Ping Liu Thesis

Optical Coherence Tomography
for Material Characterization
Ping LIU

Optical Coherence Tomography
for Material Characterization
Proefschrift
ter verkrijging van de graad van doctor

aan de Technische Universiteit Delft,
op gezag van de Rector Magnificus prof. ir. K.C.A.M. Luyben,
voorzitter van het College voor Promoties,
in het openbaar te verdedigen op Maandag 8 December 2014 om 10:00 uur
door
Ping LIU
Master of Science in Mechanical Engineering,

Nanjing University of Aeronautics and Astronautics, China
geboren te Jiangsu, China
Dit proefschrift is goedgekeurd door de promotor:
Prof. dr. ir. R. Benedictus
Copromotor: Dr. R. M. Groves
Samenstelling promotiecommissie:
Rector Magnificus voorzitter
Prof. dr. ir. R. Benedictus Technische Universiteit Delft, promotor
Dr. R. M. Groves Technische Universiteit Delft, copromotor
Prof. dr hab. P. Targowski Nicolaus Copernicus University
Prof. dr. A.G.J.M. van Leeuwen Universiteit van Amsterdam
Prof. dr. ir. Lucas J. van Vliet Technische Universiteit Delft
Prof. dr. ir. S. van der Zwaag Technische Universiteit Delft
Dr. P. H. Tomlins Queen Mary University of London
ISBN: 978-94-6186-393-5
Copyright 2014 by Ping Liu
All rights reserved. No part of the material protected by the copyright notice may be
reproduced or utilized in any form or by any means, electronic or mechanical, including
photocopying, recording or by any information storage and retrieval system, without the prior
permission of the author.
Printed in the Netherlands by Uitgeverij BOXPress, s-Hertogenbosch

To my Family
Summary
Optical coherence tomography (OCT) is a non-invasive, contactless and high resolution

imaging method, which allows the reconstruction of two or three dimensional depth-resolved
images in turbid media. In the past 20 years, OCT has been extensively developed in the field
of biomedical diagnostics, while OCT in the non-destructive testing (NDT) field is lagging far
behind.
The aim of this thesis is to use OCT as a novel NDT technique for material structure
characterization and damage detection. Besides an overview of the OCT fundamentals and
developments, the thesis is mainly driven by three tasks: instrument design, signal processing,
and applications.
An integrated OCT system working at a wavelength of 1550 nm has been built. It

combines time domain (TD) and Fourier domain (FD) OCT to make the both types of
measurement available in one instrument. TD-OCT has the advantage of a large A-scan range
and high SNR, while FD-OCT system has tremendous superiority in fast imaging due to no
movement of the reference mirror. These two kinds of measurement can be implemented
based on the user request in the developed hybrid OCT system, e.g. improved imaging depth
or speed.
In TD-OCT, the envelope detector was selected as the ideal method for the
demodulation of each axial signal. A bandpass filter and 2D median filter are applied before
and after demodulation, respectively, to reduce OCT system and speckle noise. In FD-OCT,
raw data was first processed to remove the influence from the optical source and dark noise of
the CCD detectors. It was then linearly resampled to convert to evenly spaced intervals of
wavenumber, instead of wavelength. With an inverse Fourier transform, one depth profile was
recovered and a cross-sectional image was constructed by accumulating a series of depth
profiles. The quality of cross-sectional images can be further improved by merging multiple
images with different pathlength offsets.
vii
The application of the designed OCT system was mainly focused on glass fiber
composites and the microstructure of the specimens was displayed by either cross-sectional or
volumetric images. Special attention was given to delamination growth in a glass fiber
composite for wind turbine blade applications. The glass fiber composite was tested by
incremental loading. Volumetric images obtained by OCT were further processed to
reconstruct 3D crack surface profiles, from which a full field view of the delamination crack
was given, providing substantial information for the study of crack growth in the composites.
Additionally, the study explored the use of optical coherence elastography (OCE) for the
deformation measurement of glass fiber composites, for the first time to the best of our
knowledge. The developed OCE system based on speckle tracking was first evaluated by a
test of ridge body translation. Then experiments were implemented for a set of glass fiber
composites under tensile testing and three point bending. The results show that OCE can
measure the internal displacements of a glass fiber composite in the range from a few
micrometers to hundreds of micrometers.
Besides, other applications are also presented in the thesis. These include defects and
thickness measurement of polymer coating and the microstructure characterization of a
wooden-panel painting. The results show the designed OCT system also has high potential for
these alternative applications. Recommendations for further improvement to the OCT design
and the applications are presented at the end of the thesis.
viii
Contents
Summary .................................................................................................................................. vii
1 Introduction ......................................................................................................................... 1
1.1 Non-destructive Testing ........................................................................................... 1
1.2 Research Motivation ................................................................................................ 3
1.3 Thesis Outline .......................................................................................................... 6
References .......................................................................................................................... 8
2 Fundamentals and Developments of Optical Coherence Tomography ............................ 13

2.1 Introduction ............................................................................................................ 13
2.2 Fundamentals of Optical Coherence Tomography ................................................ 14
2.2.1 Time-domain Optical Coherence Tomography .......................................... 14
2.2.2 Fourier-domain Optical Coherence Tomography ....................................... 17
2.2.3 Resolution of Optical Coherence Tomography .......................................... 19
2.3 Performance Comparison of TD-OCT and FD-OCT ............................................ 21
2.3.1 General Performance .................................................................................. 22
2.3.2 Spectral Requirement .................................................................................. 23
2.3.3 Signal to Noise Ratio .................................................................................. 24
2.4 Developments of Optical Coherence Tomography................................................ 26
2.4.1 Light Source ................................................................................................ 26
2.4.2 OCT Scanners ............................................................................................. 28
2.4.3 Swept Source Optical Coherence Tomography .......................................... 30
2.4.4 Full Field Optical Coherence Tomography ................................................ 31
2.4.5 New Imaging Concepts ............................................................................... 32
2.5 Discussion and Conclusions .................................................................................. 33
References ........................................................................................................................ 36
ix
3 Development of a Hybrid OCT System ............................................................................ 43
3.1 Introduction ............................................................................................................ 43
3.2 System Overview ................................................................................................... 43
3.3 Detail of components ............................................................................................. 46
3.3.1 Light Source ................................................................................................ 46
3.3.2 Optical Fiber ............................................................................................... 48
3.3.3 Interferometer ............................................................................................. 49
3.3.4 Polarizer and Polarization Maintaining Fiber ............................................. 50
3.3.5 Circulator and Optical Switch ..................................................................... 51
3.3.6 Collimator and Focusing Lens .................................................................... 53
3.3.7 Optical Delay Line and Reflector ............................................................... 53
3.3.8 Balanced Detector and DAQ Board ............................................................ 54
3.3.9 Spectrometer ............................................................................................... 56
3.3.10 Translation Stages ....................................................................................... 58
3.4 Instrument Control Software.................................................................................. 60
3.5 Conclusions ............................................................................................................ 61
References ........................................................................................................................ 61
4 Design of Signal Processing .............................................................................................. 65

4.1 Introduction ............................................................................................................ 65
4.2 Signal processing in TD-OCT ............................................................................... 66
4.2.1 Bandpass Filter............................................................................................ 67
4.2.2 Demodulation .............................................................................................. 69
4.2.3 Imaging smoothing ..................................................................................... 74
4.3 Signal processing in FD-OCT ................................................................................ 75
4.3.1 Reflection-mode measurement ................................................................... 76
4.3.2 Resampling ................................................................................................. 77
4.3.3 Inverse Fourier transform ........................................................................... 78
4.3.4 Axial scan smoothing .................................................................................. 78
4.4 Image Display ........................................................................................................ 80
4.5 System Evaluation ................................................................................................. 81
4.5.1 A-scan Accuracy and Linearity .................................................................. 81
x
4.5.2 Signal to Noise Ratio .................................................................................. 82
4.6 Discussion .............................................................................................................. 84
4.7 Conclusions ............................................................................................................ 85
References ........................................................................................................................ 86
5 OCT for Monitoring the Delamination Growth in Glass Fiber Composite ...................... 89
5.1 Introduction ............................................................................................................ 89
5.2 Sample Construction .............................................................................................. 90
5.3 Experimental Setup ................................................................................................ 91
5.4 Results .................................................................................................................... 92
5.4.1 Static Test.................................................................................................... 92
5.4.2 Incremental Loading Test ........................................................................... 95
5.5 Discussion ............................................................................................................ 100
5.6 Conclusions .......................................................................................................... 102
References ...................................................................................................................... 102
6 OCT Elastography for Measuring the Deformation within Glass Fiber Composite ....... 105
6.1 Introduction .......................................................................................................... 105
6.2 Methods................................................................................................................ 107
6.2.1 Experimental Setup ................................................................................... 107
6.2.2 Image Processing Techniques ................................................................... 108
6.3 Results .................................................................................................................. 109
6.3.1 Rigid Body Translation ............................................................................. 109
6.3.2 Tensile Test ............................................................................................... 111
6.3.3 Three Point Bending ................................................................................. 114
6.4 Discussion ............................................................................................................ 116
6.5 Conclusions .......................................................................................................... 118
References ...................................................................................................................... 118
7 Further Non-destructive Testing Applications ................................................................ 123

7.1 Introduction .......................................................................................................... 123
7.2 Defects and Thickness Measurement of Polymer Coatings ................................ 124
7.2.1 Introduction ............................................................................................... 124
xi
7.2.2 Detection and measurement of microcracks ............................................. 124
7.2.3 Thickness Measurement............................................................................ 126
7.2.4 Discussion ................................................................................................. 127
7.3 Study of Front and Back Faces of a Wooden Panel Painting .............................. 129
7.3.1 Introduction ............................................................................................... 129
7.3.2 Sample Description ................................................................................... 129
7.3.3 Front Face Characterization ...................................................................... 131
7.3.4 Back Face Characterization ...................................................................... 133
7.3.5 Discussion ................................................................................................. 135
7.4 Conclusions .......................................................................................................... 137
References ...................................................................................................................... 137
8 Conclusions and Recommendations................................................................................ 141

8.1 Conclusions .......................................................................................................... 141
8.1.1 Conclusions from Instrument Design ....................................................... 141
8.1.2 Conclusions from Signal Processing ........................................................ 142
8.1.3 Conclusions from Applications................................................................. 142
8.2 Recommendations ................................................................................................ 144
8.2.1 Recommendations for Instrument Design ................................................ 144
8.2.2 Recommendations for Signal Processing.................................................. 144
8.2.3 Recommendations for Applications .......................................................... 145
Samenvatting .......................................................................................................................... 147

Acknowledgements ................................................................................................................ 149
About the Author .................................................................................................................... 151
List of Publications................................................................................................................. 153
xii
1
Introduction
1.1 Non-destructive Testing

Materials are evolving faster today than at any time in history. Radical material
advances enable engineers to improve the performance of existing products and to create new
products that will make our lives safer, more convenient or more sustainable [1]. For example,
composites, a mixture of two or more different materials, are increasingly applied in
aerospace and wind energy industries for their exceptional strength and stiffness-to-density
ratios. In the newest commercial aircrafts such as Airbus A350XWB [2], up to 53% by weight
of the airframe is constructed by composite materials, which, compared with a metallic
fuselage, contributes to lower overall aircraft weight and higher fuel efficiency, along with
reduced corrosion and fatigue, as well as lower maintenance costs. Composite materials are
also boosting the development of wind energy that currently accounts for 3.5 % of the worlds
electricity demand and grows rapidly at 16% per annum [3]. In the Netherlands, wind turbines
have been installed with a total capacity of 2, 693 MW at the end of 2013 [4]. In 2020, this
should be further developed to 6,000 MW onshore and 6,000 MW of offshore wind energy
[5], which will cover nearly one third of total electricity consumption in the Netherlands.
Wind turbine blades are extensively manufactured with composites that can bear gravity and
wind loadings for long life cycle due to their low density and high stiffness [6].
During the development of materials, both scientists and engineers need to know what
properties each material has and how they might be changed. Only very rarely can the
answers be deduced using our bare hands and the naked eye, so the study of materials
inevitably involves testing materials measuring their properties and looking at their
microstructure. Also during the manufacture and in-service period, testing of materials
1
Chapter 1
becomes an indispensable step to control manufacturing processes, ensure product integrity

and reliability, and make efficient use of products during their lift time.
In science and industry, materials can be tested both in destructive and non-destructive
ways. Destructive tests are often used to determine the physical properties of materials such as
tensile strength, fracture toughness and impact resistance, while non-destructive testing (NDT)
[7] can in many cases evaluate materials and structures for discontinuities and differences in
characteristics. As the name implies, NDT does not damage or permanently alter the product
being inspected. Therefore, it is a highly valuable technique that can save both money and
time in material research, production and maintenance.
NDT can be performed by a wide group of techniques based on penetrating medium or

equipment used. Taking the aerospace industry as an example, ultrasonic inspection,
radiography, and eddy current testing are largely utilized to evaluate the health condition of
the aircraft structures [8]. Ultrasonic inspection [9] uses high frequency acoustic wave to
detect subsurface flaws. Either the reflection or the attenuation of the sound energy can be
measured to analyze the presence and location of discontinuities which could indicate defects
such as voids, cracks, and delaminations. Alternatively, radiography [10] relies on the
examination of variations in transmitted intensity of X-rays or gamma radiation arising from
difference in the absorption coefficients of a defect and surrounding structure. Radiography is
widely used to test the imperfections in welded parts of structures [11]. Lastly, eddy current
[12] testing introduces an electrical current to detect defects within conductive materials since
the flow of the eddy current is influenced by the presence of flaws. Eddy current technique is
an effective method to detect the corrosion in multilayered structures [13]. Besides, optical
NDT techniques such as shearography [14] and thermography [15] also show potential for
subsurface defects measurement of aerospace structures. However, with the development of
material science, NDT with a single method of damage determination cannot satisfy the
modern tests. Material researchers expect the ability to access the depth features of materials,
and to rebuild the internal structures virtually without damaging the specimen, to provide
substantial information for material property evaluation.
Currently, ultrasonic testing, and x-ray computed tomography (x-ray CT) [16] are the
dominant NDT techniques for structural imaging. Ultrasonic testing can estimate the internal
structure of materials by detecting the echo when ultrasonic pulse waves are directed into
specimens. Its high penetrating power allows a detection of tens of millimeters depth. But a
major drawback is the low spatial resolution that is practically on the order of hundreds of
micrometers [17]. In contrast, x-ray CT can provide 3D reconstructed images with
2
Introduction
submircometer resolution1, but it has a complex design with a large instrument size and
requires large amounts of computation, which is impossible for real time and in situ imaging.
Therefore, new NDT techniques are still needed in the field of structural imaging.
1.2 Research Motivation

In the past two decades, an imaging technique named optical coherence tomography (OCT)
[18] has become one of the hot topics in biomedical science. OCT is basically analogous to
ultrasonic B-scan except that it uses light instead of sound. Fig. 1.1 shows one example of
how OCT is performed [19]. A cross-sectional image is acquired by a succession of axial
depth scans of optical backscattering or backreflection profiles at different transverse
locations. The axial profiles provide the optical pathlength difference and intensity of
backscattered or backreflected light. For the purpose of visualization, two-dimensional OCT
data can be displayed as a gray scale or false color image. A volumetric OCT image can be
performed in a same way with two orthogonal transverse scans. The working principle of
OCT is based on low coherence interferometry which will be introduced in detail in chapter 2.
Fig. 1.1: Image generation in OCT. Cross-sectional images are constructed by measuring
depth-resolved intensity of backreflected light at different transverse positions [19].
1
Up to 0.2 m spatial resolution can be achieved by x-ray CT scanners, e.g. Phoenix nanotom S designed by
General Electric Company.
3
Chapter 1
The greatest advantage of OCT is that it has a good compromise between the spatial
resolution and penetration depth [19], as shown in Fig. 1.2. Typically, image resolutions of
1~15 m can be achieved with OCT measurement, one or two orders of magnitude higher
than the ones with conventional ultrasonic scan. Confocal microscopy [20], by using point
illumination and a spatial pinhole to eliminate out-of-focus light in specimens, can have
submicrometer resolution. But the image penetration of confocal microscopy is limited to a
few hundred micrometers in general scattering media, which is much lower than with OCT
that can penetrate 2~3 mm correspondingly. Besides, OCT has merits such as contactless
measurement, relatively simply setup and computation, fast scan and display.
Fig. 1.2: Comparison of resolution and imaging depth for ultrasound, OCT, and confocal
microscopy [19].
All the advantages make OCT an important tool for biomedical examination, and it has
developed rapidly in the past 20 years, which can be deduced from the yearly publications in
the PubMed database [21], summarized in Fig. 1.3. Besides, from the categorized publications,
it can be seen that OCT is widely used in the biomedical field, e.g. ophthalmology,
cardiovascular, gastroenterology and endoscopy, etc. Driven by these applications, OCT
techniques are evolved rapidly as well. Recent developments of OCT techniques have led to
enhanced sensitivity and imaging speed [22], to improved resolution [23], and to different
imaging concepts [24-26]. Also there is a growing number of commercial companies [27]
offering OCT scanners, showing a promising market for OCT technology.
4
Introduction
Fig. 1.3: Yearly OCT publications in PubMed categorized by application [21].
However, OCT in the NDT field, compared with the biomedical counterparts, is much
lagging behind. Only a few research groups apply the technology to material testing, and
therefore the progress in this field is slow. Two reviews of alternative OCT applications and
technique developments beyond biomedicine were published by Stifter in 2007 [28] and
Nemeth in 2013 [29]. Recent examples include the measurement of layer thickness in
multi-layered foils [30], pharmaceutical tablets [31], the characterization of paper [32],
ceramics [33], printed electronics [34], applications with polymer-based materials [35] and for
artwork conservation [36]. However, so far these results are very initial and they have not
made full use of the recent progress in OCT technology. For example, some NDT applications
were implemented with commercial OCT systems which are designed for biomedical use.
Due to the property differences between engineering materials and biomedical tissues, such as
light absorption and scattering, commercial OCT instruments designed for biomedical
applications are not exactly suitable for material characterization. In some applications which
used customized OCT systems, not enough attention was paid to signal and image processing,
which should be an important step to guarantee the image quality. Besides, rarely has research
linked OCT with mechanical testing of materials. For example, monitoring crack growth or
5
Chapter 1
evaluating displacement distributions within glass fiber composites under mechanical

loadings has never been reported.
Therefore, the aim of this study is to use OCT as a novel NDT tool for material
characterization. This includes:
1. Development of a state-of-art OCT system specialized for material

characterization.
2. Design of a series of signal processing algorithm for improved image quality.
3. To study subsurface structure and structural changes of materials using OCT
combined with mechanical loading.
4. To explore new imaging concepts, such as elastography [37], giving access to
material mechanical properties.
Since the work is carried out in Aerospace NDT laboratory, Delft University of
Technology (TU Delft), the primary testing objects are highly related with aerospace and
wind energy products, such as glass fiber composites and polymer coatings. Some alternative
specimens, e.g. oil paintings and wood were also considered for the OCT measurements.
1.3 Thesis Outline

The objective of the thesis is to design a novel OCT system for the non-destructive testing of
materials, especially for composite materials. The scope of the thesis work includes an
overview of OCT fundamentals and developments, the design of a customized setup and
signal processing algorithms, and a selection of applications.
Chapter 2 starts with an introduction of OCT theory. Two dominant OCT modalities,
time-domain OCT (TD-OCT) and Fourier-domain OCT (FD-OCT) are compared from the
view of principle and performance. Recent developments of OCT techniques are also
reviewed in this chapter.
Chapter 3 describes the customized OCT setup developed at TU Delft. The most
attractive feature of the setup is that it couples TD-OCT and FD-OCT in one system allowing
applications to select the strength of one or other of the techniques. The selection of
components used in the system, such as light source, fiber polarizer, collimator, balanced
detector, optical delay line (ODL), and translation stage, are discussed and justified in detail.
Chapter 4 focuses on signal processing algorithms for TD-OCT and FD-OCT. In

TD-OCT, three techniques are discussed for the demodulation of axial signals. Their
6
Introduction
performance is compared by using both simulated and experimental OCT data. While in
FD-OCT, a series of signal processing algorithms are applied for improved signal and image
quality. The whole OCT system, including hardware and software, has now been described.
So at the end of this chapter, a selection of experiments are made to verify the performance of
the customized OCT system.
Chapter 5 presents the results of material characterization using the customized OCT
system. The investigated object is a glass fiber composite for the manufacture of a wind
turbine blade. The measurements focus on internal microstructure and defects such as
delamination. Special attention is given to delamination growth in the glass fiber composite.
A succession of OCT volumetric images were obtained from a delaminated specimen under
incremental loadings. They were further processed to reconstructed 3D crack surfaces to
reveal the propagation of the delamination. The results are discussed and related to the
performance of the composite materials.
Chapter 6 expands OCT applications to elastography (OCE), which is an technique

typically for the measurement of volumetric deformation and strain within biomedical tissues
[38]. In this study, an OCE system for the first time was designed for deformation
measurement of glass fiber composites under different mechanical loadings. The advantages
and possible further improvements for the OCE application on glass fiber composites are
discussed.
Chapter 7 shows some further non-destructive testing applications with OCT. The
customized OCT system has been used to test polymer coatings, giving the information such
as thickness and microcracks in the coating materials. A wooden panel painting was
investigated to characterize the varnish layer, the pigments and some crack defects. The back
face of the painting was also studied by 3D reconstruction of the surface profile.
Chapter 8 discusses the research of this thesis. Conclusions and recommendations for
further study are also presented.
7
Chapter 1
References
[1] W. F. Smith, and J. Hashemi, [Foundations of materials science and engineering],

Mcgraw-Hill Publishing (2006).
[2] Intelligent airframe, http://www.a350xwb.com/advanced/fuselage/.
[3] [Half-year report 2013], The World Wind Energy Association (2013).
[4] [Wind in power: 2013 Europen statistics], The European Wind Energy Association
(2014).
[5] Wind energy facts, http://www.nwea.nl/windenergie-de-feiten.
[6] P. Brndsted, H. Lilholt, and A. Lystrup, Composite materials for wind power
turbine blades, Annual Review of Materials Research, 35, 505-538 (2005).
[7] C. Hellier, [Handbook of nondestructive evaluation], McGraw-Hill (2001).
[8] W. Staszewski, C. Boller, and G. R. Tomlinson, [Health monitoring of aerospace

structures: smart sensor technologies and signal processing], John Wiley & Sons
(2004).
[9] J. Krautkramer, and H. Krautkrmer, [Ultrasonic testing of materials], Cambridge

University Press (1990).
[10] R. Halmshaw, [Industrial radiology: theory and practice], Springer (1995).
[11] T. W. Liao, and Y. Li, An automated radiographic NDT system for weld inspection:
Part IIFlaw detection, NDT & E International, 31(3), 183-192 (1998).
[12] B. Auld, and J. Moulder, Review of advances in quantitative eddy current

nondestructive evaluation, Journal of Nondestructive Evaluation, 18(1), 3-36 (1999).
8
Introduction
[13] R. Smith, and G. Hugo, Transient eddy current NDE for ageing aircraft-capabilities
and limitations, Insight: Non-Destructive Testing and Condition Monitoring, 43(1),
14-25 (2001).
[14] D. Francis, R. P. Tatam, and R. M. Groves, Shearography technology and

applications: a review, Measurement Science and Technology, 21(10), 102001
(2010).
[15] C. Meola, G. M. Carlomagno, A. Squillace et al., Non-destructive evaluation of

aerospace materials with lock-in thermography, Engineering Failure Analysis, 13(3),
380-388 (2006).
[16] P. J. Schilling, B. R. Karedla, A. K. Tatiparthi et al., X-ray computed

microtomography of internal damage in fiber reinforced polymer matrix composites,
Composites Science and Technology, 65(14), 2071-2078 (2005).
[17] A. Ng, and J. Swanevelder, Resolution in ultrasound imaging, Continuing Education

in Anaesthesia, Critical Care & Pain, mkr030 (2011).
[18] D. Huang, E. Swanson, C. Lin et al., Optical coherence tomography, Science,

254(5035), 1178-1181 (1991).
[19] B. E. Bouma, and G. J. Tearney, [Handbook of optical coherence tomography],

Marcel Dekker (2002).
[20] J. Pawley, [Handbook of biological confocal microscopy], Springer (2010).
[21] Historical statistics of academic publications in the field of optical coherence

tomography,
http://www.octnews.org/articles/4061333/some-historical-statistics-of-academic-publi
cation/.
[22] B. Potsaid, B. Baumann, D. Huang et al., Ultrahigh speed 1050nm swept

source/Fourier domain OCT retinal and anterior segment imaging at 100,000 to
400,000 axial scans per second, Optics Express, 18(19), 20029 (2010).
9
Chapter 1
[23] N. Nishizawa, "Ultrahigh resolution optical coherence tomography," Opto-Electronics

and Communications Conference (OECC), 2012 17th, 584-585 (2012).
[24] E. Dalimier, and D. Salomon, Full-field optical coherence tomography: a new

technology for 3D high-resolution skin imaging, Dermatology, 224(1), 84-92 (2012).
[25] E. Gtzinger, B. Baumann, M. Pircher et al., Polarization maintaining fiber based

ultra-high resolution spectral domain polarization sensitive optical coherence
tomography, Optics Express, 17(25), 22704-22717 (2009).
[26] C. Sun, B. Standish, and V. X. Yang, Optical coherence elastography: current status
and future applications, Journal of Biomedical Optics, 16(4), 043001 (2011).
[27] OCT market, http://www.sweptlaser.com/OCT-market.
[28] D. Stifter, Beyond biomedicine: a review of alternative applications and

developments for optical coherence tomography, Applied Physics B, 88(3), 337-357
(2007).
[29] A. Nemeth, G. Hannesschlager, E. Leiss et al., [Optical coherence tomography

applications in non- destructive testing and evaluation], Intech (2013).
[30] G. Hannesschlger, A. Nemeth, C. Hofer et al., Optical coherence tomography as a

tool for non destructive quality control of multi-layered foils, Proceedings of the 6th
NDT in Progress, (2011).
[31] S. Zhong, Y.-C. Shen, L. Ho et al., Non-destructive quantification of pharmaceutical

tablet coatings using terahertz pulsed imaging and optical coherence tomography,
Optics and Lasers in Engineering, 49(3), 361-365 (2011).
[32] E. Alarousu, [Low coherence interferometry and optical coherence tomography in

paper measurements], University of Oulu (2006).
[33] J. Veilleux, C. Moreau, D. Lvesque et al., Optical coherence tomography for the
inspection of plasma-sprayed ceramic coatings, Journal of Thermal Spray
Technology, 16(3), 435-443 (2007).
10
Introduction
[34] J. Czajkowski, T. Prykri, E. Alarousu et al., Optical coherence tomography as a

method of quality inspection for printed electronics products, Optical Review, 17(3),
257-262 (2010).
[35] D. Stifter, K. Wiesauer, M. Wurm et al., Investigation of polymer and polymer/fibre

composite materials with optical coherence tomography, Measurement Science and
Technology, 19(7), 074011 (2008).
[36] P. Targowski, B. Rouba, M. Gra et al., Optical coherence tomography in art

diagnostics and restoration, Applied Physics A, 92(1), 1-9 (2008).
[37] J. Ophir, I. Cespedes, H. Ponnekanti et al., Elastography: a quantitative method for

imaging the elasticity of biological tissues, Ultrasonic Imaging, 13(2), 111-134
(1991).
[38] J. M. Schmitt, OCT elastography: imaging microscopic deformation and strain of

tissue, Optics Express, 3(6), 199-211 (1998).
11
2
Fundamentals and Developments
of Optical Coherence
Tomography
2.1 Introduction
Optical coherence tomography (OCT) is a contactless and non-destructive technique for
high-resolution imaging of microstructure within scattering media. Originally developed for
the imaging of the human retina in 1991 [1], nowadays OCT not only finds widespread
biomedical applications, but its techniques have also evolved dramatically. The first OCT
modality was realized in time domain (TD-OCT) based on a low coherence interferometer,
similar to those used in industrial metrology for measuring the thickness of thin films [2] and
refractive index [3]. In order to improve the imaging speed, Fourier domain OCT (FD-OCT)
was developed afterwards as a new generational OCT system which has the advantage that no
moving parts are needed to obtain axial scans [4]. The spectral intensity detected by a
spectrometer at one shot is then Fourier transformed into time domain to reconstruct the
depth-resolved optical structure of a sample. Lately, new functional OCT configurations,
such as polarization sensitive OCT (PS-OCT) [5] and Doppler OCT (DOCT) [6] were
developed to give access to additional structural information like birefringence and fluid flow.
Besides, the development of OCT components, e.g. light source and optical scanners also
promote OCT into a high resolution and high speed structural imaging technique [7, 8].
In this chapter, OCT principles are introduced first. The fundamentals of two main
categories of OCT instrumentation, TD-OCT and FD-OCT are described in detail. These
13
Chapter 2
two OCT modalities are also compared by the performance such as imaging speed, spectral
requirement and signal to noise ratio (SNR). Then the recent technique developments of OCT
are reviewed. The evolution of OCT light source, scanners and new imaging concept are
introduced. Further OCT developments for NDT application are recommended at the end of
this chapter.
2.2 Fundamentals of Optical Coherence Tomography

2.2.1 Time-domain Optical Coherence Tomography
In OCT, the first step in constructing a tomographic image is the measurement of light
backreflected from within the material or tissue. A typical TD-OCT is based on the
Low-coherence interferometer, as shown in Fig. 2.1(a). Light emitted from the optical source
is divided by a beamsplitter into reference and sample beams. The beam reflected from the
reference mirror is recombined at the beamsplitter with the one reflected from the sample, and
finally the interference signal is collected by a photo detector. A single axial profile of
optical reflectivity versus depth into the sample is obtained by rapidly translating the
reference mirror and synchronously recording the magnitude of the resulting interference
signal, as shown in Fig. 2.1(b). The envelope of the interference signal represents one axial
profile of optical reflectivity in a sample.
Fig. 2.1: (a) Schematic diagram of typical TD-OCT system and (b) one A-scan signal. The
envelope of the modulated signal represents axial profile of sample optical reflectivity that
comprises one vertical line in the tomogram.
14
Fundamentals and Developments of Optical Coherence Tomography
The optical source for OCT should be a low-coherent light that consists of a finite
bandwidth of frequency. An ideal one is a near infrared light source with Gaussian
distribution (Section 2.3.2 and 2.4.1), as shown in Fig. 2.2. Under this condition, the reference
and sample electric fields, Er and Es in an interferometer can be expressed as functions of
frequency and time t [9]:
Er (, t ) Ar () exp[ j (2r ()lr t )] (2.1)
and
Es (, t ) As () exp[ j (2s ()ls t )] (2.2)
where Ar ( ) and As ( ) are the electric field amplitudes. lr and ls are the pathlengths of
reference and sample arms respectively. t is a time during the sweeping of the reference
mirror. r ( ) and s ( ) are the propagation constants which are frequency dependent.
Fig. 2.2: A Gaussian optical source for OCT configuration.
Assuming that the sample and reference arms consist of uniform, linear, non-dispersive
material, the propagation constants in each arm are the same and can be rewrite as a
first-order Taylor expansion [10]:
r () s () (0 ) (0 )( 0 ) (2.3)
where (0 ) is the propagation constant at the center frequency 0 . Then the phase
mismatch ( ) 2 s ()ls 2r ()lr is determined solely by the pathlength mismatch
l ls lr through:
15
Chapter 2
( ) (0 )(2l ) (0 )( 0 )(2l ) (2.4)
For a 50/50 beamsplitter in an OCT configuration, the intensity collected by the photo
detector I can be expressed as [11]:
2
1 1
I Er Es
2 2
1 2 1 2
Er Es Re{Er* Es } (2.5)
2 2
1
( I r I s ) Re{Er* Es }
2
where I r and I s , are the mean intensities returning from the reference and sample arms of
the interferometer respectively, and comprise the DC components in the interference signal.
The second term of the equation, Re{Er* Es } , represents the amplitude of the interference
fringes that carry information about sample structure. As the interference signal is
proportional to the sum of the interference from each monochromatic wave component, Eq.
2.5 can be further expressed as [10]:
d d
I Re Er (, t )Es (, t ) Re S ( ) exp[ j ( )] (2.6)
2 2
where the definition S ( ) Ar * ( ) As ( ) has been used. Taking account of the propagation
constant and the phase mismatch given in Eq. 2.3 and 2.4, the detected intensity can be further
written as:
d ( 0 )
I Re exp[ j0 p ] S ( 0 ) exp[ j ( 0 ) g ] (2.7)
2
where the phase delay mismatch p and the group delay mismatch g are defined as:
0 2l
p (2l ) (2.8)
0 vp
and
0 2l
g (2l ) (2.9)
0 vg
Thus, v p 0 / (0 ) and vg 1/ (0 ) are the phase and group velocity respectively.
16
Assume that the light source has a Gaussian power spectral density, S ( 0 ) , given
by [12]:
1 ( 0 )2
S ( 0 ) exp[ ] (2.10)
2 2 2
where 0 is defined as the center frequency and is the standard deviation of the light
source frequency. Eq. 2.7 can be described as:
2 g2
I Re exp[ ]exp[ j0 p ] (2.11)
2
Eq. 2.11 indicates that the amplitude of the interference fringe contains a Gaussian
envelope and a carrier, as shown in Fig. 2.1(b). The envelope falls off rapidly with increasing
group delay mismatch g and is modulated by a carrier that oscillates with increasing
phase delay mismatch p . Thus, Eq. 2.11 defines the axial resolving properties of OCT
system. The interference fringes can only be observed when the optical pathlengths of
reference and sample arm are matched, and it reaches a maximum when the two pathlengths
are identically equal. Therefore, by continuously translating the reference mirror to change the
optical pathlength, a serial of short oscillatory pulses are detected, centered at times when the
position of the reference mirror corresponds to the locations of backreflecting interfaces along
one depth of a specimen. In this way, a single axial profile of sample optical reflectivity can
be achieved, indicated by the envelope of the axial scan signal, as shown in Fig. 2.1(b). This
axial profile is referred to as an A-scan that comprises one vertical line in the tomogram. By
laterally scanning the penetration beam in either one or two orthogonal directions, a
cross-sectional (B-scan) or volumetric image can be obtained finally. The axial and lateral
resolutions of the image are determined by the coherence length of the light source and the
spot size of the incident beam to the sample respectively, which are discussed in section 2.2.3
in detail.
2.2.2 Fourier-domain Optical Coherence Tomography
Recently the FD-OCT technique has experienced increased attention and has developed
rapidly. The measurement principle is based on spectral interferometry, as depicted in Fig. 2.3.
In FD-OCT, the reference mirror is kept still and the single light intensity detector used in
TD-OCT is replaced by a spectrometer that separates the different wavenumbers k
17
Chapter 2
( k 2 / ). With an inverse Fourier transform of the detected spectral intensity, one depth
profile can be obtained.
Fig. 2.3: (a) Schematic diagram of typical FD-OCT system and (b) one A-scan signal after
inverse Fourier transform. The object signal starts from the offset z0.
Theoretically, the spectral intensity I (k ) acquired from a spectrometer can be

expressed as [12]:
2
I (k ) G(k ) ar exp(i 2klr ) z0 a( z ) exp i 2k (lr nz ) dz (2.12)
where G(k ) is the spectral intensity distribution of the light source. The first and second term
within square brackets is the signal from the reference and the sample arm respectively. ar is
the reflection amplitude from the reference mirror which is a constant and is arbitrarily set at
1for simplicity. a( z ) is the backscattering amplitude of the object signal that needs to be
recovered. z0 is the distance mismatch between reference plane and object surface. lr and
(lr nz ) are the optical pathlength in reference and sample arm. n is the average refractive
index of the object. Since we are only interested in the pathlength difference, lr can be
defined as 0. With these assumptions the interference signal I (k ) can be written as:
2
I (k ) G (k ) 1 a( z ) exp(i 2knz )dz
z0
(2.13a)
G (k ) 1 2 a( z ) cos(2knz )dz a( z )a( z ) exp i 2kn( z z ) dzdz

z0 z0 z0
Since a(z) = 0 for all z < z0, a(z) can be replaced by the symmetrical expansion
a ( z ) a( z) a( z ) . Now Eq. 2.13a becomes:
18
1
I (k ) G (k ) 1 a ( z ) cos(2knz )dz a ( z )a ( z ) exp i 2kn( z z ) dzdz (2.13b)
4

It can be seen that the third (autocorrelation) term describes the mutual interference of all

elementary waves. By defining R(a ( z )) a ( z )a ( z)dz , Eq. 2.13b can be further written

as:
1
I (k ) G(k ) 1 a ( z ) exp(i 2knz )dz R[a ( z )]exp(i 2knz )dz (2.13c)
4

The second and third terms in Eq. 2.13c can be treated as the expressions of Fourier transform
of a ( z ) and R(a ( z )) respectively. Therefore the spectral signal is finally deduced as:
1 1
I (k ) G(k ) 1 F a ( z ) F R[a ( z )] (2.14)
2 8
where F denotes the Fourier transform. Making use of the convolution property of Fourier
transform x( z ) y( z )
F
X (k )Y (k ) , the inverse Fourier transform (F-1) of Eq. 2.14 can be
calculated as:
1 1
F1 I (k ) F1 G(k ) ( z ) a ( z ) R a ( z )
2 8 (2.15)
A ( B C D)
where indicates convolution and ( z ) is the delta function. From this result the
symmetric scattering amplitude a ( z ) and therefore a(z) can be deduced. However, besides
the signal term C, there are three disturbing terms, A, B and D. Since A B and A D are
located around z=0, a small offset in z0 (usually 200 m is sufficient) can separate them from
the signal term. Taking the spectral characteristic of the light source into account, the object
signal can finally be deconvoluted from A C . Fig. 2.3(b) shows an example of the
interferogram recorded with FD-OCT after inverse Fourier transform.
2.2.3 Resolution of Optical Coherence Tomography
In spite of the different image principles, the spatial resolution provided by TD-OCT and
FD-OCT can be determined in the same way. In contrast to a conventional microscope, OCT
has the advantage of independent axial and lateral resolution. The axial resolution of OCT is
the Full-Width-Half-Maximum (FWHM) of the coherence length, as shown in Fig. 2.4(b).
19
Chapter 2
Thus the axial resolution, z , is determined by light source parameters and the specimen
refractive index, given by:
2 ln 2 0
2
z (2.16)
n
where 0 and refer to the center wavelength and FWHM bandwidth of the optical source
respectively, as shown in Fig. 2.4(a). n is the average refractive index of the object. The axial
resolution is inversely proportional to the bandwidth of the light source, and thus broad
bandwidth optical sources are required for high axial resolution. Following this definition, a
typical superluminescent diode (SLD) centered at 800 nm with 60 nm FWHM, has a
theoretical axial resolution of approximately 5 m in the air.
Fig. 2.4: (a) OCT light source and (b) corresponding interference signal. The axial
resolution z is decided by the center wavelength 0 and bandwidth of the light
source.
The lateral resolution for OCT imaging, x , is the same as for conventional microcopy
and is determined by the focusing properties of an optical beam, given by:
0
x 0.61 (2.17)
NA
where NA is the effective numerical aperture, which can be approximated as:
d
NA (2.18)
2f
20
Here d is the spot size on the objective lens and f is its focal length. High lateral resolution
(e.g. 1 m) can be achieved by using a large numerical aperture and focusing the beam to a
small spot size. However there is always a tradeoff between the lateral resolution x and
the depth of focus b, given by:
20 n
b (2.19)
NA2
Fig. 2.5 shows schematically the relationship between the focused beam size and depth of
focusing field with low and high numerical aperture focusing.
Fig. 2.5: The tradeoff between focused spot size and depth of focus for low and high
numerical aperture focusing.
2.3 Performance Comparison of TD-OCT and FD-OCT

As introduced above, TD-OCT translates the reference mirror in a low-coherence
interferometer to obtain the optical reflectivity versus depth, allowing the generation of
structural maps within a specimen. Differently FD-OCT measures the spectrum of an
interferometer output and recovers one depth profile by a Fourier transform. The spectrum
intensity is usually obtained by a spectrometer which contains a detector array. It can also be
realized by using a single detector and a tuneable light source in the interferometer, which
refers to swept source OCT (SS-OCT) that will be introduced in section 2.4.3. The
21
Chapter 2
following discussions mainly compare the performance of TD-OCT and FD-OCT using a
spectrometer.
2.3.1 General Performance
A major advantage of a TD-OCT system is its simple design and hence the low cost of the
instrument. It is also capable of unlimited axial scan which depends only on the translation
range of the reference mirror. However, on the other hand, the movement of the reference
mirror slows down the data acquisition process. To achieve real-time imaging, a complex
design such as grating based optical delay line [13] (see section 2.4.2) or an area detector for
full-field measurement (see section 2.4.4) is needed.
On the contrary, a main feature of FD-OCT is the lack of movable parts in the reference
arm of the interferometer. The total distribution of scattering amplitude along one depth of the
object is measured at once, which is much faster than TD-OCT and allows real-time
monitoring of structure variations [4, 14].
However, in FD-OCT the depth measurement range, , is highly reliant on the

number of pixels in the detector array [15], given by:
02
zmax N (2.20)
4n full
where full and N are the full spectral bandwidth and total pixel number of the linear CCD
detector respectively. If a detector array consists of 512 elements and the optical source is
centered at 1550 nm with 160 nm full spectral bandwidth, then the maximum axial scan depth
for a sample with 1.55 refractive index is only 1.24 mm. Eq. 2.20 indicates that the
measurement range can be improved with an increased number of pixels in the detector array,
but this brings an additional cost as well, a factor which is important to consider during an
instrument design.
Besides, as discussed in section 2.2.3, a typical OCT system cannot provide a high
lateral resolution with a long depth of focus that covers the whole axial scan range. To solve
this problem, TD-OCT can track the focal depth by translating the focal lens along with the
axial scan (movement of the reference mirror) [16]. However in FD-OCT this method is not
possible since all the depth signal is acquired simultaneously. A possible solution could be to
use advanced signal processing that acquire multiple images with different zones of focus and
to fuse these images to create a single one with an extended focus depth [17].
22
2.3.2 Spectral Requirement
Another interesting difference is that of the spectral quality of the source. In TD-OCT, Eq.
2.11 represents the autocorrelation function for a Gaussian source. According to
Wiener-Khinchine theory [18], the autocorrelation function is the inverse Fourier transform of
the source power spectrum. Therefore, a source with a Gaussian spectrum ideally yields a
Gaussian autocorrelation function, as shown in Fig. 2.4. This is critical for performing
high-resolution imaging. If the light source has a noisy Gaussian or a non-Gaussian spectrum,
the modulation or ripples in the light source can cause side peaks or echoes around the main
reflection peaks in an axial signal. Two examples, a noisy Gaussian spectrum (SNR=30 dB)
and a Gaussian spectrum with side lobes are shown in Fig. 2.6(a) and (d). The corresponding
envelope of the axial signals, shown in Fig. 2.6(b) and (e), can be estimated by inverse
Fourier transform. Fig. 2.6(c) and (f) are the envelopes displayed in logarithm scale, from
which the deterioration of the signal can be clearly observed. The significant tails around the
main reflection peak may prevent the detection of a weaker reflector nearby. However
FD-OCT can compensate for non-Gaussian spectrum by dividing the interference spectrum by
the light source spectrum (see Eq. 2.14). The light source spectrum G(k ) can be easily
measured by the spectrometer in a FD-OCT system.
23
Chapter 2
1 1
(a) (d)
Normalized Intensity
0.5 0.5
0 0
Frequency Frequency
1 1
(b) (e)
0.5 0.5
0 0
Axial length Axial length
0 0
Normalized Intensity/dB
Normalized Intensity/dB
(c) (f)
-20 -20
-40 -40
-60 -60
-80 -80
Axial length Axial length
Fig. 2.6: Light sources with a noisy Gaussian spectrum (a) and a non-Gaussian spectrum
(d), and the corresponding interference signals (b) and (e) from an OCT system. (c) and (f)
are the interference signals in logarithmic scale.
2.3.3 Signal to Noise Ratio
In OCT the dominant sources of noise are shot noise, thermal noise, and relative intensity
noise (RIN) [19]. Shot noise arises from current fluctuations due to the quantization of light
and charge. The noise power is proportional to the electronic charge and the photocurrent.
Thermal noise is generated by resistors in the receiver electronics. The noise current is
approximately zero-mean white noise. The RIN, also mentioned as excess photon noise, refers
to the fluctuations in output intensity of a light source due to the beating of various spectral
components having random phase. For unpolarized broadband sources such as SLDs, the
noise is only dependent on the spectral bandwidth of light source [20].
24
Signal to noise ratio (SNR), is a critical parameter to evaluate to the performance of an

OCT imaging system. Several literature sources [12, 21-23] have deduced the theoretical SNR
of TD and FD-OCT and a common expression is given by:

SNRTD 2 Ps (2.21)
Ev
and
M
SNRFD M Ps SNRTD (2.22)
Ev 2
where and Ev are the quantum efficiency and the photon energy respectively. is the
acquisition time for a single photodetector measurement. Ps is the light power backreflected
from a sample and M is the number of detectors in a spectrometer. From Eq. 2.22, it can be
seen that FD-OCT has an enormous advantage in SNR by a factor M/2 improvement over
TD-OCT. However, the theoretical SNR is calculated based on shot noise-limited detection
whereby the thermal noise and the RIN become negligible. Achievement of this
shot-noise-limited detection requires sufficient reference arm power to assure shot noise
dominance which could have two limitations [24]. The first one is that the optical power from
the reference arm can only be increased to the detector saturation level, and the second one is
that the increase of the optical power may cause the RIN noise to dominate. An elaborated
design on the interferometer and control of the reference arm power is crucial to achieve the
optimal SNR of an OCT system.
Besides, it is also worth mentioning the mechanism to compensate for the noise. In
TD-OCT, a dual balanced detector can be used to remove the RIN noise. Due to the
movement of reference mirror, the Doppler shift is introduced, given by:
2vm
fd (2.23)
0
where f d is the Doppler frequency and vm is the velocity of reference mirror. The
Doppler shift modulates the interference signal to a high frequency, which also offers certain
advantages in noise reduction by bandpass filtering (see chapter 4). However, in FD-OCT
[25], due to a different imaging principle, balanced detection and bandpass filter are not
operable. On the contrary, the finite detector elements in FD-OCT will act as lowpass filters
during Fourier transform, which allows in thermal noise, the RIN and the dc signal. A full
evaluation of the SNR in TD and FD-OCT is presented in chapter 4.
25
Chapter 2
A full performance comparison between these two OCT modalities is summarized in

Table 2.1.
Table 2.1: Performance comparison between TD-OCT and FD-OCT
TD-OCT FD-OCT
Spatial resolution High axial and lateral Same as TD-OCT, but extended
resolution decided by light depth focusing needs advanced
source and spot size signal processing algorithms.
respectively; depth of focus
can be modified.
Imaging speed Slow. A complex design is Max. 500,000 A-scans/s [14] since
needed to reach 4000 no movement is needed for the
A-scans/s [13]. reference mirror.
Depth scan range Only decided by the moving Limited by the number of detector
range of the reference mirror. elements (normally less than 2 mm).
Spectral requirement Non-Gaussian light source Influence of the spectrum can be

brings harmonic artifacts into eliminated by division.
the signal.
Higher than TD-OCT, but an

SNR Extensive noise elimination
elaborated system design is required
by:
to reach shot-noise-limited
1. Balanced detection
detection.
2. Bandpass filtering
2.4 Developments of Optical Coherence Tomography

2.4.1 Light Source
In OCT, the light source has a large influence on the system performance. Generally, the
requirements of light source for OCT imaging are: 1) short coherence length, and 2) emission
in the near infrared. The first requirement, short coherence length determines the axial
resolution of an OCT system. From Eq. 2.16 it can be seen that the broader the bandwidth, the
better axial resolution that can be achieved. The second requirement, emission in the near
infrared, is related to the penetration depth. OCT needs to operate in a spectral range in which
26
scattering and absorption occur less so that the penetration of light into the object is adequate.
In studies carried out to date, the deepest penetration in both biomedical tissues and
engineering materials has been achieved using sources with wavelengths between 1200 and
1800 nm [26-28]. As an example, in a report by M. Wurm et al [29], two OCT systems with
center wavelengths of 840 and 1550 nm were used to characterize polymer foams with
different pore sizes. As shown in Fig. 2.7, it can be seen 1550 nm OCT could provide a better
penetration depth, compared with 840 nm OCT. However, it is also noted from Eq. 2.16 that
the longer center wavelength can greatly decrease the axial resolution. A compromise
between the axial resolution and penetration depth is needed especially when the bandwidth
of the light source is not so broad.
Fig. 2.7: OCT cross-sectional images taken from polymer foam samples with different pore
sizes. images (a) and (c) are obtained using a light source with a center wavelength of
1550 nm, while images (b) and (d) are taken at 840 nm [29]. Scale bars vertical and
horizontal in (b): 500 m.
By far, the most commonly used broadband source for OCT is a superluminescent diode
(SLD) because of its simplicity, high irradiance (1-20 mW), and relatively low cost [11, 19].
Usually these SLDs have broad spectra in the range 10-70 nm FWHM and with center
wavelengths ranging from 675 to 1550 nm. This results in coherence lengths of typically
10-30 m, which are not short enough for high axial resolution. It also explains why SLDs
with longer center wavelengths have not been considered in an OCT system due to the greatly
27
Chapter 2
deteriorated axial resolution1. Recently some studies gave the solution of coupling multiple
individual SLDs [30, 31]. For example, two SLDs with center wavelengths of 840 and 920
nm yield to a spectral FWHM of 155 nm and 3 m axial resolution in the air.
At the cutting edge of broadband light generation, femtosecond lasers [32-34] which are
titanium sapphire (Ti:Al2O3) and chromium forsterite (Cr4+Mg2SiO4) lasers, are used in OCT,
particularly when high resolution or source power is required. The Ti:Al2O3 laser usually has
a center wavelength around 800 nm, and a peak power of approximately 400 mW. The
bandwidth of the source can reach 300 nm, resulting in an axial resolution of less than 1 m.
Comparably, the Cr4+Mg2SiO4 laser offers superior penetration in a turbid object with a longer
center wavelength around 1300 nm. The source power can reach 300 mW and the bandwidth
is 200 nm, allowing an axial resolution of below 4 m. However, it is noted that femtosecond
lasers are generally complex, unstable, and expensive. Since many studies are moving toward
in-situ and even real-time imaging, a simple, compact, and robust system is essential for OCT.
Further advancements for OCT sources could be reductions in complexity, size, and cost.
Besides, other sources, such as swept laser and thermal light are used for new imaging
concepts, such as swept source OCT (SS-OCT) and full field OCT (FF-OCT), which are
introduced in section 2.4.3 and section 2.4.4.
2.4.2 OCT Scanners
OCT imaging contains axial and lateral scans. Axial scanning, as introduced before, is
performed differently in TD-OCT and FD-OCT setups. TD-OCT relies on an optical delay
line (ODL) changing the reference pathlength, to determine the depth information of an object.
The simplest version of an ODL is a mirror mounted on a linear translation stage, as shown in
Fig. 2.8(a). The reference light is directed towards the mirror, and then reflected back to the
interferometer system. A mechanical translation stage usually has more than enough working
range to provide a sufficient axial scan. However, a major drawback is the low scan velocity.
Although a commercial linear stage can move at a speed of 1m/s (LSQ Series, Zaber), the low
repetition limits the overall imaging speed. Besides, translation stages with actuators such as
dc motors, stepper motors, have some degree of nonlinearity in their motion, which could
introduce image distortion.
1
E.g., an SLD (FNPL-2000-2400-XXS, Frankfurt Laser Company, Germany) centered at 2350 nm with 35 nm
FWHM results to an OCT axial resolution only 70 m.
28
Translation stage PZT Grating Galvo mirror
(a) (b) (c)
Fig. 2.8: Examples of ODL for scanning the reference pathlength in TD-OCT systems. (a)
Linear translating mirror. (b) Parallel reflecting mirrors driven by a piezoelectric actuator.
(c) Grating based ODL.
The development of faster scan methods has been studied in numerous papers. For
example, by using a piezoelectric transducer to drive a parallel mirror system (Fig. 2.8 (b)),
the scan range and velocity are amplified due to multiple reflections of the reference light.
The maximum repetition rate is then greater than 100 scans/s and a 2-3 mm scanning range
can be achieved [35]. Another advanced technique, the grating based optical delay line, was
originally developed for femtosecond pulse measurement and has now been applied to OCT.
As shown in Fig. 2.8(c), this system employs a grating and oscillating mirror to create an
optical delay line that can achieve 4000 scans/s [13]. A major disadvantage of this ODL,
besides complexity, is the power loss associated with the grating and the chromatic dispersion
of the broadband source. Other rapid ODL scanning methods include a fiber stretcher [36], a
continuous rotating cube [37] and a polygonal mirror [38]. However, fiber stretchers suffer
from stretching hysteresis and sensitivity to temperature drift, while rotating devices have
high nonlinearity in group delay.
In FD-OCT, there is no actual scanner for the axial scan since no movement of a
reference mirror is needed. The axial scan rate of FD-OCT is only decided by the integration
time of the linear detectors. So far, ultra-high-speed FD-OCT can work at 500,000 A-lines/s
by employing two high-speed linear CMOS cameras, which enable 3D imaging within 0.72 s
for a region covering 44 mm2 (500 A-lines along both directions) [14]. However, it is noted
that the integration time is proportional to the SNR of a FD-OCT, given by Eq. 2.22.
Therefore, in an ultra-high-speed FD-OCT, high power is needed to guarantee the signal
quality [21].
For a lateral scan, TD-OCT and FD-OCT can share the same setup to move the sample
beam laterally in two orthogonal directions. Simply, a lateral scan can be realized by a pair of
translation stages. Typically they can provide a minimum step size of 0.1 m and a translation
29
Chapter 2
range to hundreds of millimeters (LSQ Series, Zaber). But the weakness of translation stages
is the low repetition as stated above. For ultrahigh-speed imaging, galvo scanners could be a
choice to provide fast movement of the sample beam. Galvo scanners consist of
galvanometer-based motors with optical mirrors mounted on shafts, as shown in Fig.
2.9. Galvanometer-based motors are limited-rotation DC motors that drive mirrors for laser
beam scanning. Due to its light weight, currently galvo scanners are able to approach 250 Hz
repetition rate within 12.5 scanning angle (GVS series, Thorlabs GmbH). However due to
the pair of deflections, the side effects of using a galvo system include [39]: 1) image
distortion caused by the arrangement of mirrors, indicated by the pillow-shaped field shown
in Fig. 2.9; 2) varying optical distance to the object surface; 3) varying spot size due to the
off-axis beam. Therefore additional work for image correction is needed. These side effects
can be reduced by using an F-theta or telecentric objective (e.g. LSM series, Thorlabs GmbH).
Signal postprocessing [40] could also provide a solution for image deterioration.
Fig. 2.9: Schematic of glavo scanning. A light beam is deflected by a pair of rotating galvo
mirrors for two dimensional scanning [39].
2.4.3 Swept Source Optical Coherence Tomography
As described in section 2.3, OCT can also be performed using a swept source and a single
detector in an interferometer system. This is called swept source OCT (SS-OCT) [41]. A
schematic diagram of SS-OCT is shown in Fig. 2.10. As in FD-OCT, SS-OCT requires no
movement of the reference mirror, however the interference spectra are measured by
30
sweeping the wavelength of the light source. Fourier transform is also utilized to recover the
depth-resolved structure of an object.
Fig. 2.10: Schematic diagram of FF-OCT system.
Compared with TD-OCT and FD-OCT, SS-OCT can be treated as a new generation
OCT modality due to its superior performance. Similar to FD-OCT, one strength of SS-OCT
is high-speed imaging. The introduction of Fourier-domain mode-locked (FDML) swept
sources made A-scan rates of 326 kHz available for the first time [42]. Besides, SS-OCT also
discards the drawbacks of FD-OCT, e.g. the spectral limitation of linear detectors and
imaging depth limitations due to finite detector elements [43]. The only obstacle of SS-OCT
is that the optical source has a high cost and so far is not so reliable in operation. Presently a
swept source is only available for limited range of wavelengths that are mostly centered at
1000nm and 1300 nm, with a bandwidth around 100 nm [42, 44]. This limits the penetration
depth and axial resolution for imaging high scattering materials. It is expected the available
spectral range and bandwidth of swept sources will be expanded in the near future.
2.4.4 Full Field Optical Coherence Tomography
Fast imaging can also be achieved by avoiding lateral scans. In an alternative OCT concept
called full field OCT (FF-OCT), a CCD area camera replaces the single detector in a TD-OCT
system, to enable the capture of two-dimensional en face images in a single exposure. A
schematic diagram of FF-OCT is shown in Fig. 2.11. As in TD-OCT, an axial scan is
implemented by translating the reference mirror. However, due to the lack of lateral scans, a
three-dimensional image can be acquired in 1 s [45]. The first FF-OCT system [46] was
embedded in a commercial microscopy, using infrared LED light. The lateral resolution,
31
Chapter 2
decided by the objective lens and camera pixel size, was 2 m. The axial resolution, limited
by the source bandwidth, was 8 m. Presently, this technique has also been investigated using
a halogen light source [45, 47]. A source of this type is inexpensive and has an ultra-broad
spectrum (around 300 nm bandwidth), resulting in an axial resolution of 0.7 m. However,
the major limitation of this OCT modality is the small lateral field of view, which is around
1x1 mm2 [48]. To achieve en face imaging of a larger object, an extra mechanical stage is
required to move the sample laterally.
Fig. 2.11: Schematic diagram of FF-OCT system.
2.4.5 New Imaging Concepts
Besides high speed and high resolution, recording more information about the object is
another direction for the development of new functional OCT modalities. For example,
polarization sensitivity OCT (PS-OCT) [49], by measuring the polarization state of
backreflected light from the object, gives access to additional physical parameters, like
birefringence, and provides enhanced structural images that are difficult to resolve with
traditional OCT techniques. One example in the NDT field is a PS-OCT system at 1550 nm
that was used to investigate structural changes within bulk polymer samples during tension
and fracture [50]. Spatially and temporally varying polarization patterns, due to defects and
material inhomogeneities, were observed and were used to finally obtain internal stress
distributions with the help of advanced image processing algorithms.
Doppler OCT (DOCT) [51], based on OCT combined with laser Doppler velocimetry
(LDV), permits the quantitative imaging of fluid flow in scattering object, such as in-vivo
monitoring of blood flow underneath the skin. Doppler OCT can provide this functional
extension of OCT by measuring the Doppler frequency shift from moving objects. This
32
frequency shift can be used to determine the velocity of an object. The velocity resolution
depends upon the detection electronics, scan angle and acquisition time. Recent developments
[52] in this field have shown that a velocity resolution of just a few micrometers per second
can be achieved.
OCT elastography (OCE) [53], is performed by speckle tracking of OCT images from
an object under loading and measures the displacement field, from which the mechanical
properties can be determined. Comparing with elastography based on techniques of
ultrasound or magnetic resonance imaging (MRI), the most important advantage is that OCE
can measure a displacement of only a few micrometers. This advantage makes OCE popular
for numerous applications in the biomedical field, e.g. for measuring deformation and strain
of atherosclerotic tissues, breast and brain tissues [54, 55], but OCE in material engineering
field has not been reported. A detailed OCE principle description and its application for
material characterization is presented in chapter 6.
2.5 Discussion and Conclusions

This chapter reviews the evolution of OCT in the past 20 years. Due to the ability to see
through opaque objects and its inherent micrometer resolution, OCT has been considered as a
vital measurement tool for tomographic imaging. Based on low coherence interferometry,
OCT can be performed in both time domain (TD-OCT) and Fourier domain (FD-OCT). The
principles of both techniques have been described and their performances were compared in
detail. Generally, TD-OCT has an unlimited depth scan range and can easily achieve a high
SNR with balanced detection and a bandpass filtering, while FD-OCT has an advantage in
fast imaging due to no movement of the reference mirror.
The developments of OCT generally follow one principle to obtain more information
in less time. Obtaining more information includes at first an improved resolution and
penetration depth. This requires a high-quality light source that can provide a wide bandwidth
for ultra-short coherence, and has a center wavelength around 1.2 - 1.8 m to avoid strong
absorption and scattering within the object. More information can also be obtained by
providing better or different contrast, such as PS-OCT and DOCT giving access to
birefringence and fluid flow in objects respectively. In less time implies high-speed
imaging, which is crucial to dynamic studies and online inspection tasks. Currently real-time
imaging is achievable by either TD-OCT with advanced ODL, or by new generational OCT
33
Chapter 2
modalities, such as FD-OCT, SS-OCT and FF-OCT. The development of OCT modalities,
together with light source and scanners are summarized in Table 2.2.
For future development of OCT, directions could be reductions in complex, size, and
cost. Due to the need of in-situ imaging for biomedical studies or industrial tests, a compact
and robust system, such as handheld OCT [56], is essential for this type of applications. It is
also noted the most technique advances so far in OCT are driven by biomedical applications.
Although some of them can be transferred or used directly for NDT applications, OCT
specialized for NDT is needed for better measurement. For example, the incident power on a
human tissue is limited to the microwatt range for safety reasons [57]. However in the NDT
field, higher powers can be used for improved SNR. Besides, most biomedical tissues contain
a lot of water that has strong absorption bands at, e.g. 1440 nm, 1930 nm, and generally the
water absorption increases with the longer light wavelength [58], which limits the source
selection for biomedical applications. However, this is not a problem in the NDT field [59] for
solid materials, such as glass fiber composite. Light sources with longer wavelengths can be
used for less scattering and hence better penetration depth.
34
Table 2.2: A summary of OCT development
Function Light Source Scanner
TD-OCT Structural imaging by Near infrared light with broad Slow imaging due to the sweeping
sweeping reference bandwidth: of the reference mirror.
mirror in a
1. SLD: simple and low cost. Advanced scanners such as gating
low-coherence
based ODL can reach 4000
interferometer. 2. Femetosecond: high power,
A-scans/s.
wide bandwidth but complex.
FD-OCT Structural imaging Same as TD-OCT. Fast imaging decided by the CCD
using spectrometer integration time.
combined with
500,000 A-scans/s achieved.
Fourier transform for
A-scan in one shot.
SS-OCT Structural imaging Swept source such as FDML, Fast imaging decided by the
using swept source mostly centered at 1 and 1.3 sweeping rate of light source.
combined with m.
326,000 A-scans/s achieved.
Fourier transform for
A-scan.
FF-OCT Structural imaging Mostly halogen light which Fast imaging decided by the
using area CCD for has a bandwidth more than integration time and the sweeping
en face imaging in 300 nm, providing of reference mirror. A volumetric
one shot. submicrometer resolution. image can be acquired in 1s.
PS-OCT Enhanced structural Decided by the basic OCT Decided by the basic OCT
imaging with modality used (TD, FD, SS or modality used.
birefringence FF-OCT).
information by
measuring
polarization state.
DOCT Flow velocity Decided by the basic OCT Decided by the basic OCT
measurement using modality used. modality used.
OCT combined with
LDV
OCE Mechanical Decided by the basic OCT Decided by the basic OCT
properties modality used. modality used.
measurement using
OCT combined with
elastography
35
Chapter 2
References
[1] D. Huang, E. Swanson, C. Lin et al., Optical coherence tomography, Science,

254(5035), 1178-1181 (1991).
[2] T. Li, A. Wang, K. Murphy et al., White-light scanning fiber Michelson

interferometer for absolute position-distance measurement, Optics Letters, 20(7),
785-787 (1995).
[3] H. Maruyama, S. Inoue, T. Mitsuyama et al., Low-Coherence Interferometer

System for the Simultaneous Measurement of Refractive Index and Thickness,
Applied Optics, 41(7), 1315-1322 (2002).
[4] M. Wojtkowski, R. Leitgeb, A. Kowalczyk et al., In vivo human retinal imaging

by Fourier domain optical coherence tomography, Journal of Biomedical Optics,
7(3), 457-463 (2002).
[5] J. T. Oh, and S. W. Kim, Polarization-sensitive optical coherence tomography for

photoelasticity testing of glass/epoxy composites, Optics Express, 11(14),
1669-1676 (2003).
[6] J. A. Izatt, M. D. Kulkarni, S. Yazdanfar et al., In vivo bidirectional color

Doppler flow imaging of picoliter blood volumes using optical coherence
tomography, Optics Letters, 22(18), 1439-1441 (1997).
[7] N. Nishizawa, "Ultrahigh resolution optical coherence tomography,"

Opto-Electronics and Communications Conference (OECC), 2012 17th, 584-585
(2012).
[8] W. Drexler, and J. G. Fujimoto, State-of-the-art retinal optical coherence

tomography, Progress in Retinal and Eye Research, 27(1), 45-88 (2008).
[9] M. R. Hee, [Optical coherence tomography of the eye], Massachusetts Institute of

Technology (1997).
36

[11] J. M. Schmitt, Optical coherence tomography (OCT): a review, IEEE Journal of

Selected Topics in Quantum Electronics, 5(4), 1205-1215 (1999).
[12] M. E. Brezinski, [Optical coherence tomography: principles and applications],

Academic Press (2006).
[13] A. Rollins, S. Yazdanfar, M. Kulkarni et al., In vivo video rate optical coherence
[14] L. An, P. Li, T. T. Shen et al., High speed spectral domain optical coherence
tomography for retinal imaging at 500,000 Alines per second, Biomedical
Optics Express, 2(10), 2770-2783 (2011).
[15] L. Sang-Won, J. Hyun-Woo, A. Yeh-Chan et al., Optimization for Axial

Resolution, Depth Range, and Sensitivity of Spectral Domain Optical Coherence
Tomography at 1.3 m, Journal of the Korean Physical Society, 55(6), 2354
(2009).
[16] J. Czajkowski, T. Fabritius, J. Ulaski et al., Ultra-high resolution optical

coherence tomography for encapsulation quality inspection, Applied Physics B,
105(3), 649-657 (2011).
[17] W. Drexler, U. Morgner, F. Krtner et al., In vivo ultrahigh-resolution optical

coherence tomography, Optics Letters, 24(17), 1221-1223 (1999).
[18] J. W. Goodman, [Statistical optics], Wiley-Interscience (1985).
[19] P. H. Tomlins, and R. K. Wang, Theory, developments and applications of optical

coherence tomography, Journal of Physics D: Applied Physics, 38(15),
2519-2535 (2005).
[20] D. Derickson, [Fiber optic test and measurement], Prentice Hall (1998).
37
Chapter 2
[21] J. F. De Boer, B. Cense, B. H. Park et al., Improved signal-to-noise ratio in

spectral-domain compared with time-domain optical coherence tomography,
Optics Letters, 28(21), 2067-2069 (2003).
[22] R. Leitgeb, C. Hitzenberger, and A. F. Fercher, Performance of fourier domain vs.

time domain optical coherence tomography, Optics Express, 11(8), 889-894
(2003).
[23] J. G. Fujimoto, and D. Farkas, [Biomedical optical imaging], Oxford University

Press (2009).
[24] M. Wojtkowski, High-speed optical coherence tomography: basics and

applications, Applied optics, 49(16), D30-D61 (2010).
[25] B. Liu, and M. E. Brezinski, Theoretical and practical considerations on

detection performance of time domain, Fourier domain, and swept source optical
coherence tomography, Journal of Biomedical Optics, 12(4), 044007 (2007).
[26] B. E. Bouma, L. E. Nelson, G. J. Tearney et al., Optical coherence tomographic

imaging of human tissue at 1.55 m and 1.81 m using Er-and Tm-doped fiber
sources, Journal of Biomedical Optics, 3(1), 76-79 (1998).
[27] U. Sharma, E. W. Chang, and S. H. Yun, Long-wavelength optical coherence

tomography at 1.7 m for enhanced imaging depth, Optics Express, 16(24),
19712-19723 (2008).

developments for optical coherence tomography, Applied Physics B, 88(3),
337-357 (2007).
[29] M. Wurm, K. Wiesauer, K. Nagel et al., "Spectral-domain optical coherence

tomography: a novel and fast tool for NDT," IV NDT in Progress, (2007).
[30] H. Wang, M. W. Jenkins, and A. M. Rollins, A combined multiple-SLED

broadband light source at 1300nm for high resolution optical coherence
tomography, Optics Communications, 281(7), 1896-1900 (2008).
38
[31] T. Ko, D. Adler, J. Fujimoto et al., Ultrahigh resolution optical coherence

tomography imaging with a broadband superluminescent diode light source,
Optics Express, 12(10), 2112-2119 (2004).
[32] B. Bouma, G. Tearney, I. Bilinsky et al., Self-phase-modulated Kerr-lens

mode-locked Cr: forsterite laser source for optical coherence tomography, Optics
Letters, 21(22), 1839-1841 (1996).
[33] H. Kuroda, M. Baba, M. Suzuki et al., A high speed three-dimensional spectral

domain optical coherence tomography with< 2 m axial resolution using wide
bandwidth femtosecond mode-locked laser, Applied Physics Letters, 102(25),
251102 (2013).
[34] B. J. Kaluzny, M. Gora, K. Karnowski et al., Imaging of the lens capsule with an
ultrahigh-resolution spectral optical coherence tomography prototype based on a
femtosecond laser, British Journal of Ophthalmology, 94(3), 275-277 (2010).
[35] Y. Pan, E. Lankenou, J. Welzel et al., Optical coherence-gated imaging of

biological tissues, IEEE Journal of Selected Topics in Quantum Electronics, 2(4),
1029-1034 (1996).
[36] L. Yan, X. S. Yao, L. Lin et al., Improved fast scanning delay line in optical
coherence tomography applications utilizing fiber stretcher, Optical Engineering,
47(10), 100501-100501-3 (2008).
[37] J. Szydlo, N. Delachenal, R. Gianotti et al., Air-turbine driven optical

low-coherence reflectometry at 28.6-kHz scan repetition rate, Optics
Communications, 154(1), 1-4 (1998).
[38] N. Delachenal, R. Wlti, R. Gianotti et al., Robust and rapid optical

low-coherence reflectometer using a polygon mirror, Optics Communications,
162(4), 195-199 (1999).
[39] Laser scanners, http://www.zamisel.com/SSpostavka2.html.
[40] V. Westphal, A. Rollins, S. Radhakrishnan et al., Correction of geometric and
39
Chapter 2
refractive image distortions in optical coherence tomography applying Fermat's

principle, Optics Express, 10(9), 397-404 (2002).
[41] S. Chinn, E. Swanson, and J. Fujimoto, Optical coherence tomography using a

frequency-tunable optical source, Optics Letters, 22(5), 340-342 (1997).
[42] R. Huber, D. Adler, V. Srinivasan et al., Fourier domain mode locking at 1050
nm for ultra-high-speed optical coherence tomography of the human retina at
236,000 axial scans per second, Optics Letters, 32(14), 2049-2051 (2007).
[43] P. Targowski, M. Gra, and M. Wojtkowski, Optical coherence tomography for

artwork diagnostics, Laser Chemistry, 2006, 1-11 (2006).
[44] Overview of polarization-maintaining single mode optical fiber,

http://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=1596.
[45] A. Dubois, G. Moneron, K. Grieve et al., Three-dimensional cellular-level

imaging using full-field optical coherence tomography, Physics in Medicine and
Biology, 49(7), 1227 (2004).
[46] E. Beaurepaire, A. C. Boccara, M. Lebec et al., Full-field optical coherence

microscopy, Optics Letters, 23(4), 244-246 (1998).
[47] E. Dalimier, and D. Salomon, Full-field optical coherence tomography: a new

technology for 3D high-resolution skin imaging, Dermatology, 224(1), 84-92
(2012).
[48] A. Dubois, and A.-C. Boccara, [Full-field optical coherence tomography],

Springer (2008).
[49] J. F. de Boer, and T. E. Milner, Review of polarization sensitive optical

coherence tomography and Stokes vector determination, Journal of Biomedical
Optics, 7(3), 359-71 (2002).
[50] D. Stifter, E. Leiss-Holzinger, Z. Major et al., Dynamic optical studies in

materials testing with spectral-domain polarization-sensitive optical coherence
40
[51] V. Yang, M. Gordon, B. Qi et al., High speed, wide velocity dynamic range
Doppler optical coherence tomography (Part I): System design, signal processing,
and performance, Optics Express, 11(7), 794-809 (2003).
[52] R. A. Leitgeb, R. M. Werkmeister, C. Blatter et al., Doppler Optical Coherence

Tomography, Progress in Retinal and Eye Research, 41, 26-43 (2014).
[53] J. M. Schmitt, OCT elastography: imaging microscopic deformation and strain

of tissue, Optics Express, 3(6), 199-211 (1998).
[54] J. Rogowska, N. Patel, J. Fujimoto et al., Optical coherence tomographic

elastography technique for measuring deformation and strain of atherosclerotic
tissues, Heart, 90(5), 556-562 (2004).
[55] C. Sun, B. Standish, and V. X. Yang, Optical coherence elastography: current

status and future applications, Journal of Biomedical Optics, 16(4), 043001
(2011).
[56] W. Jung, J. Kim, M. Jeon et al., Handheld optical coherence tomography scanner
for primary care diagnostics, IEEE Transactions on Biomedical Engineering,
58(3), 741-744 (2011).
[57] W. Drexler, U. Morgner, R. K. Ghanta et al., Ultrahigh-resolution ophthalmic

optical coherence tomography, Nature Medicine, 7(4), 502-507 (2001).
[58] G. M. Hale, and M. R. Querry, Optical constants of water in the 200-nm to

200-m wavelength region, Applied Optics, 12(3), 555-563 (1973).
[59] S. Zhong, Y.-C. Shen, L. Ho et al., Non-destructive quantification of

pharmaceutical tablet coatings using terahertz pulsed imaging and optical
coherence tomography, Optics and Lasers in Engineering, 49(3), 361-365
(2011).
41
3
Development of a Hybrid OCT
System
3.1 Introduction
As discussed in Chapter 2, current commercial OCT instruments are mostly designed for
biomedical applications. These systems use light sources centered typically at 800 nm or 1300
nm, with an incident power in the microwatt range. These parameters are not suitable for a
reasonable penetration depth of 2-3 mm or good SNR in scattering materials such as glass
fiber composite. Recent reports [1, 2] have shown the detectable depth in polymer-based
materials is enhanced or even doubled when a light source around 1550 nm is used, compared
with one around 840 nm. Besides, in some case it is necessary to tailor an OCT system
towards one particular application, e.g. inline characterization of specimens at the production
site, or combining the OCT test with mechanical loadings. Most commercial OCT systems are
desktop instruments that can have few degrees of freedom.
In this chapter, a customized OCT system for non-destructive testing, especially for
polymer-based material characterization is described. An overview of the design is introduced
first, including a schematic of the system. Then each component used in the system is
described in detail, together with a discussion of their influence on the overall OCT system
performance. The specific implementation of the measurement using the customized OCT
system is given at the end of this chapter.
3.2 System Overview

For this study, a compact and robust fiber-optic OCT system was built as shown in Fig. 3.1
and Fig. 3.2. A key feature of the instrument is that it integrates TD-OCT and FD-OCT into
43
Chapter 3
one system so the individual strengths of each design can give an advantage to different
applications. Fig. 3.1 shows a schematic of the developed OCT system [3], and Fig. 3.2 is a
photography showing the experimental setup on the optical table.
Fig. 3.1: Schematic setup of a hybrid TD-FD-OCT. Abbreviations: SLD,

superluminescent diode; C, Circulator; P, polarizer; PM, polarization maintaining optical
fibers; FC, fiber coupler; ODL, optical delay line; R, reflector; CO, collimator; FL,
Focusing Lens; TS, translation stage; SP, sample; OS, optical switch; BD, balanced
detector; DAQ, data acquisition board; SM, spectrometer; PC, personal computer. X and Y
are the two orthogonal lateral directions indicating the movement of the translation stages,
and Z is the axial direction indicating the movement of ODL.
44
Development of a Hybrid OCT System
Fig. 3.2: Experimental setup on an optical table.
The technique is based on low coherence interferometry and is implemented in fiber

optics. The light beam from a superluminescent diode (SLD, FESL-1550-20-BTF, Frankfurt
45
Chapter 3
Laser Company, Germany), passes through a circulator (6015-3-APC, Thorlabs GmbH,

Germany) and is linear polarized by an optical polarizer (FOP-21-11-1550, OZ Optics Ltd,
Canada). It is then coupled with a fiber coupler (FUSED-22-1550-8/125-50/50-PM, OZ
Optics Ltd, Canada) and divided into reference and sample arms with equal optical power. In
the reference arm, light travels through an optical delay line (ODL-650-MC, OZ Optics Ltd,
Canada) and arrives in front of a reflector (FORF-11P-1550, OZ Optics Ltd, Canada), where
it is reflected back along the same path to the fiber coupler. In the sample arm, Light passing
through a second polarizer is collimated by a collimation lens (F260APC-1550, Thorlabs
GmbH, Germany) and then focused to the specimen by a achromatic doublet lens
(AC254-030-C-ML, Thorlabs GmbH, Germany). It is then reflected from the sample and
travels back to the same fiber coupler, where it is combined with backreflected reference light
and is split again. Half of the light goes back to the circulator and is directed to one input of
the balanced detector (PDB420C, Thorlabs GmbH, Germany). Another half goes to an optical
switch (OSW22-1310E, Thorlabs GmbH, Germany), which based on the testing request could
either direct the recombined light to the second input of the balanced detector for time-domain
measurement or feed the light into a spectrometer (NIRQuest512, Ocean Optics Inc, US) for
Fourier-domain measurement. During an experimental test, this can also be realized by simply
connecting the output fiber with the balanced detector or spectrometer manually without the
optical switch, as shown in Fig. 3.2.
For TD-OCT, using the scan of the reference mirror in an optical delay line to change
the reference pathlength, a set of interferometric signals will be detected as a function of the
reference mirror position. This set corresponds to the axial distribution of scattering interfaces
within the specimen and is referred to as an A-scan. While for FD-OCT, spectral intensities
are measured at one shot without the movement of reference mirror. In FD-OCT the A-scan
signal is recovered by inverse Fourier transform, as described in section 2.2.2. By laterally
shifting the probe beam with a pair of orthogonal translation stages (T-LS28M, Zaber Inc,
Canada), a set of consecutive A-scans can be obtained. These A-scans are then combined to
form a cross sectional image of the structure, which is named a B-scan. A volumetric image
can be reconstructed as well by combining a succession of B-scans.
3.3 Detail of components

3.3.1 Light Source
The OCT system uses a fiber-coupled low-coherence superluminescent diode (SLD,

FESL-1550-20-BTF, Frankfurt Laser Company, Germany) as the light source. The use of an
46
SLD is due to its advantages of simplicity, stability, and low cost, compared with its
counterpart, e.g. femtosecond lasers, as discussed in chapter 2. SLD is also ideal for the
design of a portable OCT system which enables the measurement in situ. Besides, the selected
SLD can satisfy the power and wavelength requirement for the customized OCT system, as
described below.
As shown in Fig. 3.2, the SLD is located in a diode mount (LM14S2, Thorlabs GmbH,
Germany) which is driven by a temperature controller (TED200C, Thorlabs GmbH, Germany)
and a current controller (CAB400, Thorlabs GmbH, Germany). The current controller can
increase the current from 0 to 500 mA, resulting in a proportionally growing output power of
the SLD, as shown in Fig. 3.3(a) [4]. The maximum power of light source is 20 mW, and the
power of light finally emitted on the specimen is 2 mW. Besides the fiber coupler that directs
half the power of the light to the reference arm, the power loss of the sample beam is mainly
from the polarization control and the optical fiber connectors. However, a 2 mW sample beam
is much stronger than the biomedical counterparts, resulting in a higher SNR that allows weak
reflections from deep (e.g. 2-3 mm) of a scattering sample to be detected. A quantitative
comparison of SNR with regard to source power is given in chapter 4.
The SLD has a center wavelength of 1550 nm, which is less scattering and gives
deeper penetration in polymer-based material than the shorter wavelength used for biomedical
applications [1]. The light source has a FWHM (Full Width at Half Maximum) of 60 nm.
According to Eq. 2.16, it can provide a theoretical axial resolution of 17 m in the air, and
11m in the polymer material with a 1.55 average refractive index taken into account.
However, it is noted that the spectral distribution of the light source is not exactly a Gaussian
shape but has small spectral ripples , as shown in Fig. 3.3(b) and (c) [4]. This type of
imperfect light source could bring artifacts into the interference signal, which is evaluated in
the next chapter.
47
Chapter 3
Fig. 3.3: Performance of the SLD [4], including (a) optical power, (b) spectrum
distribution, and (c) spectrum ripple. ASE: amplified spontaneous emission.
3.3.2 Optical Fiber
Light emitted from SLD is conducted through single-mode fibers in the OCT system. The use
of fiber optics instead of bulk optics in free space is due to their convenience and ease of
alignment. The difference between single-mode and multimode fiber is shown in Fig. 3.4(a)
[5]. Single-mode fiber only support one propagation mode due to its relatively
small core (approximately 10 m) and a small refractive index difference between core and
cladding (for a larger critical angle during total internal reflection). Therefore single- mode
fiber enables the spatial coherence of light to be retained, which is crucial for an
interferometer system.
48
Fig. 3.4: (a) schematic diagrams for light transmitting in multimode and single mode fiber
[5] and (b) a FC/APC fiber connector and a mating sleeve used in the OCT system to
connect components.
To connect each component, FC/APC connectors and mating sleeves are used, as
shown in Fig. 3.4(b). FC stands for ferrule connector while APC stands for angled physical
contact. These connectors allow the surfaces of two connected fibers to be in direct contact
with each other and the fiber end is polished at an angle. The physical contact can reduce the
light power loss while the angled fiber end (with a 8angled ceramic ferrule) can prevent
reflected light traveling back to the fiber, which can eliminate artifacts in the interference
signal. Additionally, to match the pathlengths of the reference and sample arms, a customized
length of fiber is needed. A manual describing the customization of fibers with connectors is
introduced in [6].
3.3.3 Interferometer
The interferometer configuration employed in the OCT system is a fiber-optic Michelson

interferometer. The core of the interferometer system is a fiber coupler
(FUSED-22-1550-8/125-50/50-PM, OZ Optics Ltd, Canada), which is equivalent to an optical
beam splitter in free space. As shown in Fig. 3.2, a fiber coupler is generally a four-port
device consisting of two fibers. These fibers are fused together, etched or polished over a
small interaction region where the exchange of power between each fiber core occurs [7]. In
the designed Michelson interferometer, the fiber coupler splits the power of the input light
source equally between reference and sample arms and combines the backreflected reference
49
Chapter 3
and sample light. It then divides the power of combined light into two halves again to the
balanced detector or spectrometer.
It is noted that some interferometer designs based on a Mach-Zehnder interferometer

have been proposed for OCT [8]. As shown in Fig. 3.5, OCT systems of this type utilize two
fiber couplers to divide the optical source and to combine the backreflected reference and
sample beams. The first fiber coupler could be unbalanced to direct more light power to the
sample arm. This is done to optimize the imaging performance since the SNR of an OCT
image is proportional to the power of sample backreflection (Eq. 2.21), which is substantially
less than reference signal. However, this design requires an additional fiber coupler, compared
with the Michelson interferometer. The velocity of the ODL, introduced in section 3.3.7,
would also be cut to half due to the one-way reflection.
Fig. 3.5: Schematic diagram of a Mach-Zehnder interferometer. denotes the light splitting
ratio.
3.3.4 Polarizer and Polarization Maintaining Fiber
An electromagnetic wave, such as light, is composed of an electric field and a magnetic field
that travel together at the same velocity and direction, but oscillate perpendicularly. Once we
know the electric field we can easily determine the magnetic field. Thus we usually deal only
with electrical field and define light polarization as the oscillation orientation of its electrical
field. In an interferometer system, the matching of polarization states of the combined light is
one of the elements that decide the interference quality. The interference signal is strongest
when the two light beams have some polarization states. It decays with the increase of the
polarization difference and varnishes if the two polarizations are perpendicular.
It is worth mentioning that standard optical fibers always exhibit random birefringence
due to mechanical stress and geometric effects such as bending [9]. Birefringence could also
occur in the tested specimens. It is a phenomenon that light travelling in certain types of
50
materials is split into two orthogonal polarization components with different velocities. As a
consequence, in OCT, interferences could occur under both polarization states and be
recorded at different depth locations, which brings in artifacts in the final structural image.
To obtain an interference signal with good quality, the light backreflected from the
reference mirror and the specimen have to have the same polarization state. Therefore, a novel
fiber polarizer (FOP-21-11-1550, OZ Optics Ltd, Canada) is employed after the SLD to
restrict light to a linear polarization before going to the interferometer. A same fiber polarizer
is used in the sample arm to polarize the backreflected sample light to its original state. As
depicted in Fig. 3.6, a fiber polarizer of this type allows only the light components with one
linear polarization state to be transmitted. The light components with an orthogonal
polarization state are absorbed by a crystal or a thin film placed over the fiber core [10].
Fig. 3.6: A linear polarizer filters an unpolarized beam into one with a single linear
polarization. Light travels from left to right and the arrows depict the electric field vector.
Besides, to retain the linear polarization state, all the single-mode fibers in the
interferometer have to be polarization maintaining (PM). Generally PM fiber
(PMF-1550-8/125-0.25-L, OZ Optics Ltd, Canada) has a pair of stress rods running parallel to
the fiber core. These stress rods intentionally creates consistent birefringence pattern along its
length, prohibiting coupling between the two orthogonal polarization directions [11].
3.3.5 Circulator and Optical Switch
A fiber optic circulator (6015-3-APC, Thorlabs GmbH, Germany) is a passive, three-port

device that acts as a signal router. As shown in Fig. 3.7, light from the input fiber 1 is directed
from port 1 to port 2, where light is coupled into the interferometer system, returned back, and
51
Chapter 3
redirected to a third port. Light input into port 1 will not be coupled into the port 3 fiber, and
light input into port 2 will not be coupled into the port 1 fiber. Therefore, the fiber circulator
used in the system could play two functions. It directs the interference signal to the second
input of the detector for balanced detection. It could also act as an optical isolator to resist
light returning back to the source and protect the SLD module from being damaged by optical
feedback.
Fig. 3.7: Schematic diagram of an optical circulator.
Fig. 3.8 shows the fiber optical switch (OSW22-1310E, Thorlabs GmbH, Germany)
used the hybrid OCT system. The light from the input fiber can be directed to either of the
two output fibers by an optical switch with an integrated control circuit. The switching
mechanism is based on silicon Micro-Electro-Mechanical Systems (MEMS) technology,
which ensures exceptionally low crosstalk between channels, and is inherently fast (switching
time <1ms) [12]. The switching can be easily operated by toggle button on the circuit board or
be remotely controlled by software from a PC. In the final design of the integrated OCT
prototype, the optical switch decides if TD-OCT or FD-OCT will be applied for the
measurement by directing the interference signal to either the balanced detector or the
spectrometer.
Fig. 3.8: A fiber optical switch with controlling circuit [12].
52
3.3.6 Collimator and Focusing Lens
At the end of the fiber in the sample arm, the light coming to free space is collimated by a
collimator lens (F260APC-1550, Thorlabs GmbH, Germany), resulting in an output beam size
of 3 mm diameter. The collimated light is then focused to the specimen by an achromatic
doublet lens (AC254-030-C-ML, Thorlabs GmbH, Germany) with a 30 mm focal length.
According to Eq. 2.17 to Eq. 2.19, the lateral resolution of this OCT system, decided by the
spot size, is 20 m and depth of focus is 1.24 mm. Therefore, to measure a specimen with 2-3
mm thickness, it is better to align the focus point in the middle depth to minimize the image
deterioration caused by an out-of-focus measurement.
3.3.7 Optical Delay Line and Reflector
The optical delay line (ODL) in the hybrid OCT system has two functions: changing the
reference pathlength for an axial scan in TD-OCT, and for placing a small offset z0 for
separating the disturbing terms in the FD-OCT signal as discussed in chapter 2. The selected
ODL (ODL-650-MC, OZ Optics Ltd, Canada), as shown in Fig. 3.9 [13], is an electrically
controlled, reflector-style, variable device consisting of an input fiber collimator and an output
fiber coupler. The collimator projects light into free space inside the device and the coupler
directs it to the output fiber. The width of the gap between the components determines the
delay time through the device. The distance the light travels in free space is precisely
controlled by reflecting the light from a pair of moveable mirrors, which are driven by a servo
motor with a microprocessor. The microprocessor enables remote communication through an
electrical device or PC.
The servo motor provides a resolution of 0.4 m and a linear translation speed of 0.85
mm/s with maximum 25 mm travel range. In the customized OCT system shown in Fig. 3.1
and Fig. 3.2, the output of the ODL is connected to a reflector (FORF-11P-1550, OZ Optics
Ltd, Canada) that enables the light to be reflected back to the entrance port of the ODL.
Therefore, due to this dual path travel of light in the ODL, A-scan range and speed are
doubled, which correspond to 50 mm and 1.7 mm/s respectively in air. This A-scan speed and
its linearity are evaluated in chapter 4. According to Eq. 2.23, the Doppler frequency induced
by the ODL is 2193 Hz. For a specimen with 2-3 mm thickness, each A-scan in TD-OCT will
cost approximately 2 s.
53
Chapter 3
Fig. 3.9: Schematic diagram of the ODL used in the OCT system [13]. The output of the
ODL is connected to a reflector that let the light travel back along the same path, which
doubles the A-scan velocity and range during an OCT measurement.
In the fiber optical reflector (FORF-11P-1550, OZ Optics Ltd, Canada) shown in Fig.
3.10, a collimation lens and a static mirror are used to reflect partial light back to the fiber.
The blocking screw in front of the mirror decides the attenuation of the backreflected light.
Due to the strong backreflected light from the reference arm, it is sometimes necessary to
attenuate the reference power to avoid saturation of the detectors.
Fig. 3.10: Schematic diagram of the variable fiber optic reflector. The bidirectional arrow
indicates light travelling ways.
3.3.8 Balanced Detector and DAQ Board
In TD-OCT, the interference signal is measured by a balanced photo detector (PDB420C,

Thorlabs GmbH, Germany). The balanced detector consists of two photodiodes and an
ultra-low-noise, high-speed transimpedance amplifier that generates an output voltage
proportional to the difference between the photocurrents of the two photodiodes [14]. A
schematic of the detector is shown in Fig. 3.11. The balanced photo detector acts as a
balanced receiver by subtracting the two optical input signals from each other, resulting in the
cancellation of common mode noise and hence gives an improved SNR of the interferometric
54
signal. As an example, Fig. 3.12 compares one A-scan signal acquired by the unbalanced and
balanced detection (MONITOR+ and RF OUTPUT in Fig. 3.11) simultaneously. It can be
seen clearly that the noise level is decreased significantly by balanced detection. A
quantitative SNR of the system is calculated in the next chapter.
Fig. 3.11: Schematic diagram of the selected balanced detector [14]. The two optical input
signals are subtracted from each other, resulting in the cancellation of common mode noise.
Abbreviations: TIA:, transimpedance amplifier, OPAMP, operational amplifier.
Fig. 3.12: A-scan signal acquired by (a) unbalanced and (b) balanced detection. The
improvement of SNR with balanced detection can be clearly observed.
The analog voltages generated by the balanced detector need to be further captured
and digitized by a data acquisition board (DAQ) before being sent to a computer. The selected
DAQ board (NI 9233) can provide a 50 kS/s maximum sampling rate, which is suitable for
measuring an interference signal with a modulation frequency around 2 kHz. The captured
55
Chapter 3
signal is digitized with 24-bit resolution. Thus the maximum dynamic range for the displayed
TD-OCT images is 72 dB, given by:
dy 10log(2M ) (3.1)
where dy is the displayed image dynamic range and M is the bits for AD conversion. It has to
be pointed out that the typically reported SNR of an OCT system above 80 dB refers to an
analog signal [15]. In practice, the displayed image has a substantially reduced dynamic range
due to the signal loss associated with the limitation of the AD conversion. From Eq. 3.1, it can
be seen the displayed image dynamic range is proportional to the bit number M in a DAQ
board.
3.3.9 Spectrometer
For FD-OCT, spectral irradiances are collected by a spectrometer (NIRQuest512, Ocean

Optics Inc, US). Fig. 3.13 shows the schematic of the selected spectrometer [16]. The light
source is brought into the instrument with a fiber optic cable connected to the optical probe.
The amount of entering light is controlled by an adjustable slit and its wavelength range is
restricted by a filter. After collimation by a mirror, the light is scattered for different
wavelengths by a diffraction grating. The dispersed light falls onto a focusing mirror and is
directed to an array of photodiodes, each of which responds only to the narrow range of
wavelengths impinging on it. The photodiodes are connected to a charge-coupled device
(CCD) that produces voltages. The voltages from each diode are converted to digital counts
and sent to a dedicated computer with spectrum analysis software. The acquired spectrum is
immediately displayed on the screen and saved in standard computer files.
56
Fig. 3.13: Light enters the spectrometer from a fiber optic cable connected to a probe (1). It
is then conditioned by an adjustable slit (2) , a filter (3) , a collimating mirror (4) , and a
diffraction grating (5). The grating scatters the different wavelengths onto a focusing
mirror (6) that directs the light via a collector lens (7) onto the detector array (8) [16].
The spectrometer used comprises a 512-elment InGaAs array and an adjustable full
spectral range for detection which is decided by the overall performance of the OCT system.
In FD-OCT, the interval of image pixels in the axial direction z is determined as the depth
range zmax (Eq. 2.20) divided by half of the total detector elements, given by:
02
z (3.2)
2 full
The pixel interval z should be smaller than half of the theoretical axial resolution z from
Eq. 2.16 to satisfy the Nyquist sampling theory [17]. Therefore, the full spectral range full
detected by the spectrometer should satisfy:
(3.3)
Since the FWHM of the light source, is 60 nm, full is decided as 160 nm. Combined
with 512-element detectors, a maximum detectible depth of 1.92 mm in the air can be
achieved, according to Eq. 2.20. The detectible depth can be improved with more detectors
(e.g. 1024 elments), but the cost would much higher as expensive InGaAs photodiodes are
needed for the detection of near infrared light.
57
Chapter 3
The minimum integration time set by this spectrometer is 1 ms, which means a
maximum rate of 1000 A-scans/s can be achieved in the FD-OCT. For AD conversion, 16-bit
resolution is used, resulting in a displayed image dynamic range of 48 dB, based on Eq. 3.1.
3.3.10 Translation Stages
For lateral scans, a pair of translation stages (T-LS28M, Zaber Inc, Canada) are used to shift
the OCT probe beam in two orthogonal directions parallel to the specimen surface. These
motorized stages provide 0.1 m resolution and 28 mm travel range. The maximum
translation speed is 6.5 mm/s. The units can be controlled by a PC via a single RS-232 serial
port.
The use of translation stages instead of galvo mirrors could be for two reasons. Firstly, as
reviewed in chapter 2, translation stages do not generate the image distortion that occurs in a
galvo mirror system. So they are easy to use and no extra cost or work, e.g. telecentric lens or
signal postprocessing is needed. The other reason is due to the fact that the A-scan rate in the
designed OCT system is not ultra-high. The translation stages are fast enough to couple with
the A-scan system. For example, considering the highest 1000 A-scans/s in FD-OCT, a
cross-sectional image with 6.5 mm length can be acquired in 1s. The pixel interval in lateral
direction is 6.5 m, which enables a display of all the detailed structure of a specimen,
considering 20 m spot size of the sample beam.
A summary of the components and their performance in the OCT system are
summarized in Table 3.1.
58
Table 3.1: Summary of the OCT Components and their performance in the integral OCT system
Components Model and Specifications Performance in OCT

Manufacturer
Light Source FESL-1550-20-BTF, - 20 mW - 2-3 mm penetration
Frankfurt Laser - =1550 nm , =60 nm - 17 m axial resolution
Company, Germany - Close to Gaussian shape spectrum - Small artifacts in the
with small spectral ripples interference signal
Interferometer FUSED-22-1550-8/1 - Michelson interferometer - Simple setup
25-50/50-PM - fiber optics - Easy for alignment
(coupler) - FC/APC connectors - Low backreflection
Polarizer & FOP-21-11-1550, OZ Linear polarization state - High quality
PM Fibers Optics Ltd, Canada interference signal
Circulator & 6015-3-APC & - Circulator: one directional light - Isolation of light back
Optical OSW22-1310E, transmission to the source
Switch Thorlabs GmbH, - Optical switch: fast switching - Direct interference
Germany (<1ms) and remote control. signal to the detectors
Collimator F260APC-1550, and - Collimator: 3 mm collimating beam - 20 mm spot size
and Focusing AC254-030-C-ML, size (lateral resolution)
Lens Thorlabs GmbH, - Focusing lens: 30 mm focus length - 1.24 m depth of focus
Germany
Optical Delay ODL-650-MC & - 0.4 m step resolution - 1.7 mm/s A-scan speed
Line & FORF-13A-1550-S, - 0.85 mm/s with maximum 25 mm with 50 mm scan range
Reflector OZ Optics Ltd, travel range in TD-OCT.
Canada - Variable backreflection - Optimized imaging
- Polarization maintaining quality
Balanced PDB420C, Thorlabs Balanced detection - Improved SNR

Detector GmbH, Germany
DAQ NI 9233, National - 50 KS/s Sampling rate - 72 dB displayed image

Instruments Corp., - 24-bit AD conversion dynamic range
US
Spectrometer NIRQuest512, Ocean - 512-elment detectors, 160 nm - 1.92 mm axial
Optics Inc, US detection bandwidth. detection range
- 1 ms minimum integration time - Maximum 1000
- 16-bit AD conversion A-scans/s in FD-OCT
- 48 dB displayed image
dynamic range
Translation T-LS28M, Zaber Inc, - Maximum 6.5 mm/s - Maximum 6.5 mm/s
Stages Canada - 28 mm travel range for lateral scan
- 28x28 mm maximum
imaging area.
59
Chapter 3
3.4 Instrument Control Software

The implementation of the OCT measurement is based on a customized software coupled with
the hybrid OCT setup. Fig. 3.14 describes the procedure for instrument control and data
acquisition during the measurement. Generally, the measurement starts with instrument
initialization, including locating the ODL and translation stage positions. If TD-OCT is
selected for the testing, the ODL will sweep forward and back for A-scans with a range of
approximately the thickness of the specimen. After finishing each A-scan, the translation
stages will shift the objective beam one step laterally until it goes through the required
imaging area. Simultaneously with the scanning, the DAQ system records the data which is
processed afterwards. While in FD-OCT, the ODL is controlled to create a small offset
(around 200 m) between the sample and reference pathlength to separate the signal from DC
noise. Due to its fast scanning property, the translation stages can move continuously and
record the spectrum data simultaneously. With the signal processing algorithm described in
the next chapter, the OCT image can be recovered and finally displayed.
Fig. 3.14: The flow chart for the implementation of OCT measurement.
60
3.5 Conclusions
In summary, a hybrid TD-FD-OCT system has been built. Generally, from the hardware point
of view, TD-OCT has the advantage of a large A-scan range (50 mm in the air) and
substantial noise reduction by balanced detection, while FD-OCT system owns its tremendous
superiority in fast imaging (1000 A-scans/s) due to no movement of the reference mirror.
These two kinds of measurement can be implemented based on the user request, e.g. imaging
depth and speed.
The detail of the components and their performance in the OCT system were
summarized in Table 3.1. Generally, fiber optic interferometry with a light source centered at
1550 nm is used, expecting a high SNR and thus deep penetration in scattering material. In
TD-OCT, an electrical ODL is equipped in the reference arm to sweep the pathlength at 1.7
mm/s, and a balanced detector is applied to receive signal from two inputs for noise reduction.
In FD-OCT, a spectrometer with 160 nm spectral range and 512 detector elements is
employed, reaching 1000 scans/s and 1.92 mm axial scan range. For both TD and FD-OCT, a
pair of translation stages with 6.5 mm/s velocity and 28 mm travelling range, are used to shift
the sample beam for the lateral scan. However, it is necessary to point out that the overall
selection of the components is based on the cost, specifications, availability in the lab, and
operability for coupling into the OCT system. The performance of the OCT system could be
much improved by replacing selected components with the higher-performance counterparts,
but this would greatly increase the total cost of the OCT system.
References
developments for optical coherence tomography, Applied Physics B, 88(3), 337-357
(2007).
[2] M. Wurm, K. Wiesauer, K. Nagel et al., "Spectral-domain optical coherence

tomography: a novel and fast tool for NDT," IV NDT in Progress, (2007).
[3] P. Liu, R. M. Groves, and R. Benedictus, Signal processing in optical coherence

tomography for aerospace material characterization, Optical Engineering, 52(3),
033201 (2013).
61
Chapter 3
[4] [Ultra high power cooled fiber-coupled superluminescent diode FESL-1550-20-BTF],

Frankfurt Laser Company (2008).
[5] Using ethernet with fiber optics - achieving longer distances and superior noise
immunity, http://www.industrialethernetu.com/courses/301_2.htm.
[6] [Guide to connectorization and polishing of optical fibers ], Thorlabs Inc. (2013).
[7] V. J. Tekippe, D. R. Moore, D. K. Paul et al., "Production, performance, and

reliability of fused couplers," International Conference on Fiber Optics and Photonics:
Selected Papers from Photonics India'98, 56-61 (1999).
[8] A. M. Rollins, and J. A. Izatt, Optimal interferometer designs for optical coherence
tomography, Optics Letters, 24(21), 1484-1486 (1999).
[9] R. Ulrich, S. Rashleigh, and W. Eickhoff, Bending-induced birefringence in

single-mode fibers, Optics Letters, 5(6), 273-275 (1980).
[10] P. Perumalsamy, [In-line Fiber Polarizer], Virginia Polytechnic Institute and State
University (1998).
[11] Overview of polarization-maintaining single mode optical fiber,

http://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=1596.
[12] [OSW12(22)-xxxE series operation manual], Thorlabs GmbH (2012).
[13] [Operating instructions for optical delay line ODL-650-MC], OZ Optics Ltd. (2011).
[14] [Operation manual for balanced amplified photodetectectors PDB400 series], Thorlabs
GmbH (2010).
[15] B. Liu, and M. E. Brezinski, Theoretical and practical considerations on detection

performance of time domain, Fourier domain, and swept source optical coherence
tomography, Journal of Biomedical Optics, 12(4), 044007 (2007).
[16] Ocean optics, http://www.oceanoptics.com/.
62
[17] M. Wojtkowski, V. Srinivasan, T. Ko et al., Ultrahigh-resolution, high-speed, Fourier

domain optical coherence tomography and methods for dispersion compensation,
Optics Express, 12(11), 2404-2422 (2004).
63
4
Design of Signal Processing
This chapter is partly reproduced from:
P. Liu, R. M. Groves, and R. Benedictus, Signal processing in optical coherence

033201 (2013).
4.1 Introduction
In chapter 3, the hybrid TD-FD-OCT system was described. The measurement data from this
instrument needs to be further processed to reconstruct a cross-sectional or volumetric image.
For example, during a TD-OCT measurement, the axial interferogram is modulated by
Doppler frequency shifting, induced by the movement of the reference arm. Typically this
demodulation is carried out by fixed electronics in an analog way. The general AM
(Amplitude Modulated) demodulation method, mixing, is the most used technique [1]. It
implements signal multiplication with a reference sinusoidal signal of the correct phase,
followed by lowpass filtering. However, sometimes this approach does not generate a perfect
signal envelope due to the variation of the carrier frequencies induced by nonlinear sweeping
of the reference mirror. Digital signal processing techniques for OCT signal demodulation are
rarely studied and the only detailed publication that can be found was published by D. C.
Adler in 2004 [2]. Methods such as boxcar averaging, coherence, and Hilbert transform were
investigated with simulated data in this paper. However, they did not consider the noise
coupled in the signal, which has a large influence on the accuracy of the demodulation.
65
Chapter 4
On the contrary, instead of containing a signal modulation, A-scan data from FD-OCT
measurement is in the spectral domain. To recover a depth profile, an additional Fourier
transform is needed. Considering that in FD-OCT, the data is normally collected using evenly
spaced intervals of wavelength , a resampling is necessary, based on Eq. 2.15, to shift the
spectral data into an evenly spaced data in the wavenumber k domain, where k 2 / .
Besides, as introduced in chapter 2, the lowpass filtering property of the Fourier transform
allows in a high level of classical noise and the use of CCD detectors also brings in extra
noise, e.g. dark noise. These additional noise source could deteriorate the quality of FD-OCT
images.
Therefore, this chapter aims at developing a series of signal processing methods to

recover and improve the quality of OCT images so that they clearly reflect the interior
structure of a specimen. Due to the different imaging principles, two signal processing
strategies were built for the data postprocessing after TD and FD-OCT measurements. Both
simulated and experimental data were used to test the developed algorithm. By far, the whole
OCT imaging system, including hardware setup, operation system and signal processing, has
been improved. Therefore, at the end of this chapter, an evaluation of the integrated OCT
system is demonstrated. Two of the most important factors, the axial measurement accuracy
and SNR are characterized for both TD and FD-OCT measurement.
4.2 Signal processing in TD-OCT

In TD-OCT a single axial profile of optical reflectivity within the sample is obtained by
rapidly translating the pathlength of the reference arm and synchronously recording the
magnitude of the resulting interference fringes, as described in chapter 2. From Eq. 2.11 and
Eq. 2.23, it can be seen that the amplitude of the interference fringe contains a Gaussian
envelope and a cosine carrier with a Doppler shifting frequency, decided by the scanning
velocity of reference mirror. However, in an actual OCT system there are two major obstacles
for the acquisition of the Gaussian envelope that contains the structural information of the
tested sample. Firstly translation stages carrying the reference mirror are commonly available
with several types of actuators [3], e.g. DC motor, stepper motor, and server motor that is
used in this thesis. All of these motors have some degree of nonlinearity in their motion due to
jitter and/or quantization of the scan motion (e.g., a stepper motor moves in discrete steps,
giving approximately constant motion.). Therefore due to the nonlinear sweeping of the
reference mirror, the carrier frequency, also known as the Doppler frequency, is not perfectly
constant but contains small variations during the depth scan. Since the general AM
66
demodulation requires a reference signal with the same frequency and phase [4], it is not
applicable for an OCT signal. The second obstacle is noise. Typically there are two kinds of
noise in an OCT system. One is the intrinsic noise from the OCT system [5], e.g. shot noise,
thermal noise and relative intensity noise, as discussed in chapter 2. The other is speckle
noise, which originates from distorted wavefronts of light returning from the sample that
result in randomly appearing constructive and destructive features in the interferogram [6].
These noise contributions are mixed and can acutely degrade the OCT image quality.
In order to extract the OCT envelope and improve the image quality, a succession of
signal processing methods have been developed, as shown in Fig. 4.1. Once an axial signal is
obtained, bandpass filtering is performed first to reduce excess noise and to remove the DC
component. Then three different solutions were evaluated and the optimum one for the
customized OCT application was identified. After accumulating a series of processed axial
signals, a cross-sectional image can be constructed. To further smooth the image, a 2D
median filter is applied before the final image displayed. The detailed signal processing
algorithms are introduced below.
Fig. 4.1: Flowchart for signal/image processing of OCT data from the customized TD-OCT
system.
4.2.1 Bandpass Filter
Although a balanced detector is used in the system to improve the signal to noise ratio,
residual noise that degrades the quality of OCT images still exists. Since the OCT signal is
modulated by a carrier with the Doppler frequency, a bandpass filter can be used as the first
67
Chapter 4
step to remove both the noise and the DC component of the signal. For this application a
fourth-order Butterworth filter was designed. Comparing with its counterparts, e.g.
Chebyshev filter and the Elliptic filter, the advantages of Butterworth filter are that it has the
easiest configuration and the maximally flat magnitude response [7], as shown in Fig. 4.2(b).
The fourth order is selected for a fast roll off around the cutoff frequency where the
attenuation of magnitude response is accelerated by the increased order. The higher order is
not considered since it would largely increase the computation and thus decrease the imaging
speed.
Fig. 4.2: (a) shows the frequency distribution of an A-scan signal from a reflective mirror.
(b) is the magnitude response function of a 4th order Butterworth bandpass filter. The
passband is set from 2000 Hz to 2500 Hz, indicated by the green box in (a)
The center frequency of the bandpass filter is the Doppler frequency f d . The FWHM of
the filter was chosen initially to match the bandwidth of the light source, , using the
following simple relationship [3]:
f d
FWHM (4.1)

In practice the filter bandwidth is chosen to be larger than the signal bandwidth [8]. A filter
bandwidth that is too narrow will widen the signal in the time domain and arbitrarily limit the
axial point spread function, however a bandwidth that is too large will let in more noise and
reduce the minimum detectable reflectivity from a specimen. Fig. 4.2 (a) shows the frequency
distribution of an A-scan signal from a reflective mirror. As indicated by the green box, the
interference signal is modulated in a frequency range from 2000 Hz to 2500 Hz, which is thus
68
set as the passband of the filter. The filtered signal and SNR of TD-OCT measurement are
shown in the following sections.
4.2.2 Demodulation
After bandpass filtering, the axial OCT signal can be simply expressed by [2]:
I [ z] a[ z ]cos([ z]z [ z ]) (4.2)
where a[ z ] is the message signal which contains the backscattering profile of a sample
with regard to depth z. cos([ z]z [ z ]) is the carrier that modulates the message signal to
high frequency. As discussed above, due to the nonlinear sweeping of the reference mirror,
the frequency [ z ] of the carrier has small variations, which makes the general AM
demodulation unsuitable. Therefore three demodulation methods which are largely insensitive
to variations in the carrier frequency have been investigated. Both simulated and experimental
data were used to evaluate the accuracy of the three methods.
The first technique, a median filter, is not a true demodulation method but is extremely
rapid due to its low computational complexity. In this method, the envelope can be extracted
by collecting the median of the absolute value of the modulated signal in the sliding window.
However, the accuracy of this method is quite sensitive to the choice of window length. A
small window length can preserve the detailed information in one axial signal but could also
allow in noise that deteriorates the signal quality, while the performance of larger window
length is the contrary. In the following discussion, the ideal window length is selected based
on least mean square error compared with the simulated result.
The second technique, Hilbert demodulation is based on the Hilbert Transform (HT).
The analytical signal I a [ z ] of a purely real signal I [ z ] can be expressed as:
I a [ z ] I [ z] jI h [ z] (4.3)
where I h [ z ] represents the Hilbert signal which has a -/2 phase shift compared with the
original signal:
I h [ z ] a[ z ]cos([ z ]z [ z ] / 2)
(4.4)
a[ z ]sin([ z ]z [ z ])
Thus
69
Chapter 4
I a [ z ] a[ z ]e j ([ z ] z [ z ]) (4.5)
where the axial profile a[ z ] can be recovered by taking the absolute value of the analytic
signal. The flowchart for the Hilbert demodulation is shown in Fig. 4.3.
Fig. 4.3: Flowchart for the Hilbert demodulation.
The third technique is an envelope detector, which in analog circuits typically consists of
a rectifier and a lowpass filter. In software, it is easy to realize by lowpass filtering the square
of the modulated signal. Here the square of the signal can be calculated as:
I 2 [ z ] a 2 [ z ]cos 2 ([ z ]z [ z ])
1 cos(2[ z ] z 2[ z ])
a2[ z] (4.6)
2
1 1
a 2 [ z ] a 2 [ z ]cos(2[ z ] z 2[ z ])
2 2
After squaring of the signal, it can be seen that half the energy of the modulated signal is
pushed up to higher frequencies and half is shifted down towards DC. The axial profile a[ z ]
can be recovered by a lowpass filter and the square root function. The flowchart for the
envelope detector is shown in Fig. 4.4.
70
Fig. 4.4: Flowchart for the envelope detector.
To evaluate the accuracy of the three demodulation techniques, a simulated OCT signal
was utilized for the testing. According to Eq. 4.2, the envelop signal a[ z ] was specified as a
Gaussian pulse with a peak located in the middle of the signal. Three separate carrier
frequencies ( [n] =/8, /6, /4,) were present consequently during the whole period to
simulate the nonlinear sweeping of reference mirror. The simulated signal is shown in Fig.
4.5. In order to be closer to an actual OCT signal, [n] , a narrowband noise with a power
specified in the range 0-25% of the modulated signal, was added. It was acquired by bandpass
filtering of the white noise, where the design of the filter follows the principles mentioned in
section 4.2.1 (a 4th order Butterworth filter with a bandwidth from /16 to 5/16).
71
Chapter 4
Fig. 4.5: Simulated OCT signal (without noise) and its envelope extracted by median filter
(a), Hilbert demodulation (b), and envelope detector (c).
To quantitatively compare the three methods, the mean-square error (MSE) between a[ z ]
and the demodulated message a[ z ] was calculated as:
N 1
1
MSE
N
(a[ z] a[ z])
z 0
2
(4.7)
Considering the time delay after lowpass filtering within the envelope detector, a 4th
order FIR (finite impulse response) filter with Hanning window was selected, providing a
constant group delay with a value of 2. To calculate the MSE, the data after the envelope
detector was shifted by 2.
Fig. 4.5 shows the extracted envelop of simulated OCT signal without noise and Fig. 4.6
shows how the MSE of each algorithm varies as a function of the noise percentage added in
72
the simulated data. It can be seen that although the optimized window length (25 points each)
is chosen, the method with a median filter gives the worst accuracy except in the high noise
range, more than 20%. The results from the two other methods are quite similar at the very
first phase when the noise is less than 2%. However, with an increase of the noise, the method
with the envelope detector gives a better performance than that with the Hilbert transform due
to the benefit of the lowpass filter which removes the high-frequency noise from the envelope.
Fig. 4.6: MSE vs. noise percentage for the various demodulation algorithms applied to
simulated OCT data. Abbreviations: MF, median filter. HT, Hilbert transform. ED,
envelope detector.
Experimental data was also used to evaluate the three methods using a reflecting mirror
as the sample. After bandpass filtering, the signals are shown in Fig. 4.7. The red lines in the
each image of the figure are the envelopes extracted by the three algorithms. Similar to the
results from the simulated data, the performance of the envelope detector and the Hilbert
transform are better than that of median filter, and the envelope extracted by envelope
detector is smoother than that extracted by the Hilbert transform. Therefore finally the
envelope detector was selected as the demodulation method to match with the customized
OCT setup.
73
Chapter 4
Fig. 4.7: Demodulation of experimental signal by (a) median filter, (b) Hilbert transform
and (c) envelope detector. The window size for median filter is optimized to 25. The filter
in the envelope detector is a 4th order Hanning window FIR filter.
4.2.3 Imaging smoothing
To suppress the speckle noise, a 2D median filter was used for smoothing the cross-sectional
image (a 3D median filter was used for smoothing the volumetric image correspondly). The
value of each pixel was replaced by a median value in a local neighborhood (a NN window
where N = 3, 5, 7, etc.). The median filter, unlike the mean filter, while smoothing out small
structural variations, does not blur the edges of regions larger than the window size used [1].
It has been applied to OCT images of different structures [9-11], showing an easy and
promising way to remove the speckle noise and improve the image quality.
As an example, an 8-layer woven glass fiber composite was investigated. Fig. 4.8(a)
shows the original cross-sectional image of the specimen. The fiber tows are observable as
74
bright ellipses. However speckles degrade the quality of the image. Fig. 4.8 (b), (c), and (d)
are the enhanced images by 2D median filtering with different widow sizes, 33, 55, and
1111 pixels respectively. It can be seen that the image contrast is substantially improved and
the structure of the sample is much more obvious. However it is also worth noting that as the
window size increases, the image resolution is decreased and some detail structural
information is lost. In order to reduce the speckles while at the same time preserving the
structural information, an intermediate window size of 55 pixels was selected for this
application. A detailed investigation of glass fiber composites with OCT is presented in
chapter 5.
Fig. 4.8: Cross-sectional images taken from a unidirectional tape made of GFRP. (a) Original
OCT image. Others are processed by 2D median filter with window size (b) 33, (c) 55, (d)
1111. The colorbars indicate the optical reflectivity in logarithm (dB).
4.3 Signal processing in FD-OCT

In FD-OCT system the interference spectrum acquired by the detector array is processed to
create a depth dependent profile. As shown in Fig. 4.9, the acquired spectrum is first
processed under reflection mode of a spectrometer, in order to remove the influence from
the optical source and dark noise from the CCD detectors. It is then resampled with regard to
evenly spaced intervals of wavenumber, instead of wavelength. With an inverse Fourier
transform, one depth profile is recovered and a cross-sectional image can be constructed by
75
Chapter 4
accumulating a succession of depth profiles. The quality of the cross-sectional image can be
further improved with a specially designed image smoothing technique. The detailed signal
processing algorithms are introduced below.
Fig. 4.9: Flowchart for signal/image processing of OCT data from the customized FD-OCT
system. I() and Re() are the original and denoised spectral data. R() is the reference
intensity and D() is the dark noise. and k are the wavelength and wavenumber
respectively.
4.3.1 Reflection-mode measurement
To remove the influence from the optical source and dark noise the data is measured under the
reflection mode of the spectrometer expressed as:
I ( ) D ( )
Re( ) (4.8)
R ( ) D ( )
where R( ) is the reference intensity which is acquired by placing a diffuse reflectance

standard (WS-1, Ocean Optics Inc, US) in front of the sample beam. D( ) is the dark noise
thermally generated within CCD detectors, which can be obtained by shutting down the
optical source. I ( ) and Re( ) are the original and denoised spectral data, with examples
shown in Fig. 4.10(a) and (b) respectively.
76
Fig. 4.10: An example of signal processing in FD-OCT by measuring a reflective mirror. (a)
One dataset from the spectrometer was processed to obtain (b) reflection mode signal
based on Eq. 4.8. It was then linearly resampled in (c) with regard to wavenumber. Finally
by inverse Fourier transform, (d) one depth profile was created.
4.3.2 Resampling
From Eq. 2.15, it can be seen that the depth profile can be recovered from an inverse Fourier
transform of the spectral data. But one requirement is that the spectral data should be evenly
spaced in frequency, specifically referring to k-space domain. Since the data is collected using
evenly spaced intervals of wavelength rather than wavenumber k , which is inversely
proportional to wavelength ( k 2 / ), a resampling process is necessary before Fourier
transformation. A comparison of different resampling techniques for OCT can be found in
[12]. To balance processing time and image accuracy, a linear resampling was selected for the
application. Fig. 4.10(c) shows the results after resampling the denoised spectral data, shown
in Fig. 4.10(b).
77
Chapter 4
4.3.3 Inverse Fourier transform
By inverse Fourier transform of the resampled spectral data, one depth profile was finally
created, as shown in Fig. 4.10(d). However it should be noted that due to limited spectral data
(N=512), a direct Fourier transform may result in lobe leakage in an OCT depth signal, as
shown by the blue line in Fig. 4.11. Expanded main lobes and side lobes can be observed
clearly, which could bring artifacts into an OCT image and prevent weaker reflectors nearby
from being detected. To reduce the artifacts, a window function during Fourier transform
could be applied. Here a Hanning window was selected for Fourier transform due to its ability
to suppress the side lobes and simultaneously to provide a good frequency resolution that
decides a narrow width of main lobe [13]. The improved results are indicated by the green
lines in Fig. 4.11. It can be seen that with the window function, the main lobe is quickly
delayed with regard to the axial distance and the side lobes are greatly reduced, which is
required for a high-quality OCT image. The improved SNR of the FD-OCT measurement is
calibrated in section 4.5.2.
Fig. 4.11: A comparison of recovered depth profile by inverse Fourier transform with no
additional processing and Hanning window function respectively.
4.3.4 Axial scan smoothing
In FD-OCT, the reflectors from different depths in a specimen are captured by the
spectrometer as signal components with different frequencies. The deeper reflectors
contribute to higher frequency components due to larger optical pathlength difference
between reference and sample arms, thus an inverse Fourier transform recovers deeper
reflectors and localizes them at longer axial distances. However there are some artifacts in the
78
signal which almost have constant frequency components and show up at the constant
locations in a depth profile. An example is shown in Fig. 4.12. Three depth profiles are
acquired from an epoxy coating at same spot but with an axial shift of the specimen each time.
The two highest peaks in each color indicate the light reflections from the front and back
surface of the specimen. It can be seen that the three measurements have axial shifts, which
are 300, 338m, 413m respectively. However, the obvious artifacts, marked by 1-4 in Fig.
4.12, keep their same locations. The same phenomenon can be also found at other positions in
depth profiles with careful observation.
Fig. 4.12: Depth profiles of an epoxy coating with different axial shift of the specimen. The
two highest peaks in each color indicate the light reflections from the front and back
surface of the specimen. The markers 1-4 are examples of noise components that maintain
the same locations.
Therefore, this reminds us that images from FD-OCT could be improved by employing
a shifting average. As shown in Fig. 4.13(a), (b), and (c) are cross-sectional images of an
epoxy coating recorded with different axial shifts of the specimen. In each image, the first two
bright red lines indicate light backreflections from the front and back surfaces respectively.
The third light line appears due to the multiple reflections within the sample. Apart from these
components, the other features in these images are artifacts as described in Fig. 4.12. To
improve the image contrast, the three images in Fig. 4.13(a), (b), and (c) were first aligned so
the signals are overlapped but the artifacts are staggered. Then by taking an average of the
three aligned images, the artifacts are significantly reduced, as shown in Fig. 4.13(d). The
image quality could be further improved by averaging more datasets, but the time cost for
additional images should be taken into account.
79
Chapter 4
Fig. 4.13: Noise reduction by shifting average. (a)-(c) are cross-sectional images of an
epoxy coating with different axial shifts of the specimen. These images are shifted back to
align the signals for averaging. The averaged result is shown in (d). The colorbars indicate
the optical reflectivity in logarithm (dB) scale.
4.4 Image Display

As shown in Fig. 4.8 and Fig. 4.13, the jet colormap [14], a commonly used colormap in
Matlab software (Mathworks Inc., USA), can be used to display the cross-sectional images
and this will be used in the rest of the thesis. The increase of the optical reflectivity from the
specimen is indicated by the colors from dark (blue) to bright (red). An RGB color image has
a 24-bit color level provided by the computer monitor, which is three times higher than an
8-bit gray-scale image. Therefore the use of RGB color images can improve in the
differentiation of different structures. Generally an OCT signal has a large variation which
could be five orders of magnitude [3], so the logarithm of the signal is normally used to
expand the dynamic range of the display. The only drawback in doing this is that the relative
variations in signal would be compressed, making it difficult to observe small relative
changes in backreflections. Further examples of image display options are introduced in the
following chapters.
80
4.5 System Evaluation

4.5.1 A-scan Accuracy and Linearity
To verify the axial linearity of the TD-FD-OCT configuration, a displacement test was
implemented. A reflective mirror, in front of the scanning head, was driven by an actuator
(NA23C60-T4, Zaber Inc, Canada) to change the optical pathlength of the sample arm. At the
same time, both TD and FD-OCT A-scans were performed continuously. The displacements
were calculated by locating interference peaks in the successive A-scans, as shown in Fig.
4.14. It can be seen that with the axial movement of the sample mirror, the displacement
calculated by FD-OCT is much more consistent than that with TD-OCT. The measurement
accuracy was also quantitatively evaluated by computing the root mean square error (RMSE)
by:
1 n
RMSE (di di )
n i 1
(4.9)
where d i is the axial displacement of the sample mirror precisely controlled by the actuator.
di is the calculated displacement by the TD or FD-OCT system.
During this test, the RMSE for TD and FD-OCT are 12.4 m and 10.4 m, respectively.
For FD-OCT the error mostly comes from the misalignment of the movement of the sample
mirror and the incident light beam. For TD-OCT there is an extra error from the inconsistent
sweeping of the reference mirror in the ODL, which explains the nonlinearity between the
practical and measured displacement with TD-OCT, as shown in Fig. 4.14. Further example
results of inconsistent A-scan in TD-OCT are shown in the Chapter 5.
81
Chapter 4
Fig. 4.14: Displacements calculated by TD and FD-OCT during the axial movement of a
sample mirror.
4.5.2 Signal to Noise Ratio
SNR is another equivalently important parameter in the evaluation of an OCT system. The
practical SNR of the integrated OCT system, SNRp , was calibrated by A-scans of a
reflective mirror using TD and FD-OCT respectively, and can be expressed as:
As
SNRp 10 log( ) (4.10)
An
where As and An are the amplitude of the signal peak and root-mean-square of the noise.
The noise bins were collected from a signal bandwidth away from the signal peak in the
A-scan data.
Fig. 4.15 (a) and (b) are A-scan data from a TD-OCT measurement with an output
source power of 10 mW and 20 mW respectively. The calculated SNR increases from 39 dB
and 50 dB, which verifies that the higher source power can give a better SNR and thus enables
it to detect weak reflections from deeper into the specimen. Fig. 4.15 (c) shows one A-scan
data from FD-OCT measurement with 1 ms integration time. The practical SNR is 23.8 dB,
and it could be improved to 27.5 dB by increasing the integration time to 10 ms, as shown in
Fig. 4.15 (d). The SNR of the FD-OCT should be able to keep increasing with longer
integration time, however the increased signal power is not accessible due to the limitation of
the photoelectron capacity of the CCD (saturation value).
82
Fig. 4.15: SNR evaluation for the designed OCT system. (a) and (b) are A-scans by
TD-OCT measurement of a reflective mirror, with a source power of 10 mW and 20 mW
respectively. (c) and (d) are A-scans by FD-OCT measurement of a reflective mirror, with
integration times of 1 ms and 10 ms respectively.
Besides, it is worth noting that there is a decay of signal peak with increasing axial
distance during FD-OCT measurement, as shown in Fig. 4.16. The test was implemented by
shifting the ODL to increase the sample pathlength and acquiring each A-scan simultaneously
which are displayed together in Fig. 4.16. The depth-dependent signal drop is caused by
optical resolution limits, nite pixel width aliasing at high spatial frequencies, and interpixel
cross-talk [15]. Therefore, a compensation for this decay may be needed for measuring highly
scattering and thick specimens.
83
Chapter 4
Amplitude 0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2
Axial distance/mm
Fig. 4.16: A decay of signal peak with increasing axial distance during FD-OCT
measurement of a reflective mirror. The change of axial distance is created by the ODL in
the reference arm.
4.6 Discussion
The extraction of each axial envelope signal is the first important step that decides the quality
of a TD-OCT image. In this chapter three demodulation techniques, median filtering, Hilbert
transform and envelope detector were investigated. A Median filter has minimum
computational complexity but the least accuracy since it gives underestimation on each
envelope amplitude, as shown in Fig. 4.5 and Fig. 4.7. To eliminate this error during
simulation, the underestimated amplitude bias can be calculated and compensated. However,
it would be difficult to realize in the experimental data due to the changing signal amplitude
and noise under different measurement conditions, and the variations in the signal frequency.
To stay the same, bias compensation was not considered during both simulation and
experiment. The Hilbert transform can give better performance but is quite sensitive to the
noise. The MSE value increases dramatically with the noise growth. The ideal one is the
envelope detector, which is realized by lowpass filtering of the autocorrelation of the
modulated signal. At first (less than 2% noise) the accuracy of envelope detector is even a
little bit lower than that of Hilbert transform due to the influence of the filter (e.g. passband
ripple and stopband attenuation), as shown in Fig. 4.5(c). However it shows an advantage as
the noise level increases because the lowpass filter in the envelope detector can remove the
high-frequency noise and guarantee the demodulation accuracy.
84
Another equivalently important signal processing aspect is noise reduction. Image

averaging is a widely used technique to suppress random additive noise. However it is not
suitable for TD-OCT image processing because of its low efficiency caused by the repeated
scanning. In the customized design, a bandpass filter and 2D median filter were used before
and after the demodulation respectively. The bandpass filter acted as a signal preprocessing to
remove the noise and DC component that can largely affect the result of demodulation. The
2D median filter was applied as the final step for further image enhancement. However from
Fig. 4.8 the tradeoff between the resolution and speckle reduction using 2D median filter is
observable with increased window size. It is noted that to date only limited research has been
done in the area of image enhancement and speckle reduction in OCT imaging [9, 16-18].
Most techniques that have been developed trade resolution for reduction of speckle contrast.
Further work in this area is needed.
In FD-OCT, although the measurement is under reflection mode to get rid of the
influence of the light source and dark noise, the SNR is much lower than TD-OCT, as
evaluated in section 4.5.2. A major factor, besides shorter acquisition time, is the Fourier
transform function used in FD-OCT, which allows in the low-frequency noise and aliased
signal of high spatial frequencies due to limited detector pixels. The shifting average
technique introduced in section 4.3.4 could be an effect way to increase the SNR and the
image contrast can be further improved with increased averaging times. However, this
processing requires additional operations to apply different pathlength offsets and to acquire
multiple datasets, which slows down the measurement efficiency. This disadvantage in SNR,
combined with the limited depth scan range, narrows down the application range of FD-OCT.
In this thesis, examples of FD-OCT applications include characterization of polymer coatings
as described in chapter 7.
4.7 Conclusions
This study explored simple but highly effective methods for signal processing in the
customized TD-FD-OCT system. In TD-OCT, after comparison, the envelope detector was
selected as the ideal method for the demodulation of each axial signal. A bandpass filter and
2D median filter were applied before and after the demodulation, respectively, to reduce OCT
system noise and speckle noise. The calculated SNR of TD-OCT reaches 50 dB, which allows
the detection of weak reflectors from deep in the specimen. However one drawback is a small
nonlinearity during the axial scan.
85
Chapter 4
In FD-OCT, the raw data is first processed under reflection mode of a spectrometer, in
order to remove the influence from the optical source and dark noise form the CCD detectors.
It is then linearly resampled with regard to evenly spaced intervals of wavenumber, instead of
wavelength. With an inverse Fourier transform, one depth profile is recovered and a
cross-sectional image can be constructed by accumulating a succession of depth profiles. The
quality of the cross-sectional image can be further improved with shifting average, as
described in section 4.3.4. Although the SNR of FD-OCT is not comparable with that of
TD-OCT, the imaging speed and high axial linearity provides FD-OCT with applications in
the thickness measurement and microcrack detection of thin films, as discussed in chapter 7.
References
[1] M. E. Brezinski, [Optical coherence tomography: principles and applications],

Academic Press (2006).
[2] D. C. Adler, [Digital signal processing techniques for optical coherence tomography:
OCT and OCT image enhancement], Massachusetts Institute of Technology (2004).

[4] U. Meyer-Baese, and U. Meyer-Baese, [Digital signal processing with field

programmable gate arrays], Springer (2007).
[5] P. H. Tomlins, and R. K. Wang, Theory, developments and applications of optical

coherence tomography, Journal of Physics D: Applied Physics, 38(15), 2519-2535
(2005).
[6] J. M. Schmitt, S. Xiang, and K. M. Yung, Speckle in optical coherence tomography,

Journal of Biomedical Optics, 4(1), 95-105 (1999).
[7] G. Bianchi, and R. Sorrentino, [Electronic filter simulation & design], McGraw-Hill
(2007).
86
[8] M. R. Hee, [Optical coherence tomography of the eye], Massachusetts Institute of

Technology (1997).
[9] J. Rogowska, and M. E. Brezinski, Image processing techniques for noise removal,
enhancement and segmentation of cartilage OCT images, Physics in Medicine and
Biology, 47(4), 641 (2002).
[10] G. J. Tearney, H. Yabushita, S. L. Houser et al., Quantification of macrophage

content in atherosclerotic plaques by optical coherence tomography, Circulation,
107(1), 113-119 (2003).
[11] V. J. Srinivasan, D. C. Adler, Y. Chen et al., Ultrahigh-speed optical coherence

tomography for three-dimensional and en face imaging of the retina and optic nerve
head, Investigative Ophthalmology & Visual Science, 49(11), 5103-5110 (2008).
[12] C. N. Copeland, and A. K. Ellerbee, "Analysis of the effects of different resampling

techniques for optical coherence tomography," Proceeding of SPIE 8227, 82270W-1
(2012).
[13] A. V. Oppenheim, R. W. Schafer, and J. R. Buck, [Discrete-time signal processing],

Prentice-hall Englewood Cliffs (1989).
[14] Colormap, http://www.mathworks.nl/help/matlab/ref/colormap.html.
[15] B. Potsaid, I. Gorczynska, V. J. Srinivasan et al., Ultrahigh speed spectral/Fourier

domain OCT ophthalmic imaging at 70,000 to 312,500 axial scans per second, Optics
Express, 16(19), 15149-15169 (2008).
[16] M. Bashkansky, and J. Reintjes, Statistics and reduction of speckle in optical

coherence tomography, Optics Letters, 25(8), 545-547 (2000).
[17] M. Pircher, E. Go, R. Leitgeb et al., Speckle reduction in optical coherence

tomography by frequency compounding, Journal of Biomedical Optics, 8(3), 565-569
(2003).
87
Chapter 4
[18] D. C. Adler, T. H. Ko, and J. G. Fujimoto, Speckle reduction in optical coherence

tomography images by use of a spatially adaptive wavelet filter, Optics Letters,
29(24), 2878-2880 (2004).
88
5
OCT for Monitoring the
Delamination Growth in Glass
Fiber Composite
This chapter is reproduced from:
P. Liu, R. M. Groves, and R. Benedictus, 3D monitoring of delamination growth in a

wind turbine blade composite using optical coherence tomography, NDT & E
International, 64, 52-58 (2014)
5.1 Introduction
Introducing a reinforcing fiber (e.g. glass fiber and carbon fiber) within a polymer matrix
leads to composite materials with improved mechanical properties, e.g. high strength and
stiffness to weight ratio [1]. These advantages of composites attract widespread applications,
especially in aerospace and wind energy industries as introduced in chapter 1. However, this
growing market also brings the challenge of quality control and health monitoring of
composites due to the new forms of defects. For example, in a wind turbine system, the wind
turbine blades are one of the key components as a failure of the blade eventually leads to
overall breakdown of the system [2]. Currently glass fiber composites are extensively used for
the production of wind turbine blades due to their high stiffness and relatively low cost [3-5].
The blades are manufactured either by resin injection techniques or by prepreg technology.
89
Chapter 5
Both manufacturing processes may result in a variety of defects, such as porosity, cracks,
debonding, fiber breakage and delamination. Amongst these defects delamination is probably
the most frequently occurring damage and thus is of major importance [6]. It appears as a
debonding of adjoining plies in laminated composites. During the operation of the wind
turbine blade, a delamination crack can propagate under the influence of extra loading by
wind and gravity, which may lead to final failure of the structure.
A number of non-destructive testing (NDT) techniques are employed for the inspection
of delamination in wind turbine blades. Ultrasonic scanning [7] is primarily applied to study
the entire blade surface and subsurface level. However a major drawback is that it is difficult
to provide depth feature and high spatial resolution. Moreover, ultrasonic measurements are
usually complicated by the requirement of a coupling agent between the transducer and
composites. X-ray Computed Tomography (x-ray CT) [8] provides information on the
internal structure of the composite material and hence makes it possible to locate
inhomogeneities within the depth of the material. The high measurement cost and the
difficulty for inline inspection are the main obstacles for it becoming a standard NDT tool in
the field of wind turbine blades. Additionally, thermography [9] and shearography [10] have
shown potential to be fast tools for delamination inspection. However both of them have
limits in sensitivity to delamination defects in thick parts.
Optical Coherence Tomography (OCT) [11] could be an alternative technique due to its
microscale resolution and reasonable penetration depth. Whereas a carbon fiber structure
renders the composite material opaque, glass-fiber reinforced polymer poses no problem for
obtaining 2-3 mm penetration depth for OCT imaging. The aim of this chapter is to use OCT
imaging as a novel NDT tool to investigate delamination growth inside a wind turbine blade
composite. The tested samples are glass fiber composites used for the spar webs of a wind
turbine blade. A tensile test stage coupled with the customized fiber-optical OCT system was
built. Then specimens with an interlaminar crack were statically loaded and stepwise scanned
by the OCT system. With advanced signal processing, the volumetric images and the 3D
crack profiles at successive stages during the propagation of delamination were determined.
5.2 Sample Construction

The material tested was a non-crimp glass-fiber epoxy mixture which is generally used for the
spar webs in wind turbine blades. The samples were prepared by V. Sridhar, TU Delft [12]. A
total of 16 layers of glass fiber were used. The fiber orientation was +45o and -45o, relative to
the length direction of the sample, symmetrically distributed between the layers, as shown in
90
OCT for Monitoring the Delamination Growth in Glass Fiber Composite
Fig. 5.1. To initiate the delamination a non-adhesive Kapton film of about 20 m thickness
and 40 mm length was placed in the middle layer during the lay-up process. To prepare the
resin Airstone 780E epoxy and 785H hardener were used. After the vacuum infusion process,
the resin completely wetted the reinforcement material. It was then placed inside an oven to
cure at 75 oC for about 7 to 8 hours. Once it was cured the glass fiber composite was cut into
double cantilever beam (DCB) specimens for the tests. The dimensions for each specimen are
approximately 220 mm length, 25 mm width, and 4.4 mm thick. Finally each specimen was
bonded to a pair of aluminum loading blocks for the tensile test.
Fig. 5.1: Lay-up configuration of the composite specimen. +45 and 45 indicate the lay-up
directions in degrees, relative to the length direction of the sample. Kapton was inserted
between the 8th and 9th layers to initiate the delamination.
5.3 Experimental Setup

To carry out the tensile test, a mechanical stage was designed to statically load the sample to
extend delamination length. As shown in Fig. 5.2, the rectangular sample was clamped
horizontally by a pair of splitable hinges. Each hinge has a removable pin that acts as a joint
for its plate. One hinge was attached to a loading cell for tensile measurement and the other
was driven by a linear actuator (NA23C60-T4, Zaber Inc, Canada) which can provide a
maximum of 950 N thrust and 60 mm travel range. The OCT probe was located in front of the
specimen to examine the internal structure. The OCT probe was mounted on a 2D translation
stage unit (T-LS28M, Zaber Inc, Canada) by the stainless rods, which enable the distance
between the OCT probe and the specimen to be adjusted to focus the light on the specimen for
optimized performance. A CCD camera (piA2400-17gm, BASLER AG, Germany) was
connected at the back of the specimen to monitor the crack growth from the surface view,
since the glass fiber composite is semi-transparent.
91
Chapter 5
Fig. 5.2: A photograph of the loading setup. The OCT probe and CCD camera monitor the
crack growth from the two opposite surfaces of the specimen. TS denotes translation
stages.
5.4 Results
5.4.1 Static Test
OCT is a high-resolution imaging technique for examining the interior of scattering media.
Therefore of first interest is to evaluate the capability of the customized OCT system for
microstructure and defect characterization. After the specimen was clamped on the
mechanical stage, a static tensile loading was applied to grow the delamination to
approximately 10 mm length as measured from the side view. Then an OCT scan along the
delamination crack was performed in the middle of the specimen surface. Using the signal
processing algorithm developed in chapter 4, a cross-sectional OCT image was determined, as
shown in Fig. 5.3(a).
92
Fig. 5.3: OCT Images from a static test on a delaminated glass fiber composite. (a) A
cross-sectional image along the propagation of the delamination. The first bright line
indicates specimen surface location, while the second and third bright lines indicate
reflections from upper and lower surfaces of the crack. The white square indicates the
frontier of the delamination crack. (b) One depth scan signal across the middle of white
square in (a).
The width of the image is 15 mm and is composed of 300 A-scans. The first bright line
(labeled 1) in Fig. 5.3(a) indicates the location of the front surface of the specimen which was
tilted by a small angle to resist strong specular reflections. Below the surface ellipses with
lighter contrast are resolved, revealing the distribution of the fiber tows from each layer. The
fiber tows are not as pronounced as the surface line. This is largely due to low power
reflectivity (R%=0.002%) from the fiber and resin interfaces which can be simply calculated
by:
R% [(n1 n2 ) / (n1 n2 )]2 (5.1)
where n1 and n2 are the refractive indexes of glass fiber and resin which are very close
(typically 1.56 for glass fiber and 1.545 for epoxy [13]).
Fig. 5.3(a) also shows that the optical penetration depth in this composite sample can
reach 2.85 mm where the fiber tows are still detectable. This depth is deeper than the crack
location and hence the OCT system is suitable for the study of this delamination crack. As
demonstrated in Fig. 5.3(a), at the lower layer of the structure, around 2.2~2.7 mm, an evident
feature is posed. Two bright lines (labeled 2 and 3) run along the image plane, indicating the
upper and lower surfaces of the delamination crack which grows from right to left. It can be
seen that the delamination does not propagate in a straight line but in a curve along the
93
Chapter 5
tow/resin boundary. Moreover, an interesting phenomenon can be found at the frontier of the
delamination indicated by the white square. Along the delamination growth direction, the two
bright lines migrate into one with a thickness of 0.3 mm. Fig. 5.3(b) displays a corresponding
A-scan signal across the middle of the square in Fig. 5.3(a). It can be seen that multiple
reflections occur at the tip of the delamination. A reasonable explanation could be the
multiple reflections from the interface between the material and a very small air gap. It also
could be the reflections from the bridging fibers in-between the crack. These results will be
analyzed in more detail in the following discussion section. Anyway the thick bright line can
be used for the frontier detection during the crack propagation, which paved the way for the
study of crack growth by the incremental loading test described in next section.
To present the whole delamination in detail, a volumetric image with two side views has
been reconstructed as shown in Fig. 5.4(a) and (b). Fig. 5.4(a) shows the start of the
delamination in the right surface and the crack propagation in the front surface. On the
contrary, Fig. 5.4(b) shows the tip of delamination and the crack propagation from the other
side. The arrows in both images indicate the direction of crack growth. Both images have a
dimension of 12 mm in length and 6 mm in width, along and vertical to the propagation of the
delamination crack respectively. The thickness of the images is approximately 2.5 mm after
filtering the signal above the specimen surface, which can be located by tracking the first
reflection peak from each A-scan. Similar processing was implemented to remove the signal
from the air gap for a better demonstration of the crack. The strong reflectivity from the
surface and the crack area are observable as the bright area, but the fiber tows are less visible
since the step of lateral scan was increased to 150 m for imaging efficiency. Nevertheless,
similar to the demonstration in Fig. 5.3(a), it can be seen the propagation of the crack is not in
a plane from both volumetric images. The bright area is detectable in the right surface of Fig.
5.4(b), indicating multiple reflections at the frontier of delamination.
94
Fig. 5.4: (a) and (b) are the volumetric images of the delamination crack from two side
views. The arrows in both images indicate the crack growth direction.
5.4.2 Incremental Loading Test
The static test described in section 5.4.1 proves that OCT has the ability to characterize a
delamination crack. To study the behavior of delamination growth, crack profiles need to be
monitored during propagation steps. Thus an incremental loading test was carried out on the
same specimen but focusing on OCT measurement of the frontier of the delamination crack.
5.4.2.1 Test 1
In Fig. 5.5, (a), (b) and (c) are the volumetric images recorded during the three stages of the
delamination growth. As in the cross-sectional image shown in Fig. 5.3(a), the first and
second lines indicate surface and the delamination crack respectively. The delamination
propagates along the y axis from right to left. The crack lengths shown in the three tomograms
are approximately 6 mm, 9mm and 11 mm respectively. However, from these tomograms it is
still difficult to understand how exactly the delamination grows within the composite due to
the axial resolution and the artifact of multiple reflections described in the previous chapter.
95
Chapter 5
Fig. 5.5: (a), (b) and (c) are the volumetric images in three stages of crack growth. The
colorbars indicate the optical reflectivity. (d), (e) and (f) are the crack surface profiles
correspondingly. The colorbars indicate the surface height. The 45o color lines indicate the
fiber orientation.
Therefore the three tomograms were further processed to reconstruct the crack surface
profiles, as shown in Fig. 5.5(d), (e), and (f) respectively. The procedure for acquiring the
crack profiles is described by the flowchart in Fig. 5.6. Firstly, each volumetric image was
divided into cross-sectional images along the delamination growth direction. In the upper
region of each cross-sectional image, the sample surface line can be extracted by peak
detection since the first and strongest reflection comes from the surface in each A-scan signal,
96
as shown in Fig. 5.3(b). While in the lower region, although multiple reflections exist, the first
and highest signal peak can be identified as the original reflection from the material and crack
interface. Therefore the crack surface can be located in the same way by peak detection in the
lower region of the cross-sectional image. Additionally, a threshold was set to find the crack
tip location and thus to determine the crack length. Since there is a large reflectivity
difference from crack and non-crack areas, as indicated by Fig. 5.3(b), the threshold was
simply set as the median reflectivity value from these two areas. By subtracting the axial
locations between the crack and sample surface, one crack thickness line was obtained. It was
further smoothed by a 1D median filter with a block size of 3 pixels. Finally a 3D crack
surface profile was reconstructed by accumulating a series of crack thickness lines from each
cross-sectional image.
Fig. 5.6: Flowchart of the procedure for extracting the crack profiles.
From the crack surfaces shown in Fig. 5.5, delamination growth within the specimen
can be easily observed. The crack lengths in the three stages are approximately 6 mm, 9 mm
and 11 mm, which match with the results from corresponding tomograms as discussed above.
However crack growth at different lateral locations does not proceed with the same velocity.
This can be revealed by the delamination tip which is not in a vertical plane perpendicular to
the crack growth orientation, as shown in Fig. 5.5(d), (e), and (f). Besides, a more interesting
97
Chapter 5
observation is the 45o lines in the crack profile images. These 45o lines indicate the fiber
orientation in the middle layers. It can be seen that they do not run on the same height but
have a variation around 0.35 mm, indicated by the colorbar. The delamination crack grew first
2 mm at a lower layer of 2.25 mm in depth and then went to a high laminate layer with a
maximum depth of 1.9 mm. However, it is difficult to estimate if the crack propagated in both
the two middle layers since they have the same fiber orientation. A more accurate judgment
should combine the results from more pronounced volumetric tomograms that clearly show
the fiber layers. However, the cost is the dramatically increased imaging time due to the
denser scanning of the sample in the same area.
5.4.2.2 Test 2
A more complete test was performed on a second specimen with the same size and material
specification. Four stages of the crack growth were monitored. The force and displacement of
the linear actuator are recorded in Table 5.1 and Fig. 5.7. Correspondingly four photographs
were captured by the CCD camera at the back surface of the specimen, as shown in Fig. 5.8.
The crack profiles at the four stages were achieved by OCT measurement in front of the
specimen, as shown in Fig. 5.9. The measurement regions are of the area indicated by the
white frames in Fig. 5.8. Each image size is 12 mm along the crack growth orientation and 9
mm in the perpendicular direction.
Table 5.1: Force and displacement record of the actuator during the tensile test
Stage 1 2 3 4
Force /N 36.3 51.0 54.9 51.9
Displacement/mm 0 10.5 14.9 17.4
55
50
Force/N
45
40
35
0 5 10 15 20
Displacement/mm
Fig. 5.7: Force and displacement relationship during the tensile test.
98
Due to the semi-transparency of the specimen, the crack can be detected by the area
CCD camera, as shown in the lighter areas in Fig. 5.8. The 45o lines indicate the fiber
orientation while the vertical and horizontal lines are the stitches that hold each fiber tow.
Besides, the crack growth (from left to right) can be seen via Fig. 5.8(a) to (d) with a roughly
3 mm increase each time. For each stage with nearly same increased crack length, it is
noticeable from Table 5.1 and Fig. 5.7 that the relative displacement of the actuator is
decreased while the force reaches a maximum at stage 3. The maximum force indicates the
critical loading to the fracture toughness [14] of the specimen.
Fig. 5.8: Photographs captured by CCD Camera for four stages of crack growth indicated
by the lighter areas. The white squares indicate the OCT measurement area.
Fig. 5.8 could be a good reference to monitor the crack growth from the surface view.
However, to precisely track the delamination crack and reconstruct the crack profile, an OCT
measurement is necessary. From Fig. 5.9 it can be seen that the crack lengths at the four
stages are 2 mm, 5 mm, 8 mm, and 11 mm respectively, which is in good agreement with the
results from the CCD camera shown in Fig. 5.8. Fig. 5.9 also clearly shows the tip of the
crack is not in a vertical plane, which indicates the speed of the propagation at different
locations are not the same. Moreover, the behavior of crack growth under the tensile loading
99
Chapter 5
can be determined. Firstly within a 0 to 3 mm crack length, the delamination crack propagated
in a -45o fiber layer. This can be revealed by the -45o lines with blue color. Afterwards the
crack at the lower part of the specimen went to the +45o layer underneath the -45o one. This
can be revealed by the +45o lines with red color. The deduction also can be proved from the
depth range of the crack which is more than 0.35 mm, as indicated by the scale bar in Fig. 5.9.
Considering the layer thickness of the specimen is 0.275 mm, it is probable that the crack
propagated from one layer to a neighboring one.
Fig. 5.9: 3D crack profiles at the frontier of delamination measured in four stages of crack
growth. The 45o color lines indicate the fiber orientation. The colorbars indicate the
surface height.
5.5 Discussion
The results show how an integrated OCT system and image processing technique could be
used for structural imaging and 3D monitoring of delamination propagation within a glass
fiber composite. The OCT system can perform 3D volumetric images with a resolution of 11
m (axial) by 20 m (lateral), which is much higher than an ultrasonic scan. Due to the
relatively simple setup and the fiber optic used, a portable OCT prototype may be developed
in future. The optical beam travels in the fiber and emits infrared light onto the sample finally
100
with a power is around 1~2 mW. Therefore unlike X-ray techniques, an in-situ OCT
measurement would be safe for the instrument user.
Image processing techniques were developed to show 3D crack surfaces. With

incremental loading, the crack trended to migrate from the delamination plane, which is
usually referred to as a crack jump [15]. Similar results [12] came out from the parallel study
of structural nonlinearity at TU Delft on the same specimen type. Both show that the crack
can easily propagate into another layer under monotonic loading and in a fatigue test. The
crack jump is one of the most common failure modes recorded when testing multidirectional
DCB specimens. Recent studies [16, 17] suggest that a crack jump is mainly decided by the
bending stiffness of the crack beam and it can be prevented by optimizing the stacking
sequence of the fiber layers. Another observation is the slight difference from the same
imaging area in a series of crack profiles. This could be due to the influence from fiber
bridging behind the crack tip, as shown in Fig. 5.10. Fiber bridging occurs when fibers are
pulled from one side of the delamination plane to the other [18]. With incremental loading,
the bridging fibers could shift the fracture surface and finally break. Besides, strong
reflections can be detected from intensive bridging fibers closing at the crack surface. This
could bring an error into the crack surface search algorithm. Therefore, during the image
processing, a median filter was used as the last procedure to suppress the outlying points and
minimize the error in the detected crack surface position.
Fig. 5.10: Fiber bridging at the tip of the crack in the tested glass fiber composite.
The type of information found in this study is valuable for the study of delamination
growth in glass fiber composite. So far the measurement system is based on TD-OCT that
101
Chapter 5
relies on the movement of reference mirror. FD-OCT has the advantage of high speed
imaging but limited depth scan range. Further improvement of the customized OCT system
could either modify the current ODL to an advanced one (e.g. grating based ODL as
introduced in chapter 2) for fast TD-OCT imaging, or increase the CCD pixel elements for
deep FD-OCT imaging. Also the glass fiber composite used for the wind turbine blades are in
practice thicker than the specimen tested in this chapter. Therefore an increase in penetration
depth is necessary for detecting the delamination that occurs deep in the composite. Optical
sources at longer wavelengths, which are less scattering in the composite structure, could be
selected to optimize the OCT system for enhanced imaging depth [19] .
5.6 Conclusions
This chapter demonstrates the potential advantages of using optical coherence tomography for
the study of delamination and its propagation in a glass fiber composite used for a wind
turbine blade. OCT can perform high-resolution cross-sectional and volumetric images to
characterize the microstructure and the crack within a specimen. With the advanced signal
processing techniques, the OCT images can be further processed to reconstruct 3D crack
surface profiles, which provide substantial information for the study of crack growth in the
composites.
References
[1] R. E. Shalin, [Polymer matrix composites], Springer (1995).
[2] I. Amenabar, A. Mendikute, A. Lpez-Arraiza et al., Comparison and analysis of

non-destructive testing techniques suitable for delamination inspection in wind turbine
blades, Composites Part B: Engineering, 42(5), 1298-1305 (2011).
[3] P. Brndsted, H. Lilholt, and A. Lystrup, Composite materials for wind power
turbine blades, Annual Review of Materials Research, 35, 505-538 (2005).
[4] D. A. Griffin, and T. D. Ashwill, Alternative composite materials for megawatt-scale

wind turbine blades: design considerations and recommended testing, Journal of
Solar Energy Engineering, 125(4), 515-521 (2003).
102
[5] S. Habali, and I. Saleh, Local design, testing and manufacturing of small mixed
airfoil wind turbine blades of glass fiber reinforced plastics: part I: design of the blade
and root, Energy Conversion and Management, 41(3), 249-280 (2000).
[6] Y. Zou, L. Tong, and G. P. Steven, Vibration-based model-dependent damage

(delamination) identification and health monitoring for composite structures a
review, Journal of Sound and Vibration, 230(2), 357-378 (2000).
[7] E. Jasinien, R. Raiutis, R. literis et al., Ultrasonic NDT of wind turbine blades
using contact pulse-echo immersion testing with moving water container, Ultrasound,
63(3), (2008).
[8] P. J. Schilling, B. R. Karedla, A. K. Tatiparthi et al., X-ray computed

microtomography of internal damage in fiber reinforced polymer matrix composites,
[9] C. Meola, G. M. Carlomagno, A. Squillace et al., Non-destructive evaluation of

aerospace materials with lock-in thermography, Engineering Failure Analysis, 13(3),
380-388 (2006).
[10] Y. Hung, Shearography for non-destructive evaluation of composite structures,

Optics and Lasers in Engineering, 24(2), 161-182 (1996).
[11] A. F. Fercher, Optical coherence tomography, Journal of Biomedical Optics, 1(2),

157-173 (1996).
[12] V. Sridhar, [Combined Numerical & Experimental Methodology for the Detection of
Structural Nonlinearity in Composite Wind Turbine Blades], Delft University of
Technology (2012).
[13] G. Elert, [The physics hypertextbook], (2006).
[14] R. W. Hertzberg, [Deformation and fracture mechanics of engineering materials],

Wiley (1996).
103
Chapter 5
[15] J. Schn, T. Nyman, A. Blom et al., A numerical and experimental investigation of

delamination behaviour in the DCB specimen, Composites Science and Technology,
60(2), 173-184 (2000).
[16] T. Sebaey, N. Blanco, C. Lopes et al., Numerical investigation to prevent crack

jumping in Double Cantilever Beam tests of multidirectional composite laminates,
[17] T. Sebaey, N. Blanco, J. Costa et al., Characterization of crack propagation in mode I

delamination of multidirectional CFRP laminates, Composites Science and
Technology, 72(11), 1251-1256 (2012).
[18] W. S. Johnson, and P. Mangalgiri, [Investigation of fiber bridging in double cantilever

beam specimens], Defense Technical Information Center (1986).
[19] U. Sharma, E. W. Chang, and S. H. Yun, Long-wavelength optical coherence

tomography at 1.7 m for enhanced imaging depth, Optics Express, 16(24),
19712-19723 (2008).
104
6
OCT Elastography for
Measuring the Deformation
within Glass Fiber Composite
This chapter is reproduced from:
P. Liu, R. M. Groves, and R. Benedictus, Optical coherence elastography for

measuring the deformation within glass fiber composite, Applied Optics, 53(22),
5070-5077 (2014).
6.1 Introduction
In the area of experimental solid mechanics, deformation and strain measurement of materials
and structures subjected to various loadings (e.g. mechanical loading or thermal loading) are
important tasks. Aside from the widely used pointwise strain gauge technique, various
full-eld non-contact optical methods, such as holography [1], speckle interferometry [2],
moir interferometry [3], and digital image correlation (DIC) [4] have been widely accepted
and commonly used as powerful tools to analyze the mechanical properties of structures.
However, all these techniques mentioned are developed for surface deformation measurement.
Due to anisotropy or inhomogeneity of materials such as composites, the measurement of
microscopic deformation within the structure is of interest for providing new information
about the internal structural behavior under loading. So far, embedded fiber Bragg grating
(FBG) sensors [5, 6] have been proved to be well suited for accurate, localized or distributed
105
Chapter 6
strain measurement in composite structures. However a major problem is that this type of
technique is not able to provide a full field view of the strain distribution within the material.
Elastography [7] is a non-invasive medical image technique that can map the elastic
properties of soft tissue. The local variations of the stiffness inside the tissue can be
determined from the elastography maps, known as elastograms, which provide additional
clinical information for diagnosis of various diseases. In the past 20 years, elastography has
been extensively developed, largely based on ultrasound [8] and magnetic resonance imaging
(MRI) [9]. However, the major drawback of ultrasound scanning and MRI are their
relatively low spatial resolution. So far with the advance of these techniques, high-frequency
ultrasound has improved the imaging resolution to 25 m [10] and MRI can reach 44 m [11].
Nevertheless, these methods cannot provide high-resolutional measurement of
micromechanical properties in a specimen.
Optical coherence tomography (OCT) is a contactless and non-destructive imaging

modality based on the principle of low coherence interferometry. OCT has an inherent spatial
resolution of 1~15 m [12], which is much better than ultrasound and MRI techniques.
Optical coherence elastrography (OCE), an OCT based elastrography was first introduced by
Schmitt [13] in 1998. It was used to detect the depth-resolved sample deformation induced by
quasi-static compression. Benefiting from the high resolution of OCT, OCE has the potential
to evaluate the elasticity of the sample with a spatial resolution as well as sensitivity that are
out of the reach of ultrasound and MRI.
Inspired from the OCE applications on biological tissues [14-16], this chapter aims to
explore OCE to investigate glass fiber composites under mechanical loadings. The chapter is
structured as follows: section 6.2 describes a customized OCT setup, loading system, and the
speckle tracking algorithm for displacement measurement. In section 6.3, an evaluation of the
designed OCE system is reported for rigid body translation. Then displacement maps are
determined by applying the verified OCE technique on glass fiber composites under different
mechanical loadings. In section 6.4, the advantages and possible further improvements for the
OCE application on glass fiber composites are discussed. Finally, section 6.5 concludes the
work of this chapter.
106
OCT Elastography for Measuring the Deformation within Glass Fiber Composite
6.2 Methods
6.2.1 Experimental Setup
The specimens investigated in this chapter are glass fiber composites which are also prepared
at TU Delft using the same manufacture processing as introduced in chapter 5. These
specimens contain 8 layers of 0/90woven E-glass fiber in an epoxy resin. The dimensions
of each specimen are 100 mm length, 10 mm width and 2.15 mm thickness. Therefore
TD-OCT that has a larger depth scan range is selected for providing the cross-sectional
images of the composite specimen. Besides the OCT setup, two loading systems, tension and
three point bending were designed, as shown in Fig. 6.1(a) and (b) respectively. During the
tensile test, the rectangular specimen was clamped horizontally by two pairs of steel plates
and fixed by screws that penetrate both the laminate and specimen. The distance between the
two fixture points was 90 mm. Then one pair of plates was attached to a fixed stage, while the
other was connected to the post of a linear actuator (NA23C60-T4, Zaber Inc, Canada) that
provides tensile loading. In the three point bending test, two rods with a separation distance of
50 mm were in contact with the back surface of the specimen. The same actuator provided the
loading from the middle point along the length of the specimen. In both loading systems,
OCT lateral scans were symmetrically distributed around the middle of the two fixture points.
Pairs of tomograms acquired before and after deformation were then analysed to determine
the displacement maps.
Fig. 6.1: Setup for the OCE application. (a) and (b) are the tensile and three point bending
tests on the specimen respectively. The actuator provides linear loading and the OCT probe
scans the specimen to acquire pairs of tomograms under the different loading conditions.
107
Chapter 6
6.2.2 Image Processing Techniques
Similar to their ultrasonic counterpart, most OCE techniques rely on speckle tracking [17] to
estimate the relative motion of a bulk structure under different loading forces. Phase-sensitive
detection [18] is another methodology to measure the internal displacement, but it only works
for the Fourier-domain OCT (FD-OCT) platform and cannot measure the lateral displacement.
Typically, during speckle tracking, the pair of OCT B-scan images from pre-stressed and
stressed specimens are processed with the cross correlation technique. The cross correlation
coefficient, Rl ,k , is defined by:
m1 m2
( X
i 1 j 1
i, j X )(Yi l , j k Y )
Rl ,k (6.1)
m1 m2 m1 m2
( X
i 1 j 1
i, j X) 2
(Y
i 1 j 1
i l , j k Y) 2
where X and Y are the mean pixel values in a predefined subset, m1 m2 , on images X and Y
respectively. i and j are the axial and lateral locations of a given pixel in image X, while l and
k are the corresponding pixel displacements in image Y. Over the search region, an array of
correlation coefficients is calculated and its peak value identifies the target destination.
Fig. 6.2 defines the procedure for internal displacement measurement based on the
cross-correlation approach. The displacements are estimated pixel by pixel until the
processing window moves over the entire pre-stressed image. However, to guarantee the
accuracy of the speckle tracking, regions where the magnitude of the interference signals fall
below twice the level of the electronic noise are excluded from the analysis. The average of
the electronic noise is obtained from the area where the structure is not imaged (e.g. above the
specimen surface). A threshold for maximum correlation coefficient is also set to eliminate
the pixels which are not able to be tracked correctly. The threshold value is optimized to 0.25
in this chapter by minimizing the difference between the calculated mean displacement and
the actual one deduced from the translation of the actuator. Besides, the size of the processing
window, m1 m2 , is an important parameter in the tracking procedure. An optimization of the
window size is discussed in section 6.3.2. Finally a 2D median filter is applied to remove the
random noise and thus improve the contrast of the displacement images.
108
Fig. 6.2: Image processing steps for internal displacement estimation.
6.3 Results
6.3.1 Rigid Body Translation
A test of rigid body translation was introduced to evaluate the OCE system for the glass fiber
composite application. A pair of cross-sectional images were obtained before and after the
specimen translation and these are shown in Fig. 6.3(a) and (b). Both images are composed of
500 A-scans and image a region of 10 mm in length. Internal structures in the composite are
shown by the optical reflectivity in logarithmic scale, as indicated by the colorbar. Two bright
lines as well as some lighter irregularly shaped regions can be seen from both images, which
indicate the surfaces and the fiber bundles within the specimen respectively. From Fig. 6.3 (a)
to Fig. 6.3(b), the entire specimen was translated 6.8 m axially and 40 m laterally,
equivalent to 2 pixels in two orthogonal directions in the images. However the difference of
109
Chapter 6
the two structural images can be barely observed by eye. With the speckle tracking algorithm,
the axial and lateral displacements were calculated, as shown in Fig. 6.3(c) and (d)
respectively. Displacements at the edges of the structure images were neglected due to the
lack of data for the processing window. The areas without any displacement in Fig. 6.3(c) and
(d) indicate the locations without enough tracking speckles due to the weak reflectivity. These
mostly occur at resin rich areas.
Table 6.1 summaries the OCE system accuracy by evaluating the percentage error (PE)
and root mean square error (RMSE), given by:
|d d |
PE (6.2)
d
1 M N
RMSE
MN i 1 j 1
(di , j d ) 2 (6.3)
where di , j and d are the calculated local and mean displacement respectively. d is the
rigid body translation precisely controlled by the translation stages. M N is the pixel size
of the displacement maps. The areas with zero displacement because of untraceability were
excluded from the statistical analysis. From the results summarized in Table 6.1, it is
noticeable that the mean displacements in both axial and lateral direction are extremely close
to the actual physical translation. The percentage errors are 2.9% axially and 1.5% laterally,
and the RMSEs are 3.3 m and 10.2 m respectively. The higher RMSE in the lateral
direction is due to the larger pixel interval (20 m) in this direction. Actually the
corresponding RMSEs in pixels are 1 axially and 0.5 laterally, which indicates a larger
variation of the measured displacements in the axial direction. This can also be deduced from
the inhomogeneous displacement map, as shown in Fig. 6.3 (c). The higher axial
measurement error could be attributed to the nonlinear sweeping of the reference mirror in the
OCT setup, as evaluated in chapter 4. The nonlinear sweeping mainly exists at the startup of
the reference motor where acceleration occurs, so the A-scans could have some misalignment,
which brings an extra measurement error in the axial direction.
110
Fig. 6.3: OCE evaluation with a test of rigid body translation. (a) and (b) are a pair of
structural images before and after translation respectively. (c) and (d) are the calculated
displacement maps in the axial and lateral directions respectively.
Table 6.1: OCE system accuracy analysis
Orientation / m d / m PE RMSE / m
Axial 6.6 6.8 2.9% 3.3
Lateral 39.4 40 1.5% 10.2
6.3.2 Tensile Test
In the tensile test experiment, the specimen was stretched by 500 m between the clamps,
precisely controlled by the actuator (99.94% accuracy). Two cross-sectional images were
recorded before and after the loading, as shown in Fig. 6.4. Both images were obtained from
500 A-scans in the same spatial region of 10 mm length, in the middle of the specimen along
the stretching direction. The pair of images were processed to calculate the lateral
displacement. For this research, the influence of processing window size for cross correlation
was also investigated. Fig. 6.5 shows the calculated lateral displacement with processing
window sizes of 20*20, 30*30, 40*40, 60*30, 50*50 and 80*40 pixels (depth by lateral). The
111
Chapter 6
unsquared window sizes were selected considering the structural images are 1000*500 pixels
(depth by lateral) overall. Larger window sizes were not used in this study due to the fact
that they would not able to track the small particles in the images [19].
Fig. 6.4: A pair of cross-sectional images of a glass fiber composite before (a) and after (b)
deformation. Both images were acquired from the same spatial location over a 10 mm
length, composed by 500 A-scans.
Fig. 6.5:Lateral displacement maps with a processing window size (axial by lateral) of (a)
20*20, (b) 30*30, (c) 40*40, (d) 60*30, (e) 50*50, and (e) 80*40. The areas with
background color indicate where the signal-to-noise ratio is too low for the displacement
calculation.
112
The mean displacements at different lateral locations can be calculated by averaging

each axial line of the displacement maps. As an example, Fig. 6.6(a) shows the average
displacement when a 50*50 window size was applied for the correlation process. It can be
seen that the calculated displacements are quite close to 250 m, which is the theoretical
average displacement of the imaging area located in the middle of the sample. The theoretical
displacement is calculated based on the assumption that the specimen is isotropic and that the
stress is applied uniformly in the lateral direction, so the local lateral displacement dl should
be proportional to its physical distance to the fixed end of the specimen l , given by:
500
dl l (6.4)
90000
where 500 m and 90,000 m, are the overall loading displacement and distance respectively.
To compare the measurement accuracy, the mean displacement and the RMSE were
calculated for each window size, as shown in Fig. 6.6(b) and (c). It can be seen that with a
processing window size of 50*50, the calculated mean displacement is closest to the
theoretical value and the RMSE reaches minimum. In this way the ideal window size can be
determined and the same parameters were used for the rest of this study.
However, it is also noted, from Fig. 6.6(a), that it is difficult to observe the tendency of
the variation in displacement with regard to the lateral distance. A major factor is the small
displacement variation (approximately 55 m based on Eq. 6.4), relatively low lateral
resolution and pixel interval (20 m) in the OCT images. Displacements of a fraction of a
pixel are impossible to be precisely tracked with current approach. More detailed
considerations of the measurement accuracy are discussed in section 6.4.
113
Chapter 6
Fig. 6.6: The mean displacement with regard to the lateral location, calculated by averaging
each axial line of the displacement maps. (b) and (c) are the mean and RMSE of the lateral
displacement with regarding to the increased window size. Specifically, the block sizes 1 to
6 represent window sizes of 20*20, 30*30, 40*40, 60*30, 50*50, and 80*40 respectively.
6.3.3 Three Point Bending
A three point bending experiment [20] could be an ideal example to test the axial deformation
of structures with OCE. In this study, incremental transverse loadings with 250 m steps were
applied. The cross-sectional images, with 15 mm length symmetrically distributed around the
loading point, were acquired for each loading condition. They were processed further to
calculate axial displacement maps, as shown in Fig. 6.7. The white frames indicate the
location of the loading post. Fig. 6.8 displays the average axial displacement under each
loading condition. It can be seen, from Fig. 6.7 and Fig. 6.8, that both the local and the
integral axial displacements increase to around 200 ~ 250 m each time with incremental
loading. The brightest areas in Fig. 6.7, corresponding to the peak regions in Fig. 6.8, indicate
the locations with the largest displacement. They are in the middle of the imaging area, which
is in a good agreement with the lateral location of the loading post where the largest
displacement theoretically occurs. However, it is worthy of note that the distribution of the
displacement is not exactly symmetrical. Also as shown in Fig. 6.7, the maximum axial
displacements measured by OCE under the incremental loadings are 238 m, 476 m, and
680 m respectively, which are slightly less than the physical translations (250 m, 500 m,
and 750 m respectively) of the loading actuator. A most reasonable explanation is due to a
114
small symmetrical error from the setup and the loading. The OCE accuracy and the
anisotropic structure of the composite could also contribute to the phenomenon.
Fig. 6.7: Axial displacement maps from one cross-sectional structure of the glass fiber
composite under a three point bending test. The white frames indicate the location of the
loading actuator. The incremental axial translation increases by 250 m for each successive
image (a-c).
115
Chapter 6
Fig. 6.8: The average axial displacements along the lateral direction obtained from the
three loading conditions with 250 m incremental axial translation of the actuator.
6.4 Discussion
OCE provides information on the internal deformation of materials based on OCT structural
images. Therefore the additional contrast provided by elastography may improve the ability of
OCT to characterize the structure of materials. In this study, OCE was applied to a glass fiber
composite under three different translation or mechanical loadings. From the results shown in
section 6.3, the first observation is that compared with OCT images, the displacement maps of
OCE have better image contrast which more clearly show the internal structure of the
composite specimen. The OCT images rely on the depth-resolved optical reflectivity from the
specimen. Due to the extremely close refractive indexes of the glass fiber and the resin (both
of which are close to 1.55 [21]), the reflections from inside the glass fiber composite have a
low intensity, resulting in an interference signal just above the noise level, as shown in Fig.
6.3(a) and (b), as well as in Fig. 6.4. While in OCE, the speckle pattern, instead of the
reflectivity is tracked to quantify the displacement from the OCT images. Therefore in Fig.
6.3(c) and (d), as well as Fig. 6.5, the structure of the glass fiber composite is much more
clearly shown. So with this enhanced image contrast, OCT based elastography can be further
applied to characterize defects, such as microcracks and delaminations within the glass fiber
composite.
However, it has to be pointed out that the OCE spatial resolution is lower than OCT due
to the use of the processing window and this resolution decreases with the growing size of the
window, as shown in Fig. 6.5. The larger window sizes tend to average out differences in
116
displacements in a region, resulting in less noisy displacement maps but with reduced spatial
resolution. From Fig. 6.6(b) and (c) it can be seen that statistically the measurement accuracy
improved dramatically when increasing the window sizes from 20*20 to 40*40. The
measured mean displacement improved from 231 m to 243 m, equivalently to a drop in
percentage error from 7.6% to 2.8%. The RMSE decreases from 31 m to 20 m
correspondingly. Afterwards, there is no significant improvement from further increasing the
window size. Finally a size of 50*50 was found to be a good compromise between
measurement accuracy and spatial resolution.
Besides processing window size, the measurement accuracy could also be affected by
the OCT image noise, out-of-plane displacement, and speckle decorrelation [22]. For the OCT
image noise, a threshold was set to filter out the areas with low signal-to-noise ratio (SNR),
which indicate resin-rich areas and areas outside the specimen. These areas in the OCT
images are excluded from the displacement calculation. The out-of-plane displacement is not
studied in this chapter, but it is possible to determine it using a similar speckle tracking
technique or digital volume correlation (DVC) on pairs of volumetric OCT images [23]. The
most troublesome factor is speckle decorrelation. The cross-correlation approach of OCE has
a limited ability to track the deformations which are either too small or too large. For example,
when the speckle motion is only a fraction of a pixel, the cross-correlation approach cannot
track the shift robustly even with various smoothing techniques [24]. At the other extreme, the
relative positions of the scatterers that form the speckle patterns change gradually when the
deformation increases, which results in gradual speckle decorrelation. However, unlike
speckles in soft tissues that are easily decorrelated by a small deformation on the order of the
OCT spatial resolution [25], speckle patterns from composites can endure much larger
displacements as presented in this chapter. An explanation could be that the areas with fiber
reinforcement are stiffer than the polymer matrix areas. So the areas with fiber bundles, where
speckles occur, behave more like rigid bodies that translate without deformation. That is why
speckles can still be tracked after a displacement of hundreds of micrometers. Nevertheless,
speckle decorelation is an inevitable phenomenon during the loading test of a specimen. An
improvement to the tracking accuracy and measurement range [26, 27] could be a direction
for future work.
As the results in Fig. 6.7 illustrate, a sequence of displacement maps from a glass fiber
composite under gradually increasing loading can provide sufficient information to observe
the structural variations or even to deduce the regional stiffness by a knowledgeable observer
without further processing. However, to estimate a samples depth-resolved elastic properties
117
Chapter 6
quantitatively, the local strain needs to be calculated based on the displacement maps.
Currently, the method of least squares (LS) has proved to be a robust approach for the strain
estimate in OCE [28]. A set of displacement data in a neighboring region is used to calculate
local strain from least-squares fitting. However, unlike biological tissues that can provide
consecutive scatters in a cross-sectional area, optical reflections within glass fiber composite
only occur at the fiber-resin interfaces, resulting in discrete speckle patterns separated by the
resin-rich areas in OCT images. Consequently, displacement maps, as shown in section 6.3,
can only display the local results discretely, which makes it difficult to estimate the strain
distribution. A possible solution is to mix small inclusions with epoxy to track the
displacements from the resin as well, but care must be taken to minimize the influence of the
inclusions on the composite elastic properties.
6.5 Conclusions
This study explored the use of optical coherence elastography for the deformation
measurement of glass fiber composites, for the first time to the best of our knowledge. The
developed OCE system can track internal displacements in the range from a few micrometers
to hundreds of micrometers. The cross-sectional displacement maps clearly show the
structural features and structural variations, which could provide supplementary
understanding of the mechanical properties of glass fiber composites.
References
[1] R. Rowlands, and I. Daniel, Application of holography to anisotropic composite plates,

Experimental Mechanics, 12(2), 75-82 (1972).
[2] J. Butters, and J. Leendertz, Speckle pattern and holographic techniques in engineering
metrology, Optics & Laser Technology, 3(1), 26-30 (1971).
[3] D. Post, B. Han, and P. Ifju, [Moir interferometry], Springer (1994).
[4] T. Chu, W. Ranson, and M. Sutton, Applications of digital-image-correlation techniques to

experimental mechanics, Experimental Mechanics, 25(3), 232-244 (1985).
118
[5] Y. Okabe, S. Yashiro, T. Kosaka et al., Detection of transverse cracks in CFRP composites
using embedded fiber Bragg grating sensors, Smart Materials and Structures, 9(6), 832
(2000).
[6] J. Botsis, L. Humbert, F. Colpo et al., Embedded fiber Bragg grating sensor for internal strain
measurements in polymeric materials, Optics and Lasers in Engineering, 43(3), 491-510
(2005).
[7] J. Ophir, I. Cespedes, H. Ponnekanti et al., Elastography: a quantitative method for imaging
the elasticity of biological tissues, Ultrasonic Imaging, 13(2), 111-134 (1991).
[8] C. L. De Korte, G. Pasterkamp, A. F. Van Der Steen et al., Characterization of plaque

components with intravascular ultrasound elastography in human femoral and coronary
arteries in vitro, Circulation, 102(6), 617-623 (2000).
[9] A. Manduca, T. E. Oliphant, M. Dresner et al., Magnetic resonance elastography:

non-invasive mapping of tissue elasticity, Medical Image Analysis, 5(4), 237-254 (2001).
[10] Q. Zhou, S. Lau, D. Wu et al., Piezoelectric films for high frequency ultrasonic transducers in
biomedical applications, Progress in Materials Science, 56(2), 139-174 (2011).
[11] L. Massey, M. Miranda, L. Zrinzo et al., High resolution MR anatomy of the subthalamic
nucleus: Imaging at 9.4 T with histological validation, Neuroimage, 59(3), 2035-2044
(2012).
[12] B. E. Bouma, and G. J. Tearney, [Handbook of optical coherence tomography], Marcel

Dekker (2002).
[13] J. M. Schmitt, OCT elastography: imaging microscopic deformation and strain of tissue,
Optics Express, 3(6), 199-211 (1998).
[14] J. Rogowska, N. Patel, S. Plummer et al., Quantitative optical coherence tomographic

elastography: method for assessing arterial mechanical properties, British Journal of
Radiology, 79(945), 707-711 (2006).
119
Chapter 6
[15] H.-J. Ko, W. Tan, R. Stack et al., Optical coherence elastography of engineered and
developing tissue, Tissue Engineering, 12(1), 63-73 (2006).
[16] B. F. Kennedy, K. M. Kennedy, and D. D. Sampson, A Review of Optical Coherence

Elastography: Fundamentals, Techniques and Prospects, IEEE Journal of Selected Topics in
Quantum Electronics, 20(2), 1-17 (2014).
[17] D. D. Duncan, and S. J. Kirkpatrick, Processing algorithms for tracking speckle shifts in
optical elastography of biological tissues, Journal of Biomedical Optics, 6(4), 418-426
(2001).
[18] R. K. Wang, S. Kirkpatrick, and M. Hinds, Phase-sensitive optical coherence elastography

for mapping tissue microstrains in real time, Applied Physics Letters, 90(16), 164105 (2007).
[19] J. Rogowska, N. Patel, J. Fujimoto et al., Optical coherence tomographic elastography

technique for measuring deformation and strain of atherosclerotic tissues, Heart, 90(5),
556-562 (2004).
[20] A. W. Christiansen, J. Lilley, and J. B. Shortall, A three point bend test for fibre-reinforced
composites, Fibre Science and Technology, 7(1), 1-13 (1974).
[21] G. Elert, [The physics hypertextbook], (2006).
[22] C. Sun, B. Standish, B. Vuong et al., Digital image correlation-based optical coherence
elastography, Journal of Biomedical Optics, 18(12), 121515-121515 (2013).
[23] J. Fu, M. Haghighi-Abayneh, F. Pierron et al., [Assessment of corneal deformation using

optical coherence tomography and digital volume correlation], Springer (2013).
[24] K. Parker, M. Doyley, and D. Rubens, Imaging the elastic properties of tissue: the 20 year
perspective, Physics in Medicine and Biology, 56(1), R1 (2011).
[25] C. Sun, B. Standish, and V. X. Yang, Optical coherence elastography: current status and
future applications, Journal of Biomedical Optics, 16(4), 043001 (2011).
120
[26] S. J. Kirkpatrick, R. K. Wang, and D. D. Duncan, OCT-based elastography for large and
small deformations, Optics Express, 14(24), 11585-11597 (2006).
[27] V. Y. Zaitsev, L. A. Matveev, G. V. Gelikonov et al., A correlation-stability approach to

elasticity mapping in optical coherence tomography, Laser Physics Letters, 10(6), 065601
(2013).
[28] B. F. Kennedy, S. H. Koh, R. A. McLaughlin et al., Strain estimation in phase-sensitive

optical coherence elastography, Biomedical Optics Express, 3(8), 1865 (2012).
121
7
Further Non-destructive Testing
Applications
This chapter is partly reproduced from:
P. Liu, R. M. Groves, and R. Benedictus, Optical Coherence Tomography for the

Study of Polymer and Polymer Matrix Composites, Strain, 50(5), 436-443(2014).
7.1 Introduction
Optical coherence tomography (OCT) is a non-destructive, contactless and high-resolution
imaging method, which allows the acquisition of one-, two- or three-dimensional
depth-resolved image in situ and in real time. Currently OCT is not only an already
well-established diagnostics technique for biomedicine applications, it has also received
increasing attention in non-destructive testing (NDT) field. For example, in industry, OCT has
shown potential to improve productivity by inline quality control in the monitoring of the
manufacture of multiplelayered plastic foils [1]. The detected parameters include thickness,
homogeneity, and possible inclusion of air or impurities. In the field of artwork conservation,
OCT also starts to play an important role for scientific research. The structural imaging of
artworks, e.g. paintings and porcelain, provides substantial information for historical research
and restoration treatment [2]. Nevertheless, as stated in chapter 1, the development of OCT in
the NDT field is much lagging behind of its biomedical counterparts. Both the OCT
techniques and its applications in the NDT field need to be further explored.
123
Chapter 7
In chapters 5 and 6, glass fiber composites have been studied with the designed OCT
system. This chapter aims to demonstrate some further NDT applications. It is organized as
follow: section 7.2 focuses on the characterization of polymer coatings, including microcracks
and thickness measurement. Section 7.3 studies both front and back faces of a wooden panel
painting. This includes structural imaging, defects examination in paint layers and 3D surface
reconstruction of a wooden knot. Finally, section 7.4 summarizes the chapter.
7.2 Defects and Thickness Measurement of Polymer Coatings
7.2.1 Introduction
Polymer coatings have been widely used in engineering fields due to their advantages
including light weight, good processibility and chemical stability [3]. For example, epoxy
coatings are extensively employed for corrosion protection of steel pipes and metal containers
used in the gas and oil industry, water pipelines, beverage cans, etc. In aerospace engineering,
polymer coatings are used to protect aircraft from corrosion, impact, and lightning [4, 5].
Traditional coatings for the aerospace industry were developed for aluminium substrates.
Since composites substrates play an increasingly important role in new airplane designs, new
challenges arise for the coating industry. One of the main challenges concerning coatings on
composites is the appearance of cracks in the coating which occur at much lower strain levels
then on aluminum substrates [6]. Those cracks occur due to the ductile-brittle transition [7] of
the coatings and the substrates. Another initiator of the crack forming is known to be residual
stresses due to differences in expansion coefficients. Therefore, this section aims to
demonstrate the quality assessment of polymer coatings using OCT, including the detection
microcracks and the measurement of the thicknesses of polymer coatings.
7.2.2 Detection and measurement of microcracks
The first sample investigated was an epoxy coating with a painted carbon fiber composite as
the substrate. As shown in Fig. 7.1, the dimension of the specimen is 100 mm long and 10
mm wide. The light stripers indicate the microcracks within the coating. These cracks have
different width and some of them are hardly visible by naked eye. The image at the bottom of
Fig. 7.1 shows an enlarged view from the area indicated by the red square. The arrow which
crosses two microcracks indicates the OCT scan line.
124
Further Non-destructive Testing Applications
Fig. 7.1: (a) an epoxy coating on a carbon fiber composite substrate and (b) an enlarged
image from the area indicated by the red square. The dark stripes indicate microcracks in
the polymer coating. The arrow indicates the OCT scanning area.
During this test, both TD-OCT and FD-OCT measurements were implemented. As
shown in Fig. 7.2, the cross-sectional structural images were obtained from the same area
which has a size of 1.1 mm in depth by 3 mm in width, comprising to 150 A-scans. The
sample was tilted by a small angle during the measurement to avoid the strong surface
specular reflection, partially indicated by the upper bright lines in the both images. The lower
bright area in each image indicates reflectors from the surface and in the subsurface of the
substrate. By a comparison analysis of the two images from different measurements, two
cracks, indicated by arrows in the Fig. 7.2 can be identified based on the discontinuity of the
surface reflection lines. The first one is more obvious due to the larger size of crack, which is
approximately 40 m in width. The second crack is more ambiguous since weak reflections
from the surface of the coating still can be detected. This is because the lateral resolution,
decided by the beam size in the customized OCT system is 20 m, and this limits the ability
to detect smaller cracks.
125
Chapter 7
Fig. 7.2: Cross-sectional images taken from an epoxy coating on a carbon fiber composite
measured by (a) TD-OCT and (b) FD-OCT. The arrows indicate the locations of
microcracks within the coating.
7.2.3 Thickness Measurement
The second sample is a transparent Epikote Resin 1001 coating on a substrate of 99.5% pure
AA1050 aluminum alloy. Fig. 7.3 (a) is a cross-sectional image acquired by fast FD-OCT
measurement. The two bright lines indicate reflections from the front and back surfaces of the
coating respectively. Compared with the carbon fiber composite shown in Fig. 7.2, it can be
seen that the substrate of aluminum alloy in this application is more reflective rather than
scattering. Therefore the thickness of this coating is easier to measure as it is determined from
the distance between the two reflections, taking the coating refractive index of 1.55 into
account. The refractive index was determined by calculating the ratio of the optical thickness,
measured by OCT, to the real physical thickness measured by a thickness gauge (PosiTector
6000 FS, DeFelsko Corp, US). Fig. 7.3 (b) is one A-scan signal taken from the location of the
dashed line in Fig. 7.3 (a). By calculating the distance between the two peaks the estimated
thickness is 121 m, very close to a thickness of 117 m measured by a thickness gauge at the
same spot.
126
Fig. 7.3: (a) A cross-sectional and (b) an A-scan signal (taken from the dashed line in (a))
from an epoxy coating with alumina alloy substrate. The thickness of the coating can be
determined by measuring the distance between the two reflection peaks from the front and
back surfaces of the coating.
7.2.4 Discussion
The results presented in section 7.2.2 show that OCT is able to detect microcracks with a
minimum width of 20 m in a polymer coating. The detection of smaller cracks would be
possible with an improved lateral resolution using an objective with high numerical aperture
(NA), which is usually referred to as an Optical Coherence Microscope (OCM) [8]. However
one should bear in mind that there is always a tradeoff between the beam size and the focal
depth. The focal depth should be sufficient to encompass the entire axial scan range.
Otherwise, tracking the focal depth by translating the focal lens along with the axial scan is
necessary, as introduced in chapter 2.
The performance of TD-OCT and FD-OCT can be evaluated and compared based on
results presented in Fig. 7.2. The imaging speed is always one of the most important
characteristics of a measurement system. For TD-OCT it takes 3 seconds to record each
A-scan and more than 7 minutes to record the complete cross-sectional image. While in
FD-OCT even with a 20 ms integration time set for the CCD detectors to maximize the
sensitivity, only 3 seconds was needed to display the whole layer structure, which is 150
times faster than TD-OCT. Moreover, due to the nonlinear sweep of the reference mirror
discussed in chapter 4, TD-OCT suffers from image distortion in the axial direction. This is
revealed by the relatively rough coating surface in Fig. 7.2 (a) imaged by TD-OCT, compared
with Fig. 7.2 (b) obtained by FD-OCT. In FD-OCT this is not an issue as no mechanical
movement is needed to record the A-Scan.
127
Chapter 7
Besides, it is worth noting there is high amplitude signal at the beginning of the A-scan
as shown in Fig. 7.3 (b). It is a disturbing term besides the source spectrum, denoted by the
mutual interference of all backreflected waves and is located around the depth z 0 , as
introduced in chapter 2. Therefore in FD-OCT an offset of 200 m to the surface of the
sample is necessary to separate it from object signal [9]. The drawback of this processing is
that it further limits the detectable depth in FD-OCT. Taking the customized OCT system as
an example, only a depth of only 1.11 mm is achievable in a polymer-based material with
1.55 average refractive index, which limits the application of FD-OCT for relatively thick
materials such as glass fiber composites as discussed in chapters 5 and 6.
For the thickness measurement, the slightly difference in results between OCT and the
thickness gauge could be caused by measurement errors in both instruments. The thickness
gauge (PosiTector 6000 FS, DeFelsko Corp, US) has an measurement uncertainty of 2 m
[10] and OCT could have an error up to 11 m, which is decided by the axial resolution of the
customized OCT system in the polymer coating. The prime step for thickness measurement in
OCT is to locate the signal peaks, as shown in Fig. 7.3(b). Thus a higher axial resolution (for
a narrower bandwidth of the interference signal) and enhanced SNR (for fewer ripples in the
signal peaks) should improve OCT accuracy for thickness measurement.
It is needed to point out that currently most thickness gauges used for coating thickness
measurement are based on magnetic or eddy current principles [10] which require a metal
substrate and are difficult to use for the measurement of multiple-layer coatings. Some
advanced thickness gauges1 based on ultrasonic methods can break this limitation [11], but
they cannot provide high spatial resolution as one point measurement needs a surface area in
millimeter range. However OCT does not have all these limitations. It only requires the
coating to be penetrable for light and a small refractive index difference between the coating
and substrate or two neighboring coatings. Besides, the thickness distribution can also be
easily observed from a cross-sectional OCT images. These advantages could push OCT to
become an effective tool for semi-transparent coating thickness measurement.
1
E.g., QuintSonic 7 (ElektroPhysik GmbH, Germany) enables measurement of up to 5 coatings in a single
operation. The measurement range is from 10 to 7500 m with an accuracy of 4m [11].
128
7.3 Study of Front and Back Faces of a Wooden Panel Painting

7.3.1 Introduction
The conservation of artwork is an important topic in preserving cultural heritage. Therefore,

diagnosis techniques for structural and chemical deterioration are needed to monitor the
condition of objects for their preservation. Currently, increasing attention is being given to
OCT in artwork diagnostics and restoration. Besides high resolution and reasonable
penetration depth, OCT is non-invasive and can be used in situ, which are probably the two
most important advantages over sampling collection due to conservation ethics [12]. A
review paper in this field has been published in 2011 by Targowski [2]. OCT has been used to
evaluate the structure of paintings [13-15], historical glass [16, 17], and other
semi-transparent objects, e.g. jade [18], porcelain and faience [19]. During the examination of
paintings, the thickness of the varnish layer is interesting and is easy to detect due to its
transparence. However, because of high absorption or scattering, the paint and substrate layers
are difficult to analyze. Szkulmowska et al. [20] performed a systematic study on the
applicability of OCT to the imaging of paint layers. They analyzed 47 commonly used
pigments with two OCT systems working at center wavelengths of 1550 nm and 823 nm
respectively. As expected, the OCT system working at longer wavelength is better suited for
this type of application since 22 of the selected paints are penetrable for the 1550 nm light
source while only 16 paints are penetrable for the 823 nm light source.
The aim of this section is to characterize a wooden panel painting using OCT with a
1550 nm light source. The study focuses on both front and back faces of the painting. This
includes structural imaging of vanish and paint layers, the evaluation of defects, e.g. crack and
loss of structural layers, and 3D reconstruction of a knot with drilled holes. The achievements
and deficiencies of the current OCT system for this painting application, together with further
improvements are discussed at the end of this section.
7.3.2 Sample Description
The painting shown in Fig. 7.4 is a specimen for research, manufactured by conservators at
the by National Gallery of Athens, Greece. As shown in Fig. 7.5, the painting has a wooden
panel as the substrate. Materials such as rabbit skin glue, canvas, and gesso were prepared
consequently above the wood before reaching the paint layer. Nearly half of the painting
surface was covered by a thin layer of gold lying on a bole, while the other half area was
overlaid by red and green pigments mixed with egg yolk, as shown in Fig. 7.4. Above the
129
Chapter 7
pigment and gold layers, a vanish layer was added for protection from dust and other
atmospheric contaminants.
Fig. 7.4: Photograph of a wooden panel painting for OCT measurement.
Fig. 7.5: Stratigraphy of a wooden panel painting for OCT measurement.
To simulate the degradation during aging, the painting was thermally treated following
the oxidative type of ageing, in which the sample was heated in an oven with maximum air
recycling. Currently the defects of the painting include crack and loss of pigment, as shown in
the two enlarged images at the right of Fig. 7.6. The OCT testing will focus on these areas and
on areas with different paints, indicated by the three arrows with labels (a), (b) and (c) shown
in Fig. 7.6.
130
Fig. 7.6: A thermally aged wooden panel painting. The white arrows indicated by (a), (b)
and (c) are the OCT scanning regions. The two enlarged images on the right are the areas
indicated by white squares of two defects of the painting: a crack and loss of pigment.
7.3.3 Front Face Characterization
The three white arrows in Fig. 7.6 indicate the OCT scanning regions. Scanning line (a) and
(b) with a length of 8 mm, go through the areas which are covered by red and green pigments
and have the defects stated above. Scanning line (c) on the other edge of the painting travels
across the areas of gold, green and red pigments respectively, with a length of 15 mm. The
OCT cross-sectional images from these three regions are shown in Fig. 7.7. The optical
reflectivity along the depth of the specimen is displayed by false color with logarithm scale.
131
Chapter 7
Fig. 7.7: Cross-sectional images from the regions indicated by white lines in Fig. 7.6.
The first observation regards the varnish layer, shown as the strip between the first and
second bright lines in Fig. 7.7(a). These bright lines indicate the light reflections from the
front and back surfaces of the varnish layer. Taking an average refractive index of 1.5 into
account [21], the estimated thickness of the varnish layer is approximately 20-40 m, which is
not evenly distributed. However, this type of observation becomes difficult in Fig. 7.7(b) and
(c). It is noticed that the first light line indicating the reflectivity of the varnish front surface in
132
Fig. 7.7(a) has disappeared in Fig. 7.7(b) and is discontinuous in Fig. 7.7(c). One possible
reason is that the varnish layer was lost during conservation. It could also be that the thickness
of the varnish layer in these areas is below the axial resolution of the OCT imaging.
The second observation, by comparing the different imaging areas, is that the gold layer
is more reflective than green and red pigments. This can be seen from Fig. 7.7(c). The gold
surface is indicated by the brightest line in the area pointed out by the black dashed square.
Under the gold layer, the image gives no sign of detectable reflection. This means the light
from the current OCT system is strongly reflected by the gold, which limits the penetration to
deeper layers of the sample. However, at the areas with green and red pigments as shown in
the black solid square in Fig. 7.7(c), light not only strongly reflects on the surface of the
pigment, but also goes into the substrate layers. However, due to the dense speckles, it is
difficult to tell the actual penetration depth and the layer interfaces underneath the pigments.
The last observation is about the defects of the painting. At the arrow in Fig. 7.7(a), a
notch with 800 m width at the paint surface and 100 m depth can be observed. This is
coincident with the fact that no pigment layer exists at the area in the micrograph shown in
Fig. 7.6. Another defect is a crack observed in Fig. 7.6. Indicated by the arrow in Fig. 7.7(b),
the crack can be identified where a strong reflection comes from deep in the painting structure,
instead of the paint surface. The depth of the crack is approximately 300 m and can be easily
determined from the cross-sectional image. However the width of the crack is difficult to
measure quantitatively since weak reflections still can be detected above the crack at the
height level of the paint surface. This indicates the crack width could be smaller than the
lateral resolution of the OCT image which is 20 m.
7.3.4 Back Face Characterization
The painting tested in section 7.3.3 is based on a wooden panel. On the back surface of the
panel as shown in Fig. 7.8, the textile and a knot can be clearly seen. In order to release
residual stresses in the knot region of the painting, 7 drill holes were made in the knot. The
task of this research is to reconstruct the 3D surface of this knot area and to check if the drill
holes damage other layers of the painting.
133
Chapter 7
Fig. 7.8: A wooden panel used as painting substrate. The white arrow and square indicate
the OCT scanning line and area where a knot with drilled holes exists.
Fig. 7.9 shows one cross-sectional image of the wooden panel at the region where the
arrow in Fig. 7.8 is located. It can be seen that the wooden sample is partly penetrable by
OCT, but the interior structure is barely observed due to the dense speckles. However, the
brightest line from left to right can show the surface profile of the wooden panel. The label 1
indicates the area of a notch with a maximum 2 mm width and 1 mm depth. The labels 2 and
3 indicate two drill holes which the scanning line passes across. It is noted that the hole with
label 2 has a width of 1 mm and a depth of 1.5 mm, while the hole with label 3 has a width
of 1.5 mm and a depth that is out the range of the cross-sectional image. A more detailed 3D
surface profile is needed for a comprehensive understanding of the knot area of the wooden
panel.
1 2 3
500 m
Fig. 7.9: A cross-sectional image obtained from the region where the arrow in Fig. 7.8 is
located. The label 1 indicates the notch of the knot, and the label 2 and 3 indicate two
drill holes.
134
Fig. 7.10 demonstrates the 3D surface profile from the area indicated by the square in
Fig. 7.8. Due to the relatively large axial scanning range, TD-OCT was used for this
application. The image size is 24 mm in length and 12 mm in width. It was obtained by 3D
OCT imaging followed by locating the first interference signal peak in each A-scan signal. In
Fig. 7.10, besides the surface texture, the notch and 7 drill holes can be clearly seen. The
deepest notch and drill hole with regard to the highest knot surface are 2 mm and 3.7 mm in
depth respectively. Considering the 10 mm thickness of the wooden panel, the drill holes have
not penetrated the wooden layer and should not affect the painting directly. However the
residual stress in this area and its influence for the painting conservation is still an unknown
question.
Fig. 7.10: A 3D surface profile from the area indicated by the square in Fig. 7.8.
7.3.5 Discussion
As the results in section 7.3.3 show, the current OCT system is able to detect the varnish layer
and some defects, e.g. crack and loss of varnish and pigment layers. However structural
imaging of the paint layers could still be difficult. Even with the longer wavelength (1550 nm)
light source employed for the OCT system, the gold layer of the painting is not penetrable.
The layers of green and red pigments can be penetrated, but the speckles caused by strong
135
Chapter 7
interior scattering bring artefacts and blur the layer edges. Recently a paper published by
Liang et.al [22] shows that the best spectral window for the OCT imaging of subsurface
structure of paintings is around 2.2 m where a good balance between light scattering and
absorption is found. However, one of the challenges in building an OCT system in this
spectral window is looking for a stable and cost-effective light source. According to Eq. 2.16,
OCT working at longer wavelengths needs a light source with much broader bandwidth to
achieve the same depth resolution as the ones using visible sources.
Fig. 7.11 demonstrates one A-scan signal from the cross-sectional image shown in Fig.
7.7(a). Obviously it is difficult to correspond reflection peaks with the stratigraphy of the
painting since the amount of signal peaks shown in Fig. 7.11 is more than the amount of the
painting layers. Moreover, some of these signal peaks appear randomly during each A-scan
and construct high-contrast and dense speckles in the cross-sectional images. These speckles
largely deteriorate the OCT image quality and hinder the structure detection of the painting. In
chapter 4, a median filter is selected as the last step to supress the speckle noise in glass fiber
composites. However, it is not effective in the painting application due to the dense speckles.
Further improvement in speckle reduction is still needed.
Fig. 7.11: One A-scan signal from the cross-sectional image shown in Fig. 7.7(a).
OCT can be used to reconstruct the surface profile of a specimen due to a fact that no
matter whether the specimen is penetrable by light, the reflection from the surface of the
specimen can always be detected. Therefore the axial location of the first reflection peak in
each A-scan decides the height of each surface spot and thus a surface profile can be built
with 3D OCT scan. The accuracy of the surface height measurement is decided by the axial
136
resolution of the implemented OCT system, which is 17 m in the air. The results presented
in section 7.3.4 show that OCT could be an effective 3D scanner for the reconstruction of an
object surface profile. However, in the current OCT system, both TD and FD-OCT
measurement is implemented pointwisely on the surface. For real-time 3D surface profile
measurement, full-field OCT (FF-OCT), which acquires an en face image in one shot could
be more suitable for this type of application.
7.4 Conclusions
In summary, this chapter has demonstrated OCT as a novel NDT tool to investigate polymer
coatings and a wooden panel painting. Both quantitative and qualitative information of the
coating specimens such as thickness, microstructure and defects have been obtained
successfully. Improvements in axial and lateral resolutions are still needed to increase the
accuracy in coating thickness measurement and to detect microcracks with less than 20 m
width, respectively. For the characterization of the painting, the varnish thickness and defects
such as crack and loss of pigments were determined from the cross-sectional images. To
obtain a structural image that can clearly show the painting stratigraphy, an OCT system
working at a longer wavelength combined with speckle reduction techniques are requested. At
the back face of the painting, the wooden panel was characterized by 3D surface
reconstruction of a knot with drill holes. The information such as depth of the notch and drill
holes was obtained. To realise real-time 3D surface profile measurement, a full-field OCT is
needed due to its fast en face imaging.
References
[1] G. Hannesschlger, A. Nemeth, C. Hofer et al., Optical coherence tomography as a

tool for non destructive quality control of multi-layered foils, Proceedings of the 6th
NDT in Progress, (2011).
[2] P. Targowski, and M. Iwanicka, Optical Coherence Tomography: its role in the
non-invasive structural examination and conservation of cultural heritage objectsa
review, Applied Physics A, 106(2), 265-277 (2011).
[3] R. A. Ryntz, and P. V. Yaneff, [Coatings of polymers and plastics], CRC Press (2003).
137
Chapter 7
[4] J. H. Osborne, K. Y. Blohowiak, S. Taylor et al., Testing and evaluation of

nonchromated coating systems for aerospace applications, Progress in Organic
Coatings, 41(4), 217-225 (2001).
[5] G. Bierwagen, Next generation of aircraft coatings systems, Journal of Coatings

Technology, 73(915), 45-52 (2001).
[6] M. Westwood, J. Webster, R. Day et al., Oxidation protection for carbon fibre
composites, Journal of Materials Science, 31(6), 1389-1397 (1996).
[7] G. Swallowe, [Ductile-brittle transition], Springer (1999).
[8] J. Schmitt, S. Lee, and K. Yung, An optical coherence microscope with enhanced
resolving power in thick tissue, Optics Communications, 142(4), 203-207 (1997).

[10] PosiTector 6000 FS coating thickness gauge,

http://www.defelsko.com/p6000/og/p6000.php?g=FS&m=1.
[11] QuintSonic 7 coating thickness gauge,

http://www.elektrophysik.com/products/coating-thickness/quintsonic-7.html.
[12] H. Liang, M. Cid, R. Cucu et al., En-face optical coherence tomography: a novel
application of non-invasive imaging to art conservation, Optics Express, 13(16),
6133-6144 (2005).
[13] P. Targowski, M. Iwanicka, L. Tymiska-Widmer et al., Structural examination of

easel paintings with optical coherence tomography, Accounts of Chemical Research,
43(6), 826-836 (2009).
[14] H. Liang, B. Peric, M. Hughes et al., "Optical coherence tomography for art
conservation and archaeology," Optical Metrology, 661805-661805-12 (2007).
138
[15] H. Liang, M. Gomez Cid, R. Cucu et al., "Optical coherence tomography: a

non-invasive technique applied to conservation of paintings," Proceeding of SPIE,
5857, 58570W-58570W-9 (2005).
[16] P. Targowski, B. Rouba, M. Gra et al., Optical coherence tomography in art

diagnostics and restoration, Applied Physics A, 92(1), 1-9 (2008).
[17] H. Liang, B. Peric, M. Hughes et al., "Optical coherence tomography in

archaeological and conservation science - a new emerging field," Proceeding of SPIE,
7139, 713915-713915-9 (2008).
[18] M. L. Yang, C. W. Lu, I. J. Hsu et al., The use of optical coherence tomography for
monitoring the subsurface morphologies of archaic jades, Archaeometry, 46(2),
171-182 (2004).
[19] M. L. Yang, A. M. Winkler, J. K. Barton et al., Using optical coherence tomography

to examine the subsurface morphology of chinese glazes, Archaeometry, 51(5),
808-821 (2009).
[20] A. Szkulmowska, M. Gra, M. Targowska et al., [Applicability of optical coherence

tomography at 1.55 m to the examination of oil paintings], Springer (2007).
[21] R. S. Berns, and E. R. de la Rie, The effect of the refractive index of a varnish on the
appearance of oil paintings, Studies in Conservation, 48(4), 251-262 (2003).
[22] H. Liang, R. Lange, B. Peric et al., Optimum spectral window for imaging of art with
optical coherence tomography, Applied Physics B, 111(4), 589-602 (2013).
139
8
Conclusions and
Recommendations
As stated in chapter 1, the aim of this thesis is to use OCT as a novel NDT technique for
material structure characterization and damage detection. Besides an overview of the OCT
fundamentals and developments, the thesis is mainly driven by three tasks: instrument design,
signal processing, and applications. The following conclusions and recommendations are
organized with these three aspects.
8.1 Conclusions
8.1.1 Conclusions from Instrument Design
A hybrid TD-FD-OCT system working at 1550 nm wavelength has been built. TD-OCT has
the advantage of a large A-scan range (50 mm) and high SNR (50 dB), while FD-OCT system
has tremendous superiority in fast imaging (1000 A-scans/s) due to no movement of the
reference mirror. These two kinds of measurement can be implemented based on the user
request in the integrated OCT system, e.g. improved imaging depth or speed.
In more detail, a fiber-optic interferometry was used with a light source centered at
1550 nm, where light is less scattering in polymer-based materials and thus a deeper light
penetration can be expected. The axial resolution, decided by the wavelength and bandwidth
of the light source is 17 m in the air and 11 m in the polymer-based material (considering
1.55 average refractive index of the specimen). The lateral resolution, determined by the beam
size, is 20 m with 1.24 mm depth of focus. In TD-OCT, an electrical ODL was equipped in
the reference arm to sweep the pathlength at 1.7 mm/s, and a balanced detector was applied to
141
Chapter 8
remove the excess noise in an interference signal. In FD-OCT, a spectrometer with 160 nm
spectral range and 512 detector elements was employed, which can reach 1000 scans/s and
1.03 mm depth scan range. For both TD and FD-OCT, a pair of translation stages with 4
mm/s velocity and 28 mm travelling range, were used to shift the sample beam for the lateral
scans.
8.1.2 Conclusions from Signal Processing
In TD-OCT, A bandpass filter was designed to separate the interference signal from the DC
photocurrent and noise. Afterwards, the envelope detector was selected as the ideal method to
extract the envelope of the band-limited interference signal. As the last step, a 2D median
filter was implemented to reduce the speckle noise. The calculated SNR of TD-OCT reaches
50 dB, which allows the detection of weak reflectors from deep in the specimen.
In FD-OCT, the raw data was firstly processed to remove the influence from the optical
source and dark noise from the CCD detectors. It was then linearly resampled to evenly
spaced intervals of wavenumber, instead of wavelength. With an inverse Fourier transform,
one depth profile was recovered and a cross-sectional image was constructed by accumulating
a series of depth profiles. The quality of cross-sectional images can be further improved by
merging multiple images with different pathlength offsets. Although the SNR of FD-OCT is
not comparable with that of TD-OCT, the imaging speed and high axial linearity targets
FD-OCT applications in thickness measurement and microcrack detection in thin films.
8.1.3 Conclusions from Applications
The designed OCT system was first applied to the study of glass fiber composite. TD-OCT
was used for a large depth scan and its penetration depth in the tested glass fiber composites
could be as deep as 2.85 mm. The fiber tows are observable and could be as pronounced as
the microscope image if there is an obvious difference of refractive index between the glass
fiber and polymer matrix. To test the defects in a glass fiber composite for a wind turbine
blade, a delamination was first observed in both 2D cross-sectional and 3D volumetric images.
The delamination crack was found to propagate initially along the fiber tow/resin boundary.
Multiple reflections detected at the tip of the delamination could be caused by the bridging
fibers that exist between the delamination crack and still connect the upper and lower layers
of the composite. For a better understanding of the delamination and its propagation, the glass
fiber composite has also been tested by incremental loading. High-resolution volumetric
142
Conclusions and Recommendations
images obtained by OCT were further processed to reconstruct 3D crack surface profiles,
from which a full field 3D view of the delamination crack was given. The 45lines in the
crack surface profiles indicate the fiber orientations. They were also used for the prediction of
crack jumps.
Besides, the study explored the use of optical coherence elastography (OCE) for
deformation measurement of glass fiber composites, for the first time to the best of our
knowledge. The developed OCE system based on speckle tracking was first evaluated by a
test of rigid body translation. Then experiments were implemented for a set of glass fiber
composites under tensile test and three point bending. The results show that OCE can
measure internal displacements of a glass fiber composite in the range from a few
micrometers to hundreds of micrometers. The cross-sectional displacement maps clearly show
the structural features and structural variations, which could provide supplementary
understanding of the mechanical properties of glass fiber composites.
Additionally, some further NDT applications were selected. The integrated OCT system
was used to characterize polymer coatings. Both quantitative and qualitative information from
the specimens, such as thickness, microstructure and defects were obtained successfully. By
comparing analysis results from TD and FD-OCT measurements, two microcracks in an
epoxy coating were identified from the discontinuity of the surface reflection lines in the OCT
cross-sectional images. The superiority of FD-OCT in imaging speed 150 times faster than
TD-OCT was also observed during this testing, and the speed ratio could reach 3000 when
using the minimum spectrometer integration time of 1 ms. With a cross-sectional image from
an FD-OCT measurement, the thickness distribution of another epoxy coating was
characterized by measuring the distance between the two reflection peaks from front and back
surface of the coating. Compared with the traditional coating thickness gauge, OCT has the
advantages in high spatial resolution and multiple-layer coating measurement.
The OCT system was also used to investigate a wooden panel painting. At the front face
of the painting, the varnish thickness and defects such as cracks and loss of pigment were
determined from the cross-sectional images. Additionally, it was noted that the gold layer is
much more reflective than the green and red pigments. At the back face of the painting, the
3D surface profile of a knot of the wooden panel was reconstructed. Information such as
depth of the notch and drill holes was obtained.
143
Chapter 8
8.2 Recommendations
8.2.1 Recommendations for Instrument Design
OCT developments in NDT field, similar to its biomedical counterparts, follow one general
principle: to obtain more information in less time. More information from a sample first
requires a deeper penetration and a higher resolution, which are mainly dependent by the light
source. As reviewed in chapter 2, OCT systems using near-infrared light source centered at
1.55 m or even longer wavelength have the advantage of deeper penetration due to less light
scattering in a material. However they are mostly at the cost of lower axial resolution, which
is proportional to the square of the light center wavelength and inverse proportional to the
bandwidth of the light source. Following this direction, further OCT developments should use
novel light sources at 2m or even longer wavelengths with ultra-broad bandwidth.
Furthermore, if a swept light source in the demanded wavelength range is reliable, swept
source OCT (SS-OCT) is recommended to obtain the information in less time. As new
generation technology, SS-OCT has a similar high imaging speed to FD-OCT, but without the
drawbacks inherent to FD-OCT, e.g. imaging depth limitation due to the limited number of
pixels in CCD devices, and loss of sensitivity with depth.
Another direction of recommendation could be reductions in complexity, size, and cost

of an OCT system. Currently most commercial OCT systems are bulky and table mounted,
limiting the portability of the instruments. Due to the need of in situ imaging, e.g., for an
industrial test or artwork conservation, a compact and robust system, such as handheld OCT is
essential for this type of applications. As the first step, the OCT probe could be separately
designed as a handhold device that can easily reach a specimen in a production line or under
some mechanical testing.
8.2.2 Recommendations for Signal Processing
Speckle noise reduction is still the most important task during the signal processing of OCT
data. Although OCT is characterised by superb spatial resolution, it is sometimes difficult to
reveal the layered structure due to speckle noise. A median filter could be an easy way to
remove some speckles, but it is not effective when the speckles densely exist in an OCT
image. Besides, the loss of some detailed information provided by the original OCT image is
inevitable as the image resolutions are decreased during the filtering processing.
An additional research on speckle reduction is required to obtain OCT images with high
structural contrast. One direction is to understand the relationship between the scattering
144
Conclusions and Recommendations
properties of materials and the speckles in OCT images. Light scattering within a sample can
be simulated and analysed comparably with corresponding OCT images, which could help us
to understand the origin of speckles. Another direction is looking for new speckle reduction
techniques. Advanced signal processing algorithm should be introduced.
8.2.3 Recommendations for Applications
With the development of SS-OCT as recommended above, OCT applications in NDT field
could be in real time and in situ. Based on the results presented in this thesis, one suggestion
for further applications could be combining OCT with dynamic mechanical testing for
material characterization. For instance, an OCT prototype can be placed next to a fatigue
machine to monitor the delamination growth in glass fiber composites. Another example is
utilizing OCT to monitor the interior 3D displacement and strain of a specimen under loading
test. However, this requires further development of the elastography technique that enables
fast tracking of both small and large displacements. Another suggestion for further application
is related with some in situ measurement. The portable OCT system can be bought into a
museum for artwork conservation. It can also be used in industry on a production line for the
quality assessment of materials such as polymer coatings and glass fiber composites.
Additionally, more experimental work is still needed to explore OCT applications to a variety
of scattering materials. OCT could benefit more materials as a novel NDT tool for structural
imaging and health diagnosis.
145
Samenvatting
Optical coherence tomography (OCT, vertaling: optische coherentietomografie), is een

niet-invasieve en contactloze beeldvormingsmethode met een hoge resolutie, waarmee twee-
of driedimensionale dwarsdoorsnedes gemaakt kunnen worden van troebele media. In de
afgelopen 20 jaar is OCT uitgebreid toegepast en doorontwikkeld voor de biomedische
diagnostiek, maar toepassingen in non-destructive testing (NDT, vertaling: niet-destructief
onderzoek) zijn tot heden achtergebleven.
Het doel van deze thesis is om OCT te gebruiken als een nieuwe niet-destructieve
techniek om materiaalstructuren te karakteriseren en schade te detecteren. Naast een overzicht
van de fundamenten en ontwikkelingen van OCT, richt deze thesis zich ook op drie andere
onderwerpen: instrumentontwerp, signaalverwerking en toepassingen.
Allereerst is er een gentegreerd OCT systeem ontwikkeld waarin gebruikt wordt

gemaakt een bron met een golflengte van 1500nm. Dit systeem combineert tijdsdomein (TD)
en frequentiedomein (FD) OCT zodat beide type metingen met n instrument gedaan kunnen
worden. Het voordeel van TD-OCT is dat diepere A-scans maakt kunnen worden met een
hoge signaal-ruisverhouding, terwijl FD-OCT-metingen veel sneller gedaan kunnen worden
door het ontbreken van een bewegende referentiespiegel. En van deze twee meettechnieken
kan gekozen worden afhankelijk van de eisen van de gebruiker van dit hybride systeem, zoals
goede diepteresolutie of hoge meetsnelheid.
Voor de TD-OCT is een envelop-detector gekozen als de ideale methode voor het
demoduleren van het signaal. Een bandfilter en een tweedimensionale mediaanfilter worden
respectievelijk voor en na de demodulatie toegepast om de speckle of korrelige ruis van het
systeem te verminderen. De FD-OCT data worden eerst bewerkt zodat de invloed van de
lichtbron en van zogenaamde donkere ruis van de charge-coupled device detectoren (CCD,
vertaling: ladinggekoppeld component) verminderd worden. De data worden daarna lineair
verschaald om uniforme intervallen van golfgetallen in plaats van golflengtes te verkrijgen.
Met behulp van een fouriertransformatie kan het diepteprofiel achterhaald worden en kan een
dwarsdoorsnede geconstrueerd worden, door een serie van diepteprofielen te combineren.
147
De voornaamste toepassing van het ontworpen OCT-systeem betrof
glasvezelcomposieten: voor modellen van dat materiaal kan de microstructuur als twee- of
driedimensionale dwarsdoorsnedes in beeld worden gebracht. In het onderzoek werd nadruk
gelegd op de ontwikkeling van delaminatie in glasvezelcomposieten die gebruikt worden voor
windturbinewieken. Het materiaal werd onderworpen aan stapsgewijze belasting.
Afbeeldingen van een volume van de monsters werden gemaakt en verder bewerkt om de
driedimensionale oppervlakteprofielen van de barsten in beeld te brengen. Hiermee kon de
groei van breuken in composieten bestudeerd worden. Daarnaast werden de OCT metingen
aangevuld met zogenaamde optical coherence elastography (OCE, vertaling: optische
coherentie-elastografie) om deformaties in glasvezelcomposieten te onderzoeken; voor zover
bekend is dit de eerste keer dat dit gedaan is. Dit systeem, gebaseerd op de beweging van
speckle, werd getest door eerst een referentiemeting te doen met een bekende lineaire
verplaatsing van een testobject. Daarna zijn er metingen gedaan op glasvezelcomposieten die
werden onderworpen aan trekproeven en driepuntsbuigproeven. De resultaten lieten zien dat
OCE gebruikt kan worden om interne verplaatsingen in glasvezelcomposieten van groottes
varirend van enkele micrometers tot honderden micrometers te meten.
Naast de genoemde toepassingen worden ook andere toepassingen gepresenteerd in de

thesis, zoals metingen aan de dikte en de beschadigingen van polymeercoatings en de
karakterisering van de microstructuur van een schilderij op een houten paneel. De resultaten
van deze metingen laten zien dat het ontworpen OCT systeem ook potentiele andere
toepassingen heeft. Aanbevelingen over verdere ontwikkelingen van het ontwerp van het
OCT systeem en mogelijke verdere toepassingen worden aan het einde van de thesis
gepresenteerd.
148
Acknowledgements
The research presented in this thesis was carried out in the Aerospace NDT Laboratory of TU
Delft, Faculty of Aerospace Engineering of TU Delft. It is a multidisciplinary work which
cannot be completed without the support of many people. I wish to thank all of those who
contributed to this research.
I want to give my deepest thank to my promoter Prof. dr. ir. Rinze Benedictus who gave
me the freedom and encouragement to perform my research. I very much appreciated that he
trusted my judgement and approved the budgets I needed for the optical components, which
was curial during the instrument development. I would like to express my sincere gratitude to
my copromotor Dr. Roger M. Groves for his daily supervision. Honestly it was a huge
challenge for me to start this novel project almost from scratch. Thanks to Roger who showed
a lot of patience with me, discussed every question I had, and continuously encouraged me on
every progress I made. I had pleasure and privilege to work with a supervisor who always can
give me timely suggestions and feedback on my work and manuscripts. Thank you, Roger.
This research project could not have been accomplished without the sponsorship of
China Scholarship Council (CSC). Special thanks go to CSC and its staff for the valuable
support.
During my early stage of the instrument development, I contacted several optical and
mechanical component suppliers and received a lot of practical advices on how to improve
my system. So I would like to thank the technical support particularly from Thorlabs GmbH,
OZ Optics Ltd., Ocean Optics Inc., and Zaber Technologies Inc. I am also very grateful that
Dr. David Stifter (Johannes Kepler University, Austria) could share his experience with me on
building a first OCT instrument. Although we did not know each other before, he kept
supporting me with his knowledge. I appreciate his selfless help.
I would like to express my gratitude to all committee members who have dedicated part
of their time to review this thesis and participated in the defence. I am sincerely grateful to
Prof. P. Targowski and Prof. L. J. van Vliet for their corrections and suggestions on how to
improve the thesis.
149
The Aerospace NDT Laboratory probably is the most mysterious place within the
Aerospace Structures and Materials Laboratory since it is usually sealed for safety reasons.
However, people working inside are quite active and new ideas come up all the time to make
the material testing possible with different NDT techniques. I am proud to be one member in
the lab and to work together with my labmates: Nick, Eduardo, Pedro, Henk, Beril, Raf,
Maria, Andrei, Vassilis, Donghang, Victor, Guyonne, Rafael. Thank you for all the help in the
lab and your feedback on my presentations and manuscripts. Especially, I am thankful to
Vishal for the preparation of my specimens and Liesbeth for the translation of my thesis
summary. My thanks also go to lab technicians, Berthil, Hans, Frans, and staff working in the
workshop who gave necessary help on my experimental setup.
The 4-years study at TU Delft could not have been fun and successful without my
colleagues in the Structure Integrity & Composites Group. First I want to thank the group
secretary Gemma van der Windt for managing all the paper work in extremely efficient
manner. I also thank her and Mayank for organizing all the wonderful sports and dinner
events. My officemates: Rafiullah, Salim, Mehdi, Macos, Chongxin, Hao, Yao, Nikos, Lucas,
Leila, Adrian F, and Cornelis, thank you for all the support and fun. I enjoyed the nice
atmosphere in the office. My thanks also go to Gustavo, Adrian L and Ilias who have been
managing the GS meeting. I am grateful to all the colleagues who paid attention on my work
and gave me suggestions on the GS or other occasions.
In the past 4 years, my life in Delft could not be more satisfying, which mainly thanks
to a large Chinese community here. I want to thank all of my colleges and friends used to live
in Professor streets where we started our life in the new environment nearly at the same time
and had a lot of happy memories. Special thanks to Linfeng, Yong, Jiang, Yihui, Shuhong,
Lilan, Jianbin, Tiantian and Wuyuan for all the parties, BBQs and travelling together. I am
also thankful to the people who live in the doctor building where I have been living for two
years. I thank Xiaoyu, Mo Li, Xinyuan, Haiqiang, Meng Wei, Zhou Zhou, Yongjia, Jia Yan,
Lei, Huajie, Zhang Lu and Wandong for all the dinners, jokes and even beers, although I
always drink the least:). Meanwhile, my thanks go to Tian, Chu Jing, Liguo, Huayang,
Bishang, Jacqueline, Wang Lu, Xiaojia, Shuanghou... for all your friendship.
My last but certainly not least thank goes to my family. I am grateful to have parents
who support me in all my choice. They have always cared about my work and life a lot,
shared my joys and sorrows, and pushed me forward. Dr. Yan Ni, it was amazing to meet you
in Delft and we now build a small but a warming family, which I believe is the most
dreamlike experience in my life. Thank you for your love, patience and encouragement,
which make this journey easy and enjoyable.
150
About the Author
The author was born on the 10th September 1985 in Jurong, Jiangsu province of China. After
graduating from the high school in June 2004, he moved to Nanjing to study at the Nanjing
University of Aeronautics and Astronautics, Faculty of Mechanical and Electrical Engineering.
The author obtained his B.S. degree in June 2008. With the granted University scholarship, he
subsequently started his master study in the same Faculty under the supervision of Prof. dr. L.
Yue. He graduated on June 2010 on the topic Virtual Instrument for Vibration Analysis on
Rotating Machinery. This work has been published in the Journal of Mechanical Science and
Technology for Aerospace Engineering.
In September, 2010 the author moved to the Netherlands to start his PhD research in TU
Delft with the sponsorship of China Scholarship Council. His work was carried out in
Aerospace NDT Lab, Structural Integrity and Composites group, under the supervision of
Prof. dr. ir R. Benedictus and Dr. R. M. Groves. During the 4-years period of PhD study, the
author conducted activities on optical instrumentation and its application on material
characterization, which resulted in the present PhD dissertation.
On October 1, 2014, the author was appointed as a researcher in the Structural Integrity
and Composites group. His work is focused on fiber-optic instrumentation for
Non-destructive testing.
151
List of Publications
Journal
[1] P. Liu, R. M. Groves, and R. Benedictus, Optical coherence elastography for measuring
the deformation within glass fiber composite, Applied Optics, 53(22), 5070-5077
(2014).
[2] P. Liu, R. M. Groves, and R. Benedictus, 3D monitoring of delamination growth in a

wind turbine blade composite using optical coherence tomography, NDT & E
International, 64, 52-58 (2014).
[3] P. Liu, R. M. Groves, and R. Benedictus, Optical Coherence Tomography for the Study
of Polymer and Polymer Matrix Composites, Strain, 50(5), 436-443(2014).
[4] P. Liu, R. M. Groves, and R. Benedictus, Signal processing in optical coherence

033201 (2013).
Conference Proceedings
[1] P. Liu, R. M. Groves, and R. Benedictus, "Non-destructive evaluation of delamination

growth in glass fiber composites using optical coherence tomography," SPIE Smart
Structures and Materials + Nondestructive Evaluation and Health Monitoring, San Diego,
US (2014).
[2] P. Liu, R. M. Groves, and R. Benedictus, "Quality assessment of aerospace materials

with optical coherence tomography," SPIE Photonics Europe, Brussels, Belgium (2012).
153
Conference Presentations
[1] P. Liu, R. M. Groves, and R. Benedictus, Optical coherence tomography of the front
and back Faces of a wooden panel painting," Photon 14, London, UK (2014).
[2] P. Liu, R. M. Groves, and R. Benedictus, P. Liu, R. M. Groves, and R. Benedictus,

Optical coherence elastography for measuring the deformation and strain within glass
fiber composite, 16th International Conference on Experimental Mechanics, Cambridge,
UK (2014).
[3] P. Liu, R. M. Groves, and R. Benedictus, Optical coherence tomography for the study
of polymer and polymer matrix composites, PhotoMechanics, Montpellier, France
(2013).
[4] P. Liu, R. M. Groves, and R. Benedictus, Comparative analysis of digital demodulation

approaches in optical coherence tomography," Photon 12, Durham, UK (2012).
154

Ping Liu Thesis

Uploaded by

Copyright:

Available Formats

Ping Liu Thesis

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ping Liu Thesis

Uploaded by

Copyright:

Available Formats

Optical Coherence Tomography

for Material Characterization

ter verkrijging van de graad van doctor

Master of Science in Mechanical Engineering,

Prof. dr. ir. R. Benedictus

Copromotor: Dr. R. M. Groves

Copyright 2014 by Ping Liu

Printed in the Netherlands by Uitgeverij BOXPress, s-Hertogenbosch

Optical coherence tomography (OCT) is a non-invasive, contactless and high resolution

An integrated OCT system working at a wavelength of 1550 nm has been built. It

Summary .................................................................................................................................. vii

2 Fundamentals and Developments of Optical Coherence Tomography ............................ 13

4 Design of Signal Processing .............................................................................................. 65

7 Further Non-destructive Testing Applications ................................................................ 123

8 Conclusions and Recommendations................................................................................ 141

Samenvatting .......................................................................................................................... 147

1.1 Non-destructive Testing

becomes an indispensable step to control manufacturing processes, ensure product integrity

NDT can be performed by a wide group of techniques based on penetrating medium or

1.2 Research Motivation

Fig. 1.3: Yearly OCT publications in PubMed categorized by application [21].

evaluating displacement distributions within glass fiber composites under mechanical

1. Development of a state-of-art OCT system specialized for material

1.3 Thesis Outline

Chapter 4 focuses on signal processing algorithms for TD-OCT and FD-OCT. In

Chapter 6 expands OCT applications to elastography (OCE), which is an technique

[1] W. F. Smith, and J. Hashemi, [Foundations of materials science and engineering],

[2] Intelligent airframe, http://www.a350xwb.com/advanced/fuselage/.

[5] Wind energy facts, http://www.nwea.nl/windenergie-de-feiten.

[7] C. Hellier, [Handbook of nondestructive evaluation], McGraw-Hill (2001).

[8] W. Staszewski, C. Boller, and G. R. Tomlinson, [Health monitoring of aerospace

[9] J. Krautkramer, and H. Krautkrmer, [Ultrasonic testing of materials], Cambridge

[10] R. Halmshaw, [Industrial radiology: theory and practice], Springer (1995).

[12] B. Auld, and J. Moulder, Review of advances in quantitative eddy current

[14] D. Francis, R. P. Tatam, and R. M. Groves, Shearography technology and

[15] C. Meola, G. M. Carlomagno, A. Squillace et al., Non-destructive evaluation of

[16] P. J. Schilling, B. R. Karedla, A. K. Tatiparthi et al., X-ray computed

[17] A. Ng, and J. Swanevelder, Resolution in ultrasound imaging, Continuing Education

[18] D. Huang, E. Swanson, C. Lin et al., Optical coherence tomography, Science,

[19] B. E. Bouma, and G. J. Tearney, [Handbook of optical coherence tomography],

[20] J. Pawley, [Handbook of biological confocal microscopy], Springer (2010).

[21] Historical statistics of academic publications in the field of optical coherence

[22] B. Potsaid, B. Baumann, D. Huang et al., Ultrahigh speed 1050nm swept

[23] N. Nishizawa, "Ultrahigh resolution optical coherence tomography," Opto-Electronics

[24] E. Dalimier, and D. Salomon, Full-field optical coherence tomography: a new

[25] E. Gtzinger, B. Baumann, M. Pircher et al., Polarization maintaining fiber based

[27] OCT market, http://www.sweptlaser.com/OCT-market.

[28] D. Stifter, Beyond biomedicine: a review of alternative applications and

[29] A. Nemeth, G. Hannesschlager, E. Leiss et al., [Optical coherence tomography

[30] G. Hannesschlger, A. Nemeth, C. Hofer et al., Optical coherence tomography as a

[31] S. Zhong, Y.-C. Shen, L. Ho et al., Non-destructive quantification of pharmaceutical

[32] E. Alarousu, [Low coherence interferometry and optical coherence tomography in

[34] J. Czajkowski, T. Prykri, E. Alarousu et al., Optical coherence tomography as a

[35] D. Stifter, K. Wiesauer, M. Wurm et al., Investigation of polymer and polymer/fibre

[36] P. Targowski, B. Rouba, M. Gra et al., Optical coherence tomography in art

[37] J. Ophir, I. Cespedes, H. Ponnekanti et al., Elastography: a quantitative method for

[38] J. M. Schmitt, OCT elastography: imaging microscopic deformation and strain of

2.2 Fundamentals of Optical Coherence Tomography

Er (, t ) Ar () exp[ j (2r ()lr t )] (2.1)

Es (, t ) As () exp[ j (2s ()ls t )] (2.2)