- Email: [email protected]
Defects inspection of the solder bumps using self reference technology in active thermography
Infrared Physics & Technology 63 (2014) 97–102
Contents lists available at ScienceDirect
Infrared Physics & Technology journal homepage: www.elsevier.com/locate/infrared
Defects inspection of the solder bumps using self reference technology in active thermography Xiangning Lu a, Tielin Shi a,b, Jiguang Han a, Guanglan Liao b,⇑, Lei Su b, Suya Wang a a b
School of Mechanical & Electrical Engineering, Jiangsu Normal University, Xuzhou 221116, China State Key Laboratory of Digital Manufacturing Equipment and Technology, Huazhong University of Science and Technology, Wuhan 430074, China
h i g h l i g h t s We proposed a method for reducing the impact of emissivity unevenness and heating non-uniformity. The method using the self reference technology is based on the source distribution image. Three thermograms captured right after heat pulse are averaged to create SDI. For the missing bump the summation of thermal difference is no more than 14 K. It is effective using the method to identify the defects in active infrared test.
a r t i c l e
i n f o
Article history: Received 11 November 2013 Available online 30 December 2013 Keywords: Solder bump Defects inspection Self reference Source distribution image
a b s t r a c t With the decrease of solder bumps in dimension and pitch, defects inspection of the solder bumps become more difficult. A nondestructive detection system based on the active thermography has been developed for solder bumps inspection. However, heating non-uniformities and emissivity differences may impede the defects recognition. In this paper, we propose a method using a self reference technology based on a source distribution image (SDI) to eliminate the influence of unevenness in emissivity values and heating power distribution. Three thermograms captured right after the heat pulse are averaged to create the SDI. Then the SDI is subtracted from the original thermograms, and we get the thermal contrast images, in which eight points on the edge of each hot spot are selected as the feature points for the corresponding bump. Thermal difference between the feature points and the central point are adopted to quantify the thermal behaviors of the solder bumps, by which the missing bump is distinguished from the reference bumps. The results show that it is effective using the method to eliminate the impacts of emissivity unevenness and heating non-uniformities on defects identification in the active infrared test. Ó 2013 Elsevier B.V. All rights reserved.
1. Introduction The miniaturization and multifunction desire of the IC devices have promoted the higher density in microelectronics packaging. The chip I/O pins are arranged in grid array instead of locating on the periphery. Surface mounting components, such as flip chips (FC), ball grid array (BGA) and chip scale packages (CSPs), using solder bumps to realize interconnection between chips/packages and substrates or printed circuit board (PCB), are extensively used in microelectronics packaging due to the decreased package size, greater I/O density and faster speed of signal propagation [1]. However, common manufacturing defects including open, cracked, or missing solder bumps are always existed. Since the solder bumps are hidden in the package after assembly, defects inspections become more and more difficult as the solder bumps develop ⇑ Corresponding author. Tel.: +86 27 87793103; fax: +86 27 87792413. E-mail address: [email protected] (G. Liao). 1350-4495/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.infrared.2013.12.019
towards high density and ultrafine pitch, which seriously hinders the development of surface mount technology. Defect inspection of the solder bumps becomes one of the key issues in IC manufacturing technology [2]. Current nondestructive testing methods for solder bump inspection can be divided into five categories [3]: (1) electrical testing, (2) optical visual testing, (3) X-ray inspection, (4) acoustic inspection, and (5) thermal inspection. These techniques are suitable only for specific defects inspection due to their disadvantages respectively. For electrical testing, it is effective for detecting the short and open circuit in chips, but it would pass the cracks and cold joints, which provide partial or intermittent electrical connections between chip and substrate. It is difficult to locate the solder defects and manufacturing the test pads also increases the cost. Optical visual testing is always used for inspecting visible solder joints or solder bumps before assembly. It becomes almost infeasible for the chip assembly as the hidden solder bumps deny the access of light beams [4]. X-ray inspection techniques [5,6]
98
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
including radiography, laminography, and tomography, have been used for defects inspection of solder bumps. X-ray radiography applies the transmitted X-ray energy to indicate the inner situation of the package. It fails to detect fine cracks and open bump, because these defects with small air gap do not attenuate X-ray energy. X-ray laminography and tomography focusing the X-ray beam on one plane at a time and slicing the board horizontally, are effective 3D techniques which can detect almost all the defects in solder bumps. However the data acquisition and interpretation are time consuming and the equipment and operation cost are unaffordable. Acoustic techniques are effective for inspection of voids and interfacial delaminations in electronic packages, while the main drawbacks are poor resolution caused by sound scattering at object edges and the requirement for a coupling medium [7,8]. Thermal imaging inspection operates on the premise that defects exhibit different thermal behaviors from the reference, which provides a fast, non-contact and full field measurement. But surface conditions and emissivity differences may influence the resolution and infrared image contrast adversely [9]. Other inspection methods have also been investigated. Liu and Ume developed a nondestructive system using laser ultrasound and interferometer techniques to measure the transient out-ofplane displacement response of the electronic package under pulsed laser excitation [10,11]. It is effective to detect the solder bumps with large diameter and pitch. Chai utilized the hot spots in thermal images to identify the defective solder bumps when the electrical current passing through the daisy chained chips [12,13]. Defects such as a partial solder joint crack with an increased electrical resistance are identified. In our previous work, a non-destructive detection system based on the active thermography has been developed to inspect defects of the solder bump [14], in which the test chip is excited by a thermal pulse and the consequent transient response is monitored by an infrared thermal imager. The defects are distinguished by the abnormal thermal behavior. However, the heating source plays a significant role on the infrared testing. Heating non-uniformities may impede the defects recognition. In this paper, we propose a method to eliminate the effects of emissivity unevenness and heating non-uniformities on defects identification. The heating source distribution image has been constructed and the self reference technology is adopted to quantify the thermal behaviors of the solder bumps.
2. Experiment description The inspection system using active thermography has been developed, in which a commercial thermal imager of VH 680 is used to measure the transient response of the test chip under the thermal excitation of the heating source. Temperature resolution of the thermal imager is better than 80 mK, and the frame size is 640 480 pixels. A microscopic lens with the pixel resolution of 25 lm is equipped for the thermal imager to improve the spatial resolution. The fiber-coupled semiconductor laser with the center wavelength of 808 nm is used as the heating source. The optical image of the test vehicle SFA2 before assembly is shown in Fig. 1. There are 16 solder bumps arranged in an array of 4 4 pattern, one of which in up-left corner has been removed deliberately in order to introduce the defect of missing bump when the chip is bonded to the substrate. The solder bumps are 300 lm in diameter and spaced with a pitch of 600 lm. The thickness of the silicon substrate is about 300 lm. The experimental setup was deployed as depicted in Fig. 1, where the samples were inspected in transmission way during the test procedure. The silicon substrate is excited by a laser pulse heating, which is 200 ms in duration and the maximum pulse
Fig. 1. Experimental deployment and solder bumps arrangement of the SFA2.
energy is 240 mJ. Temperature evolutions of the die surface are monitored by the thermal imager and then are processed to identify the defects in the package. 3. Theory analysis and methodology 3.1. Heat conduction analysis Fig. 2 illustrates the structure with single layered chip and substrate, in which the silicon die is attached to the substrate by a solder bumps array. When a thermal excitation is activated, the heat exchange would occur in the package [15]. As known, heat transfer in the solid body is governed by the Fourier diffusion equation, which describes the distribution and variation of temperature in a given region and time,
@T k @2T @2T @2T ¼ þ þ @t qcp @x2 @y2 @z2
! þ
Q
qcp
ð1Þ
where T = T(x, y, z, t) depicts the temperature at the point (x, y, z) and time t, k denotes the thermal conductivity, q is the density and cp the specific heat capacity, and Q is defined as the internal heat generation per unit volume in unit of W/m3. There are no internal heat sources in the package because the chip is not in working condition [16]. We will put emphasis on the investigation of the heat conduction via solder bumps and the estimation of defects influence on thermal behavior in the active infrared inspection. In the transmission way, a heating pulse is imposed on the bottom surface of the substrate, and the thermal front is launched and propagates inside the structure. As the thermal front reaches the top surface of the substrate, it would propagate into the chip via solder bumps, or be hampered by the defects and then flow to adjacent solder bumps, as depicted in Fig. 2. When observed from
Fig. 2. Schematic of heat conduction via solder bumps.
99
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
Temperature evolution 20
28.4
40
28.3
34 Denoising & Smoothing
28.2
80 28.1 100 28 120 27.9
140 160 180 50
100
Temperature/ °C
33
60
32
30
27.8
29
27.7
28
150
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
31
P1(Missing bump) P2(Reference bump) P3(Reference bump) P4(Reference bump)
0
1
2
3
4
5
Time/s
Fig. 3. Thermal image of SFA2 before the thermal excitation (t = 0 s).
Fig. 5. Temperature evolution of four points.
Original thermal image 34
Thermal contrast image with contour
20 40
33.5
P1
P2
P3
P4
60 80 100
33
20
32.5
40
32
60
31.5
80
120 140 160 180 50
100
31
100
30.5
120
150
140 Fig. 4. Original thermal image (t = 0.217 s). 8
160
the die surface, the temperature over the missing bumps will be slightly lower than that of the reference bumps because of the insufficient heat transfer path. However the small temperature difference disappears rapidly since the silicon die has a high diffusivity. Then the temperature begins to drop under the effect of the air convection. The temperature around the missing bump declines more quickly than those of the reference bumps as they are different in the thermal mass. A thermal equilibrium can be reached extremely fast due to the lateral heat conduction in the package, which reduces the thermal contrast between the defects and the reference bump. Additionally, emissivity unevenness and heating non-uniformity are also obstacles for the solder bumps inspection in the active infrared test.
3.2. Self reference technology based on the source distribution image In order to eliminate the influence of emissivity unevenness and heating non-uniformity on defects identification, a method using the self reference technology based on the source distribution image is proposed. Thermal images captured right after the heat pulse would reflect the spatial distribution of the heating source power across the sample surface [17,18]. Several subsequent thermograms are selected and averaged to create the source distribution image (SDI). Averaging is done on a pixel-to-pixel basis, which allows reducing the existence of noise in images. The influence of the heating non-uniformities can be reduced by subtracting the SDI
6
20
40
60
80
100
120
140
2
0
7
180
1
160
5
3 4
180
Fig. 6. Thermal contrast image.
from the original thermal images since higher value will be removed in the region received higher excitation energy. The ‘‘self-referencing’’ is based on Shepard and Ducar’s [19] research work, in which the behavior of a defective pixel is compared with the kernel pixels surrounding it to study the deviation in the logarithmic temperature time history. In this paper thermal behavior of the pixels around the solder bumps are investigated and the self reference technique is proposed based on the fact that the hot spot over the missing bump shrinks more quickly than the reference bumps after the heating pulse. The self reference technique is calculated as following: (1) The SDI is subtracted from the original thermal images, and thermal contrast images are obtained. (2) The thermal contrast values of the hot spots are extracted from the thermal contrast images. (3) The thermal contrast value of the center point in the hot spot is subtracted from the values of the points (i, j) surrounding the hot spot, as depicted by the formula
Cði; j; tÞ ¼ T surr ði; j; tÞ T cent ðtÞ
ð2Þ
100
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
(1)
30 20
20
10
10
0 0.2
Sum of Temper Contra Differ/K
0.4
0.6
0 0.2
0.8
(3)
30
30
20
20
10
10
0 0.2
30
0.4
0.6
30
20
20
10
10
30
0.4
0.6
(7)
30
20
20
10
10
0 0.2
0.4
0.6
0 0.2
0.8
0.6
0.8
0.4
0.6
0.8
0.4
0.6
0.8
0.4
0.6
0.8
(6)
0 0.2
0.8
0.4
(4)
0 0.2
0.8
(5)
0 0.2
(2)
30
(8)
Time/s M1(Missing) R2(Reference)
R1(Reference) R3(Reference)
Fig. 7. Summation of temperature difference along with time.
4. Results and discussions
35 M1(Missing bump) R1(Reference bump) R2(Reference bump) R3(Reference bump)
Sum of temp. differ/K
30
4.1. Experimental results
25 20 15 10 5
1
2
3
4
5
6
7
8
Feature points Fig. 8. Summation of temperature difference.
where (i, j) is the coordinates of the point and t is the time. Thermal contrast value, C(i, j, t) is then used to characterize the thermal behavior of the corresponding solder bump quantitatively. We can apply this approach to overcome the emissivity difference between the metal UBM layer and the silicon material, and reduce the influence of non-uniformity in the delivery of stimulant power.
The test vehicle SFA2 has been inspected to identify the missing solder bump. It was simulated by the laser pulse, and 600 thermograms were obtained at the sampling rate of 60 Hz. Fig. 3 shows the thermal image before the thermal excitation. Due to transparence of the silicon material at the band of wavelength longer than 1.1 lm, the front side of the silicon die, where the UBM layer is located, can be seen from the back side, as shown in Fig. 3. It is almost infeasible to distinguish the missing solder bump from the reference one at the room temperature. The temperature of the UBM layer is slightly higher than that of the gap region as the emissivity differences existed. Fig. 4 shows the original thermal image at 0.217 s. As can be seen, heating non-uniformities dramatically decreased the temperature contrast between the defective solder bump and the reference one, which may hamper the defects recognition. Temperature evolution of each pixel was extracted from the thermograms sequences. Fig. 5 shows the temperature curves of four points, P1, P2, P3 and P4 (labeled in Fig. 4), where P1 is located in the defect area (missing solder bump) and the others are in sound area (with solder bumps). All the curves can be divided into three parts. In the first part, the temperature of every point rise sharply during the heating pulse. After t = 0.2 s it increases slowly
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
and reaches the maximum at t = 0.8 s.Then it decays exponentially in the third part. As depicted in Fig. 5, the temperature of the defect point P1 is slightly lower than those of the reference points in the second part. For the temperature curve of P4, the peak noise appears at 0.2 s which is affected by the random noise and heating non-uniformities. Therefore it is critical to eliminate the influence of thermal noise and heating non-uniformities. 4.2. Defects identification To improve the signal to noise ratio, each thermogram was firstly input to an improved median filter with a 5 5 mask to remove the peak noise while preserving the edge information. Then the temperature evolution of each pixel was extracted and processed using the moving average operation with the span of 5, which can eliminate the impact of periodic and random noise on the temperature. Fig. 5 also shows the smoothed temperature evolution curve of P1, P2, P3 and P4 during the time period of 0–0.6 s. Although the peak noise is suppressed to some extent, heating non-uniformities is still a serious problem to be solved. In order to avoid the inconsistent comparison between the thermal values of the solder bumps located in different heat power region. Three thermal images captured right after the heat pulse were selected to create the SDI. The SDI was then subtracted from the consequent thermal images, and thermal contrast images were obtained. Fig. 6 shows the thermal contrast image derived from Fig. 4, in which iso-value contours are also created to indicate the temperature difference. Obviously it is less ‘‘warm’’ in the region of missing bump than that of the reference bumps in the heat power iso-value contour. Solder bumps in less heat power region get lower temperatures, and vice versa. For example the temperatures of the solder bumps in the fourth line are lower than that of the solder bumps in the first line.
The self reference technology based on the SDI is explored to characterize the defects quantitatively. As shown in Fig. 6, eight points on the edge and one in the center of each hot spot have been selected as the feature points for the corresponding solder bump. Each point contains five pixels and the temperature of the point is derived by averaging the pixels values. The thermal difference between each edge point and the central point is calculated by
ði ¼ 1; 2; . . . ; 8Þ
ð3Þ
Then the summation of the C di ðtÞ values during the period [t1, t2] is derived by
C dis ðtÞ ¼
t2 t2 X X C dis ðtÞ ¼ ðT i ðtÞ T 0 ðtÞÞ ði ¼ 1; 2; . . . ; 8Þ t¼t1
5. Conclusions In this paper, we proposed a method using the self reference technology based on the source distribution images to eliminate the impacts of emissivity unevenness and heating non-uniformities. For validation of the approach, a model of heat conduction via solder bumps was constructed and the chip SFA2 with a missing solder bump was chosen as the test vehicle. Thermal images were recorded and processed using the improved median filter and moving average operation to remove the peak noise and the periodic noise. Three thermograms captured right after the heat pulse were averaged to create the SDI. Then the SDI was subtracted from the original thermal images, and we got the thermal contrast images, in which the missing bump was identified. The self reference technology was adopted to characterize the thermal behaviors of the solder bumps quantitatively. Eight points on the edge of each hot spot were selected as the feature points for the corresponding bump, and thermal difference between the feature points and the central point are calculated. For the missing bump the summation of thermal difference on every point is no more than 14 K, while for the reference bumps it is always larger than 14 K. The results demonstrated that it is effective using the self reference technology to eliminate the influence of emissivity unevenness and heating non-uniformities on defects identification in active infrared testing. Acknowledgments This research work is supported by National Natural Science Foundation of China (Grant Nos. 51305179 and 51305177) and Natural Science Foundation of Jiangsu Higher Education Institutions (Grant No. 13KJB510009). References
4.3. Quantity of the defects
C di ðtÞ ¼ T i ðtÞ T 0 ðtÞ
101
ð4Þ
t¼t1
According to Eq. (4), we get the Cdis(t) values along with the time in the period of 0.2–1 s, as shown in Fig. 7, in which M1 denotes the missing bump, while R1, R2 and R3 are the reference bumps. Fig. 7(1)–(8) are corresponding to the eight points around the hot spot respectively. As can be seen, for every point, the summation of temperature difference for the missing bump is smaller than those of the reference bumps. To distinguish the defect more explicitly, the Cdis(t) value at t = 1 s of the eight points is shown in Fig. 8. For the missing bump, the summation of the temperature difference on every point is no more than 14 K, while for the reference bumps it is always greater than 14 K, by which the missing solder bumps can be discriminated quantitatively from the reference bumps. Therefore it is effective using the self reference technology to eliminate the influence of heating non-uniformities on defects identification.
[1] R.T. Rao, Fundamentals of Microsystems Packaging, McGraw-Hill, New York, 2001. [2] M. Pacheco, Z. Wang, L. Skoglund, et al., Advanced fault isolation and failure analysis techniques for future package technologies, Intel Technol. J. 9 (4) (2005) 337–352. [3] L. Wang, H. Ding, Y. Xiong, An overview of non-destructive inspection in flip chip packaging, Mach. Electron. 5 (2004) 45–49. [4] H.N. Yen, D.M. Tsai, S.K. Feng, Full-field 3-D flip-chip solder bumps measurement using DLP-based phase shifting technique, IEEE Trans. Adv. Packag. 31 (4) (2008) 830–840. [5] T.D. Moore, D. Vanderstraeten, P.M. Forssell, Three-dimensional X-ray laminography as a tool for detection and characterization of BGA package defects, IEEE Trans. Compon. Packag. Technol. 25 (2) (2002) 224–229. [6] R. Kovacs, X-Ray inspection of microwire bonds, in: Proceedings of the 28th International Spring Seminar on Electronics Technology: Meeting the Challenges of Electronics Technology, Austria (Wiener Neustadt), Institute of Electrical and Electronics Engineers Computer Society, 2005, pp. 448–451. [7] E. Martin, C. Larato, A.C. Ment, et al., Detection of delaminations in subwavelength thick multi-layered packages from the local temporal coherence of ultrasonic signals, NDT&E Int. 41 (2008) 280–291. [8] S. Brand, K. Raum, Signal analysis in scanning acoustic microscopy for nondestructive assessment of connective defects in flip-chip BGA devices, IEEE Ultrason. Symp. (2007) 817–820. [9] X. Maldague, Theory and Practice of IR Technology for Nondestructive Testing, first ed., Wiley-Interscience, New York, 2001. [10] S. Liu, C. Ume, Digital signal processing in a novel flip chip solder joint defects inspection system, ASME J. Electron. Packag. 125 (2003) 39–43. [11] J. Yang, C. Ume, L. Zhang, Defect detection of flip chip solder bumps with wavelet analysis of laser ultrasound signals, IEEE Trans. Adv. Packag. 33 (1) (2010) 19–29. [12] T.C. Chai, B.S. Wong, W.M. Bai, A. Trigg, Y. Lam, A novel defect detection technique using active transient thermography for high density Package and interconnections, in: Electronic Components and Technology Conference, 2003, pp. 920–925. [13] J.C. Lu, A. Trigg, J.H. Wu, T.C. Chai, Detecting underfill delamination and cracks in flip chip on board assemblies using infrared microscopy, Int. J. Microelectron. Electron. Packag. 21 (3) (1998) 231–235. [14] X. Lu, G. Liao, Z. Zha, Q. Xia, T. Shi, A novel approach for flip chip solder joint inspection based on pulsed phase thermography, NDT&E Int. 44 (2011) 484– 489.
102
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
[15] R. Kandasamy, A.S. Mujumdar, Thermal analysis of a flip chip ceramic ball grid array (CBGA) package, Microelectron. Reliab. 48 (2008) 261–273. [16] X. Lu, T. Shi, Q. Xia, G. Liao, Thermal conduction analysis and characterization of solder bumps in flip chip package, Appl. Therm. Eng. 36 (2012) 181–187. [17] M. Susa, X. Maldague, I. Boras, Improved method for absolute thermal contrast evaluation using source distribution image (SDI), Infrared Phys. Technol. 53 (2010) 197–203.
[18] M. Susa, X. Maldague, I. Boras, Application of the source distribution image (SDI) procedure for porosity detection in honeycomb structures, in: 10th International Conference on Quantitative InfraRed Thermography, Quebec, 2010. [19] S. Shepard, R. Ducar, Quantitative infrared defect detection in composite aerospace structures, in: 45th International SAMPE Symposium, 2000.
Contents lists available at ScienceDirect
Infrared Physics & Technology journal homepage: www.elsevier.com/locate/infrared
Defects inspection of the solder bumps using self reference technology in active thermography Xiangning Lu a, Tielin Shi a,b, Jiguang Han a, Guanglan Liao b,⇑, Lei Su b, Suya Wang a a b
School of Mechanical & Electrical Engineering, Jiangsu Normal University, Xuzhou 221116, China State Key Laboratory of Digital Manufacturing Equipment and Technology, Huazhong University of Science and Technology, Wuhan 430074, China
h i g h l i g h t s We proposed a method for reducing the impact of emissivity unevenness and heating non-uniformity. The method using the self reference technology is based on the source distribution image. Three thermograms captured right after heat pulse are averaged to create SDI. For the missing bump the summation of thermal difference is no more than 14 K. It is effective using the method to identify the defects in active infrared test.
a r t i c l e
i n f o
Article history: Received 11 November 2013 Available online 30 December 2013 Keywords: Solder bump Defects inspection Self reference Source distribution image
a b s t r a c t With the decrease of solder bumps in dimension and pitch, defects inspection of the solder bumps become more difficult. A nondestructive detection system based on the active thermography has been developed for solder bumps inspection. However, heating non-uniformities and emissivity differences may impede the defects recognition. In this paper, we propose a method using a self reference technology based on a source distribution image (SDI) to eliminate the influence of unevenness in emissivity values and heating power distribution. Three thermograms captured right after the heat pulse are averaged to create the SDI. Then the SDI is subtracted from the original thermograms, and we get the thermal contrast images, in which eight points on the edge of each hot spot are selected as the feature points for the corresponding bump. Thermal difference between the feature points and the central point are adopted to quantify the thermal behaviors of the solder bumps, by which the missing bump is distinguished from the reference bumps. The results show that it is effective using the method to eliminate the impacts of emissivity unevenness and heating non-uniformities on defects identification in the active infrared test. Ó 2013 Elsevier B.V. All rights reserved.
1. Introduction The miniaturization and multifunction desire of the IC devices have promoted the higher density in microelectronics packaging. The chip I/O pins are arranged in grid array instead of locating on the periphery. Surface mounting components, such as flip chips (FC), ball grid array (BGA) and chip scale packages (CSPs), using solder bumps to realize interconnection between chips/packages and substrates or printed circuit board (PCB), are extensively used in microelectronics packaging due to the decreased package size, greater I/O density and faster speed of signal propagation [1]. However, common manufacturing defects including open, cracked, or missing solder bumps are always existed. Since the solder bumps are hidden in the package after assembly, defects inspections become more and more difficult as the solder bumps develop ⇑ Corresponding author. Tel.: +86 27 87793103; fax: +86 27 87792413. E-mail address: [email protected] (G. Liao). 1350-4495/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.infrared.2013.12.019
towards high density and ultrafine pitch, which seriously hinders the development of surface mount technology. Defect inspection of the solder bumps becomes one of the key issues in IC manufacturing technology [2]. Current nondestructive testing methods for solder bump inspection can be divided into five categories [3]: (1) electrical testing, (2) optical visual testing, (3) X-ray inspection, (4) acoustic inspection, and (5) thermal inspection. These techniques are suitable only for specific defects inspection due to their disadvantages respectively. For electrical testing, it is effective for detecting the short and open circuit in chips, but it would pass the cracks and cold joints, which provide partial or intermittent electrical connections between chip and substrate. It is difficult to locate the solder defects and manufacturing the test pads also increases the cost. Optical visual testing is always used for inspecting visible solder joints or solder bumps before assembly. It becomes almost infeasible for the chip assembly as the hidden solder bumps deny the access of light beams [4]. X-ray inspection techniques [5,6]
98
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
including radiography, laminography, and tomography, have been used for defects inspection of solder bumps. X-ray radiography applies the transmitted X-ray energy to indicate the inner situation of the package. It fails to detect fine cracks and open bump, because these defects with small air gap do not attenuate X-ray energy. X-ray laminography and tomography focusing the X-ray beam on one plane at a time and slicing the board horizontally, are effective 3D techniques which can detect almost all the defects in solder bumps. However the data acquisition and interpretation are time consuming and the equipment and operation cost are unaffordable. Acoustic techniques are effective for inspection of voids and interfacial delaminations in electronic packages, while the main drawbacks are poor resolution caused by sound scattering at object edges and the requirement for a coupling medium [7,8]. Thermal imaging inspection operates on the premise that defects exhibit different thermal behaviors from the reference, which provides a fast, non-contact and full field measurement. But surface conditions and emissivity differences may influence the resolution and infrared image contrast adversely [9]. Other inspection methods have also been investigated. Liu and Ume developed a nondestructive system using laser ultrasound and interferometer techniques to measure the transient out-ofplane displacement response of the electronic package under pulsed laser excitation [10,11]. It is effective to detect the solder bumps with large diameter and pitch. Chai utilized the hot spots in thermal images to identify the defective solder bumps when the electrical current passing through the daisy chained chips [12,13]. Defects such as a partial solder joint crack with an increased electrical resistance are identified. In our previous work, a non-destructive detection system based on the active thermography has been developed to inspect defects of the solder bump [14], in which the test chip is excited by a thermal pulse and the consequent transient response is monitored by an infrared thermal imager. The defects are distinguished by the abnormal thermal behavior. However, the heating source plays a significant role on the infrared testing. Heating non-uniformities may impede the defects recognition. In this paper, we propose a method to eliminate the effects of emissivity unevenness and heating non-uniformities on defects identification. The heating source distribution image has been constructed and the self reference technology is adopted to quantify the thermal behaviors of the solder bumps.
2. Experiment description The inspection system using active thermography has been developed, in which a commercial thermal imager of VH 680 is used to measure the transient response of the test chip under the thermal excitation of the heating source. Temperature resolution of the thermal imager is better than 80 mK, and the frame size is 640 480 pixels. A microscopic lens with the pixel resolution of 25 lm is equipped for the thermal imager to improve the spatial resolution. The fiber-coupled semiconductor laser with the center wavelength of 808 nm is used as the heating source. The optical image of the test vehicle SFA2 before assembly is shown in Fig. 1. There are 16 solder bumps arranged in an array of 4 4 pattern, one of which in up-left corner has been removed deliberately in order to introduce the defect of missing bump when the chip is bonded to the substrate. The solder bumps are 300 lm in diameter and spaced with a pitch of 600 lm. The thickness of the silicon substrate is about 300 lm. The experimental setup was deployed as depicted in Fig. 1, where the samples were inspected in transmission way during the test procedure. The silicon substrate is excited by a laser pulse heating, which is 200 ms in duration and the maximum pulse
Fig. 1. Experimental deployment and solder bumps arrangement of the SFA2.
energy is 240 mJ. Temperature evolutions of the die surface are monitored by the thermal imager and then are processed to identify the defects in the package. 3. Theory analysis and methodology 3.1. Heat conduction analysis Fig. 2 illustrates the structure with single layered chip and substrate, in which the silicon die is attached to the substrate by a solder bumps array. When a thermal excitation is activated, the heat exchange would occur in the package [15]. As known, heat transfer in the solid body is governed by the Fourier diffusion equation, which describes the distribution and variation of temperature in a given region and time,
@T k @2T @2T @2T ¼ þ þ @t qcp @x2 @y2 @z2
! þ
Q
qcp
ð1Þ
where T = T(x, y, z, t) depicts the temperature at the point (x, y, z) and time t, k denotes the thermal conductivity, q is the density and cp the specific heat capacity, and Q is defined as the internal heat generation per unit volume in unit of W/m3. There are no internal heat sources in the package because the chip is not in working condition [16]. We will put emphasis on the investigation of the heat conduction via solder bumps and the estimation of defects influence on thermal behavior in the active infrared inspection. In the transmission way, a heating pulse is imposed on the bottom surface of the substrate, and the thermal front is launched and propagates inside the structure. As the thermal front reaches the top surface of the substrate, it would propagate into the chip via solder bumps, or be hampered by the defects and then flow to adjacent solder bumps, as depicted in Fig. 2. When observed from
Fig. 2. Schematic of heat conduction via solder bumps.
99
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
Temperature evolution 20
28.4
40
28.3
34 Denoising & Smoothing
28.2
80 28.1 100 28 120 27.9
140 160 180 50
100
Temperature/ °C
33
60
32
30
27.8
29
27.7
28
150
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
31
P1(Missing bump) P2(Reference bump) P3(Reference bump) P4(Reference bump)
0
1
2
3
4
5
Time/s
Fig. 3. Thermal image of SFA2 before the thermal excitation (t = 0 s).
Fig. 5. Temperature evolution of four points.
Original thermal image 34
Thermal contrast image with contour
20 40
33.5
P1
P2
P3
P4
60 80 100
33
20
32.5
40
32
60
31.5
80
120 140 160 180 50
100
31
100
30.5
120
150
140 Fig. 4. Original thermal image (t = 0.217 s). 8
160
the die surface, the temperature over the missing bumps will be slightly lower than that of the reference bumps because of the insufficient heat transfer path. However the small temperature difference disappears rapidly since the silicon die has a high diffusivity. Then the temperature begins to drop under the effect of the air convection. The temperature around the missing bump declines more quickly than those of the reference bumps as they are different in the thermal mass. A thermal equilibrium can be reached extremely fast due to the lateral heat conduction in the package, which reduces the thermal contrast between the defects and the reference bump. Additionally, emissivity unevenness and heating non-uniformity are also obstacles for the solder bumps inspection in the active infrared test.
3.2. Self reference technology based on the source distribution image In order to eliminate the influence of emissivity unevenness and heating non-uniformity on defects identification, a method using the self reference technology based on the source distribution image is proposed. Thermal images captured right after the heat pulse would reflect the spatial distribution of the heating source power across the sample surface [17,18]. Several subsequent thermograms are selected and averaged to create the source distribution image (SDI). Averaging is done on a pixel-to-pixel basis, which allows reducing the existence of noise in images. The influence of the heating non-uniformities can be reduced by subtracting the SDI
6
20
40
60
80
100
120
140
2
0
7
180
1
160
5
3 4
180
Fig. 6. Thermal contrast image.
from the original thermal images since higher value will be removed in the region received higher excitation energy. The ‘‘self-referencing’’ is based on Shepard and Ducar’s [19] research work, in which the behavior of a defective pixel is compared with the kernel pixels surrounding it to study the deviation in the logarithmic temperature time history. In this paper thermal behavior of the pixels around the solder bumps are investigated and the self reference technique is proposed based on the fact that the hot spot over the missing bump shrinks more quickly than the reference bumps after the heating pulse. The self reference technique is calculated as following: (1) The SDI is subtracted from the original thermal images, and thermal contrast images are obtained. (2) The thermal contrast values of the hot spots are extracted from the thermal contrast images. (3) The thermal contrast value of the center point in the hot spot is subtracted from the values of the points (i, j) surrounding the hot spot, as depicted by the formula
Cði; j; tÞ ¼ T surr ði; j; tÞ T cent ðtÞ
ð2Þ
100
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
(1)
30 20
20
10
10
0 0.2
Sum of Temper Contra Differ/K
0.4
0.6
0 0.2
0.8
(3)
30
30
20
20
10
10
0 0.2
30
0.4
0.6
30
20
20
10
10
30
0.4
0.6
(7)
30
20
20
10
10
0 0.2
0.4
0.6
0 0.2
0.8
0.6
0.8
0.4
0.6
0.8
0.4
0.6
0.8
0.4
0.6
0.8
(6)
0 0.2
0.8
0.4
(4)
0 0.2
0.8
(5)
0 0.2
(2)
30
(8)
Time/s M1(Missing) R2(Reference)
R1(Reference) R3(Reference)
Fig. 7. Summation of temperature difference along with time.
4. Results and discussions
35 M1(Missing bump) R1(Reference bump) R2(Reference bump) R3(Reference bump)
Sum of temp. differ/K
30
4.1. Experimental results
25 20 15 10 5
1
2
3
4
5
6
7
8
Feature points Fig. 8. Summation of temperature difference.
where (i, j) is the coordinates of the point and t is the time. Thermal contrast value, C(i, j, t) is then used to characterize the thermal behavior of the corresponding solder bump quantitatively. We can apply this approach to overcome the emissivity difference between the metal UBM layer and the silicon material, and reduce the influence of non-uniformity in the delivery of stimulant power.
The test vehicle SFA2 has been inspected to identify the missing solder bump. It was simulated by the laser pulse, and 600 thermograms were obtained at the sampling rate of 60 Hz. Fig. 3 shows the thermal image before the thermal excitation. Due to transparence of the silicon material at the band of wavelength longer than 1.1 lm, the front side of the silicon die, where the UBM layer is located, can be seen from the back side, as shown in Fig. 3. It is almost infeasible to distinguish the missing solder bump from the reference one at the room temperature. The temperature of the UBM layer is slightly higher than that of the gap region as the emissivity differences existed. Fig. 4 shows the original thermal image at 0.217 s. As can be seen, heating non-uniformities dramatically decreased the temperature contrast between the defective solder bump and the reference one, which may hamper the defects recognition. Temperature evolution of each pixel was extracted from the thermograms sequences. Fig. 5 shows the temperature curves of four points, P1, P2, P3 and P4 (labeled in Fig. 4), where P1 is located in the defect area (missing solder bump) and the others are in sound area (with solder bumps). All the curves can be divided into three parts. In the first part, the temperature of every point rise sharply during the heating pulse. After t = 0.2 s it increases slowly
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
and reaches the maximum at t = 0.8 s.Then it decays exponentially in the third part. As depicted in Fig. 5, the temperature of the defect point P1 is slightly lower than those of the reference points in the second part. For the temperature curve of P4, the peak noise appears at 0.2 s which is affected by the random noise and heating non-uniformities. Therefore it is critical to eliminate the influence of thermal noise and heating non-uniformities. 4.2. Defects identification To improve the signal to noise ratio, each thermogram was firstly input to an improved median filter with a 5 5 mask to remove the peak noise while preserving the edge information. Then the temperature evolution of each pixel was extracted and processed using the moving average operation with the span of 5, which can eliminate the impact of periodic and random noise on the temperature. Fig. 5 also shows the smoothed temperature evolution curve of P1, P2, P3 and P4 during the time period of 0–0.6 s. Although the peak noise is suppressed to some extent, heating non-uniformities is still a serious problem to be solved. In order to avoid the inconsistent comparison between the thermal values of the solder bumps located in different heat power region. Three thermal images captured right after the heat pulse were selected to create the SDI. The SDI was then subtracted from the consequent thermal images, and thermal contrast images were obtained. Fig. 6 shows the thermal contrast image derived from Fig. 4, in which iso-value contours are also created to indicate the temperature difference. Obviously it is less ‘‘warm’’ in the region of missing bump than that of the reference bumps in the heat power iso-value contour. Solder bumps in less heat power region get lower temperatures, and vice versa. For example the temperatures of the solder bumps in the fourth line are lower than that of the solder bumps in the first line.
The self reference technology based on the SDI is explored to characterize the defects quantitatively. As shown in Fig. 6, eight points on the edge and one in the center of each hot spot have been selected as the feature points for the corresponding solder bump. Each point contains five pixels and the temperature of the point is derived by averaging the pixels values. The thermal difference between each edge point and the central point is calculated by
ði ¼ 1; 2; . . . ; 8Þ
ð3Þ
Then the summation of the C di ðtÞ values during the period [t1, t2] is derived by
C dis ðtÞ ¼
t2 t2 X X C dis ðtÞ ¼ ðT i ðtÞ T 0 ðtÞÞ ði ¼ 1; 2; . . . ; 8Þ t¼t1
5. Conclusions In this paper, we proposed a method using the self reference technology based on the source distribution images to eliminate the impacts of emissivity unevenness and heating non-uniformities. For validation of the approach, a model of heat conduction via solder bumps was constructed and the chip SFA2 with a missing solder bump was chosen as the test vehicle. Thermal images were recorded and processed using the improved median filter and moving average operation to remove the peak noise and the periodic noise. Three thermograms captured right after the heat pulse were averaged to create the SDI. Then the SDI was subtracted from the original thermal images, and we got the thermal contrast images, in which the missing bump was identified. The self reference technology was adopted to characterize the thermal behaviors of the solder bumps quantitatively. Eight points on the edge of each hot spot were selected as the feature points for the corresponding bump, and thermal difference between the feature points and the central point are calculated. For the missing bump the summation of thermal difference on every point is no more than 14 K, while for the reference bumps it is always larger than 14 K. The results demonstrated that it is effective using the self reference technology to eliminate the influence of emissivity unevenness and heating non-uniformities on defects identification in active infrared testing. Acknowledgments This research work is supported by National Natural Science Foundation of China (Grant Nos. 51305179 and 51305177) and Natural Science Foundation of Jiangsu Higher Education Institutions (Grant No. 13KJB510009). References
4.3. Quantity of the defects
C di ðtÞ ¼ T i ðtÞ T 0 ðtÞ
101
ð4Þ
t¼t1
According to Eq. (4), we get the Cdis(t) values along with the time in the period of 0.2–1 s, as shown in Fig. 7, in which M1 denotes the missing bump, while R1, R2 and R3 are the reference bumps. Fig. 7(1)–(8) are corresponding to the eight points around the hot spot respectively. As can be seen, for every point, the summation of temperature difference for the missing bump is smaller than those of the reference bumps. To distinguish the defect more explicitly, the Cdis(t) value at t = 1 s of the eight points is shown in Fig. 8. For the missing bump, the summation of the temperature difference on every point is no more than 14 K, while for the reference bumps it is always greater than 14 K, by which the missing solder bumps can be discriminated quantitatively from the reference bumps. Therefore it is effective using the self reference technology to eliminate the influence of heating non-uniformities on defects identification.
[1] R.T. Rao, Fundamentals of Microsystems Packaging, McGraw-Hill, New York, 2001. [2] M. Pacheco, Z. Wang, L. Skoglund, et al., Advanced fault isolation and failure analysis techniques for future package technologies, Intel Technol. J. 9 (4) (2005) 337–352. [3] L. Wang, H. Ding, Y. Xiong, An overview of non-destructive inspection in flip chip packaging, Mach. Electron. 5 (2004) 45–49. [4] H.N. Yen, D.M. Tsai, S.K. Feng, Full-field 3-D flip-chip solder bumps measurement using DLP-based phase shifting technique, IEEE Trans. Adv. Packag. 31 (4) (2008) 830–840. [5] T.D. Moore, D. Vanderstraeten, P.M. Forssell, Three-dimensional X-ray laminography as a tool for detection and characterization of BGA package defects, IEEE Trans. Compon. Packag. Technol. 25 (2) (2002) 224–229. [6] R. Kovacs, X-Ray inspection of microwire bonds, in: Proceedings of the 28th International Spring Seminar on Electronics Technology: Meeting the Challenges of Electronics Technology, Austria (Wiener Neustadt), Institute of Electrical and Electronics Engineers Computer Society, 2005, pp. 448–451. [7] E. Martin, C. Larato, A.C. Ment, et al., Detection of delaminations in subwavelength thick multi-layered packages from the local temporal coherence of ultrasonic signals, NDT&E Int. 41 (2008) 280–291. [8] S. Brand, K. Raum, Signal analysis in scanning acoustic microscopy for nondestructive assessment of connective defects in flip-chip BGA devices, IEEE Ultrason. Symp. (2007) 817–820. [9] X. Maldague, Theory and Practice of IR Technology for Nondestructive Testing, first ed., Wiley-Interscience, New York, 2001. [10] S. Liu, C. Ume, Digital signal processing in a novel flip chip solder joint defects inspection system, ASME J. Electron. Packag. 125 (2003) 39–43. [11] J. Yang, C. Ume, L. Zhang, Defect detection of flip chip solder bumps with wavelet analysis of laser ultrasound signals, IEEE Trans. Adv. Packag. 33 (1) (2010) 19–29. [12] T.C. Chai, B.S. Wong, W.M. Bai, A. Trigg, Y. Lam, A novel defect detection technique using active transient thermography for high density Package and interconnections, in: Electronic Components and Technology Conference, 2003, pp. 920–925. [13] J.C. Lu, A. Trigg, J.H. Wu, T.C. Chai, Detecting underfill delamination and cracks in flip chip on board assemblies using infrared microscopy, Int. J. Microelectron. Electron. Packag. 21 (3) (1998) 231–235. [14] X. Lu, G. Liao, Z. Zha, Q. Xia, T. Shi, A novel approach for flip chip solder joint inspection based on pulsed phase thermography, NDT&E Int. 44 (2011) 484– 489.
102
X. Lu et al. / Infrared Physics & Technology 63 (2014) 97–102
[15] R. Kandasamy, A.S. Mujumdar, Thermal analysis of a flip chip ceramic ball grid array (CBGA) package, Microelectron. Reliab. 48 (2008) 261–273. [16] X. Lu, T. Shi, Q. Xia, G. Liao, Thermal conduction analysis and characterization of solder bumps in flip chip package, Appl. Therm. Eng. 36 (2012) 181–187. [17] M. Susa, X. Maldague, I. Boras, Improved method for absolute thermal contrast evaluation using source distribution image (SDI), Infrared Phys. Technol. 53 (2010) 197–203.
[18] M. Susa, X. Maldague, I. Boras, Application of the source distribution image (SDI) procedure for porosity detection in honeycomb structures, in: 10th International Conference on Quantitative InfraRed Thermography, Quebec, 2010. [19] S. Shepard, R. Ducar, Quantitative infrared defect detection in composite aerospace structures, in: 45th International SAMPE Symposium, 2000.