The Road to the Next Generation Power Module – 100% Solder Free Design

Posted on 28 September 2019

power module

 

The reliability of classical power modules with various solder interfaces is not sufficient to meet the demand of progressive applications in power electronics. Technologies were developed in the last decade to improve the reliability. The elimination of the base plate prevents the solder fatigue failure in the substrate solder layer and permits the implementation of arbitrary ceramic materials. Spring contacts and pressure contacts replace soldered terminals. Diffusion sintering substitutes the chip solder layer. The combination of these technologies leads to a module architecture without any solder layer that has the potential to exceed the reliability limits of classic modules.

Uwe Scheuermann, SEMIKRON Elektronik GmbH & Co. KG, Nürnberg, Germany
Peter Beckedahl, SEMIKRON International GmbH, Nürnberg, Germany

1. Introduction

Since the development of the first power module with internal basic and functional insulation by SEMIKRON in 1975, power modules have successfully penetrated into all types of power conversion applications. Today, the classical power module architecture is dominating the market for 600V, 1200V and 1700V voltage range and is increasing its share in the range of 3.3kV and up to 6.5kV.

The classical power module architecture is defined by a solid base plate on which an insulated ceramic substrate is soldered. The silicon chips are soldered to these substrates along with load current leads and leads for control and auxiliary contacts. Aluminium bond wires are used to connect the top side contact of the silicon chip to the conductive paths on the substrate.

For approximately 25 years, the improvement and optimization of the classical module concept was sufficient for most applications. But in the last 10 years, new applications and higher requirements in existing applications evolved, which demand a noticeable increase in the reliability of power modules.

Automotive applications aim to deploy the engine cooling system for the heat extraction from power modules in hybrid propulsion systems and therefore demand for extended application temperatures without reduction in lifetime. This is a major driving factor for the currently proceeding transition to an extended operation temperature range (-40°C≤Tj≤175°C).

Also new SiC devices, which physically allow much higher junction temperatures than Si devices, cannot be exploited to there capability because of the limited reliability of classical modules. These examples illustrate the need for increasing the reliability of power modules.

The lifetime of power modules is limited by the stress induced by thermal cycles. Since thermal cycles are inevitable in power electronic applications, a way must be found to reduce the stress generated in the module. One way to reduce the stress is to eliminate the rigid coupling between elements of the module construction wherever this is possible. Because the solder connection between elements of the module construction result in such a rigid coupling these solder connections must be eliminated or – if that is not feasible – be replaced by a more reliable connection.

2. The base plate problem

The base plate is a general feature of the classical module architecture. It serves as a mounting platform for substrates during the assembly process and delivers a flat interface for mounting the module to a heat sink. As this base plate is in the thermal path from the chip to the heat sink it must be thermally high conductive. Therefore Cu has become the standard base plate material. But the difference in thermal expansion between the copper base plate and the ceramic DBC substrate causes a bending of the base plate when the joined base plate/substrate system is cooled down from the solidus temperature of the solder material. This leads to a curvature of the system resulting in a hollow gap in the centre of the mounting surface to the heat sink.

Base plate-substrate system before and after

Figure 1. Base plate/substrate system before and after [1]

Fig. 1 illustrates this effect along with possible countermeasures. The traditional solution is shown in the centre schematics: the base plate is bow stamped to maintain a bow better suitable to establish a good thermal contact to the heat sink after mounting. The drawback of this design is that the solder interface between the base plate and the substrate is now convex shaped, which is not desirable for an optimized thermal resistance. The schematic on the right side of Fig. 1 is an improved solution with a machined bow. This design establishes a solder layer of even thickness between base plate and substrate while still maintaining a desirable bow of the base plate after the solder process.

The bow machined base plate seems to solve the problem, but the stress induced into the solder interface is way beyond the elastic range of the solder layer, as can be seen in Fig. 2. The plastic creep of the solder layer leads to a relaxation of the system over time. Therefore, it is only possible to define the optimal bow of the base plate for a selected time after the solder process. Above all, the mounting of the base plate to a heat sink surface changes the stress in the solder interface and the solder will react again with a relaxation process, once more changing the geometrical shape of the interface over time.

Relaxation of the base plate as a function of time after the solder process

Figure 2. Relaxation of the base plate as a function of time after the solder process

This perception leads to the idea to build a module without a base plate. A simulation illustrates the consequence on the stress induced in a module with and without a base plate. A 3-dimesional thermal model was calculated for an IGBT of 13.6 x 13.6mm² area for a dissipated power of 270W [3]. Assuming a homogeneous temperature in a centre column of 3 x 3mm², the temperature rise of the layers relative to the heat sink temperature was calculated for a system with and without base plate and the linear thermal expansion for this column was calculated with the coefficients of thermal expansion (CTE) in both systems. Figure 3 shows the simulated temperatures and layer elongations inside the centre column. The junction temperature of the base plate module is lower than the junction temperature of the system without base plate.

Simulation of a base plate module and a module without base plate using Al2O3 substrates

Figure 3. Simulation of a base plate module and a module without base plate using Al2O3 substrates

This effect is caused by the missing thermal spreading of the 3mm thick Cu base plate and the difference in ceramic thickness used for this comparison. But the difference in thermal expansion is lower in the system without base plate, because the substrate with the higher CTE has a lower temperature (because it is closer to the heat sink) and the silicon chip with the low CTE has a higher temperature. Therefore, the stress induced in the solder layer between chip and substrate is reduced, even though a higher junction temperature is observed.

This comparison shows that the value of the absolute junction temperature is not sufficient to evaluate the reliability of different system designs. For stress conditions, where solder fatigue is the dominating failure mechanism, the system without base plate will have a higher lifetime because of the reduced stress, even though the maximum junction temperature is higher. But the module design without base plate shows its real advantage, when ceramic materials other than Al2O3 are implemented. Figure 4 shows a simulation comparison of two base plate designs with a design without base plate for AlN substrates. The same geometry and power dissipation as in Figure 3 was applied. The higher thermal conductivity of AlN yields lower junction temperatures for all models. The base plate module with a Cu base plate shows an extreme thermal mismatch between the substrate and the base plate, resulting in reduced lifetime of this construction.

Simulation of a base plate module with different base plate materials

Figure 4. Simulation of a base plate module with different base plate materials and a module without base plate using AlN substrates [3]

By replacing Cu with AlSiC as base plate material this mismatch can be reduced [4] at the cost of a higher junction temperature due to the smaller thermal conductivity of AlSiC. The design without base plate yields the lowest junction temperature and the smallest mismatch between the chip and the substrate and is therefore the superior solution. This was confirmed by end-of-life power cycling tests [5].

The design without base plate also has a penalty for the application: the thermal capacity of the base plate is not available anymore. The thermal impedance for the three systems simulated in Fig. 4 is displayed in Fig. 5.

Thermal Impendances

Figure 5. Thermal Impendances for the simulated models displayed in Figure 4 [3]

The Cu base plate module with AlN substrate does not meet reliability requirements. The comparison with the AlSiC system shows that the thermal impedance of the system without a base plate is higher in the range between 50ms and 500ms. But it must be emphasized that this advantage of the base plat module is only observed for single events or events with a very low duty cycle, so that the base plate has a chance to dissipate the heat before the next pulse occurs. Sine waves between 2 and 20Hz do not fulfil this requirement. Only for applications where single overload events with long pauses in this time range are the dominating stress factor base plate modules should be preferred.

The elimination of the base plate has another important consequence: a major failure mechanism that is responsible for the limited temperature cycling lifetime of classical base plate modules is eliminated. So modules without base plate have proven to survive more than 1000 temperature cycles (-40/+125°C).

3. The problem of contact leads

After removing the large area solder layer between the base plate and the ceramic substrate by simply eliminating the base plate completely, there are still contact leads for the load and control contact which are attached to the substrate by solder interfaces in the classical module design. The load contacts are primarily made of copper to provide the high electrical and thermal conductivity required by the high currents – therefore, they are subject to the same CTE mismatch even though they feature smaller footprints than the base plate solder connection.

A method of optimization for the load terminals is to reduce the cross section of the solder contact to a minimum. This is illustrated in Fig. 6a, where the solder interface is even smaller than the cross section of the load terminal as realized in the SKiM5 module. This contact design provides a good thermal contact for the heat dissipated in the load terminal.

Load and Control terminal constructions to increase the module reliability

Figure 6. Load and Control terminal constructions to increase the module reliability a) reduction of interface cross section (SKiM5) b) wire bond connection (Econopack)

A solder-less solution is displayed in Fig. 6b as realized in the Infineon Econopack module, where Aluminium wire bonds are used to connect the load terminals injected into the thermoplastic case to the substrate. The major handicap of this architecture is the poor thermal conductivity through the wire bonds, resulting in a limited current capability of the leads. A second problem is the sensitivity to micro-vibrations, which can destroy the wire bond contact and therefore restricts the reliability of the module.

Load and Control terminal constructions to increase the module reliability

Figure 7. Load and Control terminal constructions to increase the module reliability a) spring contacts (MiniSKiiP) and b) pressure contacts (SKiiP)

An improved method to connect the load and control terminals to the substrate is the implementation of spring contacts. These springs are made from an electrically high conductive spring material. Load and control contacts are achieved with identical springs with a capability of 20A continuous current in the MiniSKiiP II module (Fig. 7a). Higher currents are conducted by paralleling of springs.

Even though there is no rigid connection between the spring and the substrate, the high contact force in the interface – resulting in a contact pressure comparable to a nut-and-bolt connection – establishes a gas tight contact interface which is stable even in high corrosive environments [6].

The current capability of springs is limited, because the mechanical and conductive properties cannot be optimized independently. This problem is solved in Fig. 7b, where the load terminals are designed as massive Cu bars, while the pressure is generated by a pressure system above the load terminals. A highperformance silicone foam is compressed by a rigid moulded steal plate and distributes the pressure homogeneously on the load terminals. This construction maintains the sufficient contact pressure during high temperature storage and temperature cycling tests [5].

The avoidance of a rigid solder connecting between the substrate and the load and control leads allows a relative movement between the terminals and the substrate and impedes fatigue phenomena in the interface.

4. Pressure contact reliability

Concerns relating the electrical contact stability of spring and pressure contacts lead to an intensive investigation program to verify the reliability of electrical pressure contact under various stress condition in application environments.

Since the pressure system gains its tension by mounting the module with a stiff pressure plate to a solid heat sink, high contact forces can be achieved. The contact pressure reaches values of 20-100N/mm² and is therefore more comparable to a nut-and-bold connection with typically 50N/mm² than to a conventional plug connector with a contact pressure of 10N/mm² and below, so that experiences with plug connectors cannot be transferred to the high pressure contacts. Still a relative movement in the contact interface cannot be avoided under thermal stress.

Micro-vibration test result

Figure 8. Micro-vibration test result on 2 pairs of springs in a MiniSKiiP module: Resistance change during 4.65 million cycles with 50μm displacement (1Hz)

The impact of micro-movements inside the contact interface was investigated in extended micro-vibration tests. The test result in Fig. 8 shows, that after an initial improvement of the interface resistance the resistance remains stable for a current in the range of 20mA with a voltage limited to 20mV for millions of cycles.

While the contact stability is not affected by mircovibrations of different amplitudes and even contamination of the PCB pads with dust or silicone soft mould, it is more sensitive to temperature cycling, where along with the mechanical movement generated by the differences in thermal expansion, additional temperature gradients are present.

In temperature cycling tests (-40/+125°C) an impact of the material combination at the contact interface was confirmed. Ag as the optimal surface plating for the spring had already been identified after temperature cycling tests [7]. Further tests revealed that the contact can be improved by furnishing the DBC substrate surface with a Ni/Au flash. Fig. 9 displays the result of a temperature cycling experiment. The contact system was cycled without current. After every 50 cycles the contact resistance was re-measured under ‘dry-circuit-conditions’ (i.e. with a current in the range of 10mA and with a voltage limitation of 50mV). The results show that the Ag spring on a Ni/Au substrate surface yields a high contact stability independent of the PCB surface plating. Therefore, almost any leadfree PCB plating can be used for spring contacts.

Temperature cycling

Figure 9. Temperature cycling (-40/+125°C) of Mini-SKiiP II contact system. No current during cycling – intermediate resistance measurement (10mA/<50mV)

In power electronic applications, corrosive environmental conditions have to be taken into account. A variety of different corrosive atmosphere tests has been conducted with spring contact systems. High concentrations of H2S assisted by a high relative humidity were investigated as well as mixed corrosive gas atmospheres with combined H2S, SO2, NOx and Cl2 pollution. All results verified the high contact reliability in a corrosive atmosphere. Even establishing a spring contact to an extremely aged PCB which was additionally stored in a highly corrosive atmosphere proved that the high contact force results in a good electrical contact, which became indistinguishable from a new contact interface after a few temperature cycles.

Concerns related to a possibility of silver migration in relation with corrosive atmosphere conditions and the silver plating of the spring have lead to the electromigration tests. Fig. 10 shows the landing pads of a Ni/Au plated PCB after a typical test. The discoloured spots on the landing pad consist of Ag2S and Cu2S. The source of the Cu is the tarnish protection of the spring. These corrosion products spread on the gold surface of the pad by surface diffusion. These spots show no dissymmetry in the direction of the electric field, nor could any traces of silver migration be detected in energy dispersive X-ray analysis (EDX).

A not discoloured area can be seen directly in the centre of the spots, where the electrical contact area of the spring is located. In these contact areas, no corrosion products can be detected by EDX. This suggests, that the high pressure spring contact establishes a virtually gas tight contact region, where no corrosive gas penetrates. This is confirmed by the resistance measurements, which show no degeneration of the contact interface after corrosive atmosphere tests [6].

Pads of a Ni-Au-flash

Figure 10. Pads of a Ni/Au-flash PCB after corrosive atmosphere test (3ppm H2S, 40°C, 80% RH, 2000h, 15V bias). Bias polarity indicated

All given examples discussed the contact stability for small currents <10mA, because low currents are the most critical condition for the contact stability. At higher currents and voltages, as applied on control and load contacts, the electrical contacts are far less susceptible for contact instabilities [8].

5. Advantage of pressure contact modules without base plate

A construction principle of a pressure contact module without base plate is shown in Fig. 11. The DBC substrate with the soldered and bonded power dies is pressed directly to the heat sink without the need for a rigid base plate. The large area joint between DBC substrate and heat sink is not soldered and the substrate has the ability to ‘move’ on the heat sink with virtually no limitation in terms of temperature cycling. Since the DBC is soldered to the base plate in the traditional designs it becomes necessary to limit the individual DBC size and place several small DBCs inside the power module in order to overcome the CTE mismatch problems.

Pressure Contact power module without base plate

Figure 11. Pressure Contact power module without base plate (Housing not shown )

Here the non-rigid connection from DBC to heat sink allows for large DBC designs with multiple chips in parallel and complex circuits on the same substrate. Additional internal DBC interconnections known from modules with base plate for power and gate traces are avoided, leading to a better switching performance, less parts and a lower module internal parasitic resistance RCC'-EE'.

Solution without base plate on a single DBC

Figure 12. Solution without base plate on a single DBC

As an example for a large DBC design Fig. 12 is showing the complete 350A sixpack circuit on a single 115mm x 80mm AlN substrate of the SKiM4 module. The same configuration and power range of a design with base plate consisting of 6 small DBCs is displayed in Fig. 13 for comparison.

To achieve a good thermal resistance multiple pressure contacts next to each die are used, keeping the DBC flat and no bimetal effect known from modules with base plates is disturbing the thermal performance. This construction also allows for a thinner layer of thermal grease (typically 20μm) compared to modules with base plate (typically 100μm). Since the thermal conductivity of thermal grease is only about 1 W/mK, this has a significant impact on the over all thermal performance, compensating well for the limited thermal spreading of the missing copper base plate [9].

Base plate solution with 6 small DBCs and internal connectors

Figure 13. Base plate solution with 6 small DBCs and internal connectors

In comparison modules without base plate have an equal or in most cases better thermal resistance than modules with base plate, especially for substrates with a high thermal conductivity like AlN.

In modern high current rating pressure contact designs the pressure distribution to achieve a good thermal resistance is realized by stamped and folded copper bus bars that contact the DBC next to each IGBT and freewheeling diode, eliminating the need of an additional plastic pressure spreading element [10]. The pressure is applied by a rigid pressure plate and a high performance silicone spring foam that clamps the bus bar sandwich and the DBC to the heat sink as described before.

This unique bus bar design (Fig. 14) of a DC positive, DC negative and AC bus bar sandwich with electrical contacts next to each chip provides a low inductive, low resistive symmetrical connection between main terminals and chips. The high contact force assures a low contact resistance, no additional solder joints are necessary.

Low inductive pressure contact bus bar

Figure 14. Low inductive pressure contact bus bar

The main advantage if this design is the low inductive bus bar construction on top of the paralleled dies. It seams that the chips on the right side, close to the DC terminals have a lower commutation inductance to the DC Link capacitor than the chips at the left side close to the AC terminal. During switching transients this should lead to an inductive voltage drop across the device and an asymmetrical switching performance along the row of paralleled IGBTs. But, due to the close magnetic coupling of the current carrying bus bar sandwich the voltage drop across the bus bar can be neglected and in fact each die sees almost the same effective inductance. It is insignificant, if the chip is located close to the DC terminal or at the AC terminal. This corresponds well with the minimum measured voltage difference across the module displayed in Fig. 15. The internal peak voltage at a turn-off with IC=400A, VCC=700V and di/dt=6600A/μs was measured between 929V and 936V at chip positions close to DC terminal, in middle position and close to AC terminal, which correlates to an effective inductance of only 1nH [11].

IGBT turnoff

Figure 15. IGBT turn-off Voltage measurement at DCterminal screws and at different Chip positions

In addition, the multiple electrical bus bar contacts next to each die cause a further reduction of the thermal impedance due to the additional copper heat path next to each die. This effect leads to an additional reduction of the module thermal resistance of about 5%. A final advantage of a pressure contact design is the implementation of spring contacts, which can be placed at any position of the DBC independent from the DBC layout and bond wires. The spring contact goes straight up to the driver PCB, only limited by high voltage spacing restrictions to the adjacent potentials. This gives the module designer the freedom to choose the ideal gate and auxiliary emitter position on the DBC layout to further improve the switching behavior and reduce module losses.

The combination of these technologies merged in one module design illustrates very clear the synergy, where the advantage of the combination is greater than the sum of the benefits of the single technologies.

6. Sinter technology

New fields of high power inverter systems – such as hybrid cars – require new ways of power electronics integration and packaging. The stringent requirements in size and weight are driving the operation temperatures of power electronics beyond the limits of today’s industrial applications. As described before the traditional soldered die attachment and bond wires are reaching their reliability limits for chip operation temperatures above 125°C. The low temperature diffusion sinter technology has been developed to replace the solder connection between die and DBC substrate already over 10 years ago [12], but only until recently SEMIKRON has developed a high volume multi chip attachment process that allows the simultaneous sinter attachment of IGBT, diodes and temperature sensors in one process step [13].

This proven high reliable joining technique extends the power cycling capability of modern power modules to those values that are required for high temperature applications and enable the full utilization of 175°C rated silicon chips without compromising today’s established reliability standards for lower operation temperatures.

Currently, solder is the material of choice for attachment and interconnection of semiconductor die to the substrate. However solder alloys are limited to reach higher die operating temperatures due to these low melting points. Silver is a desirable material for high temperature packaging applications, but the high process or sinter temperatures (>400°C), would destroy or change the performance of various silicon chips. General investigations about the use of pressure to lower the sintering temperature of silver powder compounds for attaching power semiconductor devices have been performed by independent organizations. The studies confirmed that pressure assisted low temperature sintering of silver pastes is a feasible alternative to traditional die attachment processes [14]. In the pressure assisted sinter process the temperature can be reduced to about 200-250°C dependent on the silver particle size, the particle coating material and the applied pressure. Higher operation pressure will generally lead to lower operation temperatures.

Once the sinter process is finished, it is not possible to separate the joint unless it is heated above the Ag melting temperature of 962°C.

A drawback of the sinter process is the need of noble metal on all joining surfaces. While power semiconductor chips typically have a silver backside metallization anyway, the copper layer on the substrate (DBC) needs an additional treatment which is typical a Ni-plating with an Au-flash. However for pressure contact modules this does not add additional cost since these modules anyway require a noble metal surface on the DBC to assure a high reliable electrical contact of the auxiliary spring contacts.

6.1 Material properties

In Table 1 the advantage of the Ag sinter layer over SnAg(3) solder is shown. The melting temperature of 962°C is the key to achieve higher reliability at high operating temperatures.

The great potential of the diffusion sintering technology can be illustrated best by the concept of the ‘homologous temperature’ as utilized by metallurgists. The homologous temperature of a material is defined as the ratio of the operation temperature to the melting temperature, both temperatures given in Kelvin. A comparison between the Ag diffusion sinter layer and available solder materials is shown in Figure 16.

Material properties of sintered Ag-layer

Table 1. Material properties of sintered Ag-layer compared to a standard SnAg(3) solder

For homologous temperatures below 40% material properties are considered little affected by temperature; between 40% and 60% in the so called creep range the material properties are sensitive to strain, whereas above 60% materials a considered unable to bear engineering loads in a structure. The comparison shows, that standard solder material SnAg(3) as well as high melting AuGe(3) with a liquidus temperature of 363°C are considered not capable of reliably bearing stress, while an Ag diffusion sinter layer is expected to be little affected by thermal stress. This expectation was confirmed by various power cycling tests [15].

Homologous temperature of Ag diffusion sinter layer

Figure 16. Homologous temperature of Ag diffusion sinter layer compared to selected solder materials for an operation temperature of 150°C

The thermal conductivity of the sinter layer of 240 W/mK together with the layer thickness of only 20μm makes the thermal resistance chip to DBC about 15 times better than a standard solder layer. This is a large improvement mainly for the short time transient thermal impedance of the die in short circuit and overload conditions. In steady state the thermal resistance of a sintered power module is only about 5% better since the layer between chip and DBC has only a small impact on the over all thermal resistance. Finally the lower CTE together with the higher tensile strength make the sintered Ag layer a good choice to achieve higher power cycles than a solder joint.

Although this has not a large impact on the over all module resistance it is worth to mention that the electrical resistance of the silver layer is about 23 times better than the soldered layer.

6.2. Sinter process and equipment

First the silver paste is screen printed on the DBC substrate. Afterwards the dies are assembled with a pick and place equipment and the complete card is moved into the sinter press. The process time is similar to a vacuum solder process. A final washing process – as in the flux assisted solder paste process – is not required.

Ag layer before and after pressure sintering

Figure 17. REM image of a screen-printed Ag layer before and after pressure sintering with a connection inhibiting polished Si-wafer [12]

The processed sintered layer has a thickness of 20μm and the flakes have a remaining porosity of approximately 15%.

Sintered 5”x7” Substrate with IGBTs, diodes, gate resistors and temperature sensors connected in a single process step

Figure 18. Sintered 5”x7” Substrate with IGBTs, diodes, gate resistors and temperature sensors connected in a single process step

SEMIKRON has developed a hydraulic press that can adjust specified pressures and temperatures. With this equipment complete 5”x7” DBC cards with various silicon chips can be processed in one step. Die sizes from 2mm2 to 600mm2 have been successfully processed.

The process parameters of pressure, temperature and time are well controlled and verified, ready for high volume production.

7. The solder-free module

With the replacement of the solder layer by an Ag diffusion sinter layer in the die attach, the design of a 100% solder-free power module is completed. For two 1200V modules of this family, the design process is terminated and the qualification process has started. The SKiM63 module is a sixpack configuration with 300A current rating and the SKiM93 module is also a sixpack configuration with 450A rated current. Both modules are equipped with generation 4 IGBTs and CAL4 freewheeling diodes. All modules are produced with sinter technology [10].

Fig. 19 shows an explosion view of the final module design with all the technological features discussed above implemented.

power module

Figure 19. Explosion view of the SKiM93 power module – the first 100% solder-free sixpack power module

The qualification program has already started, but final results of the test program are not available today. A preliminary result of the first power cycling test on a SKiM63 module is depicted in Fig. 20. In this test, the centre phase leg of the sixpack module is cycled with a constant DC current of 209A. The heating time averages to 95s, while the cooling cycle lasts in the range of 60s. The module successfully survived already more than 25,000 cycles. The test will be continued.

power cycling test

Figure 20. Preliminary result of the first power cycling test with a SKiM63 power module – test in progress.

8. Conclusion

The reliability of the classical module design will not suffice for many of the new evolving applications in power electronics. Therefore, a new architecture had to be developed, capable of meeting the high reliability requirements of modern applications.

A major limitation of power module lifetime is the solder fatigue problem in the classical construction. It has been shown that the solder fatigue phenomenon is contributing to the end-of-life failure of power modules especially in the range of lower temperature swings, which are predominant in most applications [16].

The elimination of all solder interfaces requires the combination of several new technologies, each of them have been discussed in detail. Each new technology has a small advantage compared to the classical design. But the real progress lies in the combination of these technologies in a single architecture and the resulting synergy effects.

The potential of implementing large size substrates in modules without base plate reduces the number of interconnects required in solutions with several small size substrates on a base plate. The need for numerous pressure posts can beneficially be combined with the advantage of multi-contact pressure bus bar designs, which reduce the parasitic inductance to the optimum [17] and thus minimize the switching imbalance in highly paralleled modules. The application of springs adds an additional freedom of design in positioning the gate and auxiliary contacts exactly where they are best placed from the electrical point of view. Additionally, the substrate plating with a Ni/Au surface is not only a precondition for the Ag diffusion sinter process, but also increases the contact reliability of pressure and spring contacts. And finally, the implementation of the thermally high conductive Ag sinter interface is capable of reducing the unfavourable effect of the reduced chip thickness in modern chip designs on the thermal overload capacity and therefore the short circuit capability.

All these improvements were not in the centre of focus, when these new technologies were developed, but only surfaced when the technologies were combined. Therefore, the 100% solder-free module has a high potential to improve the electrical and the reliability performance of power modules considerably.

The qualification results and the practical experience in real application have to reveal, if these high expectations can be fulfilled.

9 Literature

[1] Schütze, T.; Berg, H.; Hierholzer, M.: Further Improvements in the Reliability of IGBT Modules. Proc. IEEE Industry Applications Conference IAS 33 (1998), pp. 1022-1025

[2] Poech, M. H.; Eisele, R.: A Modelling Approach to Assess the Creep Behaviour of Large-Area Solder Joints, Microelectronics Reliability 40 (2000), pp. 1653-1658

[3] Scheuermann, U.; Lutz, J.: High voltage power module with extended reliability, Proc. EPE99, Lausanne, CD-ROM (1999)

[4] Lefranc, G.; Licht, T.; Schultz, H.J.; Beinert, R.; Mitic, G.: Reliability Testing of High-Power Multi-Chip IGBT Modules, Microelectronics Reliability 40 (2000), pp. 1659-1663

[5] Scheuermann, U.: Advanced Power Modules with AlN-substrats – extending current capability and lifetime, Proc. PCIM, PE 12.5, pp. 309-314, Nürnberg, 2003

[6] Lang, F.; Scheuermann, U.: Reliability of Spring Pressure Contacts under Environmental Stress, Microelectronics Reliability 47 (2007), pp. 1761-1766

[7] Hornung, E.; Scheuermann, U.: Reliability of low current electrical spring contacts in power modules, Microelectronics Reliability 43/9-11 (2003), 1859-1864

[8] Slade, P.: Electrical Contacts – Principles and Applications, Marcel Dekker, New York, 1999 [

9] Beckedahl, P.; Tursky, W.; Scheuermann, U.: Packaging Considerations of an Integrated Inverter Module (IIM) for Hybrid Vehicles, Proc. PCIM, S6b1, pp. 778-783, Nürnberg, 2005

[10] Beckedahl, P.; Grasshoff, T.; Lederer, M.: A new power module concept for automotive applications, PCIM, Nürnberg, 2007

[11] Wintrich, A.; Beckedahl, P.; Wurm, T.: Electrical and thermal optimization of an automotive power module family, APE Conference Paris, 2007 Power cycling test DTj=110K 1.800 1.900 2.000 2.100 2.200 2.300 0 5.000 10.000 15.000 20.000 25.000 30.000 power cycles VCE,sat [mV] 110 120 130 140 150 160 Temperature [°C] VCE TOP VCE BOT Tj,max TOP Tj,max BOT

[12] Scheuermann, U.; Wiedl, P.: Low Temperature Joining Technology – a High Reliability Alternative to Solder Contacts, Workshop on Metal Ceramic Composites for Functional Applications, 181-192, Wien, 1997

[13] Goebl, C.; Beckedahl, P.; Braml, H.: Low Temperature Sinter Technology – Die Attachment for Automotive Power Electronic Applications, APE Conference Paris, Session 2b, 2006

[14] Mertens, C.: Die Niedertemperatur-Verbindungstechnik der Leistungselektronik, Ph.D. Thesis, 2004, Fortschritt-Berichte VDI, Reihe 21, Nr. 365

[15] Amro, R.; Lutz, J.; Rudzki, J.; Thoben, M.; Lindemann, A.: Double-Sided Low-Temperature Joining Technique for Power Cycling Capability at High Temperature, EPE 2005, Dresden, CDROM

[16] Scheuermann, U.; Hecht, U.: Power Cycling Lifetime of Advanced Power Modules for Different Temperature Swings, Proc. PCIM, PE4.5, pp. 59-64, Nürnberg, 2002

[17] Mourick, P.; Steger, J.; Tursky, W.: 750 A, 75 V MOSFET Power Module with Sub-nH Inductance, Proc. PCIM, PE4.5, pp. 59-64, Nürnberg, 2002

 

For more information, please read:

A new 3D power module packaging without bond wires

Intelligent Power Modules Drive Public Transport

Sinter Technology for Power Modules

From Packaging to Unpackaging - Trends in Power Semiconductor Modules

 

VN:F [1.9.17_1161]
Rating: 0.0/6 (0 votes cast)

This post was written by:

- who has written 92 posts on PowerGuru - Power Electronics Jobs & Basics & e-Car.


Contact the author

Leave a Reply

Newsletter

Tell us your opinion

What do you find most useful on PowerGuru?

View Results

Loading ... Loading ...

Related sites

  • Bodos Power
  • Epcos
  • LEM
  • Mersen
  • Proton Electrotex
  • SEMIKRON
  • SindoPower
  • Weidmüller

Reader/Author Log-in