170 Giant Magnetoresistance Based Eddy-Current Sensor for High-Speed PCB Defect Detection Ravindra Koggalage, K. Chomsuwan, S. Yamada, M. Iwahara, and Udantha R. Abeyratne* Institute of Nature and Environmental Technology Kanazawa University, 2-40-20 Kodatsuno, Kanazawa 920-8667, Ishikawa, Japan Email: KOGGALAGE@YAHOO.COM *School of Information Technology and Electrical Engineering The University of Queensland, St Lucia, Brisbane, QLD4072, Australia Abstract Recently Eddy Current Testing (ECT) has become popular in many industries, perhaps due to its major advantages such as non contacting nature, low cost, portability, high sensitivity etc. In this paper, an ECT probe which consists of a meander coil and a spin valve giant magnetoresistance (SV-GMR) sensor is used for defect detection of printed circuit board (PCB). The output signal from the ECT probe is captured using two methods. First one is using a lock-in amplifier and the second is using an Analogue to Digital Converter (ADC) card. Both methods managed to detect the defects of the test PCB specimen; however the latter was much faster in operation. I. INTRODUCTION Electrical reliability of printed circuit board (PCB) is important because it ultimately determines the reliability of the final product. The PCB manufacturing process is based on chemical and mechanical actions that may damage the intended design and hence affect the reliability. Various PCB defects such as cuts, opens, nicks, protrusions, mouse bites, pin holes, missing or extra copper, incomplete drills, and narrow or wider conductors can be occurred during production. Many commercially available PCB defect detectors are based on Automatic Visual Inspection (AVI), which use visual image of the test piece to compare with standard PCB with no defects. However, these methods cannot differentiate the actual defects and other marks which are non defects appear on the surface, such as pen marks. This paper presents a high-speed PCB defect detection method based on Eddy Current Testing (ECT). Eddy Current Testing (ECT) probe can be used to detect ferromagnetic material or conductive material which influence the magnetic field generated by the excitation current of the ECT probe. Applications may vary from locating hidden metallic objects such as underground pipes, buried bombs or ore bodies, measuring the precise dimensions of bearings and bearing races, small mechanism components, detecting material discontinuities, to identifying or separating materials by composition or structure, where these influence electrical or magnetic properties of the test material [1]. Eddy Current (EC) inspection is popular for many reasons [2] such as; its non contacting nature, low cost, portability, high sensitivity to a broad range of geometric and material parameters, ability of use in dangerous environments etc. Among the electromagnetic methods for EC detection, sensors based on either Hall effect, anisotropic magnetoresistance (AMR), giant magnetoresistance (GMR) effect, have been successfully used for crack detection [3]- [6]. Both AMR and GMR sensors have small dimensions, high sensitivity over a broad range of frequency (from hertz to megahertz domains), low noise, operate at room temperature, and are inexpensive. However, GMR sensors have better directional property than AMR sensors [7]. GMR sensors have greater output than conventional AMR or Hall effect sensors and are able to operate at fields well above the range of AMR sensors [8]. It is possible to detect very small changes of material properties using GMR sensors due to its characteristics [9]. An Eddy Current Testing (ECT) probe which consists of a meander coil and a spin valve giant magnetoresistance (SV-GMR) sensor can be used to inspect high density PCB [10]. However the signal capturing process is time consuming due to the limitations of the lock-in amplifier. The method proposed in this paper is to improve the signal capturing time by using a high speed Analogue to Digital Converter (ADC) card using Fourier s transform. It also compares the results with lockin amplifier results. II. ECT PROBE CONSTRUCTION WITH SV-GMR SENSOR Fig. 1 shows the structure of the proposed high-frequency ECT probe designed for this experiment. It consists of an exciting coil and a SV-GMR sensor.
171 Another advantage of using the planer meander coil is that its ability to minimize the distance between the PCB surface and the magnetic sensor. The SV-GMR sensor is mounted on the planar meander coil between two conductors (Fig. 3), so that the effective distance between the sensor and PCB surface can be minimized. Fig. 1. High frequency eddy current probe using SV-GMR sensor Fig. 3. Cross-sectional structure of the proposed ECT probe and the PCB conductor The exciting coil is a planar meander coil which can carry a high-frequency exciting current and generate uniform magnetic field distribution only in x and y directions. This is important as SV-GMR sensor is sensitive to all directions. Following experimental results show the effects on SV- GMR sensor with magnetic fields in different directions (Fig. 2). Three constant external magnetic fields along x direction (B x ) were tested with variable magnetic fields in z- direction (B z ) at the same frequency of 500 khz. Three plots in the Fig. 2 indicate that the constant magnetic fields in other directions have no effect with the detection of magnetic field B z. Fig. 4. SV-GMR sensor characteristics Fig. 2. Effect of constant external magnetic field B x and variable magnetic field B z on SV-GMR sensor The SV-GMR sensor consists of 4 strips and each strip has dimensions of 100 µm 18 µm. Therefore, total effective area of the SV-GMR sensor is 100 µm 93 µm with 7 µm gap between strips. Normal resistance of SV- GMR sensor used in this paper is around 400 Ω. Small signal characteristics of the SV-GMR at each of its axis are shown in Fig. 4. The sensitivity of SV-GMR sensor in sensing axis B z is around 0.5 % per 100 µt and it is lower than 0.15 % per 100 µt and 0.05 % per 100 µt in x- and y- axis, respectively.
172 Reference signal (5 MHz) Sensing axis Function generator Power amplifier 93 µm 100 µm SV-GMR sensor Reference signal (5 MHz) Exciting current Lock-in amplifier Collecting data and controlling the PCB position during the scanning process and displaying the scanning results Detecting signal form SV-GMR sensor Printed circuit board Constant dc current of 5 ma Exciting coil (meander coil) Position Stepping motor controller PCB position control (a) Signal flow diagram with lock-in amplifier Input from SV-GMR High pass filter (-3dB at 500kHz) LMH6738 400MHz BW 40dB Gain High frequency Low-noise LMH6626 20dB Gain software sel. Bandpass filter (2MHz-15MHz) software sel. AD-1210-PCI Board 12 Bits 100MHz A/D conveter DSP-FPGA & Memory 512 ksamples X Integration Cos(ωt+φ) Sin(ωt+φ) Reference frequency Compute Amplitude R X Integration PC calculation (Fourier transfrom) (b) Signal flow diagram with ADC card Fig. 5. Signal flow diagrams from SV-GMR sensor to the computer with two capturing methods.
173 III. LOCK-IN AMPLIFIER The main purpose of the lock-in amplifier is to detect and measure even very small AC signals down to few nanovolts. It has the capability of taking accurate measurements even when the small signal is obscured by noise sources many thousands of times larger. It uses a technique known as phase sensitive detection to single out the component of the signal at a specific reference frequency and phase. Noise signals at frequencies other than the reference frequency are rejected and do not affect the measurement [11]. The block diagram for signal capturing process using the lock-in amplifier is shown in Fig. 5(a). IV. ADC CARD The analogue to digital converter (ADC) card used in this experiment was high speed AD-IPR-1210 by NDT Automation. This extremely low noise PCI-bus card is designed for wide bandwidths and fits into one standard PCI slot on the personal computer (PC). It has 12 bits resolution and supports sampling rates 100, 50, 25, 10, 5 MS/s. In this experiment, 100 MS/s is used as the sampling rate. The block diagram for signal capturing process using ADC card is shown in Fig. 5(b). Input signal from the SV-GMR sensor is fed through a high pass filter to remove unnecessary low frequency signals before capture through ADC card. Fourier s transform (Fig. 6) is used to determine the magnitude of the ECT signal obtained from SV-GMR sensor at fundamental frequency (exciting frequency) instead of Lock-in amplifier. The signal generally can be expressed as, a f t a n t b n t where 0 () = + ncos( ω ) + nsin( ω ), 2 n= 1 2 t 2 t an = f( t)cos( n t) dt and b ( )cos( ) 0 n f t n t dt t ω = ω t 0 2 2 Therefore the magnitude, R = an + bn, n = 1 for fundamental frequency so that it can move in x-y plane where, x is the horizontal direction from left to right and y is the vertical direction from bottom to top of the PCB shown in figure. The length of the scan line along the x-direction, and the pitch (distance to next line of scanning along the y-direction) can be specified using the interface of the position control software. Fig. 7. Marked square is the scanned area of the PCB As shown in Fig. 8, PCB is pasted horizontally and fixed to the precise position controller so that it can move on horizontal plane. The SV-GMR sensor is fixed in a way, such that the vertical distance from it to the PCB can be adjusted. However, SV-GMR sensor is kept stationary during the scanning time. Fig. 6. Fourier s transform calculation Equations used for Fourier s transform calculations are shown in Fig. 6. The signal can be expressed as a summation of sine and cosine components. Therefore, the magnitude of the selected frequency calculated as R can be used to reconstruct the image. V. EXPERIMENTAL SETUP A 200 µm single layer PCB is used in this experiment and the area investigated is a 20mm 20mm square as depicted in Fig. 7. The PCB is fixed to a precise position controller
174 Fig. 8. Experimental setup for PCB inspection High-frequency (5 MHz) sinusoidal current is fed to the planar meander coil to generate eddy-current flowing in the PCB conductor. Sensing direction of the SV-GMR sensor was set to detect magnetic field B z that is parallel to scanning direction. The magnetic field B z, usually, occurs at the defect point or PCB conductor boundary that is perpendicular to the scanning direction as shown in Fig. 9. Fig. 10. Scanned area of PCB for this experiment where five vertical defects to be detected are marked as D1 to D5. Fig. 9. Basic principle of proposed ECT probe According to Fig. 3, the minimum distance between the SV-GMR sensor and the PCB surface is 135µm. The PCB can move horizontally on xy-plane. The output signal of SV-GMR sensor is captured through ADC card, which is connected to the PCI slot of the computer. Fourier s transform (Fig. 6) is applied to measure the fundamental amplitude of ECT signal at exciting frequency of 5MHz, and the sampling rate of 100 MS/s. As a comparison, same test area of the PCB is scanned with the similar setup, using the lock-in amplifier instead of the ADC card. The ECT signals captured in both methods are shown in Fig. 11 (a) and (b) respectively. VI. RESULTS AND DISCUSSION As shown in Fig. 8 and Fig. 3, the SV-GMR sensor is placed on a horizontal plane close to the scanning area of the PCB. The scanning direction is parallel to the conductor line so that it can detect defects which are perpendicular to the scanning direction. Test area of the PCB is selected such that a 20mm x 20mm square which consists of five defects D1 to D5 as depicted in Fig. 10. Direction of scan is horizontally from left to right. D1 to D5 are the vertical defects to be detected. (a) ECT signal from ADC card
175 1. The results indicate that the ADC card method is much faster than the lock-in amplifier method. Signal Capturing Method ADC card Lock-in amplifier TABLE 1 TIME TAKEN FOR SCAN Time taken to scan 20mmx20mm area 3 minutes and 24 S 49 minutes and 33 S (b) ECT signal from Lock-in amplifier Fig. 11. ECT signal from the SV-GMR sensor corresponding to the scanned area of PCB captured with two different devices Two peaks in opposite direction can be seen at each discontinuity of the conductor. This is due to the change of eddy current direction at discontinuities. There is a scale difference in the magnitude of the ECT signal because of the gain applied at ADC card. According to the figures, fine details of the captured signal can be observed in the case of lock-in amplifier, when compared to the ADC signal. This is due to the data loss occurred during data transfer from ADC card to the PC. However, this data loss is insignificant in our experiment as all five defects were visible in the reconstructed 2-d images of both cases. Fig. 12. Reconstructed image using the ECT Signal captured by the ADC card VII. CONCLUSION The method of using the ADC card is much faster when compared to the lock-in amplifier, and still managed to produce similar results. Both methods managed to detect all five defects on the PCB area tested. Hence the high-speed ADC card is a better option for faster PCB defect detection. ACKNOWLEDGMENT This work was supported in part by Innovation Plaza Ishikawa, Japan Science and Technology Corporation. REFERENCES [1] R. C. McMaster, P. McIntire, and M. L. Mester, Nondestructive Testing Handbook, Electromagnetic Testing, American Society for Nondestructive Testing, Vol.4, 1986, p 16. [2] Peter J Shull, Nondestructive Evaluation: Theory, Techniques, and Applications, Chapter 5, Marcel Dekker Inc., New York, Basel, 2002. [3] B. Lebrun, Y. Jayet, and J. C. Baboux, Pulsed Eddy Current Application to the Detection of Deep Cracks, Mater. Eval., vol. 53, no.11, 1995, pp1296-1300. [4] E.S. Boltz and T. C. Tiernan, New Electromagnetic Sensors for Detection of Subsurface Cracking and Corrosion Rev. Prog. Quant. Non-Destruct. Eval., vol. 17,1998, pp. 1033-1038. [5] W. W. Ward and J. C. Moulder, Low Frequency, Pulsed Eddy Currents for Deep Penetration, Rev. Prog. Quant. Non-destruct. Eval. Vol.17, 1998, pp.291-298. [6] T. Dogaru and S. T. Smith, Edge Crack Detection Using a Giant Magnetoresistance Based Eddy Current Sensor, Non-dest. Test. Eval., vol. 16, 2000, pp. 31-53. [7] T. Dogoru and S. T. Smith, Giant Magnotoresistance-Based Eddy- Current Sensor, IEEE trans. on magnetics, vol. 37, No. 5, Sept. 2001. [8] Nonvolatile Electronics, Inc. GMR sensor application notes. Available online: http://www.nve.com. [9] Tomasz Chady, Evaluation of Stress Loaded Steel Samples Using MR Magnetic Field Sensor, IEEE sensors journal, vol. 2, no. 5, Oct. 2002. [10] S. Yamada, K. Chomsuwan, Y. Fukuda, M. Iwahara, H. Wakiwaka, and S. Shoji, Eddy-Current Testing Probe With Spin-Valve Type GMR Sensor for Printed Circuit Board Inspection, IEEE Trans. of Magnetics, vol. 40, no. 4, July 2004. [11] Users Manual; Model SR844 RF Lock-In Amplifier, Revision 2.5, SRS Inc., 1997. The reconstructed image using the output signal of the SV-GMR sensor captured by ADC card is shown in the Fig. 12. The time taken to scan in each case was listed in Table