Final Report Summary - MATPLAN (CONSTRUCTION OF BESPOKE EVALUATION POWER MODULES~(MATPLAN))
Executive Summary:
Recently, silicon carbide (SiC) has been widely considered as an excellent alternative for the high power semiconductor devices in the application of energy-saving fields such as EV/HEV, wind energy, smart grid and aviation. As is well known, the operation temperature of standard semiconductor devices, originating from the material properties, has been a bottleneck for most power module with the range of temperature of -60 to 200°C. This can be overcome with the use of wide bandgap materials for example Silicon carbide. The advantages of silicon carbide include high temperature operation, high voltage blocking capability, high speed switching, and high energy efficiency compared with silicon transistor technology. However, the reliability of the SiC semiconductor and the surrounding package material is still under investigation. In order to take full advantage of wide bandgap semiconductor devices, a fully sintered SiC power module with low stray inductance has been investigated to meet the requirement of high temperature operation. A 10 kW four leg SiC power module using Sandwich Technology and SiC-JFET dies was designed in this project
The Project consisted of two main areas :-
1) Low voltage 1700V and below (work packages 1,3 and 4)
The module was designed to the following target specification. work package 4 .
Current & voltage:-1200 V, 25 A (phase-leg rating)
Cooling:-Single side cooled
Coolant temp / back pressure etc:-cold plate, max temperature 115 °C
Choice of SiC die:-SiC JFET plus Schottky diode
Electrical and thermal specification:-Half bridge (leg) per substrate junction temperature range -60 to 200 °C
The resultant module consisted of two single switches to form a half bridge configuration, each of which had four SiC-MOSFET and four matched SiC-SBD. The module outlines used new packaging concept which included silver sintering , a silicon nitride substrate, no wire bonds , no baseplate and a flexible PCB foil with integrated terminals. By using the newly developed SiC-based chip set and the new packaging technonlogy the temperature of the SiC power module could be operated at 200 ºC as per the specification .
1) High voltage 3300V and above (Work package 2) .
The second is being focussed at a higher voltage option where a larger standoff is necessary, this used technology, based on an electroforming process, Selective Layer Additive Manufacture Process. This novel approach is based on the 2000 year old lost wax process incorporating a low temperature plating process to create an electroformed shell.
Project Context and Objectives:
The Project consisted of two main areas :-
1) Low voltage 1700V and below (Work packages 1,3 and 4) :- this resulted in 10 planar power modules
2) High voltage 3300V and above (Work package 2):- this resulted in a process suitable for the standoffs required for high voltage modules.
Objective 1 :- Low voltage 1700V and below (Work packages 1,3 and 4),
The following target specification was given to the project from the topic manager
Table 1 Original project Specification
From this specification a fully Silicon carbide wire-less half-bridge power switch module was designed based on the Ag sintering and flexible PCB technologies which can be seen in figure 1 and 3. The module accommodates 8 SiC devices ,4 SiC junction gate field-effect transistors (JFETs) and 4 SiC Schottky diodes and 2 Cu conductive supports. The 2 Copper conductive supports were used to achieve the interconnection between the substrate and the flexible PCB, these would be replaced in a high voltage module with the SLAM process supports. Consideration was given to a) the material used for the packaging due to the temperature specification of -60 to 200°C A list can be seen in table 2 b) inductance of the module :- The inductance of the package was considered very important since silicon carbide has higher switching frequencies than silicon and therefore the inductance has a greater influence of the module performance. The electrical layout and the inductance calculation cab be seen in figures 5 &6
Figure 1 Power Module
Figure 3 Cross Section of Power Module
Table 2 Material selection list
Figure 5 Electrical Layout
Figure 6 Equivalent circuit with extracted inductance at (a) 1kHz (b) 100kHz
One of the most unique features of the power module was the flexible PCB foil with integrated interconnections instead of wire bonds and busbars inside the module. This design (see Figure 10) of the PCB has a range of benefits both electrically, thermally ,manufacturability and reliability compared to standard terminal attachments. Various problems were encountered due the very small size of the Silicon carbide die and the manual processing equipment available to Dynex. After investigation using 3D Tomography, it was observed that the gate connection had a weak /void like connection , this connection was strengthen by the addition of springs to the lid which due to the pressure connection of the lid would result in a better connection in service. In mass production of the sintered subassembly a pick and place machine could be used to eminlate this issue
Figure 10 Cross section of PCB
Figure 30 below shows the final process route adopted to manufacture the prototype power module.
Figure 30 Final process route
Objective 2:- High voltage 3300V and above (Work package 2)
The second activity was to develop a process to produce compliant pillars for interconnecting between the die and top DCB. The process considered was based a variant of the lost wax process described in the schematic shown in Figure14
Figure 14 Schematic of Selective layer additive Manufacture process
A number of trials were performed by developing a low temperature process for
coating wax mandrels. The first samples were too granular and too thin, burnt, crystalline and brittle. Follow up trials showed some improvement (Figure 8 ) but plating thickness were still insufficient to provide strength for handling - this was manifest in rupture of the skin during wax removal with an air gun.
Figure 15 Silver plated wax coupons
A number of exercises have been performed to improve this situation:
a) Acquisition and development of a high build copper plating process which will enable plating thicknesses of 200 micron per hour as opposed to the previous 15 micron per hour.
One important facet of this plating solution is the inclusion of micro-macro leveler additives to ensure even build up over irregular shapes.
b) Autoclave techniques have also been developed to enable a stress free extraction of the wax from the produced shell.
A third activity which has been essential to the future plating activity within this programme has been the design and build a number of exercises have been performed to improve this situation:
c) of a dedicated plating tank) The downside of this activity however has been a delay to the completion of work package.
Project Results:
Work Package 1: Module design, die and optimisation control
Dynex activity within work package 1 was to design a planar module to be manufactured in work package 4.
Project specification
The module was designed to the target specification in table 1
Table 1 Original project Specification
Overview of MATPLAN module outline
From the original specification a fully Silicon carbide wire-less half-bridge power switch module was designed based on the Ag sintering and flexible PCB technologies which can be seen in figure 1. Figure 2 and 3 show the construction of the Module which had a compact structural volume of 54 mm × 53 mm × 10 mm.
Figure 1 3D View of the Power module
Figure 2 View of power module without lid
Figure 3 Cross section of the whole module
Module components
It was crucial to identify which materials could be utilized or improved for the high temperature operation. Based on a comprehensive survey and reliability testing of component parts, the materials for each part of the power module package were compared and selected. The finally selected materials for each part are listed in Table 2
Table 2 Material selection list
Substrate Outline Design
As is well known, Silicon nitride has higher fracture strength than alumina and aluminium nitride. The coefficient of thermal expansion (CTE) of Si3N4 is ~2.9 ppm/oC and the thermal conductivity ranges from 50~70 W/m-K. Active metal brazed Cu- Si3N4 substrates are commercially available for power modules. Thus Si3N4 substrate is considered for enhancing high thermal reliability of SiC module.
Figure 4 substrate outline with dimensions. The Si3N4-based substrate is 30 mm × 27 mm × 0.92 mm in size and consists of 0.3 mm thick Cu sandwiched on both side of 0.32 mm thick Si3N4 tile. As shown in Fig. 5, it accommodates 8 SiC devices ,4 SiC junction gate field-effect transistors (JFETs) and 4 SiC Schottky diodes and 2 Cu conductive supports. The 2 Copper conductive supports were used to achieve the interconnection between the substrate and the flexible PCB.
Figure 4 Substrate outline
To enable silver sintering of the Silicon carbide devices the substrate was silver plated to the specification in table 3
Table 3 substrate material specification
Figure 5 gives the electrical layout of the substrate, a typical half-bridge configuration.The inductance of the package was considered very important since silicon carbide has higher switching frequencies than silicon and therefore the inductance has a greater influence of the module performance.
Generally, basic design rules for lower inductance were taken into account:
• Current loop as small as possible
• Paralleled sub-loops preferred
• Parasitic parameters should be matched
Fig. 6 shows the parasitic inductance values extracted with the FastHenry. It can be seen that the extracted values are indeed extremely low. Compared to standard substrate/ module layout .
Figure 5 electrical layout
Figure 6 Equivalent circuit with extracted inductance at (a) 1kHz (b) 100kHz
PCB layout design
One of the most unique features of this power module is a flexible PCB foil with integrated interconnections instead of wire bonds and busbars inside the module. This design ( see Figure 7 ) of the PCB has a range of benefits both electrically , thermally manufacturability and reliability compared to standard terminal attachments.
Figure 8 and 9 shows the topside and backside metal of PCB layout for sintering, respectively.
Figure 10 shows the cross section of the flexible PCB. The bottom metal layer is 0.1 mm thick Cu with Ag finish and directly bonded on the front sides of the SiC devices and the Cu conductive supports. The polymer1 layer is 0.1 mm thick polyimide embedded with a number of metal vias of 0.2 mm in diameter for forming conductive paths to carry currents or gate signals. The upper metal layer is also 0.1 mm thick Cu with Ag finish. This metal layer is designed as the main conductive layer to carry both the currents and gate signals. The polymer2 layer is 0.05 mm thick polyimide and mainly for facilitating the manufacturing of the flexible PCB and helping with tracking of the module. Table 4 explains the material considerations and geometric dimensions of PCB.
Table 4 PCB material specification
Figure 7 PCB Layout
Figure 8 Top metal of PCB Layout
Figure 9 PCB Backside metal layout
Figure 10 Cross section of PCB
Plastics Design
The plastic was designed for a high temperature plastic with special consideration being taken to radius to ensure the pcb did not fracture when assembled however the plastic designed in this work package was designed with a latch mechanism which was problematic to machine in work package 4 and had to be redesigned in work package 4
Figure 11-13 shows the 3d profile of the plastic frame and the lid.
Figure 11 3D profile of the plastic frame
Figure 12 3D profile of the plastic lid
Figure 13 Backside view of the plastic pillars
Work Package 2 Investigation technologies for planar module substrate fabrication
AMT activity, within work package 2, is to develop a process to produce compliant pillars for interconnecting between the die and top DCB.
The process being considered is based a variant of the lost wax process described in the schematic shown in Figure 14 below
Figure 14 Schematic of Selective layer additive Manufacture process
A number of trials were performed by developing a low temperature process for coating wax mandrels.
The first samples were too granular and too thin, burnt, crystalline and brittle.
Follow up trials showed some improvement (Figure 15) but plating thickness were still insufficient to provide strength for handling - this was manifest in rupture of the skin during wax removal with an air gun.
Figure 15 Silver plated wax coupons
A number of exercises have been performed to improve this situation:
a) Acquisition and development of a high build copper plating process which will enable plating thicknesses of 200 micron per hour as opposed to the previous 15 micron per hour.
One important facet of this plating solution is the inclusion of micro-macro leveler additives to ensure even build up over irregular shapes.
b) Autoclave techniques have also been developed to enable a stress free extraction of the wax from the produced shell.
A third activity which has been essential to the future plating activity within this programme has been the design and build a number of exercises have been performed to improve this situation:
c) of a dedicated plating tank (Figure 16) The downside of this activity however has been a delay to the completion of Deliverable D2
Figure 16 Development Plating Tank
An important demonstrator of the success of the project relates to the reliability of the modules when compared with a standard wire bonded and soldered product. It was agreed that this would be demonstrated by a thermal shock test between -60 and 200`C. AMT took delivery of a system with this potential capability (Figure 17). However some modifications and refurbishments were required to ensure this capability. This work has been completed during this period and thermal shock testing on a range of component parts is ready to start.
Figure 17 Thermal shock equipment
Work Package 3 Establish a low cost Manufacturing route
Areas addressed for optimisation of the manufacturability were:
1) AMT
a. The low temperature bonding process
As well bonding the die to substrate, the interconnect pillars require bonding to both die and substrate. In line with the introduction of low temperature bonding, this technique will be used for the pillar bonding. The benefits of using low temp bonding is the improved fatigue strength and the lack of a liquid phase which will improve registration capability
Potential processes being considered are :
a) Silver-tin transient liquid phase. Using fluidised bed
b) Nano silver solid state diffusion.
Previous work has demonstrated the potential for producing a high remelt joints following bonding at temperatures at approximately 250`C. Figures 18 to 20 show the mechanism for the transient liquid phase process, together with metallographic sections using single plating and multi layer plating techniques.
Figure 18 Equilibrium Diagram
Figure 19 single layer plating
Figure 20 Multi layer plating
Considerable work is reported in the technical press using nano silver with a press incorporating compliant interface to compensate for any thickness variations in the semiconductor die.
As there is potential for very large thickness differences (90 micron Igbt and 400 micron diode die) a novel solution using an integrated pick and place and bonding press was discussed and proposed. This is now well under development to bond die individually. This will allow more even pressure on each of the IGBT and diode die. A part-manufactured system is shown in fig 21
Figure 21 Photo of press/pick and place construction
During processing for transient liquid phase bonding, it is necessary to obtain bonding temperature a quickly as possible to ensure the silver-tin eutectic does not occur. A fluidised bas (fig 22) has been acquired and this is process of commissioning with extract for safe operation.
Figure 22 Fluidised Bed
b. Optimisation of manufacture of interconnect
During the first review period, development was centred on formation of pillars using the lost wax process to produce silver filled tubes for use as the interconnects.
For improved manufacturability, to make a more robust and accurate pillar and to obviate the need for time consuming production of wax mandrils, work was centred on the use of 3d printing of suitable plastic mandrils. This will enable multi structures to be manufactured in one pass.
To this end, AMT have acquired 2 systems, shown in Figure 23
Figure 23 Photo’s of Ormerod 3d printer and Eosint equipment
2) Dynex.
Dynex’s focus for work package 3 was the manufacturability of the design from work package one .There are 3 main areas which influence the manufacturability of the module , these are 1) sintering:- the bonding of the chips to the PCB and substrate, 2)the housing and 3) the encapsulation. Each one of these areas was investigated and the problems encountered are described in this section with a possible solution for large scale manufacture.
1. Sintering
1.1 Silicon carbide gate size issue
Problems were encountered due the very small size of the Silicon carbide die and the manual processing equipment available to Dynex. After investigation using 3D Tomography see figures 24 and 25 , it was observed that the gate connection had a weak /void like connection , this connection was strengthen by the addition of springs to the lid which due to the pressure connection of the lid would result in a better connection in service.
In mass production of the sintered subassembly a pick and place machine could be used to eminlate this issue
Figure 24 Xray images of the SAM
Figure 25 3D Tomography
1.2 PCB design
SiC die top side sintering to PCB is one of the most unique novelties in this project. Due to the small die size and gate area, sintering to the gate is even more challenging. A PCB with Cu filled-via structure is designed for the top side interconnect. Figure 26 shows the distribution and cross section of the PCB for top side Ag sintering.
Figure 26 Cu-filled via size and distribution & Cross section of PCB
The PCB supplier reported issues to manufacture the Cu-filled via structure at the beginning and the delivery was delayed for about 2 months. Their manufacturing processes was as follows. The Original PCB was bought from Du-pont with fully coated Cu on both sides of polyimide. They use laser to drill the Cu/Polymide/Cu sandwich structure, in which the Cu is 75 µm. Large amount of heat was generated during the top Cu drilling and the heat deformed or even melt the polyimide. Their final solution is to chemically etch the top Cu first and then use laser to drill the polyimide, which will only generate a little bit of heat. When the top Cu and polyimide were drilled off, the hole should be filled with Cu till a even top surface. The via in the final delivered PCB is not 100% Cu filled as we have asked. The via is only filled alongside the wall of the via. The via is actually a core-shell structure, as shown in Figure 27 , the centre is filled with polyimide and the wall is filled with a Cu shell of 20 µm thickness. The top polyimide and kapton layer is to protect the Cu surface. After a few iterations of the pcb the design of the pcb is now fixed and the manufacturing issues now resolved
Figure 27 PCB Core Cross Section
2. Housing
There was two main issues with the plastics the first was the insufficient flexibility of the PCB, which did not allow the structure to be placed in the module from above this resulted in lengthy redesign from a single frame to a two part frame as can be seen in figure .
A second issue was found that the seal was not sufficient to prevent the gel leaking, due to time restraints this was not addressed but for further evaluation this would need a n increase in size of the module to approx 58*57*10mm .
3. Encapsulation
The encapsulation for the module was a silicon gel however the standard power module gel could not be used because of the thermal specification of the module (-60 to 200°C). As can be seen in figures 29, the “old “standard gel became hard and brittle at minus 60°C. Investigation took place to evaluate various gels and a solution was found. However the new gel had a problem during the evacuation process of excesses foaming, this could be eliminated in future build by gelling and curing under vacuum.
Figure 28 Frame design
Figure29 New gel / old gel
WORK PACKAGE 4
Dynex’s focus for work package 4 was the manufacture of the design and initial electrical characterisation of the module.10 modules were delivered late September 2014, their assembly process was as follows Figure 30 :-
To summaries, although the module design for Matplan project is full of challenges, the novel planar SiC power module with fully silver sintering process was developed in this project . The main issues in the manufacture has been discussed and could be overcome in a large production environment by pick and place machine and under vacuum gel filling The problems encountered with the frame and the lid could be overcome with injection moulding of the plastics parts some modification .
The main design strategy of the SiC power module is described as well as the material considerations.
Finally, there are some keypoints involved in this new SiC planar power module:
1. The Double side sintering with SiC devices resulted in a module with ultra high operation temperature (>200°C), higher blocking voltage, lower power loss, greater reliability.
2. The design of the PCB structure integrates all of the signal contact tracks which are populated on the top side of the PCB, instead of the cost-consumed terminal attachments. This enables easy processing, ultra-compact structure and enhance increased reliabilit
Potential Impact:
For objective 1
In order to take full advantage of wide bandgap semiconductor devices, a fully sintered SiC power module with low stray inductance has been investigated to meet the requirement of high temperature operation. A 10 kW four leg SiC power module using Sandwich Technology and SiC-JFET dies was designed and manufactured in this project. The main issues in the manufacture was addressed and solution for large production environment was proposed for example a pick and place machine and under vacuum gel filling. These are currently being investigated for purchase in Dynex . The problems encountered with the frame and the lid could be overcome with injection moulding of the plastics parts, some modification will be required to the design in order to meet the requirements required for production.
This project has initiated further work into wide band gap materials for example long term reliability of the package with respect to temperature cycling , power cycling and switching characterization. This will be required to enable a decision to be made as to whether this is a viable alternative to standard production modules for certain applications
For objective 2:-
As part of the development of high voltage stand-off`s, this project has initiated detailed experiments into electroform plating. This work has precipitated the purchase of 3d printing technologies using both the high and low cost end equipment. As a result, a project on “high isostatic pressure for near netshape manufacture” has been applied for and approved by UK TSB funding.
Recently, silicon carbide (SiC) has been widely considered as an excellent alternative for the high power semiconductor devices in the application of energy-saving fields such as EV/HEV, wind energy, smart grid and aviation. As is well known, the operation temperature of standard semiconductor devices, originating from the material properties, has been a bottleneck for most power module with the range of temperature of -60 to 200°C. This can be overcome with the use of wide bandgap materials for example Silicon carbide. The advantages of silicon carbide include high temperature operation, high voltage blocking capability, high speed switching, and high energy efficiency compared with silicon transistor technology. However, the reliability of the SiC semiconductor and the surrounding package material is still under investigation. In order to take full advantage of wide bandgap semiconductor devices, a fully sintered SiC power module with low stray inductance has been investigated to meet the requirement of high temperature operation. A 10 kW four leg SiC power module using Sandwich Technology and SiC-JFET dies was designed in this project
The Project consisted of two main areas :-
1) Low voltage 1700V and below (work packages 1,3 and 4)
The module was designed to the following target specification. work package 4 .
Current & voltage:-1200 V, 25 A (phase-leg rating)
Cooling:-Single side cooled
Coolant temp / back pressure etc:-cold plate, max temperature 115 °C
Choice of SiC die:-SiC JFET plus Schottky diode
Electrical and thermal specification:-Half bridge (leg) per substrate junction temperature range -60 to 200 °C
The resultant module consisted of two single switches to form a half bridge configuration, each of which had four SiC-MOSFET and four matched SiC-SBD. The module outlines used new packaging concept which included silver sintering , a silicon nitride substrate, no wire bonds , no baseplate and a flexible PCB foil with integrated terminals. By using the newly developed SiC-based chip set and the new packaging technonlogy the temperature of the SiC power module could be operated at 200 ºC as per the specification .
1) High voltage 3300V and above (Work package 2) .
The second is being focussed at a higher voltage option where a larger standoff is necessary, this used technology, based on an electroforming process, Selective Layer Additive Manufacture Process. This novel approach is based on the 2000 year old lost wax process incorporating a low temperature plating process to create an electroformed shell.
Project Context and Objectives:
The Project consisted of two main areas :-
1) Low voltage 1700V and below (Work packages 1,3 and 4) :- this resulted in 10 planar power modules
2) High voltage 3300V and above (Work package 2):- this resulted in a process suitable for the standoffs required for high voltage modules.
Objective 1 :- Low voltage 1700V and below (Work packages 1,3 and 4),
The following target specification was given to the project from the topic manager
Table 1 Original project Specification
From this specification a fully Silicon carbide wire-less half-bridge power switch module was designed based on the Ag sintering and flexible PCB technologies which can be seen in figure 1 and 3. The module accommodates 8 SiC devices ,4 SiC junction gate field-effect transistors (JFETs) and 4 SiC Schottky diodes and 2 Cu conductive supports. The 2 Copper conductive supports were used to achieve the interconnection between the substrate and the flexible PCB, these would be replaced in a high voltage module with the SLAM process supports. Consideration was given to a) the material used for the packaging due to the temperature specification of -60 to 200°C A list can be seen in table 2 b) inductance of the module :- The inductance of the package was considered very important since silicon carbide has higher switching frequencies than silicon and therefore the inductance has a greater influence of the module performance. The electrical layout and the inductance calculation cab be seen in figures 5 &6
Figure 1 Power Module
Figure 3 Cross Section of Power Module
Table 2 Material selection list
Figure 5 Electrical Layout
Figure 6 Equivalent circuit with extracted inductance at (a) 1kHz (b) 100kHz
One of the most unique features of the power module was the flexible PCB foil with integrated interconnections instead of wire bonds and busbars inside the module. This design (see Figure 10) of the PCB has a range of benefits both electrically, thermally ,manufacturability and reliability compared to standard terminal attachments. Various problems were encountered due the very small size of the Silicon carbide die and the manual processing equipment available to Dynex. After investigation using 3D Tomography, it was observed that the gate connection had a weak /void like connection , this connection was strengthen by the addition of springs to the lid which due to the pressure connection of the lid would result in a better connection in service. In mass production of the sintered subassembly a pick and place machine could be used to eminlate this issue
Figure 10 Cross section of PCB
Figure 30 below shows the final process route adopted to manufacture the prototype power module.
Figure 30 Final process route
Objective 2:- High voltage 3300V and above (Work package 2)
The second activity was to develop a process to produce compliant pillars for interconnecting between the die and top DCB. The process considered was based a variant of the lost wax process described in the schematic shown in Figure14
Figure 14 Schematic of Selective layer additive Manufacture process
A number of trials were performed by developing a low temperature process for
coating wax mandrels. The first samples were too granular and too thin, burnt, crystalline and brittle. Follow up trials showed some improvement (Figure 8 ) but plating thickness were still insufficient to provide strength for handling - this was manifest in rupture of the skin during wax removal with an air gun.
Figure 15 Silver plated wax coupons
A number of exercises have been performed to improve this situation:
a) Acquisition and development of a high build copper plating process which will enable plating thicknesses of 200 micron per hour as opposed to the previous 15 micron per hour.
One important facet of this plating solution is the inclusion of micro-macro leveler additives to ensure even build up over irregular shapes.
b) Autoclave techniques have also been developed to enable a stress free extraction of the wax from the produced shell.
A third activity which has been essential to the future plating activity within this programme has been the design and build a number of exercises have been performed to improve this situation:
c) of a dedicated plating tank) The downside of this activity however has been a delay to the completion of work package.
Project Results:
Work Package 1: Module design, die and optimisation control
Dynex activity within work package 1 was to design a planar module to be manufactured in work package 4.
Project specification
The module was designed to the target specification in table 1
Table 1 Original project Specification
Overview of MATPLAN module outline
From the original specification a fully Silicon carbide wire-less half-bridge power switch module was designed based on the Ag sintering and flexible PCB technologies which can be seen in figure 1. Figure 2 and 3 show the construction of the Module which had a compact structural volume of 54 mm × 53 mm × 10 mm.
Figure 1 3D View of the Power module
Figure 2 View of power module without lid
Figure 3 Cross section of the whole module
Module components
It was crucial to identify which materials could be utilized or improved for the high temperature operation. Based on a comprehensive survey and reliability testing of component parts, the materials for each part of the power module package were compared and selected. The finally selected materials for each part are listed in Table 2
Table 2 Material selection list
Substrate Outline Design
As is well known, Silicon nitride has higher fracture strength than alumina and aluminium nitride. The coefficient of thermal expansion (CTE) of Si3N4 is ~2.9 ppm/oC and the thermal conductivity ranges from 50~70 W/m-K. Active metal brazed Cu- Si3N4 substrates are commercially available for power modules. Thus Si3N4 substrate is considered for enhancing high thermal reliability of SiC module.
Figure 4 substrate outline with dimensions. The Si3N4-based substrate is 30 mm × 27 mm × 0.92 mm in size and consists of 0.3 mm thick Cu sandwiched on both side of 0.32 mm thick Si3N4 tile. As shown in Fig. 5, it accommodates 8 SiC devices ,4 SiC junction gate field-effect transistors (JFETs) and 4 SiC Schottky diodes and 2 Cu conductive supports. The 2 Copper conductive supports were used to achieve the interconnection between the substrate and the flexible PCB.
Figure 4 Substrate outline
To enable silver sintering of the Silicon carbide devices the substrate was silver plated to the specification in table 3
Table 3 substrate material specification
Figure 5 gives the electrical layout of the substrate, a typical half-bridge configuration.The inductance of the package was considered very important since silicon carbide has higher switching frequencies than silicon and therefore the inductance has a greater influence of the module performance.
Generally, basic design rules for lower inductance were taken into account:
• Current loop as small as possible
• Paralleled sub-loops preferred
• Parasitic parameters should be matched
Fig. 6 shows the parasitic inductance values extracted with the FastHenry. It can be seen that the extracted values are indeed extremely low. Compared to standard substrate/ module layout .
Figure 5 electrical layout
Figure 6 Equivalent circuit with extracted inductance at (a) 1kHz (b) 100kHz
PCB layout design
One of the most unique features of this power module is a flexible PCB foil with integrated interconnections instead of wire bonds and busbars inside the module. This design ( see Figure 7 ) of the PCB has a range of benefits both electrically , thermally manufacturability and reliability compared to standard terminal attachments.
Figure 8 and 9 shows the topside and backside metal of PCB layout for sintering, respectively.
Figure 10 shows the cross section of the flexible PCB. The bottom metal layer is 0.1 mm thick Cu with Ag finish and directly bonded on the front sides of the SiC devices and the Cu conductive supports. The polymer1 layer is 0.1 mm thick polyimide embedded with a number of metal vias of 0.2 mm in diameter for forming conductive paths to carry currents or gate signals. The upper metal layer is also 0.1 mm thick Cu with Ag finish. This metal layer is designed as the main conductive layer to carry both the currents and gate signals. The polymer2 layer is 0.05 mm thick polyimide and mainly for facilitating the manufacturing of the flexible PCB and helping with tracking of the module. Table 4 explains the material considerations and geometric dimensions of PCB.
Table 4 PCB material specification
Figure 7 PCB Layout
Figure 8 Top metal of PCB Layout
Figure 9 PCB Backside metal layout
Figure 10 Cross section of PCB
Plastics Design
The plastic was designed for a high temperature plastic with special consideration being taken to radius to ensure the pcb did not fracture when assembled however the plastic designed in this work package was designed with a latch mechanism which was problematic to machine in work package 4 and had to be redesigned in work package 4
Figure 11-13 shows the 3d profile of the plastic frame and the lid.
Figure 11 3D profile of the plastic frame
Figure 12 3D profile of the plastic lid
Figure 13 Backside view of the plastic pillars
Work Package 2 Investigation technologies for planar module substrate fabrication
AMT activity, within work package 2, is to develop a process to produce compliant pillars for interconnecting between the die and top DCB.
The process being considered is based a variant of the lost wax process described in the schematic shown in Figure 14 below
Figure 14 Schematic of Selective layer additive Manufacture process
A number of trials were performed by developing a low temperature process for coating wax mandrels.
The first samples were too granular and too thin, burnt, crystalline and brittle.
Follow up trials showed some improvement (Figure 15) but plating thickness were still insufficient to provide strength for handling - this was manifest in rupture of the skin during wax removal with an air gun.
Figure 15 Silver plated wax coupons
A number of exercises have been performed to improve this situation:
a) Acquisition and development of a high build copper plating process which will enable plating thicknesses of 200 micron per hour as opposed to the previous 15 micron per hour.
One important facet of this plating solution is the inclusion of micro-macro leveler additives to ensure even build up over irregular shapes.
b) Autoclave techniques have also been developed to enable a stress free extraction of the wax from the produced shell.
A third activity which has been essential to the future plating activity within this programme has been the design and build a number of exercises have been performed to improve this situation:
c) of a dedicated plating tank (Figure 16) The downside of this activity however has been a delay to the completion of Deliverable D2
Figure 16 Development Plating Tank
An important demonstrator of the success of the project relates to the reliability of the modules when compared with a standard wire bonded and soldered product. It was agreed that this would be demonstrated by a thermal shock test between -60 and 200`C. AMT took delivery of a system with this potential capability (Figure 17). However some modifications and refurbishments were required to ensure this capability. This work has been completed during this period and thermal shock testing on a range of component parts is ready to start.
Figure 17 Thermal shock equipment
Work Package 3 Establish a low cost Manufacturing route
Areas addressed for optimisation of the manufacturability were:
1) AMT
a. The low temperature bonding process
As well bonding the die to substrate, the interconnect pillars require bonding to both die and substrate. In line with the introduction of low temperature bonding, this technique will be used for the pillar bonding. The benefits of using low temp bonding is the improved fatigue strength and the lack of a liquid phase which will improve registration capability
Potential processes being considered are :
a) Silver-tin transient liquid phase. Using fluidised bed
b) Nano silver solid state diffusion.
Previous work has demonstrated the potential for producing a high remelt joints following bonding at temperatures at approximately 250`C. Figures 18 to 20 show the mechanism for the transient liquid phase process, together with metallographic sections using single plating and multi layer plating techniques.
Figure 18 Equilibrium Diagram
Figure 19 single layer plating
Figure 20 Multi layer plating
Considerable work is reported in the technical press using nano silver with a press incorporating compliant interface to compensate for any thickness variations in the semiconductor die.
As there is potential for very large thickness differences (90 micron Igbt and 400 micron diode die) a novel solution using an integrated pick and place and bonding press was discussed and proposed. This is now well under development to bond die individually. This will allow more even pressure on each of the IGBT and diode die. A part-manufactured system is shown in fig 21
Figure 21 Photo of press/pick and place construction
During processing for transient liquid phase bonding, it is necessary to obtain bonding temperature a quickly as possible to ensure the silver-tin eutectic does not occur. A fluidised bas (fig 22) has been acquired and this is process of commissioning with extract for safe operation.
Figure 22 Fluidised Bed
b. Optimisation of manufacture of interconnect
During the first review period, development was centred on formation of pillars using the lost wax process to produce silver filled tubes for use as the interconnects.
For improved manufacturability, to make a more robust and accurate pillar and to obviate the need for time consuming production of wax mandrils, work was centred on the use of 3d printing of suitable plastic mandrils. This will enable multi structures to be manufactured in one pass.
To this end, AMT have acquired 2 systems, shown in Figure 23
Figure 23 Photo’s of Ormerod 3d printer and Eosint equipment
2) Dynex.
Dynex’s focus for work package 3 was the manufacturability of the design from work package one .There are 3 main areas which influence the manufacturability of the module , these are 1) sintering:- the bonding of the chips to the PCB and substrate, 2)the housing and 3) the encapsulation. Each one of these areas was investigated and the problems encountered are described in this section with a possible solution for large scale manufacture.
1. Sintering
1.1 Silicon carbide gate size issue
Problems were encountered due the very small size of the Silicon carbide die and the manual processing equipment available to Dynex. After investigation using 3D Tomography see figures 24 and 25 , it was observed that the gate connection had a weak /void like connection , this connection was strengthen by the addition of springs to the lid which due to the pressure connection of the lid would result in a better connection in service.
In mass production of the sintered subassembly a pick and place machine could be used to eminlate this issue
Figure 24 Xray images of the SAM
Figure 25 3D Tomography
1.2 PCB design
SiC die top side sintering to PCB is one of the most unique novelties in this project. Due to the small die size and gate area, sintering to the gate is even more challenging. A PCB with Cu filled-via structure is designed for the top side interconnect. Figure 26 shows the distribution and cross section of the PCB for top side Ag sintering.
Figure 26 Cu-filled via size and distribution & Cross section of PCB
The PCB supplier reported issues to manufacture the Cu-filled via structure at the beginning and the delivery was delayed for about 2 months. Their manufacturing processes was as follows. The Original PCB was bought from Du-pont with fully coated Cu on both sides of polyimide. They use laser to drill the Cu/Polymide/Cu sandwich structure, in which the Cu is 75 µm. Large amount of heat was generated during the top Cu drilling and the heat deformed or even melt the polyimide. Their final solution is to chemically etch the top Cu first and then use laser to drill the polyimide, which will only generate a little bit of heat. When the top Cu and polyimide were drilled off, the hole should be filled with Cu till a even top surface. The via in the final delivered PCB is not 100% Cu filled as we have asked. The via is only filled alongside the wall of the via. The via is actually a core-shell structure, as shown in Figure 27 , the centre is filled with polyimide and the wall is filled with a Cu shell of 20 µm thickness. The top polyimide and kapton layer is to protect the Cu surface. After a few iterations of the pcb the design of the pcb is now fixed and the manufacturing issues now resolved
Figure 27 PCB Core Cross Section
2. Housing
There was two main issues with the plastics the first was the insufficient flexibility of the PCB, which did not allow the structure to be placed in the module from above this resulted in lengthy redesign from a single frame to a two part frame as can be seen in figure .
A second issue was found that the seal was not sufficient to prevent the gel leaking, due to time restraints this was not addressed but for further evaluation this would need a n increase in size of the module to approx 58*57*10mm .
3. Encapsulation
The encapsulation for the module was a silicon gel however the standard power module gel could not be used because of the thermal specification of the module (-60 to 200°C). As can be seen in figures 29, the “old “standard gel became hard and brittle at minus 60°C. Investigation took place to evaluate various gels and a solution was found. However the new gel had a problem during the evacuation process of excesses foaming, this could be eliminated in future build by gelling and curing under vacuum.
Figure 28 Frame design
Figure29 New gel / old gel
WORK PACKAGE 4
Dynex’s focus for work package 4 was the manufacture of the design and initial electrical characterisation of the module.10 modules were delivered late September 2014, their assembly process was as follows Figure 30 :-
To summaries, although the module design for Matplan project is full of challenges, the novel planar SiC power module with fully silver sintering process was developed in this project . The main issues in the manufacture has been discussed and could be overcome in a large production environment by pick and place machine and under vacuum gel filling The problems encountered with the frame and the lid could be overcome with injection moulding of the plastics parts some modification .
The main design strategy of the SiC power module is described as well as the material considerations.
Finally, there are some keypoints involved in this new SiC planar power module:
1. The Double side sintering with SiC devices resulted in a module with ultra high operation temperature (>200°C), higher blocking voltage, lower power loss, greater reliability.
2. The design of the PCB structure integrates all of the signal contact tracks which are populated on the top side of the PCB, instead of the cost-consumed terminal attachments. This enables easy processing, ultra-compact structure and enhance increased reliabilit
Potential Impact:
For objective 1
In order to take full advantage of wide bandgap semiconductor devices, a fully sintered SiC power module with low stray inductance has been investigated to meet the requirement of high temperature operation. A 10 kW four leg SiC power module using Sandwich Technology and SiC-JFET dies was designed and manufactured in this project. The main issues in the manufacture was addressed and solution for large production environment was proposed for example a pick and place machine and under vacuum gel filling. These are currently being investigated for purchase in Dynex . The problems encountered with the frame and the lid could be overcome with injection moulding of the plastics parts, some modification will be required to the design in order to meet the requirements required for production.
This project has initiated further work into wide band gap materials for example long term reliability of the package with respect to temperature cycling , power cycling and switching characterization. This will be required to enable a decision to be made as to whether this is a viable alternative to standard production modules for certain applications
For objective 2:-
As part of the development of high voltage stand-off`s, this project has initiated detailed experiments into electroform plating. This work has precipitated the purchase of 3d printing technologies using both the high and low cost end equipment. As a result, a project on “high isostatic pressure for near netshape manufacture” has been applied for and approved by UK TSB funding.
