Vous êtes sur la page 1sur 6

Ultrasonics 44 (2006) 93–98

www.elsevier.com/locate/ultras

Depth gauging of defects using low frequency wideband


Rayleigh waves
R.S. Edwards *, S. Dixon, X. Jian
Department of Physics, University of Warwick, Coventry, CV4 7AL, UK

Received 8 June 2005; received in revised form 10 August 2005; accepted 13 August 2005
Available online 29 September 2005

Abstract

In non-destructive testing for cracks it is not sufficient to simply detect the presence of a defect, but it is essential to have an accurate
measure of the depth. Accurate calibration of the techniques used to gauge defect size is therefore necessary. Recent progress in the field
of non-contact ultrasonic testing has led to the development of a practically viable system for generating and detecting wideband Ray-
leigh waves on electrically conducting or magnetic samples using electro-magnetic acoustic transducers (EMATs). This system has been
used to gauge the depth and position of surface breaking defects, and has many applications including metal billet testing and detecting
and sizing gauge corner cracking in rails. In this paper we report experiments calibrating the response of EMATs when a defect is present
between the generator and receiver, using a calibration sample with slots machined perpendicular to the surface to simulate surface
breaking cracks. The depth of the defect can be gauged in the time domain and frequency domain, with an accurate ÔfingerprintÕ of
the position given by an enhancement of the signal when the receiver is close to the defect. The best choice of EMAT design for different
applications is discussed, as is the best position for the receive EMAT to avoid areas of interference between the Rayleigh wave and bulk
waves diffracted from the crack tip.
 2005 Elsevier B.V. All rights reserved.

PACS: 43.38.+n; 81.70.Cv; 81.05.Bx; 42.62.Cf

Keywords: Ultrasound; Non-destructive testing; Surface defect; EMATs

1. Introduction There are several approaches to defect characterisation.


One possibility is the use of ultrasonic surface waves,
Non-destructive testing (NDT) is increasingly becoming namely Rayleigh waves on flat samples where the sample
an essential tool in the manufacturing process, especially in thickness is significantly greater than the wavelength [5–
steel and aluminium casting, and for safety testing of struc- 8]. Traditional ultrasonic testing techniques involve the
tures and rails [1–3]. An NDT technique should be able to use of bulk waves generated by contacting transducers with
locate the position of any defects present with the highest a coupling fluid between the transducer and the sample
possible detection accuracy, with defect sizing essential in under test, in which surface defects may be missed [2].
order to measure quality and safety. In many applications The scattering of Rayleigh waves from defects has been
surface defects may be present, for example gauge corner studied in the time domain by several authors, both exper-
cracking in rails [2,4]. In this case, a technique sensitive imentally and theoretically [8–15]. A sizing technique in the
to surface defects will give the highest probability of frequency domain has been suggested [14,15]. Recent ad-
detection. vances in the field of non-contact ultrasonic testing has
led to the possibility of using non-contact techniques, for
example laser generation [16,17], air coupled transducers,
*
Corresponding author. Tel.: +44 2476 574116; fax: +44 2476 692016. or electro-magnetic acoustic transducers (EMATs) [5–
E-mail address: r.s.edwards@warwick.ac.uk (R.S. Edwards). 7,18,19] to generate and detect Rayleigh waves.

0041-624X/$ - see front matter  2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.ultras.2005.08.005
94 R.S. Edwards et al. / Ultrasonics 44 (2006) 93–98

In this paper we concentrate on NDT using EMATs. the mirror current (with velocity v) opposing this current
We have developed a system generating and detecting pulse will experience a force, F,
mainly Rayleigh waves using two EMATs in a pitch-catch F ¼ qv  ðBs þ Bd Þ; ð1Þ
configuration [5–7,10]. The majority of the energy in a
Rayleigh wave lies within one wavelength below the sample due to the interaction with the static field Bs into the sam-
surface; wavelengths longer than the defect depth will ple from a magnet which may be held above the coil and
mostly be able to pass under the crack, whereas those with the dynamic magnetic field Bd from the current pulse
shorter wavelengths will be effectively blocked [8]. In order [21]. This force will accelerate the electrons, which will in
to probe to depths typical of surface breaking defects in turn drag the atoms slightly and so create a transient force
rails and billets we use a low frequency wideband signal, and resulting ultrasonic displacement in the sample. In
with each pulse containing wavelengths between about 5 magnetic materials generation by the magnetoelastic mech-
and 25 mm. The presence of a defect between the EMATs anism is also possible. This is the magnetic equivalent of
will lead to a reduction in signal amplitude and a change in the piezoelectric effect; a distortion will occur in a magnetic
the frequency content of the pulse. material when a magnetic field is applied [7].
We report calibration experiments on an aluminium Different designs of EMAT coil will generate pulses with
sample containing slots ranging in depth from 0.5 to different characteristics. For example, a linear coil pro-
20 mm, each 1 mm wide. The experimental details are duced by wrapping a wire around a magnet will generate
introduced, and the changes in signal amplitude and fre- a highly directional wideband signal, whereas a spiral coil
quency content with depth are calibrated. Comparisons will be omnidirectional for surface waves. The experiments
are made to earlier measurements by Viktorov [8]. The reported here use either a linear coil consisting of six turns
variations in calibration between different designs of of 0.315 mm diameter wire, or a meander coil, to generate,
EMAT coil are discussed, and finally the correct position- and a 2.5 mm wide linear coil of 25 turns of 0.08 mm diam-
ing of an EMAT for accurate depth gauging is commented eter insulated copper wire to detect the signal [5–7,23], with
on. typical signals shown in Fig. 1. The EMATs are held a
fixed distance apart and can be kept at a fixed liftoff by
2. Experimental details use of a trolley. The EMATs can then be scanned along
a sample with the signal recorded using an analogue to dig-
The experimental details of the EMAT technique have ital data acquisition card and saved and analysed using a
been described elsewhere [7,19–22], and we will only talk Labview routine. For online experiments, using the
briefly about the mechanism here. EMATs can generate EMATs in Ôsingle-shotÕ mode would normally be required,
ultrasound on electrically conducting or magnetic samples however, for lab based experiments each signal is averaged
via the Lorentz mechanism or a magnetoelastic mecha- five times for improved signal to noise.
nism, with the EMATs coupling electromagnetically to
the sample and the ultrasound generated directly within 3. Experimental results
the sample skin depth [22]. No couplant is required, and
liftoffs to several millimetres above the sample are possible. 3.1. Depth calibration
A pulse generator sends a current with a peak value of
approximately 300 A with a duration of 5 ls through the In order to size defects in real samples, calibration sam-
coil forming part of the generating transducer. In the Lor- ples containing well characterised simulated defects are re-
entz force mechanism the electrons in the sample forming quired. Measurements on an aluminium bar of dimensions

40 1.2
Amplitude (arb.units)

1.0
20
Peak to peak amplitude

0.8
0
0.6

-20
0.4
No crack
2.5 mm slot
-40 5 mm 0.2
15 mm

54 56 58 60 62 64x10-6 0 5 10 15 20
(a) Time (s) (b) Slot depth (mm)

Fig. 1. (a) Calibration of the linear EMATs over a clear surface or either side of the 2.5, 5 and 15 mm deep slots; (b) the reduction in transmitted signal
(open points) plus the fit to equation 2 (solid line) and simulated curve from data in [8].
R.S. Edwards et al. / Ultrasonics 44 (2006) 93–98 95

65 · 65 · 1000 mm are reported in this paper. The sample content between 100 and 300 kHz in each pulse (see
contains slots machined normal to the sample surface to Fig. 4) and calculated the expected transmitted signal
depths ranging from 0.5 to 20 mm with a width of 1 mm, amplitude as a function of slot depth by taking an average
where the separation between the slots is large enough that of the Viktorov transmission coefficient over this range of
reflections from other slots and sample ends do not inter- wavelengths (using data from Ref. [8]). This is shown as
fere with the signal of interest [6,10]. We compare our mea- a dotted line in Fig. 1(b), and is seen to agree very well with
surements with earlier experiments using narrow band the data measured in these experiments. Similar results are
transducers reported in Ref. [8]. This reference gives the found when using the meander generating coil set-up, with
transmission and reflection coefficients as a function of good agreement with the Viktorov data when the slightly
d/kR, where d is the depth of the slot and kR is the Rayleigh different frequency content of the waves is taken into ac-
wavelength. count (see Fig. 4, inset).
We report first the calibration for a pair of linear A ÔfingerprintÕ of the position of a defect when scanning
EMATs wound onto rectangular magnets, with the static slowly is an enhancement of the signal when the receive
field of the generator oriented into the sample and that EMAT is close to the defect [6,24]. As the receive EMAT
of the receiver aligned along the sample surface, with a sep- approaches the slot interference can be seen between the
aration of 162.5 mm. This set-up is primarily sensitive to surface wave passing directly from the generation to the re-
out-of-plane motion. Other experiments presented later in ceive EMAT and those waves reflected back from the slot.
the paper use a meander generating coil and a linear receive At a point very close to the crack constructive interference
coil with its static magnetic field directed into the sample, occurs and an enhancement of the signal is seen. This effect
detecting mainly in-plane motion. depends partly on the reflected Rayleigh wave, and hence
The dominant wavelength kR1 contained in the Rayleigh the reflection coefficient, and partly on mode conversion
wave pulse using the pair of linear coils described above is at the defect [25]. The enhancement of the signal as a func-
approximately 15 mm. Fig. 1(a) shows the change in signal tion of slot depth using the linear EMAT pair is shown in
amplitude as the EMATs are moved from over a clear sec- Fig. 2 (solid points), with the enhancement reaching a value
tion of sample (solid line) to either side of the 2.5, 5 or of around 1.5 for deep slots, measured at a liftoff of
15 mm deep slots (broken lines) with the slot situated close 0.1 mm. The enhancement values were measured by taking
to the mid-point between the two EMATs. The amplitude the peak to peak signal at the point where the signal is
of the signal is reduced by an amount dependent on the slot greatest and normalising to the averaged signal measured
depth as wavelengths less than the depth of the slot will be far from the defect in the scan.
effectively blocked, while the longer wavelengths are able to For a wide receive coil the dominant effect will be from
pass under and are detected by the receive EMAT. The interference between the direct and reflected Rayleigh
proportion of reflected or transmitted surface wave energy waves, as the mode converted waves only have significant
will depend on the crack depth and the wavelength of the amplitude very close to the defect and their effect will be
signal through the reflection and transmission coefficients. smoothed out by a wide coil [25]. Mode converted surface
The change in peak to peak amplitude of the signal skimming longitudinal waves will also have more effect on
transmitted under each slot has been measured and the re- the in-plane motion than the out-of-plane motion mea-
sults are shown in Fig. 1(b) as open points, with the peak to sured by this receive EMAT. It is therefore possible to
peak signal for the EMATs over a clear section of surface
normalised to one. The change in amplitude has an approx-
imately 1/depth dependence. We have fitted the data (fit
shown as a solid line) to the equation 1.5

amplitude ¼ A þ ½B=ðd þ CÞ; ð2Þ


Signal enhancement

1.4
where d is again the slot depth and A, B and C are con-
stants of the fit with measured values A = 0.3 ± 0.2, 1.3
B = 6.5 ± 4.2 and C = 4.6 ± 2.5. For the slot of depth
0.5 mm the amplitude of the signal rises when the EMATs 1.2
are either side of the slot. This is an effect due to the width
of the slot being greater than the depth. 1.1
Many authors report transmission and reflection data as
a function of d/kR rather than d [8]. As we have a wideband 1.0
pulse each Rayleigh wave will contain many wavelengths 0 5 10 15 20
and it is difficult to transform the data directly to this nota- Slot depth (mm)
tion. Alternatively, we can calculate a transmission coeffi-
Fig. 2. The enhancement of the signal with the receiver very close to each
cient using the Viktorov data for a pulse with a specified slot (points), plus simulation of the behaviour from the reflection
frequency content. As a first approximation to our Ray- coefficient assuming a constant frequency content between 100 and
leigh wave pulse we have assumed a constant frequency 300 kHz (line).
96 R.S. Edwards et al. / Ultrasonics 44 (2006) 93–98

compare these enhancements and the reflection coefficient Fig. 3(b) as a function of crack depth. These points again
measurements in Ref. [8], by assuming that the enhanced have an approximately 1/depth dependence. As a simple
signal amplitude is a simple addition of the amplitude of estimate, if all wavelengths less than the depth of the slot
the direct signal and the reflected signal. We again assume are blocked, we would expect a fit of fc = cR/d, where fc
a constant frequency content in our pulse between 100 and is the cut-off frequency in this case and cR is the Rayleigh
300 kHz, and take an average of the measured reflection wave velocity, which is 2906 m/s in aluminium. A fit to
coefficient over this range. The solid line in Fig. 2 is the ex- equation 2 (solid line) gives the constants A = 50 ± 46,
pected enhancement calculated using this approximation B = 2111 ± 96 and C = 0.31 ± 0.01 using frequency in
and assuming that the enhancement is entirely due to kHz and depth in mm, which agrees in form with the sim-
reflections from the slot. Enhancements agree well; some ple estimate. The discrepancy between B and the known
discrepancies for deeper slots may be due to the lack of value for cR shows that the cut-off frequency defined in this
data at high enough values of d/kR in [8]. Similar results paper is lower than the simple estimate given above sug-
are again found using the meander coil set-up. gests, with the fit suggesting that wavelengths of less than
The change in signal amplitude on moving the EMATs 73% of the crack depth are effectively blocked. For this
from a clear part of surface to either side of an unknown fit C gives the error in making the slot to the required
defect can be compared to the amplitude calibration for depths as 0.31 ± 0.01 mm, which is within manufacturing
the particular EMAT combination and a measurement of error.
the defect depth from the change in signal amplitude Ref. [23] reports measurements on defects on the rail-
gained. This, however, requires an initial measurement head, in particular on samples with closely spaced surface
over a defect free section of surface. When multiple cracks cracking which have been removed from service. The rail-
are present between the EMATs further portions of the sig- head is partially corroded leading to changes in liftoff. In
nal will be reflected back towards the generating EMAT these experiments the frequency calibration was used to
and the depth can be overestimated. Changes in liftoff will give a measure of the defect depths, with the deepest defect
also change the signal amplitude, whereas the frequency between the EMATs dominating. In general, a measure-
content remains fairly constant with liftoff and this can also ment of a clean sample using the change in amplitude will
be used for depth gauging [23]. give a more accurate value for the depth, however, the
Changes in the fast Fourier transform (FFT) are difficult change in frequency content is affected primarily by the
to quantify. However, they become clearer when they are deepest defect with changes in liftoff having only a small
normalised by the FFT over a clear section of surface. effect, and therefore may be more useful online.
When a defect is present the region of the normalised
FFT corresponding to the main frequency content of the 3.2. Choice of EMATs for different applications
pulse (for the linear EMATs, 100–300 kHz, Fig. 4 (inset))
shows a slope, as shown in Fig. 3(a) for several different The previous section has described the results of exper-
slot depths. We can define a cut-off frequency that is easily iments using a pair of linear coils. In some cases a different
measurable by applying a linear fit to the slope (dotted lines design of generator coil may give better results, for example
in Fig. 3(a)) and measuring the frequency at which the fit using a meander design of generator coil gives a strong sig-
crosses the frequency axis, and defining this as a cut-off fre- nal on rails and is able to be used in single shot mode after
quency. For the two deepest slots measured the normalised filtering [23]. This coil generates a highly directional signal
FFT shows a minimum in the main region of frequency with large amplitude, but has a narrower band signal than
content and this was used as the cut-off frequency. the linear coil, leading to a slightly lower accuracy in depth
The frequencies at which the fits have zero magnitude in gauging from the frequency content due to the smaller
the FFT were recorded and are plotted as solid points in range of wavelengths used to probe a crack. In other appli-

2000
1.4
Cut-off frequency (kHz)

1.2 1 mm
Normalised FFT

1500
1.0
0.8 2.5 mm 1000
0.6
10 mm
0.4 500
0.2
20 mm 0
0.0 *
100 150 200 250 300 0 5 10 15 20
(a) Frequency (kHz) (b) Crack depth (mm)

Fig. 3. Calibration by frequency content: (a) shows the normalised FFTs and the linear fits used to define a cut-off frequency; (b) shows the behaviour of
the cut-off frequency with depth.
R.S. Edwards et al. / Ultrasonics 44 (2006) 93–98 97

cations an EMAT generating signal in all directions, such is observed as both EMATs pass over the slot). When scan-
as a spiral coil, may be required in order to test a larger sec- ning slowly and using the enhancement as an indication of
tion of the sample in one pulse using an array of receive a crack, it may be more beneficial to have a difference be-
EMATs. tween these two enhancements in order to distinguish
The difference in the frequency contents of the pulses for them. When using a coil such as the meander coil the
different designs of generator coil means that the amplitude enhancement effect is spread out as the different sections
calibrations will also vary. The current experimental set-up of the generator coil pass over the crack, and the two
for the meander coil involves a cylindrical magnet held enhancements are easily distinguishable.
above the generating coil and a linear receive coil wound The meander coil set-up has been tested and calibrated
on a cylindrical magnet with the field directed into the sam- as above, and gives similar results to the linear coil set-
ple, separated by 150 mm [6,23]. We show in Fig. 4 a com- up. This calibration has been used in later experiments
parison of the amplitude depth calibration of the meander on rails to be reported in Ref. [23], where the benefit of
coil set-up (open triangles, dotted line) and the linear coil the strong signal on steel samples outweighs the need for
set-up discussed above (solid points, solid line), with the in- the widest possible frequency range for depth gauging.
set showing the frequency content of the Rayleigh waves
for each coil. It can be seen from this inset that the meander 3.3. Positioning of detection EMAT
coil is less wideband than the linear coil, and contains less
low-frequency content. This in turn leads to the amplitude- In most of the experiments reported here scans of the
depth slope tending to zero for deeper slots, due to the sample have been taken, with the EMATs moved in small
presence of fewer long wavelengths in each pulse. steps towards the defect. For this case we can take an aver-
Another consideration in choice of coil is the enhance- age of the signals measured far from the slot to gain accu-
ment effect, which may be seen both as the receive EMAT rate values for the signals with both EMATs on the same
and the generator EMAT pass over the defect (see, for side and for the EMATs on either side of each slot.
example, Fig. 5(b); when using two similar width linear We show in Fig. 5 the results of a scan using two linear
coils to generate and detect a similar strength enhancement EMATs wound onto cylindrical magnets with their fields
oriented into the sample, held at a fixed separation of
150 mm. In Fig. 5(a) a B-scan formed from scanning the
1.2 EMATs from both being on one side of the 5 mm deep slot
4 linear (distance is negative) to both EMATs being on the other
Normalised amplitude

1.0 meander
3
side (distance P150 mm) is shown, with the peak to peak
FFT

0.8 amplitude (arbitrary units) plotted in (b) as a function of


2
position. The B-scan is arranged so that time is along the
1
0.6 x-axis, scanning distance with the slot under the receive
0
0 200 400 600x10
3 EMAT set as zero is on the y-axis, and the signal level is
0.4
Frequency (Hz) shown by the brightness of the plot. The Rayleigh wave
0.2 is the feature at 55 ls, with the reflections from the slot
moving in time with distance.
0.0 Fig. 5(b) shows the measurement of the peak to peak
0 5 10 15 20
signal during the scan. With either the receive or generator
Crack depth (mm) EMAT close to the slot (at 0 or 150 mm) an enhancement
Fig. 4. Comparison of amplitude calibrations and frequency contents of of the signal is seen. With the EMATs both on the same
pulses from linear and meander generating coils. side of the slot, away from the region where interference

Fig. 5. Scanning over the 5 mm slot. (a) B-scan described in the text; (b) peak to peak amplitude of the Rayleigh wave signal (arbitrary units).
98 R.S. Edwards et al. / Ultrasonics 44 (2006) 93–98

is measured, the signal is steady. With the EMATs on structively with the signal passing directly between the
either side of the slot the signal drops, and in the region two EMATs.
where the slot is close to the centre point of the EMATs In typical online use an array of EMATs will be scanned
the signal is again steady. With either EMAT close to the across a sample, with the signals analysed and a map of the
slot, however, the peak to peak amplitude of the signal var- positions and characteristics of each defect generated. At
ies due to the waves diffracted from the slot tip interfering the crack waves are diffracted and mode conversion occurs,
with those passing under the slot, causing a modulation of and it is important to measure the changes in signal at a po-
the signal dependant on the distance from the crack (arrival sition where the detecting EMAT is sufficiently far from the
times of the new signals). The behaviour settles down once crack. In online applications an array of EMATs should
x/kR1 P 3 [11]. It is important when measuring crack overcome problems due to this effect.
depths to ensure that the EMATs are far enough away
from the crack so that the diffracted waves do not arrive Acknowledgements
at the detector EMAT at a similar time to the Rayleigh
wave. This work was funded by the EPSRC under Grant GR/
S24435/01. We thank John Reed for his assistance.
4. Conclusions References
Using EMATs to generate and detect Rayleigh waves is [1] J. Blitz, G. Simpson, Ultrasonic Methods of Non-destructive Testing,
a suitable method for detecting and sizing surface breaking Chapman & Hall, 1996.
defects, both in the time and frequency domains. We have [2] D.F. Cannon, K.-O. Edel, S.L. Grassie, K. Sawley, Fatigue Fract.
shown that the EMATs can be calibrated using an alumin- Eng. Mater. Struct. 26 (2003) 865–887.
[3] R.W. Irving, Continuous Casting of Steel, Institute of Materials
ium bar containing various depth slots, and the change in publishing, 1993.
signal amplitude and frequency content recorded and fit- [4] P.R. Armitage, Insight 44 (2002) 369.
ted. This calibration can then be used to size defects in real [5] S. Dixon, R.S. Edwards, X. Jian, Insight 46 (6) (2004) 326.
samples when a wave can first be recorded over a clear sec- [6] R.S. Edwards, S. Dixon, X. Jian, J. Phys. D: Appl. Phys. 37 (2004)
tion of the sample. 2291.
[7] S.B. Palmer, S. Dixon, Insight 45 (3) (2003) 211.
The signal amplitude follows an approximately inverse [8] I.A. Viktorov, Rayleigh and Lamb Waves, Plenum Press, 1967.
variation with depth, with slight changes in calibration be- [9] G. Hevin, O. Abraham, H.A. Pedersen, M. Campillo, NDT&E Int.
tween different EMATs due to the different frequency con- 31 (1998) 289–297.
tents. In many applications, for example gauge corner [10] R.S. Edwards, S. Dixon, X. Jian, Rev. Prog. QNDE 24 B (2004)
cracking, many surface defects may be clustered together, 1568–1575.
[11] B.Q. Vu, V.K. Kinra, J. Acoust. Soc. Am. 77 (1985) 1425–1430.
which will lead to a larger reduction of the signal amplitude [12] R.J. Blake, L.J. Bond, Ultrasonics 30 (1992) 255–265.
and an overestimation of the crack depth. However, a cut- [13] M. Munasinghe, G.W. Farnell, J. Geophys. Res. 78 (1973) 2454–
off frequency can still be defined and measured in all cases 2466.
[23]. The cut-off frequencies measured in the experiments [14] J.A. Cooper, R.J. Dewhurst, S.B. Palmer, Philos. Trans. Roy. Soc. A
reported here again follow an approximately 1/d depen- 320 (1986) 319–328.
[15] C.M. Scala, S.J. Bowles, Rev. Prog. QNDE (2000) 327–334.
dence, as expected from a simple estimate. The error in [16] C.B. Scruby, Ultrasonics 27 (1989) 195.
the slot depths measured from the fit to the frequency [17] R.J. Dewhurst, C. Edwards, S.B. Palmer, Appl. Phys. Lett. 49 (7)
cut-off is within manufacturing error. (1986) 374.
Measurements taken using wideband transducers gen- [18] S. Dixon, C. Edwards, S.B. Palmer, J. Phys. D: Appl. Phys. 35 (2002)
erating and detecting Rayleigh waves in a non-contact 816.
[19] H.M. Frost, Physical Acoustics XIV, Academic, New York, 1979,
manner have been compared to earlier experiments using p. 179.
narrow-band sources [8]. We have shown that the trans- [20] E.R. Dobbs, Physical Acoustics X, Academic, New York, 1973,
mission and reflection coefficients measured in those p. 127.
experiments reproduce our data well. The good agree- [21] X. Jian, S. Dixon, R.S. Edwards, Insight 46 (2004) 671.
ment between the reflection coefficient measurements [22] X. Jian, S. Dixon, R.S. Edwards, NDT&E, in press.
[23] R.S. Edwards, S. Dixon, X. Jian, submitted for publication.
and our enhancement measurements show that the major- [24] A.K. Kromine, P.A. Fomitchov, S. Krishnaswamy, J.D. Achenbach,
ity of the enhancement effect when using a wide receive Mater. Eval. 58 (2) (2000) 173.
coil is due to reflections from the slots interfering con- [25] R.S. Edwards, X. Jian, S. Dixon, Y. Fan, Appl. Phys. Lett., in press.

Vous aimerez peut-être aussi