Portable Eddy Current NDT Instrument Using Two
Dário Pasadas, Tiago Rocha, Helena Ramos, A. Lopes Ribeiro
Instituto de Telecomunicações, Instituto Superior Técnico
Abstract—This paper describes two implementations towards the
development of a portable low-cost system capable of detecting
defects on metallic surfaces through non-destructive testing. The
defect detection is made via an eddy-current probe using a giant
magnetoresistor (GMR) sensor. Both approaches use a computer
mouse as a position device to locate the probe and deliver a
graphical representation to facilitate the defect analysis. One
implementation is based on a planar excitation coil, embedded
power supplies and the signal processing is made by a dsPIC. The
dsPIC controls the signal generation, the measurement procedure
and the communication. The defect visualisation can be made in
a LCD in real-time or transmitted to a PC. The other
implementation uses a low-cost analog system to reduce the data
processing required to detect the surface defects. The controller
is a PIC microcontroller that does all the analog/digital
conversions and transmits the defect data to a PC via wireless
Keywords- Nondestructive Testing; Eddy Current Testing; Sine-
fitting; Planar probe; Position System; Signal Processing;
The use of non-destructive testing (NDT) plays an
increasingly important role on safety and maintenance costs of
several industries. A specific type of NDT called eddy current
testing (ECT) can be used on metallic surfaces presenting good
results when detecting metal fatigue and corrosion , while
requiring minimum preparation prior to surface testing.
When eddy currents are induced on a metallic surface,
magnetic fields are created and these fields contain the
information required to detect surface defects. There are
several types of probes with sensors capable of measuring this
magnetic field such as: giant magnetoresistor (GMR) sensors
, SQUIDS , hall sensors  and differential detection
coils . This work uses a GMR as the magnetic sensor
because it has small dimensions, can work in a wide range of
frequencies with high sensibility.
To assist the surface defect analysis, a visualisation of the
magnetic field in the vicinity of the sample is required. In order
to provide this graphical visualisation it is necessary to
associate a position system to the probe. The position system
used is a computer mouse as it is low-cost, provides enough
resolution to the purpose and requires low control processing.
This paper presents two different implementation solutions
that are compared to conclude about the optimal characteristics
of a final device.
IMPLEMENTATION USING DIGITAL SIGNAL PROCESSING
A. System Architecture
Fig. 1 shows the architecture of the implemented instrument.
(mouse + probe)
UART to USB
Figure 1. Architecture of the implemented instrument.
This system has a current generator and a DC power supply
used to power all the electronic devices in the circuit,
increasing the system portability. The probe is composed by a
planar coil and a GMR sensor, both used in ECT. The probe
position is monitored using a ball-mouse. Both probe and
positioning system require signal conditioning to allow the
dsPIC to retrieve information. The dsPIC is responsible for the
processing and the transmission of the information relative to
the eddy current readings to the LCD or to the PC and for the
control and generation of the excitation voltage by the Direct
Digital Synthesizer (DDS).
B. Probe and signal conditiong
The developed probe is presented in Fig. 2. It is built with a
GMR sensor placed in the core of a planar excitation coil with
25 turns and a inner diameter of 10 mm. The use of a planar
coil minimizes the lift-off effect (distance to the testing
surface) and can be made with a simple fabrication technique
(PCB). The GMR sensor used is AA002-02 and is made by
Non Volatile Electronics (NVE). It has four identical GMRs
assembled as a Wheatstone bridge.
Figure 2. Photo of the ECT probe assembly.
This work was supported in part by Portuguese Science and Technology
Foundation Project PTDC/EEA-ELC/67719/2006 an in part by the Instituto de
Telecomunicações – Project CLASSE. This support is gratefully acknowledged.
The signal conditioning for the GMR sensor includes an
instrumentation amplifier with a gain of 500, decoupling
capacitors for the devices power supplies (± 5V) and two high-
pass filters. The filters are connected to the differential output
from the GMR and have a cutoff frequency of 150 Hz (about 1
decade short when compared with the frequencies used in the
system – 1 kHz to 5 kHz). A voltage equal to 1.65 V is added
to the GMR output to use the maximum range (0 V – 3.3 V) of
the 12-bit ADC embedded in the dsPIC.
C. Ball-mouse based position system
In this implementation, the positioning system (attached to
the probe) is based on a roller-ball computer mouse. It enables
the defect visualization and analysis in a 2D representation.
The mouse roller ball movement causes a rotation in two
perpendicular axis (aligned to x and y directions) which
contains two perforated encoder wheels on the axis edge. The
movement and direction in each axis is monitored by two sets
of emitter (LED) and receiver (phototransistor) placed
carefully around the encoder wheel so it can interrupt the light
beam. The output signal from the phototransistors contains the
D. Signal generation
Fig. 3 depicts the circuit used to deliver the current to the
excitation coil. A transadmittance amplifier, capable of
handling current outputs up to 1 A is used. The current
controlled by the resistor R1:
The excitation sinusoidal signal is synthesized by the DDS
AD9833 from Analog Devices. The DDS is a low-power,
programmable (via SPI) waveform generator with a possible
frequency range of 0 to 12.5 MHz, 28 resolution bits and fixed
output voltage amplitude (0.6 Vpp with 0.308 V DC
Output Signal of
Figure 3. Transadmittance circuit.
The transadmittance amplifier is powered by a DC-DC
converter whose energy source is a 9 V battery or DC power
E. Software implementation
Fig. 4 shows the software flowchart used for crack
detection in this implementation. The software includes the
data processing to determine the probe position, the detection
of the crack, the excitation current control, to display (LCD)
and transmit to the PC the measurement data to the user.
The application starts by waiting for a command given by
the user. This command can set the working frequency (Ft)
and the number of periods necessary to acquire the signals
Select sampling frequency
Encodes operating frequency
Three parameters sine fitting
Estimate signal magnitude
Connected to PC?
Send information to PC
Figure 4 – Flowchart of a crack detection procedure.
When in stand-alone mode, default values are available
(2 kHz and 3 periods) and the results are visualized in a LCD.
After initialization the application maximizes the sampling
frequency for the chosen working frequency.
The application acquires the signal from the GMR using
the 12-bit ADC and processes the data with a sine-fitting
algorithm. This algorithm is used to extract the amplitude and
phase from the noisy signal, reducing the noise component
and compressing the information relative to the signal (easing
The application sends these data to the LCD, informing the
user about a crack presence in the probe proximity. The crack
proximity flag is retrieved based on the amplitude parameter
estimated. When not in stand-alone mode, the PC application
registers the current probe position and the sensor signal
amplitude. These data are transmitted from the dsPIC using a
RS232 interface supported by an embedded UART module on
the microcontroller. A UART to USB converter (FT232R
from FTDI) is used to achieve transfer speeds up to 1.25 Mbps
and to increase the connector compatibility on personal
computers. The application then returns to the signal
F. User interface and results
The computer application was developed using the
LabVIEW platform with the purpose of establishing a
graphical interface for the user. It can be divided in two steps.
In the first step, LabVIEW provides the dsPIC with the
parameters set by the user (working frequency and periods to
be processed). After the parameters transmission the
application changes to the second phase. This phase is a
reading cycle, constantly receiving and storing data from the
dsPIC containing information about the probe position and the
measurement made in that location. When the user halts the
reading state the application draws a graphical representation
with amplitude measurements in their position, as depicted in
Fig. 5. This is a representation of a measurement made several
times across the same crack (row by row). This figure
represents the result obtained when cracked sample is covered
by another 1.5 mm thick aluminum plate. The probe scanned
an aluminum crack with less than 1 mm width and 20 mm
length. A 2 kHz excitation current was used with amplitude of
Figure 5 – Scan of a 1.5 mm thick plate placed over a sample crack with 1 mm
III. IMPLEMENTATION USING ANALOG SIGNAL PROCESSING
A. Measuring System
This implementation uses an analog system to decrease the
digital processing required to retrieve the parameters related to
the defects characteristics.
Fig. 6 shows the working system block diagram.
Figure 6. System Block Diagram
The probe has both the excitation coil and GMR to
generate eddy currents on the metallic surface and retrieve
readings from the magnetic field generated by the eddy
currents. The excitation coil’s driver is a current generator set
to deliver a constant amplitude sinusoidal current.
A USB computer mouse is attached to the probe so it is
used as a position device.
After eddy current induction by the probe the analog
system filters and amplifies the signals from the GMR and
excitation coil and delivers them to two separate analog
blocks: amplitude measurer and phase detector. The amplitude
measurer’s output is a DC voltage equal to the input signal’s
amplitude. The phase detector outputs two digital signals that
are used to drive a counter in the PIC microcontroller. This
can be used to calculate the phase shift between both probe
The microcontroller is a Microchip PIC18F4550 and is the
main system controller. It gathers the parameters from the
probe signals and transmits them to a personal computer using
a radio-frequency module. This wireless link is used to
increase the portability of the whole system. The transmission
is made at 19.2 kbps.
Each time the personal computer receives data from the
microcontroller, it stores both the probe signal parameters and
probe location. When requested, it presents a graphical
representation of the various readings to the user in their
B. Probe and Signal Conditioning
The probe includes a vertical excitation coil and a GMR
magnetic sensor. The coil has 6 turns by layer and 10 layers
(60 turns) with 0.5mm in diameter and air core. The GMR is a
NVE AA002 and it is placed inside the coil’s core facing the
metallic surface (almost touching it) and its axis is
perpendicular to the excitation coil axis. This axis orientation
makes the GMR insensible to the excitation magnetic field.
However the GMR requires an external permanent magnet
placed carefully so it is biased and has a linear operation.
The probe output signals are the retrieved from the GMR
and from a current to voltage resistor (10 Ω) used in series
with the excitation coil. The GMR signal is single frequency
sinusoidal voltage with noise added to it. To reduce reading
errors both signals are filtered with a 4th order Butterworth
bandpass filter (1 kHz-10 kHz). This preserves the phase shift
between the signals and increases the signal-to-noise ratio.
The integrated circuit used in the filter design is a Texas
Instruments UAF42. It was chosen because each UAF42 can
deliver a 2nd order filter so in total 4 integrated circuits were
used and the filter implementation can be made with simple
external passive components.
C. Optical mouse based position system
The merge of a position measuring system to the probe
allows the user to visualize the GMR output sensor waveform
parameters along a position axis. This is important to conclude
on the geometry of the defect on the metal surface. Keeping in
mind the required position accuracy, portability and low-cost
of the system, an optical computer mouse was chosen as the
position device. It is inexpensive, ergonomic and has 1200
dots per inch accuracy (0.212 mm) which is enough to the
current system. The mouse is connected to a personal
computer via a USB interface and a MATLAB application
continuously updates the probe’s coordinates.
D. Signal Processing Blocks
In this implementation the processing blocks used to
retrieve the waveform parameters (amplitude and phase shift)
The amplitude measurement is made by using a precision
rectifier which includes an instrumentation amplifier from
Texas Instruments -TLE2042 followed by a low pass filter.
Being the input a sinusoidal voltage with a certain amplitude
parameter, the output will be a DC voltage that matches the
parameter being measured. The output is then converted to
digital form using the 10 bit analog-to-digital converter (ADC)
embedded in the microcontroller.
The phase detector uses two LM311 comparators (one for
each signal) connected in a Schmitt trigger design to increase
the resilience against noise. There are two Schmitt-trigger
circuits, with small hysteresis, one for each output probe
signal (two sinusoidal voltages with a certain phase shift
between them). These outputs can be interpreted by the
microcontroller as TTL signals. Hence it is possible to
measure the time between their positive transitions. A simple
timer embedded in the microcontroller can be used to measure
the time window, and a simple calculation is required to
compute the phase shift:
360º Download full-text
Timer SignalFreq TimerPeriod
The phase shift accuracy increases with the MCU clock
frequency, so to obtain the best results, the MCU clock is set
to the maximum allowed (48MHz). The microcontroller has a
RISC pipelined architecture and in this case a single
instruction takes 4 MCU cycles to complete. This means the
timer clock equals 12 MHz. The maximum time resolution is
As an example, the maximum error on the phase shift
measure of a 5 kHz signal frequency is
360 83.3 10
This value is within the expected phase shift accuracy.
The flowchart depicted in Fig. 7 shows the algorithm used
to measure the amplitude of the sensor output voltage and the
phase shift between this signal and the excitation current. It
starts by measuring the sinusoid period, followed by 256
measures to the phase shift. Each time a measure is made the
cumulative average calculation is updated and stored in
memory. After an amplitude measurement, the MCU transmits
the data and restarts.
Time measure between two positive slopes
Cumulative average calculation
Mean phase calculation
Figure 7 – Parameter acquiring algorithm flowchart.
F. User interface and obtained results
A computer is used to host the user interface software,
which is developed in a Matlab platform. This application
enables the user to define probe coordinate boundaries, to
calibrate the computer mouse pixel per mm ratio and draws
two graphs to the user with amplitude and phase shift along a
position axis. The results represented in Fig. 8 were obtained
when a through crack with 1 mm of width was detected on an
aluminum plate with 2 mm of thickness and conductivity of
Fig. 8 shows both the amplitude and the phase shift
measurement. The excitation
Iex=100 mA and the frequency was set to 5 kHz. An
amplitude variation can be observed when the probe is in the
proximity of the crack (placed on the 20 mm). The same can
be said about the phase shift.
current amplitude was
Probe Position [mm]
Figure 8 – Amplitude and phase shift along position axis.
Both implementations are capable of detecting and
analyzing cracks on aluminium surfaces based on eddy current
testing, despite several system differences. The position
system using the roller-ball mouse requires cleansing cares
and regular ball movements to achieve good results. However
this is not an issue on the optical mouse solution.
The planar coil from the first implementation showed
smaller sensibility to the lift-off effect and lower eddy current
concentration in the area of interest when compared to the
vertical coil from the second implementation. The lower eddy
current concentration is due to a lower magnetic field
concentration in the coil’s core because of the increasing
distance in each turn of the planar coil.
The processing for the waveform parameters acquisition
has shown good results on both implementations. While the
use of the sine-fitting algorithm shows great noise immunity
the analogue implementation uses bandpass filters to obtain
The GMR magnetic sensor has proven to be well suited for
both systems thanks to their small size, high sensibility in the
working frequency range and large bandwidth.
The transmission system using USB has a greater bit rate
than the wireless one, however the wireless connection can
provide better mobility while making measurements.
In the future it will be possible to use the best features
from both implementations and make a better, more portable,
flexible, smaller and versatile system.
 P. A. R. Mathieson, B. S. Hockenhull, J. Billingham, “Monitoring
corrosion fatigue cracks in an offshore structural steel by acoustic
emissions”, NDT & E International, Vol. 25, No. 6, pp. 305, 1992.
 T. Dogaru and S.T. Smith, “Giant Magnetoresistance-Based Eddy
Current Sensor”, IEEE Trans. Magnetics, vol. 37, NO. 5, pp. 3831-3838,
 C. Carr, D. Graham, J. C Macfarlane, G. B Donaldson “HTS SQUIDs
for the non-destructive evaluation of composite structures”, Inst. of
Physics, Supercond. Sci. Technol. No.16, 2003, pp. 1387–1390.
 Y. He, F. Luo, M. Pan, F. Weng, X. Hu, J. Gao, B. Liu, “Pulsed Eddy
Current Technique for Defect Detection in Aircraft Riveted Structures”,
NDT & E International, Vol. 43, No. 2, pp. 176-181, Mar. 2010.
 A. L. Ribeiro, H. G. Ramos, “Inductive Probe for Flaw Detection in
non-Magnetic Metallic Plates Using Eddy Currents”, IEEE Instr. &
Meas. Tech. Conf., I2MTC, Victoria, Canada, pp. 1447-1451, 2008.
Probe Position [mm]