EVALUATION OF DIFFERENT TEST TECHNIQUES TO ASSESS THE EFFECT OF ENVIRONMENTAL EXPOSURE ON THE ASSEMBLY OF MCGA PACKAGES FOR PRINTED CIRCUIT BOARDS

Transcription

1 EVALUATION OF DIFFERENT TEST TECHNIQUES TO ASSESS THE EFFECT OF ENVIRONMENTAL EXPOSURE ON THE ASSEMBLY OF MCGA PACKAGES FOR PRINTED CIRCUIT BOARDS Gianni Corocher (1),(2), Carole Villette (1), Antonius de Rooij (1), Mark A. Baker (2) (1) ESA-ESTEC, Materials Techniology Section, P.O. Box 299, 2200 AG Noordijk ZH, The Netherlands, Tel.: +31 (0) , (2) Faculty of Engineering and Physical Sciences, University of Surrey, Guildford, Surrey,GU2 7XH, United Kingdom ABSTRACT The use of area grid array devices for space application requires the assessment of the reliability of their assembly hen exposed to environmental testing. Verification of the assembly of surface mount devices (SMDs) is done according to the ECSS-Q-ST-70-38C [1] hich requires microsectioning of the sample representative of the assembly after specific vibration and thermal cycling tests. For area array type devices the ECSS-Q-ST-70-38C requires a dye penetrant pull test in addition to microsectioning. Here a ne method based on scanning acoustic microscopy (C-SAM) is explored and compared ith serial microsectioning and the dye penetrant pull test. The three different techniques have been applied to multilayer ceramic grid area array (MCGA) type of devices. The results sho the limits of the pull test method and indicate that the C-SAM inspection can provide more reliable results and a comparatively rapid sample preparation procedure. 1. INTRODUCTION Space activities sho an increasing interest in the use of area array types of electronic devices. These devices allo an increase in the surface density of connections ith respect to traditional surface mounted devices (SMD), leading to more densely populated and more complex electronic equipment. The reliability of SMD assembly is evaluated according to the standard ECSS- Q-ST-70-38C. As detailed in [2] this standard has been created for the traditional type of SMDs i.e. chip resistor and flat packs ith leads along the periphery. The test sequence used on these types of components consists of submitting representative assembled devices on PCBs to a vibration test and 500 thermal cycles (-55/+100 C -15 minutes dell time) to ensure their suitability in sustaining the cumulative ground testing, launch and thermal stresses during in orbit mission. The samples are then microsectioned to verify that the size of the cracks in the solder joints being assessed comply ith the criteria given in [1]. ECSS-Q-ST-70-38C has a dedicated sub-clause concerning the area array type of devices here to possible test methods are considered to assess the damage induced during environmental testing. The first test method is the dye penetrant pull test hich has been developed as described in [3]. This test method consists of soaking the assembly in a dye penetrant after environmental testing. The dye penetrant is then left to dry and the device pulled from the printed circuit board (PCB). If cracks ere present before the pull test, their surface appears coloured by the dye penetrant. The extension of the crack is measured. The maximum alloed crack extension on a single joint is 25% of the original cross section of the column. The second method is microsectioning as per the traditional type of SMDs. An alternative test method based on scanning acoustic microscopy (SAM) is considered here. This method has been applied to an assembled multi-ceramic column grid array (MCGA). The to techniques described above are compared ith an appositely developed reference method based on serial microsectioning MCGA ASSEMBLIES The multilayer ceramic grid array (MCGA) is an area grid array package hich has a solder column interposer (SCI) hich interfaces ith the PCB (Fig. 1 and 2). The SCI is a thin ceramic plate onto hich columns of Pb90Sn10 are mounted, equal in number to the I/O connections of the package. The connection of the SCI to the ceramic land grid array package (CLGA) is made using a eutectic solder (Sn63Pb37). The MCGA is provided to the users ith the SCI already assembled on the CLGA. The connection of the MCGA to the printed circuit board is made via solder joints by the users. This connection is again made using the eutectic solder (Sn63Pb37) by means of different technologies: e.g. hot air reflo, vapour phase. At ESTEC, microsections taken of different assemblies of MCGA packages have shon that the cracks nucleated during thermal cycling are generally developed at the SCI end of the columns. Although possible, the formation of cracks at the solder joint of the PCB to the column is not frequent. For this reason the development of the methods described in this paper have been focused on the SCI side of the columns.

2 SCI CLGA In this case the joints beteen the solder ball and the package ere inspected using a pulse echo from the top of the package. Defects at the solder/ceramic interface ere detected using a 15MHz transducer. CLGA Fig. 1. MCGA package Pb90Sn10 columns 2. EXPERIMENTAL 2.1. Dye penetrant pull test The dye penetrant pull test has been performed folloing the procedure described in [1]. The test has been performed on the MCGA device assembly on polyimide after the environmental test. An identical sample has been submitted to the serial microsectioning test for comparison. The extension of the cracked surface on the interposer area of the sample has been measured on the optical microscopy images using Clemex image analysis softare. SCI PCB Fig. 2. Assembly of a MCGA device on a PCB C-SAM Pb90Sn10 columns C scan- Scanning Acoustic Microscopy (C-SAM) uses an ultrasonic test system equipped ith focused transducers and a high frequency range (from 10 MHz to > 1GHz) to obtain high resolution images of features located under the surface of a sample [4]. C-SAM is a pulse echo technique hich produces images by scanning a transducer over the surface of the sample. The transducer uses a spherical lens to produce a focused beam of ultrasound. The most used frequencies in C-SAM are in the range of MHz. Semmens and Kessler [5], [6] have proved that this method can be applied to flip chip solder bumps and chip scale packaging. These devices have connections similar to those of a BGA component to a PCB but on a much smaller scale. The technique developed by Semmens and Kessler consists of examining the solder joints from the top of the device. The results obtained sho that it is possible to detect voids and complete detachment/fracture of solder joint using a 100 MHz transducer. Better resolution as obtained using a 230MHz transducer but after thinning the top of the die. This as needed due to the smaller penetration depth of the high frequency transducers. Ousten et al. [7] have applied C-SAM to the inspection of BGA assemblies to detect failures in the solder joints Serial microsectioning Serial microsectioning enables the structure of a certain volume of material to be reconstructed using optical micrograph pictures of sections taken from knon depths. The method consists of taking pictures in the same location of a micrographic sample but from different section planes (exposed through mechanical polishing). Knoing the depth to hich beteen each picture corresponds, it is possible to reconstruct ho a feature (microstructure, crack, void etc.) evolves ithin the sample bulk. To do this it is necessary to have a reference point that can be easily found after each polishing step. In the case of the MCGA assembly, serial microsectioning is aimed at reconstructing and measuring the surface of the crack developed in the interposer area of the device. This is done using the column as a reference to determine the distance beteen section planes and also to align the data acquired at each step. Fig.3 shos the concept used to reconstruct the plan vie of a crack in a column using the column idth as a reference. Fig.3A shos a section of a column on a plane perpendicular to the column axis. Fig.3B shos a microsection through a column on a plane parallel to the column axis. The to pictures have been taken at the same magnification, and aligned. If the idth of the column measured in Fig 3B is projected onto Fig. 3A to possible positions of the section in Fig. 3B are identified in Fig.3A. Knoing the radius of the column it is possible to determine the distance h of the section from the centre of the column by applying Pythagoras s theorem (Eq.1). Since the starting point and direction of polishing are knon, it is then possible to identify the position of the section in Fig. 3B on the plan vie of the sample (Fig. 3A).

3 h = r (1) x a = a (2) 2 The diameter of a column ithin a device can vary beteen 880 and 930 µm. The diameter is important in the construction of the map and the value used for each map is the maximum column idth measured in the sections. The position h of each section along the column is calculated using Eq.2. Cracks present at the interposer side nucleate at the edges of the column (Fig. 4). The extension of each crack (left and right side of the column in the section) can be measured. The positions of the tip of the cracks (x a and x b ) are calculated as the distance from the midpoint of the section using Eqs. 2 and 3. The examination of a series of microsection planes of a column, by measuring the idth of the column and length of cracks in each section, allos a map shoing the plan vie of the crack surface in the column to be built up. This is accomplished using an Excel orksheet here all the measurements taken at each section of a column are processed. On average, for each column 22 sections are examined. x b = b 2 The position of the crack is then plotted on an XY spatial chart ith the centre of the column corresponding to the axis origin. A plan vie of the crack is obtained by joining the points identifying the tip of cracks A and B on different sections (Fig. 4). The area enclosed by this line, representing the portion of column not cracked, is then calculated by graphical integration. This area subtracted from the area of the column gives the extension of the crack hich can then be expressed as percentage of the column surface. The application of this method has required the development of a reliable and repeatable polishing procedure hich does not induce artefacts or damage to the sample. (3) r h A C B Figure 3. Correlation of a picture of a microsection on a plane parallel to axis of the column (B) ith the position of the section plane on the plan vie of the column (A).

4 500 x a x b a b a b Figure 4. Measurements performed for each column at each step of serial microsectioning. Example of a map constructed using the measurements performed on serial microsections. The blue circle is the perimeter of the column, the purple line encloses the non cracked portion of the column. The purple line is constructed joining the points acquired on each microsection for each side of the crack (A and B). The map is made of four curves (one per quadrant) identifiable by the different markers for each individual measurement 2.3. C-SAM The MCGA devices investigated have a sealed package. The presence of an empty cavity under the metallic lead of the package makes the C-SAM inspection impossible from the top of the device. The approach explored here is to perform the inspection from the PCB side after having ground aay the PCB and the connections masking the interface of interest. Grinding as stopped at a distance of about 0.5 mm from the interposer. The method requires that the assembly is embedded using epoxy resin for metallographic preparation. The embedding process is done under vacuum to limit the entrapment of air bubbles in beteen the columns. The sample is then ground in order to remove the PCB and the part of the columns shielding the SCI. The portion of the assembly removed by grinding is schematically shon on a section in Fig. 5. The inspection has been performed ith a Sonoscan D6000 scanning acoustic microscope using a 50MHz transducer ith aperture F# 2. The sample has been scanned setting the electronic gate to include the hole thickness of the interposer. The C-SAM images have been further analysed using Clemex image analysis softare. 3. RESULTS 3.1 Dye penetrant test versus serial microsectioning The sample submitted to the dye penetrant pull test shoed several failures at the interposer side of the columns. The crack extension measured on columns located along the periphery of the package range from 78% to 90% of the cross sectional area of the column using this dye penetrant method. The crack extensions measured on an identical assembly submitted to serial microsectioning range beteen 55% and 66%. Figure 5. The concept of the C-SAM inspection of the interposer area of the columns, vieed as a section of a device. The diagonal pattern indicates the portion of the sample to be ground before C-SAM analysis. A visual inspection of these failure surfaces shos severe localised necking (Fig. 6). This necking affects

5 the measurements of the size of the crack surface and explains the difference beteen the dye penetrant and serial microsectioning results a) Figure 6. Detail of the failure of a column of a sample submitted to dye penetrant pull test. The failure occurred at the interposer side of the column. 3.2 C-SAM versus serial microsectioning The C-SAM procedure previously described has been applied, at first, on a fe columns of an MCGA assembly. Images ere acquired at different gate positions and ith different types of colour mapping before submitting the columns to serial microsectioning. Comparison of the C-SAM images ith the maps produced by serial microsectioning has enabled the false colour map to be chosen for visualization of the C-SAM images. In these images the non-cracked portion of the column has a lo reflection intensity and is visualised in the red to orange colour range. The crack surface is an interface hich generates a much higher echo than the rest of the column and is visualised in the yello to hite colour range. The surface of the interposer being an interface ith respect to the embedding resin is also producing a large echo and is visualised in hite. Consequentially the images do not sho the perimeter of the column but define the contour of the non-cracked part of the column. The validity of this interpretation has been proven by the comparison beteen C-SAM and serial microsectioning (Fig. 7). The method has then been validated on a larger sample of columns from a different assembly. The obtained image has been analysed using the Clemex image analysis softare to measure the extension of the noncracked portion of the column. For each column, the measurement has been performed tice ith different criteria to define the boundary of the non-cracked surface of the column. b) c) Figure 7. Comparison beteen a microsection (b) used for the construction of the serial microsection map shon in (a) and the C-SAM picture of the same column (c). The column circumference is represented in the C-SAM vie by the blue circle. The light blue line locates the microsection a) in the C-SAM picture. The red lines connecting the to pictures sho that the crack tips visible in the microsection are located in the yello portion of the C-SAM image.

6 In one case the boundary has been considered to be at the transition from red to yello; in the second at the transition from yello to green (Fig. 8). From these measurements the percentage of crack for each column has been determined on the basis of the nominal diameter of the column (0.9mm). The C-SAM picture acquired on this larger sample is shon in Fig. 9. Some columns appears linked in the image. This is due to the presence of air bubbles in the resin. The columns affected by this have not been analysed. To ros of columns of this assembly have been then serial microsectioned. The comparison of the crack percentage measured on these to ros (Table 1) shos the folloing: The values measured on the C-SAM images using to different criteria to define the crack boundary differ on average by 7.3 % points. The average difference beteen the serial microsection values and the C-SAM values is 5.0 % points in the case of the red to yello transition and 3.9 % points for the yello to green transition. Figure 8. Vie of the C-SAM image of an individual column. The to black lines encloses are the to different boundaries of the non-cracked area of the column: red to yello transition for the continuous line, yello to green for the dotted line. Table 1: Comparison of the crack percentage measured on C-SAM images ith those measured by serial microsectioning. N/A 54.0 N/A N/A 48.0 N/A N/A N/A N/A N/A N/A N/A Legend: x % of crack surface measured on C-SAM pictures considered as being not cracked (black to orange colour range). y % of crack surface measured on C-SAM pictures considered considered as being not cracked (black to yello colour range). z % of crack surface measured on the serial microsectioning maps. Figure 9. C-SAM image of the interposer area of a device. The links beteen some of the columns are due to air bubbles entrapped in the embedding resin.

7 4. DISCUSSION The serial microsectioning technique described here allos the detection of cracks in the interposer area of the column. From the resulting maps, the crack extension, as percentage of the column cross section, can be calculated. This technique is considered to give the most reliable result and hence has been used as reference for comparison ith the other methods. Microsectioning requires a high number of polishing steps and is therefore time consuming. Moreover the variety of materials in the sample requires the development of a dedicated and repeatable polishing method to obtain reliable results from each section plane. The large difference beteen the crack extension measured on the sample submitted to dye penetrant test ith respect to the serial microsection is due to the extensive necking of the failure surface. There is very good agreement ith the serial microsectioning maps and C-SAM images regarding the shape of the non-cracked portion of the column. The small differences beteen the to methods in the cracks surface area values are considered to be due to the folloing: The determination of the residual area of the column from C-SAM is performed using image analysis softare by selecting the colour corresponding to the crack tip. The values in Table 1 and the image in Fig. 8 sho that the effect of a different choice of the colour (reflected signal) to define the crack tip can affect the final result. As shon in Fig. 8 the difference beteen the to cases in terms of diameter of the selected area is minimal. The difference among surface enclosed in these to lines is nevertheless very important as shon in Table 1. A difference in the initial setting of the gain of the transducer in the acquisition of the images leads to differences in the acquired images. If the gain value is too high, the reflection from the residual column is increased and the non-cracked portion of the column ill be displayed in yello. All of the image ill be shifted toard colours indicating higher reflection. On the contrary if the gain value is too lo, the central part of the column is shon in dark red or black and the apparent extension of the crack in the image decreases. 5. CONCLUSIONS The dye penetrant pull test required by [1] has been found to have limitations in accurately measure the structural damage in the columns of MCGA devices. The ne C-SAM approach has been shon to yield a reliable measure of structural damage in the columns of MCGA devices. The removal of the printed circuit board and part of the column by grinding allos the interposer area of the column to be examined ith a resolution sufficient to detect and measure the extension of cracks. C-SAM allos the measurement of the crack extension at the interposer end of the columns. The data obtained by C-SAM shos a 3.9 % average difference to that determined using the reference microsectioning method. In comparison, the dye penetrant pull test shos a much larger difference. C-SAM allos an inspection of the hole assembly in much shorter time than serial microsectioning. C-SAM and serial microsectioning can be applied as valuable tools in the study of the reliability and the failure mechanisms of a MCGA assemblies. C- SAM is the technique exhibiting the best potential as routine method for assessing the reliability of MCGA packages. Further ork is needed to develop a calibration procedure in order to improve the precision of the measurement of crack extension by C-SAM. REFERENCES 1. ECSS-Q-ST-70-38C. High-reliability soldering for surface mount and mixed technology. ECSS secretariat ESA-ESTEC requirements and standards division Noordijk The Netherlands, Dunn B.D. Workmanship standards and their application on ESA projects. Soldering & Surface Mount Technology. Vol.20 (4), 37-44, Tegehall P-E, and Dunn B.D., Assesment of the Reliability of Ball and Column Grid Array packages fo Space Appllications, ESA STM-266, ESA Paris, AA.VV. Metals Handbook Volume 17: Nondestructive evaluation and quality control. ASM International. 5. Semmens Janet E. and Kessler Larence W.. Private communication. High frequency acoustic micro imaging for process monitoring and quality control in flip chip underfill assembly. Sonoscan Inc. Elk Grove Village IL (USA). 6. Semmens Janet E. and Kessler Larence W.. Private communication. Evaluation of chip scale packaging using acousting microimaging: an overvie of applications, limitations and directions for future developments. Sonoscan Inc. Elk Grove Village IL (USA). 7.Ousten Y., Mejdi S., Fenech A., Deletage J. Y., Bechou L., Perichaud M.G., Danto Y.. The use of impendence spectroscopy, SEM and SAM imaging for the detection of failure in SMT assemblies. Microelectronic reliability 38, , 1998.