.

lnitial

To calculate the equivalent depth in the test piece we solve Pythagorean Theorem using the acoustic
velocity and pulse duration to determine the travelled equivalent distance for the lateral wave,
correcting for the fact that it is the total distance from transmitter to receiver, and half the PCS.
The ring-time near the test surface (Dar) is then defined by the pulse-duration:

E,*,
Da,: l-l-*Sct,
*
(5.4)
",J
Where;
tr: Pulse duration to where the amplitude is 10% of peak
S: Half the PCS
c : Velocity of sound of the mode used

From the equation, we c€u:r see that reduction of the lateral wave dead zone is made by decreasing PCS
or using probes with shorter pulse duration (and to some extent a higher angle).

Back-wall signals present another source of lost information due to the ring-time of the pulse.
However, because the angle made at the back-wall is smaller, the equivalent depth for the same time
is less. The equation to calculate the back-wall dead zone uses the fact that the reflection of the pulse
uses up the equivalent depth and subtracts the wall thickness from the total pulse-duration equivalent
depth.

87
 : TOFD Calculations

The equivalent depth ring-time near the back-wall surface (Da)is defined by the equation:

c2(t,^, + t^)2
Dd*: -- 4------ - s2 - w (s.5)

Where;
fr: Back-wall time of flight
W: W all thickness of the component

Reduction of the back-wall dead zone is accomplished by decreasing PCS or using probes with shorter
pulse duration (and to some extent a smaller angle)

Spatial resolution defines the ability to resolve upper and lower tip signals (between the lateral wave
and back-wall):

c2(ta + tr)z
-------.--._-_J.
4 -d (5.6)

Where;
tr: Length of the acoustic pulse
ta: Time-of-flight at depth d

Resolution increases with increasing depth, and can be improved by decreasing the probe separation
or the acoustic pulse length. If the flaw is less than the ring-time, at the depth calculated, it will not be
possible to provide an estimate of flaw size, based on the TOFD principles of much less than the ring-
time equivalent depth.
For the most part TOFD probes are relatively broadband, i.e. have short pulse durations on the order
of a single cycle or 1.5 cycles. Since these are approximations, for ideal conditions, they are to be used
as a guide only.
As an example, we imagine a pulse from a probe at three depths in Figure 5-4. Scaled values for
imaginary distances are used from the exit point to the midpoints of the PCS. The image indicates the
distance of the PCS is 50mm and a wall thickness of 30mm.

Using the equations above for a SMHz probe with a 1.5 cycle ring-time, the equivalent depths for the
three conditions can be indicated.

The equivalent depth distance for the lateral wave to occur from its initial rise to the end of the pulse
ring would be equivalentto 6.7rrrrrr.
The equivalent depth distance for the pulse at a depth of 16mm from its initial rise to the end of the
pulse ring would be equivalent to 1.6mm.

The equivalent depth distance for the pulse at a depth of the full wall thickness of 30mm would be
equivalent to 1.1mm.

B8
Chapter (5): TOFD Calculations

Figure 5-4 Position resolution limits for TOFD

To validate the calculations, Figure 5-5 provides the calibrated cursor values for a TOFD scan, where
the scan is of a weld section 32mm thick using aTMHz probe with a 100mm PCS. This PCS provides a
crossing point deeper than the recommended 66"/"t,but the project it was used on was using the TOFD
setup to improve the assessment of the root region.

The calculated and measured values are provided on the images for the lateral wave, back-wall and
spatial resolution, at the positions indicated from left to right in Figure 5-5.
FrS::i:r=@83<re
': :::i;ailtt:-::::*

re A:*i..-i+:*:i* :]: :illf

iffi i*i+w sw
E
*. .*: .::$ari.: '
r$$$*g:::'=s:1:n:s,$..$irii
'::::
€wc$re,*+
aEt.:tfst::'- #Sffi l&tr:",.......t'....,t :ii
::::;t::i.t''
ffi
itl.=::!p:.r:
k

ry rr
*
n re
.
rc ...: a:::::ir::::::t'" .''

Figure 5-5 Confirmation of dead zone calculations for TOFD

ESBeamTool software (from Eclipse Scientific Products) solves the ring-time equations and identifies
the locations on a cross sectional view of the weld. For the identical conditions, the ESBeamTool
model in Figure 5-6 indicates the ring-times as 7.98mm for the lateral wave, 1.16mm for the back-wall
and 2.5mm at the 11.5mm depth" where the flaw is located.

89
 Chapter (5): TOFD Calculations

m
,m.8ffi

I 5fffl

&.*---
Figure generated using

Figure 5-6 Modelled dead zones for TOFD

5.2 FLAW POSITION ERRORS

Due primarily to uncertainties, in the main parameters in the equations above, there is a potential for
error in depth estimates. Uncertainties exist for timing, acoustic velocities, PCS and the lateral position
of the flaw (i.e. the actual position of the flaw between the transmitter and receiver probes). These
uncertainties can be reasonably approximated and the potential error in depth can be quantified.

5.2.1. TIMING ERROR

Timing errors result from the fact that the time base becomes digitised in order to display data on the
computer. Error in timing is primarily a function of the digitisation rate. For a 100MHz ADC rate the
error would be plus or minus one sample interval, or 10ns. For 50MHz the error band increases to
+2Ons.

The depth error due to timing inaccuracy is quantified by the equation:

c x dt,[1a4 97
6d: 2d
(5.7)

Where;
c: Acoustic velocity
d: Depth to the flaw
6t: Timing uncertainty
S: Half the PCS
This incorporates the timing uncertainty into the Pythagorean equation; thus the depth error, due to
timing uncertainty, is also a function of flaw depth.
Using the parameters in our example, in Figure 5-4, the error in depth estimate, at 16mm assumed
depth for a digitisation rate of 50MHz (i.e. 20ns per sample interval), would result in a depth
uncertainty of 0.2mm.

90
Chapter (5): TOFD Calculations

5.2.2 ACOUSTIC VELOCITY ERROR

Acoustic velocity is another source of uncertainty. It has been assumed by most users that the acoustic
velocity for the compression mode in steel is 5920mls.In fact, the acoustic velocity can vary depending
on the alloy and rolling or casting conditions. Even basic, low alloy carbon steel can vary from about
5800m/s to 6100m/s or more. This is further complicated by the anisotropy of some steels. It implies
that velocities can vary with changes in direction (e.g. . one velocity is observed at a 30o refracted
angle and a different velocity is measured at say a 70o refracted angle. Even when the assessments are
relatively accurate there is some uncertainty and +30m/s is about as accurate as most measurements
can be.

The effect on depth estimate error due to just acoustic velocity error is given by the equation:

dd = d'c
dz + .s2 - *[@1 sT (s.8)
cd
Where;
c: Acoustic velocity
d: Depth to the flaw
dc: Uncertainty of acoustic velocity
S: Half the PCS
Using the parameters in our example, in Figure 5-4, the error in depth estimate for a possible error in
velocity assumption of 50m/s from the assumed 5920mls would result in a depth uncertainty of
0.07mm at an assumed depth of 16mm.

Although velocity can vary with angle in a specimen (called anisotropy) the effect is not usually
significant in TOFD. Velocity of a material can easily be determined using the ultrasonic instrument in
pulse-echo mode with a 0" probe. This is done by accurately measuring a section with parallel faces
using a Vernier calliper and obtaining the pulse-echo time between the back-wall and next multiple
using a digitisation rate of 100MHz. Velocity is then obtained by

v--2dt (5.e)

Where;
7: Velocity
d: Sample thickness (physically measured)
t: Time interval (measured by pulse-echo)
Using these simple tools, it is reasonable to expect thickness to be accurately assessed to within +/-
0.01mm and time +/- 0.011rs. This provides velocity accuracy of approximately +/-50m/s.

5.2.3 PCS ERROIT

Probe Centre Spacing (PCS) might be considered an assumed absolute value. The setting is usually
fixed by a ruler being slid under the probes and a simple alignment of the exit point scoring with the
ruler markings. Parallax is one source of error here, as is rounding off the value. Another source arises

91
Chapter (5): TOFD Calculations

during the scanning. The probes mounted in holders are often seen to move slightly due to
mechanical springing action. They can be seen moving relative to each other. These motions can be
one of two basic forms; closer and farther apart. This is more likely when weld caps are large and can
catch the probes. In addition to the simple in and out motion there may be a slight skewing (one probe
centreline does not align with the probe opposite). The skewing may occur due to rough surfaces or
mechanical recoil as the probe pair slides along the weld cap causing the probes to be moved nearer or
farther from the weld centreline. This may result in one probe being slightly ahead or behind the
other. These effects are easily seen as a wandering of the lateral wave.

The effect on depth estimate error due to PCS variation is given by the equation:

dd=dSx (5.10)
d
Where;
d: Depth to the flaw
d.9:Error (variation) in PCS
S: Half the PCS

Here only the depth estimate and the PCS are involved in the error calculation

Using the parameters in the example, in Figure 5-4, the error in depth estimate for a possible variation
of 1mm, from the assumed 100mm, would result in a depth uncertainty of 0.14mm at an assumed
depth of 16mm. This parameter has a fairly significant effect on results. Variations of 3-4mm, due to
mechanical springing, are not uncommon. A depth variation of a millimetre or more can easily result
depending on the depth to the flaw.

5.2,4 LATERAL I]OSITION ERROR

Finally, the last source of error to be discussed occurs due to the fact that there is some uncertainty as
to the flaw's exact position between the probes. This is considered "lateral positioning uncertainty".

This lateral uncertainty is given the notation dy and is best explained by the illustration in Figure 5-7,
where it is seen as the displacement of the actual flaw from the theoretical centreline between the
probe exit points.

rigur. {e'elnlrd !.in* f5$c,1in Iool

Figure 5-7 Lateral uncertainty of flaw position

92
Chapter (5): TOFD Calculations

Depth error estimation then incorporates the lateral uncertainty using the equation:

(5.11)
46v'
vzv
In a weld inspection, there is some practical limit to how far off the centreline the flaw mighl occur.
Typically, this will not be farther than the heat affected zone (H.AZ) dimension. For the example in
Figure 5-4, the error in depth estimate for a flaw 10mm from the weld centreline would be 0.25mm at
an assumed depth of 16mm.

5.2.5 CONSTANT TIME LOCUS

Related to the lateral position uncertainty is a concept knor.tm as the "constant time locus". For a given
depth, the time taken for the sound to travel from the transmitter, to the flaw tip and then to the
receiver, is at minimum when the flaw is midway between the transmitter and receiver. However, as a
result of the lateral position uncertainty, the time valtJe "t" could be experienced by *y combination
of soundpaths (i.e. t = h I tz).
(See Figure 5-8) where t, and t2 are the path lengths from the transmitter to the flaw tip and the flaw
tip to the receiver respectively.
In Figure 5-8, it is seen that this constant time traces out a semi-ellipse with foci at the probe exit
points. Theoretically, this implies that a flaw, seen on the A-scan at a given time, can have originated
from any point along that locus. Flowever, there is a practical limit to this set of points. In order for a
flaw to be detected, it needs to be located inside the effective beam of the probe. This then limits the
lateral positional offset that the flaw might have. In Figure 5-8, the shaded region indicates the
effective beam area where a flaw might be detected. If the flaw is located at the mid-point between the
transmitter and receiver, it will have a maximum assessed depth for that time. If the flaw is offset to
one edge or another in the beam, its actual depth will be slightly less (as calculated using the depth
error estimation for the lateral uncertainty equation).

\Loo*olequat t*he
_ ilr
-,j

1:iqure q.nr.ili&1 usiilg lSRenmI ool

Figure 5-8 Constant time locus

93
 Chapter (5): TOFD Calculatiolq

The practical observation in this theory relates to how the parallel scan is used in TOFD. Lateral
position uncertainty prevents us from assessing the true depth of a flaw with just a non-parallel scan.
Therefore, it will be necessary to perform a parallel scan to improve depth estimates. Once the
operator locates the flaw using the nonparallel scan, they then position the probe pair at the flaw and
offset the pair, so that the transmitter or receiver exit point is centred over the weld centreline (if the
weld cap is not removed the extent of offset can be restricted). The encoder then needs to be rotated
90",if the positional data is to be collected. Alternatively, a free-running data collection canbe made
(i.e. time based as opposed to accurate positional based data acquisition). The operator then moves the
probe pair perpendicular to the weld centreline such that the opposite probe-exit point reaches the
weld centreline.
The result is an arc as the beam passes over the flaw. The point on the arc exhibiting the shortest time
(apex of the arc) will correspond to the probe pair position that straddles the flaw, such that the
transmitter and receiver are equidistant from the flaw. Figure 5-9 illustrates the sort of motion (left)
and the image shape that will result with a parallel scan. The B-scan on the right side of Figure 5-9
illustrates the arcs that form, from a parallel scan over 6 notches on the opposite wall, ranging in size
from 0.5 to 10mm height (0.5, 1, 2,3,5,10mm) in a 20mm thick plate.

Figure 5-9 Parallel scan arcs

The error on height estimation of small internal flaws is negligible. The operator should be cautious
and watch for small flaws situated at the back-wall. The constant-time locus will compound the
problems of dead zone due to pulse width. Even slightly larger (higher) flaws can be poorly resolved
and go undetected, if located far enough from the PCS midpoint. This shortcoming on the back-wall
can be addressed partially by increased PCS and by additional offset scans. Offset scans are those
where non-parallel scans are made with the PCS midpoint not equal to the weld centreline.

5.2.6 COUPLANT THICKNESS

Another source of depth error estimate, indicated in B5-7706, is couplant thickness variation.
Although it is possible to use immersion methods for TOFD, most practical applications use contact
testing. Couplant thickness should, however, not be a concern for contact tests. In fact, the coupling
gap should be kept to a minimum. Ideally it should be 0mm, i.e. no gap between the wedge and test
piece. Operators, who adjust wear pins or other gap control mechanisms, do so at the expense of gross

94
Chapter (5): TOFD Calculations

deterioration of the resolution capabilities of the TOFD technique. A gup between the refracting
wedge and test piece results in a portion of the sound being reflected in the couplant. This results in a
double lateral wave, due to the direct and then the couplant multiple pulses entering the test piece.
The effect is frequenry dependent but typically more than about 0.1mm gap (the thickness of a sheet
of paper) will result in an interfering multiple lateral wave.

For those concerned by the effect of local dips (areas blended by surface grinding) where the couplant
thickness might increase, there is an equation for the possible depth error:
6Wxvxd
6d: (5.12)
+cxr[1a,+s)
Wherei
dW: Change in thickness of couplant
u : Velocity in the couplant
Using the parameters, in our example in Figure 5-4, the error in depth estimate at 16mm assumed
depth for a 1mm water gap will result in a depth uncertainty of only 0.01mm.

5.3 HYPERBOLIC ARCS

Both real and modelled images of TOFD data have been used in our illustrations. Common to all of
the indications, seen between the lateral wave and back-wall, has been the curved shape of the
indications as the flaws are approached and then moved away from. The shape of these curves is
considered to be hyperbolic and the details of the arc depend on the probe and scan direction used.

The origin of the hyperbolic shape comes from the fact that we can approximate our sound field as a
cone emanating from the element. \zVhen a flaw enters the conic field, its path effectively forms a plane
that cuts along the cone and the locus forms a conic section. In mathematics, a conic section is a curve
obtained by the intersection of a cone with a plane. A conic section consists of points whose distances,
to some point called a focus and some line called a directrir; are in a fixed ratio called the eccentricity.
Three types of conic section are usually
recognised, the hyperbol4 the parabola and
the ellipse. The circle is a special case of the
ellipse and is therefore sometimes called the
fourth type of conic section. The type of conic
section is determined by its eccentricity. Those
with eccentricity less than 1 are considered
ellipses; those with eccentricity equal to 1 are
parabolas; and those with eccentricity greater
than 1 are hyperbolas. In the focus-directrix
definition of a conic, the circle is a limiting case
Figure 5-10 Three basic conic sg6fisns (zz)
with eccentricity 0. The conic sections are
illustrated in Figure 5-L0.

95
Chapter (5): TOFD Calculations

deterioration of the resolution capabilities of the TOFD technique. A gup between the refracting
wedge and test piece results in a portion of the sound being reflected in the couplant. This results in a
double lateral wave, due to the direct and then the couplant multiple pulses entering the test piece.
The effect is frequency dependent, but typically more than about 0.1mm gap (the thickness of a sheet
of paper) will result in an interfering multiple lateral wave.

For those concerned by the effect of local dips (areas blended by surface grinding) where the couplant
thickness might increase, there is an equation for the possible depth error:
SWxuxd
6d: (5.12)
+cxr[1a,+s,)
Where;
dl,4/: Change in thickness of couplant
u : Velocity in the couplant
Using the parameters, in our example in Figure 5-4, the error in depth estimate at 16mm assumed
depth for a 1mm water gap will result in a depth uncertainty of only 0.01mm.

5.3 HYPERBOLIC ARCS

Both real and modelled images of TOFD data have been used in our illustrations. Common to all of
the indications, seen between the lateral wave and back-wall, has been the curved shape of the
indications as the flaws are approached and then moved away from. The shape of these curves is
considered to be hyperbolic and the details of the arc depend on the probe and scan direction used.

The origin of the hyperbolic shape comes from the fact that we can approximate our sound field as a
cone emanating from the element. \zVhen a flaw enters the conic field, its path effectively forms a plane
that cuts along the cone and the locus forms a conic section. In mathematics, a conic section is a curve
obtained by the intersection of a cone with a plane. A conic section consists of points whose distances,
to some point called a focus and some line called a directrir; are in a fixed ratio called the eccentricity.
Three types of conic section are usually
recognised, the hyperbola, the parabola and
the ellipse. The cirde is a special case of the
ellipse and is therefore sometimes called the
fourth type of conic section. The tlpe of conic
section is determined by its eccentricity. Those
with eccentricity less than 1 are considered
ellipses; those with eccentrici$z equal to 1 are
parabolas; and those with eccentricity greater
than 1 are hyperbolas. In the focus-directrix
definition of a conic, the circle is a limiting case
Figure 5-10 Three basic conic sections (27)
with eccentri"ity 0. The conic sections are
illustrated in Figure 5-10.

95
Chapter (5): TOFD Calculations

In all cases the cone used to illustrate conic sections is a double cone; howevet, for practical Purposes/
the single cone with the probe imagined at the tip can be used. As a result, any path through the cone
simulates the hyperbolic conic section.

In TOFD scanning, the shapes of the arcs differ depending on the direction of scan. This is illustrated
for seven notches in Figure 5-10. The notches are at depths of 0.5mm to 15mm from the far surface and
in a 20mm thick plate (similar to the six notches modelled in Figure 5-9). The uPper scan is the non-
parallel scan and the scan in the lower part of the Figure is a parallel scan over the same seven
notches. It can be seen that the tips of the notches occur at the same depth for both scans; however, the
shape of the arcs is the same for each notch in the case of the non-parallel scan, but the arcs get flatter
as the notch gets closer to the test surface for the parallel scan.

Figure 5-LL Seven notches, non-parallel scan (left) and parallel scan (right)

Charlesworth and Temple (6) suggest a simplified approach to predicting the shape of the arcs. A
coordinate system is used with x indicating the displacement of the flaw from the midpoint, y is the
lateral position of the flaw, d is the depth of the flaw and s is half the PCS value. The layout is
illustrated in Figure 5-12.

Figure 5-12 Coordinate layout for calculating hyperbolic arcs

Time of flight to the flaw, for each position in the scan, can be calculated by solving the Pythagorean
Theorem for the three dimensional condition.

96
 Chapter (5): TOFD Calculations

'=ll @+,lG:*li+All (5.13)

\rVhere;
c: Material's acoustic velocity,
- s and s: Positions at which the probes are fixed,
x, y , d: coordinates of defect.

In the special case, where the flaw is at the midpoint between the probes, x : 0. This simplifies the
equation to the form:
c2t2 v' _a
4k2 'k2
Where; ka2 : s2 + d2

Using an MS Excel spreadsheet to solve the equation for a non-parallel and parallel scan, the curves
rN{

are seen for several depths using a PCS of 100mm and an assumed velocity of 5900m/s. The resulting
curves are seen in Figure 5-13.

Figure 5-L3 Hyperbolic curve plots for Non-Parallel scan (left) and Parallel scan (right)

97