Influence of Quality of Mounting Process of RF Transistors on Their Thermal Parameters and Lifetime

Featured Application

The presented results can be applied in the assembly process of RF networks.

Abstract

The paper presents the results of investigations illustrating the influence of imperfections in the soldering process on thermal properties of transistors operating in RF (Radio Frequency) systems. The methods of measurements used to determine the junction and case temperature of the tested transistors and their thermal resistance are described. The results of the performed measurements are presented and discussed, and their statistical analysis is carried out. The correlation between thermal parameters of the tested transistors, parameters of the assembly process and the measurement conditions is investigated. Some calculation results illustrating an influence of device thermal resistance and dissipated power on changing its lifetime are shown and discussed. It is also shown that no correlation between the void sizes and the device thermal resistance and lifetime is observed. Possibilities of improving the assembly process are indicated, which allows extending the expected lifetime of the tested transistors.

1. Introduction

In RF technology, a power amplifier (PA) is a system that is fed with a low-power RF signal. This circuit increases the signal power to the necessary level without causing distortions. Sometimes an additional filter is introduced between the PA input and the load to minimize unwanted signals due to design imperfections. In the RF signal path, the PA is located between the transmitter signal source circuits and the antenna [1,2].

Companies designing RF electronic modules for telecommunications applications need the appropriate power devices that are able to power supply TRX (Transmitter-Receiver) circuits in order to transmit signals with the required power level in the appropriate frequency range. The efficiency of the PA is a function of the level and shape of the signal [1,2]. For high power levels (in the range 500–1000 W) a single PA may not be able to ensure the required power level. In such cases, multiple PA modules may be switched on in parallel. Additionally, in PA circuits, energy efficiency influences the lifetime of batteries used in devices such as smartphones, in which the PA circuit is the major consumer of energy. The PA also determines the value of heat which must be dissipated by the telecommunications system [1,2,3,4,5].

Although the use of several PA modules connected in parallel can solve the power supply problem, there is a problem of power balance, current sharing, thermal matching, and preventing a failure or overheating of the components. All PAs exhibit gain compression at higher power levels, which means that the gain is not constant but it decreases when the input power increases. The above-mentioned limitations mean that we are looking for discrete power semiconductor devices that will meet a wide range of requirements, taking into account the optimization of four parameters: output power, gain, operating frequency, and energy efficiency. Power transistors included in PAs that meet these requirements belong to the group of LDMOS (Laterally Diffused Metal Oxide Semiconductor), HVLDMOS (high-voltage LDMOS) [6], or GaN transistors [7,8]. The technology used allows achieving a high gain, high efficiency, and linearity for the operation at the frequency up to 12 GHz [9,10,11,12].

One of the applications of power transistors in RF circuits is the Doherty amplifier. The Doherty amplifier topology is used to maximize the effective use of DC power, when the transmitted signal has a high peak-to-average ratio [4]. Typically, using the Doherty amplifier, PA circuits operate at the frequency of 2.14 GHz using input signals of the W-CDMA (Wideband Code Division Multiple Access) type [3,4,5]. An example of such an amplifier can be a single-band amplifier containing two PAs. The first one is efficient at the lower power (and disables the other one in this range), and the other one—with smooth transition to the one that is less efficient with the low power and better adapted to work at the higher power [2]. The principle of operation of the Doherty amplifier is presented in detail in [2,3,4,5].

All semiconductor devices lose power during their operation. The heat not dissipated outside the semiconductor device leads to an increase in its junction temperature, i.e., overheating of semiconductor devices is one of the most common causes of damage to electronic devices, because the intensity of damage to electronic components increases with an increase in the junction temperature [13,14].

Not only the reliability and the lifetime of semiconductor devices, but also material properties, mechanical and electrical strength (viscosity, loss, dielectric permittivity, etc.) depend on temperature [15]. Moreover, the permissible power load of practically all devices is severely limited by an increase in their junction temperature. For practical calculations of the cooling system, it can be assumed that in consumer electronics devices all the electricity supplied to the device, and in transmitting devices (about 75% of the supplied energy) is converted into heat [15].

The device junction temperature T_j increases proportionally to the power dissipated. The dependence of the junction temperature on the dissipated power P_D is described by the equation [1,16,17]

$${\mathrm{T}}_{\mathrm{j}}={\mathrm{T}}_{\mathrm{a}}+{\mathrm{R}}_{\mathrm{th}}\cdot {\mathrm{P}}_{\mathrm{D}}$$

(1)

where T_a is the ambient temperature and R_th is thermal resistance of the semiconductor device.

The above dependence is valid at the steady state, when power losses do not change with time or change very slowly. In transient states, thermal properties of semiconductor devices are characterized by transient thermal impedance Z_th(t) [17,18,19]. The value of thermal resistance of semiconductor devices depends, among others, on the quality and the method of their assembly [20,21,22]. In particular, the quality of the solder connection of the thermal pad to the PCB (Printed Circuit Board) is important [23].

In the soldering process, the maximum of 30% void [24] not covered with the solder is allowed under power semiconductor devices. Additionally, according to the IPC-610 [24], other criteria related to voids should be agreed on with the manufacturer. Typically, the maximum individual void sizes are set at 8%.

The paper [25] presents the results of the experimental research on the use of dies for MOSFET RF transistors in a case made of plastic materials. Unfortunately, the cited paper does not provide the results of thermal parameters measurements. Yet, it was shown that the use of plastic materials has led to a deterioration of the RF performance of these transistors.

The paper [26] presents the design and the results of experimental tests of the integrated RF amplifier modulated with the WCDMA signal, operating in mobile phones. Unfortunately, in the cited paper there is no data concerning the assembly of amplifiers and external RF power stages.

The paper [27] presents the results of measurements of thermal parameters of VDMOS transistors mounted on glass substrates. Attention was paid to the measurements of thermal resistance and thermal analyses carried out.

In the paper [28], a junction temperature rise in GaN-HEMT RF transistors was analyzed. The dimensions of voids and their number in the brazed joint were also estimated. This parameter was taken into account in the thermal analyses carried out in the ANSYS software.

In turn, the paper [29] contains the results of the research on thermal phenomena in high-power GaN RF transistors operating in the switching mode. The paper [30] presents a new non-linear electrothermal model of RF transistors of the LDMOS type. Algorithms based on neural networks are used for modeling this device.

The paper [31] describes the results of the study of the correlation between a change in electrical parameters of the N-LDMOS power RF transistor and the damage that occurred after the impulse RF lifetime tests (1500 h). The tests were first carried out in real time with the use of a test stand designed for experiments with radar systems.

The subject of the research presented in [32] is the influence of temperature on properties of N-channel LDMOS transistors operating in the radio frequency range, subjected to reliability tests in the temperature range from 10 to 150 °C.

The article [33] presents the measuring method and exemplary results of the measurements of transient thermal impedance of microwave transistors. Particular attention was paid to the measurements in very short periods of time after a power step change. The paper [34] presents a thermal model of RF transistors taking into account voids in solder joints.

The aim of this article is to investigate the influence of imperfections in the soldering process on thermal parameters of power transistors used in RF circuits. In particular, the influence of the dissipated power on the transistor case temperature is considered for different sizes of the unsoldered spaces (voids) under the tested power transistors. The influence of the size of the void spaces in the soldered joint on the temperature distribution on the surface of these transistors is also assessed. Based on the performed measurements, changes in the lifetime of transistors depending on the size of voids are estimated. The obtained values of thermal parameters are compared with the data declared by the manufacturer. The main goal is to investigate how the voids influence thermal resistance and the lifetime of an RF transistor. In contrast to the studies given in the literature, we used both: the measurement methods based on infrared measurements with X-ray imaging and the statistical analysis.

The investigated power transistors are described in Section 2. The soldering process is described in Section 3. The applied method and the measuring system are described in Section 4. The results of measurements are presented and discussed in Section 5. The influence of thermal parameters on the lifetime is analyzed in Section 6.

2. Investigated Transistors

The subjects of the investigations are power transistors LDMOS and HVLDMOS manufactured by Ampleon, type: BLC9G22LS-160VT [35] and BLC10G22LS-240PVT [36] placed in the RF circuit containing the Doherty amplifier. Both these transistors are characterized by high efficiency and excellent thermal stability [35,36]. Both the transistors are capable of operating with a load mismatch corresponding to a standing wave ratio VSWR = 10:1 (perfect match is 1:1). The manufacturer’s declared value of thermal resistance between the junction and the case R_thj-c of these transistors is 0.47 K/W [35] and 0.35 K/W [36], respectively.

At typical application of the considered transistors are RF power amplifiers for base stations and multi carrier applications in the 2110–2200 MHz frequency range. Some data given by the manufacturer on RF characteristics of the considered transistors was obtained for the test signal characterized by: 2-carrier W-CDMA; 3GPP test model 1 with 64 DPCH; PAR = 8.4 dB at 0.01 % probability on the CCDF; f₁ = 2112.5 MHz; f₂ = 2117.5 MHz; f₃ = 2162.5 MHz; f₄ = 2167.5 MHz. RF performance is obtained at V_DS = 28 V; I_Dq = 1600 mA; T_C = 25 °C; in a water-cooled class-AB test circuit. In such conditions, the following typical values of parameters are obtained [36]:

• Power gain G_p = 18.5 dB at the dissipated power P_L = 60 W;
• Drain efficiency η_D = 27%;
• Input return loss RL_in = −14 dB;
• Adjacent channel power ratio (5 MHz) ACPR_5M = −30 dBc.

The cases of the investigated transistors are shown in Figure 1.

Both the transistors have gold-plated terminals, which, according to the experience of one of the companies producing electronic equipment, has a significant impact on the quality of SMT (Surface Mount Technology) [37]. The area of the thermal pad (transistor sources) is 201.88 mm² for both the transistors. The tested transistors are mounted on the PCB shown in Figure 2.

This PCB has the dimensions of 245 × 280 mm. The PCB laminate has 12 conductive layers of the thickness 75 µm (internal layers) or 100 µm (outer layers). The insulation layers are 75 µm thick. On the PCB there are two PA circuits containing 2 pairs of LDMOS and HVLDMOS transistors, which are marked on the main page (top) of the PCB. The BLC9G22LS-160VT and BLC10G22LS-240PVT transistors are part of the transceiver TRX, where they are placed in the PA. They play the role of an amplifier in the Doherty amplifier. They are assembled in the SMT process. There are pockets in the printed circuit board in the place where each transistor is mounted. The pocket dimensions in the PCB laminate are: x = 21 mm (max 21.6 mm), y = 10.15 mm (max 10.4 mm), and depth = 1.63 mm. The area of the pocket is only 11.12 mm² larger than the area of the transistor source soldered to the pocket.

As mentioned in the Introduction, one of the applications of power transistors in RF circuits is the Doherty amplifier. In the tested modules, the considered transistors work in such amplifiers. The main amplifier, called the carrier signal amplifier (Main or Carrier), is connected to the peak amplifier (Peak). Connecting these amplifiers requires that the amplitudes of the signals from both the outputs are in the appropriate phase with regard to each other, which is ensured by the use of additional phase shifters on the inputs and outputs [4]. The phase change introduced by the impedance inverter is compensated for at the input of the peak amplifier. The load is connected to the main amplifier via an impedance inverter, which is typically a quarter-wave transmission line. In this way, only the Main amplifier works at a low signal level, up to 13 dBm—amplifier classes A and AB [3]. In the Doherty range, from 23.5 to 32 dBm, the two amplifiers generate power at a higher level. This means that the Peak amplifier goes from the switched off state to the correct operation in class C [3].

3. Solder Process

According to the IPC-610 [24] standard, 30% of the total area of voids cannot be exceeded, and the additional criterion is the limitation to the maximum of 8% of the individual empty space under the soldering surface of the investigated RF power transistors. In order to control both the indicators, 100% production of such modules undergoes the X-ray test [38] with the X-ray DAGE system: XD7600NT. This test detects any errors in the SMT process, such as voids in the soldering process or short circuits. The examples of the results of such measurements are presented in Figure 3.

The presented results are for the transistors soldered in such a way that at least one of the requirements related to the size of the void is not satisfied. In Figure 3a, the total void size is 18%, but the individual void size is as high as 12.7%. On the other hand, the largest total surface not covered with the solder is shown in Figure 3b. In this case, both the overall void area exceeds 30% and the individual largest void is 9.6%. An important observation is that the presence of voids occurred in the case of gold-plated transistors. For the other models of transistors with silver coating of the leads mounted on other PA modules no process problems were observed.

For the presented examples of the process defects, a cross-section was made, also presented in Figure 3. As a result of the cross-sectional cutting of the transistor by the area of voids occurrence, 8 regions can be distinguished. Regions 1–3 are for the solder layer of the transistor source. In the photo, there are solder voids (region 1) visible, which were previously localized by performing an X-ray test.

In order to illustrate the problem of a high number of voids in the produced modules, some information about the production process is collected in Figure 4. This figure presents the qualitative results and the level of such process errors for the discussed power transistors [35,36] installed in the described electronic modules.

Figure 4 presents values of the quality index in the successive months, the yield [38], related to the level of occurrence of empty individual spaces in relation to the production level. It was compared with the yield level associated with the occurrence of short-circuits on the transistors, also appearing in the SMT assembly process. It can be noticed that regardless of the registered number of modules tested using X-ray (blue bars in Figure 4) in each month from 20 January to 21 February, the permissible value of voids was exceeded. Products with such process defects accounted for 8% to even 34% of the production batch (red line in Figure 4). This proves that the void level indicator was not sufficiently controlled in the discussed period.

At the same time, for the same number of samples measured in the X-ray test, the level of short-circuits on the assembled RF power transistors was recorded. In this case, the results remained stable and indicated that less than 1% of the products made in individual batches required a rework process (black line in Figure 4). Figure 4 shows that the presence of voids that did not meet the criteria described above affected the quality of soldering much more than the occurrence of short circuits. Thus, it is the main procedural problem. At the same time, it is important to check the risk of voids for the operation of electronic modules.

Due to the fact that it is not known how the power devices behave when the above-described limits of non-soldered surfaces are exceeded, and it is not known how the operating temperature of these devices is distributed with the release of the set power, it is necessary to measure thermal parameters. They will allow estimating the operating conditions of the system in the natural working environment. It is necessary to determine whether the emerging voids under the soldered power devices have an influence on the service life of the component and therefore the whole module. This knowledge will help determine the risk of complaints on the manufactured products and also determine the acceptable ranges of the dimensions of void spaces of the soldered elements also for new products.

Based on the Quality House analysis [39], it was found out that the values of thermal parameters of the transistors included in the analyzed system are significantly important. The next section presents the results of the measurements of these parameters for selected test circuits containing the investigated transistors.

For testing, 4 PCBs were selected, on which the tested transistors were mounted in the manner shown in Figure 5. On each PCB there were 2 transistors of the BLC9G22LS-160VT type marked with the symbols TR2_B and TR4_B and 2 transistors of the BLC10G22LS-240PVT type marked with the symbols TR1_A and TR3_A.

For each of the transistors, the level of voids was measured using the X-ray Dage [38]. The selection of the obtained results of such measurements is shown in Figure 6.

As can be seen in Figure 6, the measured total void area in the soldering area is 15.3% and the largest individual void is 8.7%. It is marked red in Figure 6. This is an example, where the customer’s imposed maximum 8% of the individual void size in the soldering area was exceeded.

The measured values of the total voids and the largest individual voids for each transistor are summarized in

Table 1. Voids sizes for the investigated RF power transistors.

PCB	Transistor	Total Void [%]	Maximum Individual Void [%]
1	TR1_A	18	12.7
1	TR3_A	8.7	2.7
2	TR1_A	13.5	2.9
2	TR3_A	15.2	8.7
3	TR3_A	30.6	9.6
4	TR1_A	12	2.4
4	TR3_A	16.9	8.2
1	TR2_B	18.9	11.7
1	TR4_B	16.8	7.2
2	TR2_B	14.4	2.9
2	TR4_B	16.6	9.2
3	TR2_B	19.9	8.3
3	TR4_B	9.9	5.6
4	TR2_B	15	8
4	TR4_B	21.1	8.7

.

As can be seen in Table 1, the total void values range from 8.7% to 30.6% for the BLC10G22LS-240PVT transistor and from 9.9% to 21.1% for the BLC9G22LS-160VT transistor. The largest individual void ranges from 2.4% to 12.7% for the BLC10G22LS-240PVT transistor and from 2.9% to 11.7% for the BLC9G22LS-160VT transistor, respectively.

4. Measurement Method and Set-Up

In order to check if there is an influence of void spaces on thermal parameters of power transistors, values of their thermal resistance R_th were measured. According to the classical theory [40], thermal resistance between the semiconductor die and the surroundings R_th is the sum of thermal resistance R_thj-c between the semiconductor die and the case and thermal resistance R_thc-a between the case and the surroundings. The first of these components depends on the quality of the assembly process of the semiconductor die in the case and the case construction [41]. The other component depends on the external cooling system and the mounting method of the RF power transistor [20,22].

The indirect electrical method described in [42] was used to measure thermal resistance R_th. This method requires the inclusion of the tested transistor in a special measuring system and periodic switching of the transistor’s drain current. This measurement can be performed only for the transistor, which is not connected to a special application circuit. In order to determine the value of thermal resistance R_thj-c, it is necessary to measure the junction temperature of transistor T_j at a predetermined value of temperature of the case T_C. The stability of this temperature is ensured by placing the tested transistor on a cold plate, which is a part of the forced liquid cooling system of this transistor.

For the considered transistors mounted on the tested PCBs, values of thermal resistance R_thc-a were measured using the measuring set-ups shown in Figure 7. These systems allow separating the power of the fixed value in the tested transistors and measuring the case temperature T_C in the steady state using an IR camera.

Both the set-ups enable power supply of the tested transistors in such a way that it is possible to generate enough power in them to significantly increase their junction temperature. The transistor’s operating point coordinates are measured with voltmeters V₁, V₂ and an ammeter A₁. At the set voltage V_GS and voltage V_DS with flowing current I_D in the tested transistor, power P_D is dissipated. As a result of self-heating, the junction temperature of the tested transistors T_j and the case temperature T_C increase. The temperature T_C is measured with an IR camera. On the basis of the measured temperature T_C and the information about the reference temperature T_a, also indicated by the IR camera, thermal resistance R_thc-a is determined using the formula:

$${\mathrm{R}}_{\mathrm{thc}-\mathrm{a}}=\frac{{\mathrm{T}}_{\mathrm{C}}-{\mathrm{T}}_{\mathrm{a}}}{{\mathrm{P}}_{\mathrm{D}}}$$

(2)

The unit of thermal resistance is K/W.

The circuit shown in Figure 7a is prepared to measure R_thc-a of the BLC9G22LS-160VT transistor, and in Figure 7b of the BLC10G22LS-240PVT transistor. The view of the test stand is shown in Figure 8.

The following measuring instruments were used for the measurements: the Agilent E3645A series power supply, the Agilent E3633A series power supply and the Optris IR camera and the Optris PI connect software [43]. The measuring systems together with the IR camera were placed in a specially prepared chamber, isolating it from the external factors, such as blasts of air or sunlight. This allowed to minimize the measurement error and stabilize the ambient temperature during the measurements.

In order to measure thermal resistance R_thc-a, these transistors were controlled by the voltage source E₁ and voltage V_GS was measured with the voltmeter V₁. The drain current I_D was controlled by the current source G₂. The transistor output voltage V_DS was measured with the voltmeter V₂. It should be noted that the BLC10G22LS-240PVT transistor includes two LDMOS dies [36]. In the measurement system, one of the gates was disconnected and the flow of the drain current I_D was possible due to the diode present in the die and marked in Figure 7b.

During the measurements, the tested transistor should be properly controlled in order to dissipate power P_D in it. The value of this power should be sufficient to obtain an increase in the transistor case temperature above the ambient temperature exceeding 40 °C. As a result, the measurement error of thermal resistance R_th did not exceed a few percent [44].

The visible difference between the ambient and case temperatures of the transistor allowed determining R_thc-a from the Formula (2). For each of the transistors, two measurements of R_thc-a were made at two different power values.

It should be added that setting the transistor’s operating point at the given electrical parameters is a time-consuming process, because the thermally steady state is obtained after the time often exceeding 30 min, and sometimes even up to 1 h. From the thermographs obtained using the IR camera, apart from the temperature in the hottest point on the transistor case, the information about the temperature distribution along the cross-section of the transistor case was also obtained. This was to check the effect of selecting a point on the case where the T_C value was measured on the obtained R_thc-a values. It was also checked whether the changes in the temperature T_C distribution depend on the void level, which was previously measured for the tested transistors.

5. Results of Measurements

The tests were carried out for four modules described as PCB1, PCB2, PCB3, and PCB4. Each of them contained two pairs of the investigated transistors. First, indirect measurements were made with an electrical method and then with an optical method.

The measurements of transient thermal impedance Z_th(t) and thermal resistance R_th of the BLC9G22LS-160VT transistor mounted on a heat exchanger of the dimensions 150 × 70 × 10 mm in a forced cooling system [45] were performed using the indirect electrical method. This method was described, inter alia, in the paper [42] and the measuring system used in the paper [45]. The measurements were made with free cooling (the pump was turned off in the cooling system) and forced cooling (the pump was turned on in the cooling system). The investigated transistor mounted on the heat exchanger is shown in Figure 9, and the obtained results are shown in Figure 10 and Figure 11. During the measurement, the tested transistor gave off the power equal to 21.5 (blue line and red line) and 41.5 W (black line).

Figure 10 shows that the steady state Z_th(t) value of about 5.2 K/W was obtained with free cooling and the steady state was obtained after about 6000 s. In turn, with forced cooling the steady state occurred after about 100 s and the steady state value of Z_th(t) was approximately 2.6 K/W. With forced cooling, it can be seen that doubling the power dissipated in the transistor caused a decrease in the thermal resistance value by only 0.5%. It is also worth noting that the waveforms Z_th(t) determined for both the cooling conditions in the time range t < 5 s are practically indistinguishable. Therefore, it can be concluded that this part of the waveform Z_th(t) describes the heat transfer from the semiconductor die to the cold plate.

Simultaneously with the measurements performed with the indirect electrical method, the case temperature of the tested transistor was measured with the use of a pyrometer. Based on the obtained measurement results, values of thermal resistance between the transistor die and the surroundings R_thj-a and between the case and the surroundings R_thc-a were determined for both the considered sets of cooling conditions. These results are shown in Figure 11. The difference between the mentioned parameters for both the sets of cooling conditions is about 1.1 K/W and corresponds to the thermal resistance between the die of the transistor and its case R_thj-c. It is more than twice the value declared by the manufacturer.

Using the measurement method described in Section 4, the temperature of the transistor case was measured with the IR camera for selected power values. For example, the temperature distributions on the surface of the case of the investigated transistor are shown in Figure 12 and Figure 13. In these figures, the image from the IR camera is presented in its upper part. The areas with a higher temperature are lighter than those showing a lower temperature. Additionally, the temperature control lines were introduced through the transistor’s longitudinal section. At the bottom of the thermograph there is a graph of the temperature distribution: in area 1—blue line; and the transistor cross-section—green line. On the bottom line of the application, one can also read the reference temperature.

The measurements, the result of which are shown in Figure 12, were performed with the power dissipation equal to about 6.74 W, and in Figure 13 with the power of P_D = 21 W. The temperature T_C of the measured area was determined as the average value obtained from the measurements with the IR camera.

In Figure 12, the case and reference temperature difference is ∆T = 20 °C. Under such conditions, R_thc-a = 2.97 K/W. For the BLC9G22LS-160VT transistor, the lowest R_th value is 1.86 K/W at power P_D = 7.1 W (for the TR2_B transistor in the PCB3 module), while the highest R_th = 3.65 K/W was recorded for the TR4_B transistor in the PCB4 module. The difference between the highest and the lowest value of R_th is ∆R_th = 1.79 K/W, which is 32% of the mean value of R_th = 2.76 K/W.

For the BLC10G22LS-240PVT transistor measurements, the lowest R_th value measured was 2.03 K/W with power P_D = 21.67 W for the TR3_A in the PCB2. The highest value of R_th was 2.69 K/W for the TR3_A transistor of the PCB4 module at power P_D = 21.51 W. The changes in the R_th value were therefore ∆R_th = 0.66 K/W. This represents a 14% change in the R_th value compared with the average value of 2.36 K/W.

The distribution of the obtained results of the measurements of thermal resistance R_thc-a for all the tested transistors for both power values is shown in Figure 14.

For the described operating conditions (T_a = 29.5 °C and the dissipated power P_D = 22.5 W) when measuring R_th for the BLC9G22LS-160VT transistor, the lowest value was noted for the TR2_B transistor in the PCB2 module. It is R_th = 1.98 K/W. The highest value of R_th = 2.9 K/W was determined for the TR2_B transistor in the PCB4 module. In relation to these values, the maximum difference R_th in the measured transistors was as high as ∆R_th = 0.92 K/W. This is as much as 19% of the deviation from the average value of the measurements of 2.44 K/W. It should be noted that the maximum deviation R_th was obtained during the measurements for the TR2_B transistors in the tested modules.

The noticeable differences between the values of thermal resistance of each transistor may be caused by the following factors:

• Varying the size of voids in the soldering area;
• An influence of the distance between the tested transistors and the PCB edge or the differences in the distribution of copper on the PCB in the vicinity of the tested transistors;
• Differentiation in the quality of the transistors manufacture resulting in differentiation of their thermal properties.

One can see in Figure 14a that for the BLC10G22LS-240PVT transistor, a change in the power dissipated caused an average decrease in the R_th value by 0.5 K/W. For the transistor BLC9G22LS-160VT (Figure 14b), the changes in R_th at different power values were different depending on the tested module and the position of the component on the PCB.

It is worth noticing that the R_thc-a values measured by the indirect electrical method were 4 and 1.5 K/W for free and forced cooling, respectively. Comparing these results with the results presented in Figure 14, it can be observed that the PCB, to which the considered transistors were mounted, ensured a more efficient heat removal to the surroundings than the heat sink shown in Figure 9 operating in the free cooling mode. This proves a very good design of the tested PCB in terms of its effective cooling.

The statistical considerations presented in the further part of this article will show whether the changes in R_th depend on the number of void spaces in the area of the transistor soldering. Another interesting question is how, apart from the dissipated power P_D, changes in the case temperature T_C are also affected by the voids [46].

From the temperature value distribution through the transistor cross-section (Figure 15), three temperature values were determined: T_emp1, T_min, and T_emp2. For each value of the temperature thus measured, three R_th values are determined, respectively, denoted as R_thT1, R_thTmin, and R_thT2. The distribution R_th along the cross-section of the BLC10G22LS-240PVT transistor is shown in Figure 16.

The data presented in Figure 16 shows that the value of thermal resistance R_th undergoes significant changes depending on the point in the cross-section of the transistor case, where temperature T_C is determined. The highest values of the temperature described as T₁ were measured in the vicinity of the application of the supply voltage V_DS. This is not important with the highest R_thT1 value for each of the transistors in each module. In the middle of the transistor case, one can see a visible drop in temperature described as T_min. In the vicinity of the output of the other transistor, an increase in temperature defined as T₂ was observed again. However, T₂ < T₁, hence also R_thT2 < R_thT1. Changes in the R_th value determined for various points on the surface of the transistor case differed from each other by up to 10%.

6. Influence of Thermal Parameters on the Lifetime

Temperature changes affect the lifetime of electronic circuits. This lifetime is related both to the lifetime of elements used to build electronic modules and to the degradation time of solder joints.

In order to check the strength of solder joints and their degradation during the life of the devices, tests of electronic modules are performed in the properly established thermal cycles. As it is presented in [47], brazed joints are a subject to stresses resulting from thermal cycles and constant elevated temperatures (from 40 to 70 °C) during the normal operation of electronic modules, which leads to thermomechanical fatigue and the formation of intermetallic compounds. From the tests carried out by the authors of the cited paper on PV modules, the strength of solder joints at the operating temperatures of 85, 95, and 105 °C was tested in 15 min cycles [47]. The series resistance values of the PV cell obtained from the three temperature zones made it possible to calculate the activation energy for the degradation of the solder joint. The beneficial effect of Ag in the solder material in the mitigation of the solder degradation is also illustrated. The analyses show that the temperature at the ends of electronic components has a significant impact on the service life of the solder joint [48]. The strength of solder joints is also related to the size of the component. The graph of the risk of a solder joint failure due to the component size and the temperature value is shown in Figure 17.

As it is visible, the risk of damaging the soldering joint decreases with a decrease in temperature and with an increase in the component size. This relation is a result of an increase in thermal capacitance of the component with an increase in its size.

In the case of the considered RF power transistors [35,36], such a temperature, which has a significant impact on the life of the joint, is achieved with the continuous operation of the transistors and the power dissipated, depending on the type of a transistor, from 20 to 25 W without forced cooling and thermocouples. Setting the transistor’s operating point at P_D = 22 W illustrates the real nature of heating of the power devices in the environment in which they must operate in a typical application.

The junction temperature of the transistor, according to the Formula (1), varies linearly with the increasing power. The slope of line T_j (P_D) is equal to the value of R_th. Thus, it is the highest for the highest R_th values.

The changes in T_j caused by power P_D dissipation at a fixed R_th value have a significant impact on the transistor lifetime [49,50]. According to the Arrhenius formula [51], the lifetime τ to failure of a transistor is described by the formula [52]

$$\mathsf{\tau }={\mathsf{\tau }}_0\cdot \exp \left(\frac{{\mathrm{E}}_{\mathrm{a}}}{\mathrm{k}}\cdot \left(\frac{1}{{\mathrm{T}}_{\mathrm{j}}}-\frac{1}{{\mathrm{T}}_0}\right)\right)$$

(3)

where E_a is the activation energy of the damage process, k is the Boltzmann constant, and τ₀ is the lifetime of the transistor operating at temperature T₀.

In order to determine the value of τ, aging tests should be performed to determine the value of parameter τ₀. In this study, the lifetime reduction factor is determined by determining the value of the quotient τ/τ₀. In the calculations, the value of the activation energy of the damage process E_a equal to 1 eV was assumed.

Since the value of T_j depends on thermal resistance R_th and power P_D, the relationship τ/τ₀(P_D) was determined for parameters R_thmin, R_thmax, and R_{th_average} for both the tested transistors. These parameters were equal to the minimum value, the maximum value, and the average value of thermal resistance, respectively. The results for the BLC10G22LS-240PVT transistor are shown in Figure 18.

It can be seen from Figure 18 that, regardless of the R_th value, with an increase in the dissipated power P_D, the lifetime sharply shortens, and the quotient τ/τ₀ decreases as much as 10 times with an increase in the power from 1 to 10 W. At the power value of 30 W this time is reduced even 1000 times. For both the transistors it is clearly visible that for the highest values of R_th = R_thmax the changes in the value of τ/τ₀ are the biggest.

Apart from the dependence τ/τ₀(P_D), the relationship τ/τ₀(R_th) was determined. These characteristics were calculated for several values of the dissipated power equal to 8.5, 22, and 44 W and they are shown in Figure 19. One should pay attention to the slope of the characteristics and thus changes in the value of the quotient τ/τ₀ especially at intervals, where there are actual R_th values of both the transistors.

For the BLC10G22LS-240PVT transistor, the measured value R_th under the same conditions range from 2.1 to 3.1 K/W. For such a range of R_th values with power P_D = 22 W and T_j about 80 °C, the changes of ∆τ/τ₀ were 19.7 times. This proves that the lifetime in the discussed range of R_th changes with the given parameters shortens from 2.6 to 19.7 times. As for the BLC9G22LS-160VT transistor, there appeared similar dependences. For the measured R_th values of this transistor in the range from 1.9 to 3.2 K/W, the changes of τ/τ₀ were from 5.5 to even 50.7 times.

Using the measurement results presented in Figure 14, the statistical distribution of the results of thermal resistance R_th is also checked for the transistors TRA and TRB [24,37] with the dissipated power P_D ≤ 10 W and P_D ≥ 20 W. From the obtained distributions it can be observed that in terms of measuring thermal resistance R_th under the described operating conditions both the transistors have similar properties, as shown in Figure 20.

The results of the measurements of thermal resistance R_th presented in Figure 20 show that it does not statistically depend on the tested transistor types. Its value depends on the power dissipated by these transistors. For the transistor TRA with P_D ≤ 10 W, the median of the measured R_th values is 2.93 K/W, and for the transistor TRB type it is 2.99 K/W. For P_D > 20 W the values of R_th are even 20% lower than for P_D < 10 W. The values of the median of the measured R_th are presented in

Table 2. Values of the median thermal resistance R_th for the tested TRA and TRB transistors with the dissipated power P_D ≤ 10 W and P_D ≥ 20 W.

Transistor Type	Index Used in the Analysis	PD [W]	Median Rth [K/W]
BLC10G22LS-240PVT	TRA	≤10	2.93
BLC9G22LS-160VT	TRB	≤10	2.99
BLC10G22LS-240PVT	TRA	≥20	2.46
BLC9G22LS-160VT	TRB	≥20	2.56

.

From the results of the measurements presented in Table 2 it can be seen that in the tested cases, for both the types of transistors, a two-fold increase in the dissipated power from 10 to 20 W caused a decrease in the R_th value by 15%.

In order to determine how the gaps in the solder affect the parameters of the tested RF power transistors [24,37], including the value of thermal resistance and the internal temperature of the component, the 6 Sigma methodology and the Quality House analysis [39] were used. The Minitab tool includes a “Transfer Function Identification” function that allowed us to create a matrix graph showing the relationship between selected parameters. The matrix shows how the indicators depend on each other. The degree of dependence is presented by the correlation coefficients on the scale from 0 to 1, which determines whether there are statistical relationships between the parameters under consideration or not. The correlations values exceeding 0.7 [53] indicate a significant influence of one quantity on another.

It was checked whether the value of resistance R_th depends on the area of the void. In none of the cases for both the TRA and TRB transistors the effect of the individual and total void sizes on the R_th value approached the level of 0.7. The highest correlation value found was only 0.18, tested for the TRB transistor. This level of correlation is so low that it can be concluded that the R_th value does not depend on the individual or total void value. Similarly to the determined value of R_th, the components influencing the value of thermal resistance resulting from the Formula (2), i.e., the value of the temperature emitted on its case, do not depend on the measured values of the void areas in the solder joint.

Hence, when analyzing the obtained results, it was found out that there was no statistical influence of the total and individual void areas on the obtained values of thermal resistance of the tested transistors. Consequently, the size of the voids does not affect the lifetime of the tested transistors.

Although no direct correlation was found between the size of the total or individual voids and thermal resistance R_th and the junction temperature of transistor T_j, it was shown that transistors of the same type can significantly differ in thermal parameters (even by 25%). The discrepancy between the values of thermal resistance R_th in the BLC10G22LS-240PVT and BLC9G22LS-160VT transistors is reflected in their lifetime, especially in the high-power range. From the curves shown in Figure 18, you can see how drastically the lifetime of these components decreases along with the power dissipated. For the tested modules with P_D > 15 W, this lifetime decreases even several hundred times. The observed differences in R_th values reach even 40% and may indicate the existing risk of damaging some components before the deadline provided by the manufacturer. To determine the absolute lifetime, additional aging tests, which were not part of the performed measurements and analyses, should be performed.

7. Conclusions

This article analyzes selected problems related to the impact of the quality of a soldering process of RF power transistors on multilayer printed circuit boards on the efficiency of the dissipation of heat generated by these transistors. The considerations were carried out for power LDMOSFETs operating in the Doherty amplifiers. Various solutions of the cases of the transistors and a different manner of their mounting on the PCB were considered.

Thermal parameters of these transistors were measured for many transistors mounted on identical multilayer PCBs. The performed measurements included the determination of thermal resistance between the case and the surroundings, and the non-uniformity of temperature distribution on the surface of these transistor cases. The obtained measurement results were compared with the results of the measurements made by the indirect electrical method for one of the considered transistors operating in the forced liquid cooling system. It was shown that for all the tested transistors, the tested PCB ensures more effective cooling than a heat-sink in the form of an aluminum block of the dimensions 150 × 70 × 10 mm.

Thermal parameters were measured for each of the tested transistors and additionally, X-rays of the soldered joints were taken. Based on the obtained images, the area of the total and individual voids under each transistor was determined. The obtained values change in a wide range. The results of the measurements of thermal resistance R_th were compared with the determined void values and it was found out that there was no correlation between these values. This is inconsistent with the common opinion given in the literature that increasing the surface area of the void causes worsening in the heat removal efficiency of the transistors. The 6 Sigma method was used to statistically check whether there was indeed a correlation between the above-mentioned quantities. This analysis showed no correlation between the considered parameters. It was also shown that an influence in the dissipated power causes a decrease in thermal resistance. For the transistors of the same type mounted in the same PCB in the same mounting process, very big differences in the measured values of thermal resistance were obtained. They were equal even to 20%.

It was also shown that the quality of the assembly, characterized by the R_th value, has a decisive influence on the lifetime of the components under consideration. Changing this parameter by only 10% causes even a several time change in the lifetime of these elements.

The risk associated with the SMT process and the manufactured product was estimated in terms of the soldering quality requirements, product specifications, and the assumptions related to its use. The compliance of the parameters of power transistors with the specification and the main functional assumptions of the manufactured product was verified.

The activities related to the described validation of the influence of voids on the lifetime of the module may also allow defining the acceptable limits of unsoldered areas for subsequent, newly introduced products. It was shown that the information obtained using X-ray imaging cannot give the proper information concerning the estimated lifetime of the mounted devices. A better solution is to measure the device thermal resistance. The further work by the authors will be focused on investigating the correlation between the parameters characterizing the soldering process and the reliability of RF transistors.