1 Introduction

The ability to evaluate the performance of various components used in the manufacturing of PCBA (Printed Circuit Board Assembly) depends to a large extent on the possibility of evaluating the performance of these same PCBA at the output of each production stage, e.g., reflow soldering. One of the parameters that must be monitored during PCB production is strain, which must be kept within the allowable limits. Currently, strain measurements are performed through conventional electric strain gauges (SG) applied to the surface of the PCB [1, 2]. Due to the mode of operation of these devices, and the number of required monitoring points, the placement of sensors and the measurement process become complex. Firstly, when placed over a surface, the electrical resistivity of the SG varies with its temperature, introducing deviations, e.g., in warpage measurements. And secondary, the considerable dimension of the thermal ovens implies the use of an extensive cabling, and the high temperatures and machinery movements may induce inaccuracy or structural damage to the sensors. Given the current level of component miniaturization and increasingly complexity of most PCBA, driven by the high integration level, an alternative technology for strain measurements on PCBA becomes necessary. Considering that for each SG, two wires are at least required, one can easily imagine the large amount of cabling one single PCB can end up with, especially if one considers their limited area. Moreover, associated to these SG, we have the long cabling required to connect them to the data acquisition system.

In recent years, strain measurements using optical fibers (OF) have received great attention, and several sensor architectures have already been developed and studied for various applications related to medicine, architecture, aeronautics, electronics industry, among others. [3, 4]. OF measurement systems have gained great importance [5]. In addition, its small size and multiplexing capability allow the scan of a vast measurement area, using just a single fiber and its respective interrogation/acquisition system [6,7,8]. Distributed sensing has several advantages, such as, high spatial resolution, high number of detection points per OF, monitoring of punctual locations and its specific selection, and finally, a commercial OF can be used as a sensor, and therefore a low cost and space efficient detection solution can be implemented [9]. Also, strain and temperature distributed sensing has growing interests due to their cost–benefit and their fully distributed detection capacity [10]. The main advantages of using such OF technology is clear and includes the possibility to create a strain map along the entire PCB (high number measurement points), the easing of the fiber placement and routing when compared with electrical based strain gauges and its cabling.

For the assembly of OFS on the PCB, an insightful selection of the adhesive is necessary, as currently there are several alternatives. The adhesive selection process can be difficult, since many factors must be considered and there is no universal adhesive that meets all application´s requirements. The adhesive selection mainly depends upon [11]:

  • The type and nature of the substrates to be bonded.

  • The practical and available adhesive curing methods.

  • The environments expected during service.

The most suitable bonding materials for instrumenting the PCB surface with OF are based on cyanoacrylates, methacrylate, silicones, and epoxies [12]. Cyanoacrylates are characterized by having strong adhesion strength, but structural bonding problems to glass. However, they stand out for having a faster adhesion time (few seconds). Epoxy adhesives (either one or two components) are characterized by having a low retraction after curing, but the curing time is a disadvantage because it is longer than other adhesives [12]. The methacrylate adhesives are mostly of UV curing nature and present a low shrinkage after the curing process. High temperature adhesives are generally characterized by a rigid polymeric structure, high softening temperatures, and stable chemical groups [13]. Epoxy-phenolic, bismaleimide, polyimide, and polybenzimidazole adhesives can withstand long-term service temperatures higher than 175 °C. However, modified epoxy resins and even certain cyanoacrylate adhesives are resistant to higher temperature in the short term. Silicone adhesives also have excellent performances but exhibit low cut resistance [13]. In all cases, the adhesives should guarantee an efficient bond between OF and the PCB and transmit the strain of the PCB to the OF as accurately as possible. It is therefore important to evaluate the influence of the thickness and the type of bonding material with the lowest interference in the strain measurements.

In a previous work [9], a new approach for strain measurements on PCB was adopted and benchmarked against conventional foil electrical SG in order to reduce the installation complexity and the evaluation process without compromising the measurement quality. This solution is based on optical fiber (OF) distributed strain sensing using the optical frequency domain reflectometry (OFDR) technique, which has high sensitivity and inherent multiplexing capacity. A PCB was instrumented with both sensor technologies (Fig. 1) and subjected to an In-Circuit Test machine. The strain results showed a match between both technologies, with a maximum error of 3.5%, a high precision level for multiple tests, and a high sensitivity for both high and low induced strains.

Fig. 1
figure 1

Example of a PCB from BOSCH Car Multimedia prepared for strain measurements with electrical SG and cabling (in red), and with a OF (highlighted in yellow). The numbers are points of interest for strain measurements (Color figure online)

Full size image

For an effective strain measurement of PCB using OF, this work investigates the effect of different adhesives and its mode of application (e.g., layer thickness). Two main classes of adhesives are studied for different uses of OF for strain measurements on PCB: (i) environmental temperature adhesives for measurements performed at room temperature processes (e.g., assembly components); and (ii) high temperature adhesives (up to 260 °C) for measurements during thermal processes (e.g., reflow soldering). The mechanical properties of the adhesive and its effectiveness on OF strain measurements on PCB are evaluated.

2 Materials and methods

2.1 Materials

An Ø155 µm OF with polyimide (PI) coating was used in this work. A high-definition OF detection system ODiSI-B (LUNA inc®, USA) was used for data acquisition. This is an optical backscattered reflectometry system based on the Rayleigh OFDR. Luna ODiSI-B equipment was operated at 250 Hz, with a spatial resolution of 2.56 mm, with an accuracy of 1.25 μm/m. The process of installing the OF strain sensor to the PCB is performed manually, similarly to the process of installing SG. Table 1 lists the different adhesives considered in this study. The OF was glued to the top layer (solder mask) of a PCB (100 × 25 mm and 1.6 mm thickness).

Table 1 Main properties of bonding materials used
Full size table

2.2 Adhesive samples preparation

The following procedure was adopted for preparation of the adhesive samples, ensuring a minimum and constant thickness of the adhesive:

  1. 1.

    Cleaning the PCB and OF with isopropyl alcohol.

  2. 2.

    Placing the OF in the desired position and fix it with Kapton tape/dots. The curvature radius of the optical fiber must be greater than 5 mm.

  3. 3.

    Poring the adhesive around the OF; the excess of the adhesive was removed using a cotton swab.

  4. 4.

    Place the samples in a convective oven, with the recommended cure temperature and time (Table 1).

The thickness of the adhesive, i.e., the amount of adhesive between the fiber and the surface, must be minimized and a thin bond line with enough resin must be applied for a correct bonding process [14, 15].

2.3 Adhesive thickness measurements

To evaluate shapes and thicknesses of the adhesives, a reflection microscope BH-2 (Olympus, Japan) was used. Sample preparation involved the immersion of a bonded OF section into a support resin and then polishing it with sandpaper with different granulometry (starting from 120 to 4000) to obtain a smooth surface to be observed under the microscope.

2.4 Mechanical properties

The mechanical properties of room and high temperature adhesives were assessed according to ISO 4587:2003, consisting of testing bonded specimens (6 samples each) with the same thickness (of 0.2 mm) with different adhesives materials (Fig. 2). The tensile lap-shear strength is calculated from the maximum load during the test carried until the rupture of the specimen, at a crosshead velocity of 2 mm/min to 50 mm/min, so that the average joint will be broken in a period of 65 s ± 20 s, at RH of 50 ± 2% at 23 °C ± 1 °C.

Fig. 2
figure 2

Specimen prepared for shear test: sample according to ISO 4587:2003 (dimensions in mm) and final bonded sample under test

Full size image

The bonding failure pattern observation is complementary to the mechanical properties and was done in accordance with ISO 10365. There are different types of failure that can occur, either on the substrate or on the adhesive under some specifications (Fig. 3). If occurring on the substrate in one or both adherends and outside the joint, the failure is designated as substrate failure (SF). A cohesive failure occurs (CF) if it is in adherend joint, and a delamination failure occurs if the substrate delamination happens (adhesive failure, AF). A mixed adhesive and cohesive failure (MACF) can also occur.

Fig. 3
figure 3

Failure modes of an adhesive joint (references on text)

Full size image

2.5 OF signal quality with room temperature adhesives

The signal quality of an OF assembled on a PCB (300 mm length, 25 mm width, and 1.6 mm thickness) with different adhesives was evaluated with the ODiSI-B OF interrogator under tensile and bending tests. The OF was bonded to the PCB with 3 different adhesives with segments 5 mm apart (Fig. 4). For each OF segment, a different adhesive was used. Close distance between segments were kept guaranteeing the 3 segments suffer the same mechanical conditions. The tensile test based on ISO 527:2012 (5 samples) was performed at a crosshead velocity of 0.8 mm/min and maximum extension of 0.4 mm. The bending test was based on ISO 178:2010 (5 samples) and was conducted with a span length of 200 mm, at a crosshead velocity of 20 mm/min and maximum displacement of 8 mm.

Fig. 4
figure 4

PCB sample instrumented with OF and 3 adhesives

Full size image

2.6 OF signal with different thicknesses adhesive

The study of the effect of adhesive thickness on the strain measurements was performed using M-Bond 200. The OF was bonded to the PCB in 8 straight lines with a distance of 2.5 mm between them, as depicted in Fig. 5. Four of the OF segments were bonded on the PCB and the excess of material removed with the help of a swab. The other four OF segments were bonded with a controlled bonding thickness, namely, of 50 µm, 100 µm, 150 µm, and 200 µm. Standard and controlled bonded thickness segments are intercalated. Due to PCB fabrication tolerances and inherent warpage, a strain gradient may be induced on the PCB resulting in reading errors between OF segments. To compensate for this error, a reference sensor (standard bonding segment) was bonded near each test sensor (bonding layer segment).

Fig. 5
figure 5

PCB sample instrumented with OF with different bonding layers thickness

Full size image

The tensile test based on the ISO 527:2012 (5 samples) was conducted at a crosshead speed of 0.8 mm/min and maximum displacement of 0.4 mm, and the bending test based on the ISO 178:2010 (5 samples) was conducted with a span length of 200 mm, at a crosshead speed of 20 mm/min, and maximum displacement of 8 mm.

2.7 OF signal quality with high temperature adhesives

The high temperature adhesive with the best properties from previous tests was selected to assess the optical signal quality under strain. The OF was mounted on a PCB sample 150 mm long, 25 mm wide, and 1.6 mm thick. The assembly of the samples (Fig. 6) is similar to the case of the room temperature adhesives, previously presented.

Fig. 6
figure 6

Instrumented PCB sample using high temperature adhesives

Full size image

To characterize the contribution of thermo-induced strain 3-point flexural tests were carried out inside a temperature chamber at temperatures in discrete levels of 23, 100, and 250 °C. The bending test, following ISO 178:2010, was performed with a span length of 100 mm, at a crosshead velocity of 2 mm/min, and maximum displacement of 2 mm, within each temperature level. Strains were measured using ODiSI-B OF interrogator.

3 Results discussion

3.1 Mechanical properties

Figure 7 shows representative curves of the shear stress vs strain for tested adhesives, and Table 2, the average values of the maximum shear stress, elastic modulus, and strain at maximum stress. The maximum stress corresponds, in general, to the breakdown value. The 60 s Universal Glue Loctite® shows the lowest shear stress and strain. The Permabond® 630 adhesive has the lowest elastic modulus, and a high strain at break with low stress. However, this adhesive has the particularity of presenting UV curing and short cure times.

Fig. 7
figure 7

Representative curve of the shear stress vs. strain for the different adhesives

Full size image
Table 2 Maximum shear stresses, elastic moduli, and strains at maximum stress for the different adhesives
Full size table

Conversely, Loctite® 495 has the best mechanical properties, with a high elastic modulus, high maximum shear stress, and the highest deformation capability. The M-Bond 200 adhesive has similar properties, but with lower shear strain at break. Figure 8 shows the appearance of the samples after the tests.

Fig. 8
figure 8

Failure surfaces of the tested adhesive samples: a 60 s Universal glue from Loctite®; b Loctite® 495; c Araldite® Instant 90 s; d M-Bond 200

Full size image

The 60 s Universal Glue shows a uniform fracture of the adhesive layer, namely a cohesion failure. M-Bond 200 also has fracture on the adhesive layer; however, it presents as MACF, due to the presentation of a more irregular fracture. On the other hand, Loctite® 495 presents a failure in the adherend, more precisely, between the solder mask and FR4 joint. The Permabond® 630 adhesive shows a failure mode between the adhesive and the surface of the PCB (AF). The Araldite® adhesive shows a failure mode identical to The Permabond® 630 adhesive (AF). Associating the maximum stress to the respective type of failure, one observed that the Loctite® 495 has excellent adhesion properties because it does not fracture in the adhesive joint, but instead, in the solder mask and FR4 joint. The M-Bond 200 and 60 s Universal Glue exhibit CF failures, i.e., adhesion appears to be good but presented low stress and strain properties. The M-Bond 200 failure type allows us to speculate that the value of the shear stress of 5.5 MPa might show signs of fracture of the solder mask and of the FR4 layer.

Of a pre-selection of 5 adhesives, 60 s Universal Glue Loctite® and Araldite® Instant 90 s adhesives were excluded. The former because has the lowest mechanical properties, and the latter since it does not show mechanical properties that are so differentiated from the other adhesives and presents longer curing times. Higher curing times may lead to more problems of fixing and positioning the OF on a PCB, such as fiber displacement when routing has curvilinear movements. Three adhesives are selected for the next test campaign: M-Bond 200, Loctite® 495, and Permabond® 630.

Figure 9 shows representative curves of the shear stress vs strain for the high temperature adhesive. Table 3 presents the values of the maximum shear stresses, the elastic moduli, and the strains at maximum shear stress. The maximum shear stress corresponds, normally, to the breakdown value. The silicone-based adhesive shows the most different mechanical behavior: the lowest elastic modulus and shear stress, but the highest strain at maximum stress. The other 3 adhesives show similar mechanical properties. The adhesives M-Bond 610, PI-32, and P-Adhesive present values of the maximum shear stress in the order of 3–4 MPa, which are similar to the room temperature adhesives (Table 2).

Fig. 9
figure 9

Representative curve of the shear stress vs strain for the different adhesive’s high temperature

Full size image
Table 3 Maximum stresses, elastic moduli, and strains at maximum stress for the high temperature adhesives
Full size table

Figure 10 shows the final appearance of the failure surfaces of the samples after mechanical tests. Silicone shows a uniform fracture of the adhesive layer (CF type), with adhesive on both sides of the PCB, while cyanoacrylate shows an AF. In the case of M-Bond 610, PI-32, and P-Adhesive, they present a fracture in the adhesive layer of a mixed type (MACF), and there were no adhesion and cohesion problems. It is verified that they remove material from the PCB.

Fig. 10
figure 10

Failure surfaces of the samples tested with a M-Bond 610; b PI-32; c P-Adhesive; d Silicone; e Cyanoacrylate

Full size image

From a pre-selection of 5 adhesives, silicone and cyanoacrylate adhesives were excluded. The latter has low mechanical properties with no adhesion to the substrate. The former, despite having good mechanical properties, was excluded due to the difficulty in controlling its application process.

P-Adhesive, despite having good mechanical properties and easy application, has high curing temperatures. On the other hand, PI-32 is used as support where small dots are placed to help attach the PTFE to the PCB, as it is more difficult to apply and to control its thickness when applied to OF. M-Bond 610 will be the adhesive used in the next testing campaign.

3.2 OF behavior with different ambient temperature adhesives

Three ambient temperature adhesives (M-Bond 200, Loctite® 495, Permabond® 630) were used to bond a OF to a PCB sample. OF measurements were taken under tensile and bending tests to access the quality of the signal. The measured strain and the average strain deviations along the OF for the tensile tests are presented in Fig. 11. The Loctite 495 and the Permabond 630 had low strain average error, 1.3% compared to the segment of M-Bond 200 bonding material. For the Loctite® 495, the strain average error was 0.7% compared to M-Bond 200. At 1300 μe, 1.3% and 0.7% correspond to 16.9 and 9.1 μe, respectively. The variations of strain along the fiber length (Fig. 11b), i.e., along the specimen length, are due to the imposed deformation of the specimen by a single movable grip, typical of universal testing machines. The similitude of the strains over the three fiber sections, placed along the fiber width, means a uniform strain distribution over the fiber cross-section, as expected.

Fig. 11
figure 11

a 2200 mm OF strain measurements at tensile test with 3 adhesives; b comparison of the 3 bonded segments

Full size image

The OF strain measurements for the bending tests are presented in Fig. 12 for the 3 selected adhesives. For 8 mm deflection, the M-Bond 200 reaches a maximum strain peak of 1266 μe, the Loctite® 495 reaches 1285 μe, and the Permabond® 630 1273 μe. The Loctite® 495 and the Permabond ® 630 have low strain average error. It is verified an average strain error of 1.7% and 1.8% for the Permabond® 630 and Loctite® 495 segments, respectively. The corresponds 22.1 µe and 23.4 µe, respectively, which are very small errors.

Fig. 12
figure 12

a 2200 mm OF strain measurement at bending test with the 3 adhesives; b comparison of the 3 bonded segments

Full size image

Analyzing the bonded segments 1–4 (Fig. 13), the strain errors are of 2%, which are small. They may be linked to differences in the adhesive thickness, slight PCB warpage, adjustment of the supports in the bending test or some intrinsic variations of the PCB structure. M-Bond 200, Permabond® 630, and Loctite® 495 are suitable for OF bonding, being feasible alternatives for OF placement and routing.

Fig. 13
figure 13

Bonding segments strain measurement in the flexural test (M-Bond 200)

Full size image

3.3 Effect of adhesive thickness on the OF signal acquisition

OF strain measurements are affected by the adhesive bond geometry. Adhesive thickness and length are the most important factors affecting the bonding performance [16]. The M-Bond 200 was considered. Figure 14 shows cross-sections of the OF bonded to the PCB. The thicknesses of each bonding layer were measured, showing deviations from the initial defined ones. Figure 15 shows the strains measured for different adhesive thicknesses.

Fig. 14
figure 14

Cross-sections of the adhesive layers: a typical segment; b 50 μm; c 100 μm; d 150 μm; e 200 μm thick (note that the measured values deviate from the nominal ones)

Full size image
Fig. 15
figure 15

a 2200 mm OF strain measurement under tensile test; b OF segments with different adhesive thickness

Full size image

The variations of strain along the fiber length (Fig. 15b) are representative of the strain distribution during a tensile test in a universal testing machine, as already mentioned. It appears that there is no representative influence of the bonding material thickness on the strain measurement during a tensile test, as shown in Fig. 16. Having as reference the bonding segments, the strain average error is uniform and independent of the thickness of the adhesive layer. At 1300 µm strain, 0.5% represents only 6.5 µm of strain. Nevertheless, and as shown in Fig. 15b), a thicker adhesive reduces the strain variations along the fiber/specimen length, meaning that the adhesive is also deforming and not fully transmitting the deformation to the OF.

Fig. 16
figure 16

Strain average error as a function of the adhesive thickness in the tensile test

Full size image

The OF strain measurement and the average strain deviations for the bending tests are shown in Fig. 17. The thickness of the adhesive material influences the strain measurements under bending. As the bonding material thickness increases, the measured strain also increases, and the strain error also increases (Fig. 16). While 58.2 µe represents a 5.6% average strain error, 207.1 µe corresponds to a 26% error.

Fig. 17
figure 17

a 2200 mm OF strain measurement under bending test; b OF segments with different thicknesses

Full size image

There is a linear relationship between the strain average error and the adhesive layer thickness. In fact, it is known that in a bending loading the strain values from OF measurements need to take into account the adhesive layer thickness and fiber diameter [17]. For a given OF diameter, the strain at the component surface, ec, is given by:

$$\varepsilon_{c} = \frac{0.5h}{{0.5h + d}}\varepsilon_{OF}$$
(1)

where, εOF is the OF measured strain, h is the thickness of the PCB, and d is the distance between the OF and the PCB surface (adhesive layer thickness). As shown in Eq. (1), for d ≪ h, if the adhesive layer increases the error on the strain varies almost linearly. The adhesive thickness should be the lowest as possible particularly in the case of the bending tests. For any predicted increase in bonding material thickness, a correction of the measured strain value must be made. Such behavior is the result of the shifting of the strain plane under analysis as the thickness increases. With a thicker layer of adhesive, the strain measurement is performed farther from the PCB surface, which under a bending effort result in a higher strain level.

3.4 OF signal behavior with high temperature adhesive

The OF signal behavior with a high temperature adhesive (M-Bond 610) was assessed by 3-point bending tests, under different temperature levels. An instrumented PCB was loaded with a 2 mm deflection at its mid span at temperatures of 23, 150, 200, and 250 °C. Figure 18 shows the strain along the length of the OF under the 3-point flexural tests.

Fig. 18
figure 18

Strain along the length of the OF under a 3-point flexural tests at different temperatures

Full size image

In fact, at each temperature level, the response to the mechanically induced deformations is similar as the loading is also equal. The OF sensor response to strain is constant at each temperature level (in this case, 2 mm of bending corresponds to a frequency shift of approximately − 150 ± 20 GHz, and a maximum strain of 980 me). However, the force required to apply 2 mm deflection decreases with the increasing of temperature, displaying a step-down after 100–130 °C, corresponding to the glass transition temperature, \({T}_{g}\), of the FR4, the major component of the PCB (\({T}_{g}\) of FR4 is c.a. 135 °C). The OF signal seems to be reproducible along the temperature variation range, showing the adhesive applicability to high temperature measurements, without major influence.

4 Conclusion

This work investigates the effect of the adhesive on optical fiber-based strain measurements on PCB. Room temperature and high temperature adhesives were considered, for different applications. Regarding room temperature cured adhesives, cyanoacrylate based one are the most suitable for bonding and OF to a PCB (e.g., Loctite 495 or M-Bond 200). Nevertheless, not all ethyl-cyanoacrylate adhesives are recommended (60 Sec Universal Glue Loctite® showed the lowest mechanical performance). Araldite based adhesive (Araldite® Instant 90 s) showed good mechanical properties but presented longer curing times. Methyl-methacrylate UV cure adhesive (Permabond® 630) showed lower mechanical properties, but the curing method and time are very appealing from the assembly point of view. Regarding high temperature adhesives, the silicone-based one showed the less attractive mechanical behavior (with low Young´s modulus and high strain at break), being difficult to apply to the PCB surface. Cyanoacrylate adhesives also showed low mechanical properties with no adhesion to the PCB. Epoxy and polyimide-based adhesives (M-Bond 610, PI-32, and P-Adhesive) have similar mechanical properties, suitable for high temperature measurements (up to 260 °C).

Manual bonding results on variations of the adhesive material thickness that affects the OF strain measurements. The results indicate that adhesive thickness should be the lowest possible and uniformly applied, mainly for the case of bending tests. A linear relationship between adhesive thickness and strain error was found under bending and tensile loadings, with a much higher dependence in the former case.

For any increment upon the adhesive thickness, a correction of the measured strain value must be made. Finally, due to the high sensitivity of the OF and its distributed nature (i.e., function as a sensor in its entire length), it is key to ensure a proper bonding between it and the PCB surface.

The selection of the adhesive and its application are crucial for accurate strain measurements using OF. The use of the OF strain measurements to measure the thermal-induced strains of PCB, e.g., during their fabrication processes, requires an adequate procedure and accurate calibration method, as the response of the OF is sensitive to both temperature and strain. This will be the subject of other publications.