Packaging and Integration Strategy for mm-Wave Products

The fifth Generation (5G) mobile communication era is expected to address the insatiable need for data communication by introducing mm-Wave technology and protocols.

The unprecedented latencies offered by 5G Networks will enable users to indulge in gigabit speed immersive services regardless of geographical and time dependent factors. Cost effective ways of packaging these complex systems present a challenge that are being addressed. [1, 2, 3]

Different “flavors” of 5G are emerging, each with unique requirements and cost. For example,

• 5G for IoT: frequency below 1 GHz requiring low data rates, along with long-range coverage and at a low cost. Technologies are already available. Designed for low power and long battery life, they can use existing packaging technologies.

• 5G Sub-6GHz, 4G++: frequency below 6GHz marking the actual 4G technology's evolution. This includes massive (multiple input, multiple output) MIMO, more carrier aggregation and more front-end module integration will push the limits of current technologies.

• 5G mm-Wave: typically 28GHz, 39 GHz, or 60 GHz for short-range, high data-rate exchanges. Today, two specific use-cases are observed in the US and Korea. For example, fixed wireless access for Verizon and AT&T in the US. Trials are progressing, and deployment phases are scheduled in around 15 US cities. Also, 5G trial services have successfully supported the February 2018 Winter Olympics Games in Korea for KT.

These platforms are schematically represented in Figure 1 below.

Typically 4G electronic integration platforms involved densely packed PCB or substrate based “modules or SiP” with multiple wire-bond or flip-chip mounted bare or packaged dies and many SMT passives (such as resistors, capacitors and inductors, oscillators, crystals etc.). Majority of these modules use single sided assembly with advanced design rules. However, the antenna and matching circuitry are “external” to the module, typically on the main PCB. This requires major system level optimization for performance enhancement.

For 5G systems, especially for mm-Wave (>30GHz) applications, integrating the antenna in the package/SIP becomes a key enabler for most enhanced performance optimization. The types of implementation can include wire-bond, flip-chip, Fan out wafer level packaging (FOWLP), such as eWLB and the traditional LTCC based modules as shown in Figure 2 below. At these frequencies, wire-bonded packages will typically have challenges to meet the performance. LTCC based systems have cost constraints, so the majority of the next generation wireless systems is looking at flip chip or eWLB packaging.

We have previously discussed and compared the advantages and disadvantages of the various packaging solutions. In the following sections, we will provide several examples and case-studies for high frequency products.

Managing the performance requirements for next generation networking infrastructure equipment, especially for 5G frequencies, in a cost-effective manner, presents unique challenges. Unlike the mobile and consumer segments, the solutions here are generally not challenged by form factor and can be enhanced by addition of more antenna arrays and RFICs. However, care has to be taken to address the thermal solutions for effective heat dissipation as well as manufacturability issues as the package size may target ~400mm² for Gen 1 and double or triple the area for subsequent generations.

There are several possible ways of managing the diverse system requirements. Most system designers want to have the flexibility of providing ways of increasing the performance of the system by adding additional antenna arrays as well as RFICs. As a result, many are in pathfinding options with several different package form factors.

Since 5G cellular networks are still in the very early phase of rollout, the majority of the suppliers are embarking on “phased” approach for their product/sample rollout. In phase I, where they will provide mostly “demo” samples to their network customers for field trials, they want to have “fully tested” antenna as well as RFIC. As a consequence, the first set of demo products may have discrete antennas as well as fully tested, packaged ICs. In Phase 2, their plans may be to integrate either the antennas in the package or have direct IC attached to the package substrate. In the final phase they have a fully integrated AiP SiP or Module.

Additionally, the choice of substrate technology may widely vary with both the market segment and the performance needs. Prior to the recent advancement of cellular standards, previous implementations of mm-Wave RF technology have been focused on using high performance ceramic substrates for military, aerospace and defense. However, for commercial applications, we see a strong trend of using traditional laminate based substrate technology.

A key enabler for meeting the antenna propagation requirements at these high frequencies are the materials used in the substrate technologies. Many material suppliers have been developing low dk (dielectric constant) and low df (dissipation factor or loss-tangent) core, pre-preg and build up materials and are now in early commercialization stages with these materials.

Figure 3 (a) and (b) below show the Dk and Df values for several pre-preg materials as a function of frequency.

As shown in the graphs, materials D and E are adequate for high frequency applications. We have qualified both the pre-preg and core materials from vendor E. We also are continuing active collaboration with additional materials and substrate suppliers to qualify more materials for future product categories. Table 1 summarizes the properties at 10GHz

The other major contributor to antenna propagation loss and hence the overall performance of the product, is the roughness of copper surface. This phenomenon is well known. However, in the context of mm-wave commercial package development, this has to be studied and optimized for specific material set. Data for low loss build-up material, using the same substrate manufacturing process is shown in Figure 4 below.

For the lowest insertion loss and hence optimized signal propagation, the combination of flat copper surface and low loss substrate materials are needed to be optimized. However, for high-yield substrate manufacturing process, roughness of copper is needed for adhesion to the laminate film. Hence careful process development is needed to best optimize the substrate manufacturability and reliability and also maintain the performance requirement. This is shown in the following chart in Figure 5.

As seen in Figures 5 and 6, untreated copper is equivalent to treatment A. However, process development for adequate adhesion may require adhesion promoter.

One of the early demonstrator vehicles has been successfully assembled at our Korea SiP/Flip-chip operations. A typical schematic is shown below. The Antenna-on-Package (AoP) requires double-sided surface mount assembly of the parts, with the packaged RFIC (multiple flavors, with Tx and Rx ICs) mounted on the marsupial format (BGA-side).

One representative flavor can be summarized as follows:

• 23×23mm FBGA

• 10L substrate: ~1.0mm thick with low loss materials

• Packaged IC: FCCSP in marsupial configuration

• Discrete stamped antennas

• Passives for matching

• 1.0mm Ball pitch

• 5G cellular infrastructure

• Freq: 28 GHz

A typical assembly flow is shown in Figure 8. Due to the marsupial format of the RFIC design, the SiP packages will be subjected to 3 reflows during the assembly process, the 3^rd reflow being associated with the solder ball mount (SBM) process.

Figure 8 shows an optical inspection image after assembly.

Designing for mobile and consumer segments create additional challenges of form factor and also cost. However, in general, the performance requirements could be significantly less challenging and in many cases, the antenna elements may be designed external to the transceiver package. That is antenna can be implemented on the system board PCB.

Fan out wafer level packages, such as the eWLB packages pioneered by StatsChipPac Singapore factory provides one of the best performing mm-Wave solutions [4, 5]. eWLB structures inherently provide the shortest routing length and hence the lowest loss signal propagation path. Also the copper roughness for eWLB process is very smooth (typically the RMS value can be as low as 0.3mm and can be an order of magnitude smaller than the roughness of typical copper plating used in laminate based processes in substrate technology). A high performance 60GHz WiGig (802.11ad compliant system) using 2L eWLB package structure is shown below as a representative case study. The structure is shown in Figure 9 and is summarized below.

• eWLB 13.0mm × 13.0mm package

• 2RDL layers

• 0.5mm ball pitch

• 0.8mm total package thickness

• WiGig compliant

• Freq: 60GHz

Sample lots of WiGig compliant packages have been built and also successfully passed component level qualification (CLR).

With the push toward autonomous driving and automated vehicles (AV), the research and development programs for sensors enabling the above mentioned technologies have gone into high gear over the last several years. Short, mid and long range radar (SRR, MRR, LRR) sensors are essential components for the overall sensors needed for complete implementation of AVs. Many packages in these segments use advanced packaging options such as fan out eWLB, with lower parasitics at 77GHz.[6]

Many of the next generation radar ICs operate at 77GHz and hence managing overall insertion loss and signal integrity provides a challenge. eWLB packages provide the same advantages in this market segment as discussed in section III. We provide below a representative example of a Dual die MRR SiP eWLB package that is currently going through assembly and characterization builds.

• eWLB ~7.0mm × 9.0mm package

• Dual die

• 1RDL layer

• 0.5mm ball pitch

• 0.8mm total package thickness

• Radar module

• Freq: 77GHz

Many different packaging choices are emerging for commercial, cost-effective implementation of mm-Wave system/packages. There are many trade-offs that need to be considered for the specific application and the performance/cost targets. As discussed in the paper, we see a diversity of packaging platforms, depending on the end market requirements, such as networking, mobile or automotive radar systems. However, the industry has clearly driven many innovations in commercializing “low cost/high performance” mm-Wave packages based on commercial material sets, which are very different from traditional ceramic/LTCC substrate based military and defense systems.

Significant development by substrate and materials vendors, especially for low loss materials that can minimize the transmission loss has resulted in substrates that can be used in production ready in 2018. Continuous optimization is an ongoing part of the development.

A key requirement for the success of rapid commercialization of mm-Wave systems in the commercial sector is the close collaboration in the supply chain. The system design houses need to work closely with the packaging solution provider to optimize the implementation. Key enablers include (a) low loss material development, (b) commercialization and reliability of substrate technologies using those materials (c) effective integration of muli-layer RDL antenna structures in a FOWLP format and (d) reliability of such package configurations.

There are several key challenges still to be addressed before widespread commercialization. One area that needs significant development is testing. Establishing commercial test infrastructure and rolling it out to assembly houses (such as OSATs) need sustained development. The test community is trying to establish whether contact based testing will be sufficient or there will be need for over the air (OAA) non-contact testing for mm-Wave packages.[7]

The design of integrated antenna-in-package solutions also continues to be challenging. Using multiple substrate layers, complicated arrays of patch and phased array antennas can be fabricated. However, having multi-port antenna structures fabricated in fan out technology with inductance matching and efficient ground ports, continue to be problematic. Hence adoption of 3D structures with inductance matching mainly with Cu RDL structure, i.e., “Balun” structures is possible in fan out eWLB layers. These structures can be effectively used for performance gain as well as directional beaming. Inductors can also be built into the eWLB structure using the RDL as well as in the laminate packages using substrate embedded thin film cores.