Microwave Journal
www.microwavejournal.com/articles/32947-piezo-on-insulator-engineered-substrates-for-4g5g-rf-front-end-filters

Piezo-On-Insulator Engineered Substrates for 4G/5G RF Front-End Filters

October 10, 2019

The deployment of advanced 4G and 5G sub-6 GHz networks requires the introduction of new features and technologies by operators and phone makers. In order to provide access to the larger data bandwidth that the new network promises to deliver, the RF communication between the base station and user equipment must rely on a more complicated band setup. The complexity of the RF front-end module therefore increases dramatically and will have to integrate more than 100 filters to support all communication modes.

Different technologies are available to address the growing filter market but most struggle to meet the more stringent requirements demanded by the 5G networks. New Piezo-On-Insulator (POI) substrates, however, allow the manufacturing of high performance, integrated surface acoustic wave (SAW) filter components that can meet the requirements of 5G networks. These filters can be used in smartphone front-end modules along with the power amplifiers, switches and antenna tuners devices that are already manufactured using RF-SOI substrates or others.

5G Challenges of Front-End Modules

With 5G, larger RF spectrum provides access to data rates twenty times higher than 4G data rates. Devices connected simultaneously will multiply, resulting in a connection density of a thousand times higher than what is available today. All devices using the mobile network will be impacted by the arrival of this new standard.

To provide data rate in excess of 20 Gb/s, acoustic filters need to adapt to the complex challenges related to 5G networks: more bands, bands with larger bandwidths, higher frequency bands, many band combinations to support the different carrier aggregation (CA) modes and MIMO antenna design.

In order to achieve these new requirements, the signal selectivity needs to be more precise. For this, it is important to enable resonators that have an extremely low temperature coefficient factor (TCF), typically lower than 10 ppm/K, while providing a high Q-factor, Bode Q typically greater than 2000. Also, out of band rejection must be considered much more carefully in order to support the different carrier aggregation and MIMO features.

The optimization of energy consumption in the front-end module remains a key concern. The components must also limit insertion losses so that at equal power levels, the signal travel as far as possible and at the same time the device must dissipate the power very efficiently.

A proliferation of components inside the smartphone front-end modules are greatly constraining the available space. There are already more than 60 filters in the current high-end phones, and we should expect to see more than a hundred in the next generation of high-end phones. Each filter addresses a specific RF band and requires unique design and performance characteristics. Integrating such a high number of different components in a very limited space causes many challenges for design and manufacturing teams. For these reasons, form factor, thermal dissipation and improved performance are becoming critical characteristics of the filters inside the front-end modules.

Market Needs

Until recently, there were two main filter technologies to select the signal in smartphones. The piezoelectric material would generate acoustic waves that could propagate freely on the surface of the material (SAW), or through the bulk of the active layers (BAW: bulk acoustic wave).

Current SAW filters are very well suited for low and medium 4G frequency bands but are limited when addressing the more difficult 5G requirements (high TCF, low Q, low coupling factor) and higher frequencies. The SAW filter frequency response is sensitive to temperature variations due to the high thermal expansion of the substrate (usually lithium tantalate or lithium niobate). This issue of temperature sensitivity can be compensated partially by adding a layer on top of the metal layer at the end of the device fabrication process, but this layer affects the coupling efficiency and the final performance of the filter.

The BAW filter technology allows filters to operate at higher frequencies with good performance but cannot be thinned down as much as SAW filters, creating module integration challenges. In addition, it requires more complicated manufacturing processes and offers limited integration of a filter multiplexer or filter duplexer on the same die.

Thin Film Piezo-on-Insulator

Figure 1

Figure 1 Structure of the POI engineered substrate (a) and resulting SAW propagation (b).

Because it is no longer possible to compromise on some of the performance criteria and in response to the more stringent requirements demanded by the new 5G network features, Soitec has developed a new engineered substrate that enables operators and phone makers to respond to these challenges. POI engineered substrates consist of a thin layer of single-crystal piezo material (today single-crystal lithium tantalate) on top of a SiO2 layer and a high resistivity silicon substrate as shown in Figure 1a. The top lithium tantalate thickness typically ranges from 0.3 to 1 μm. This thin film POI engineered substrate is built using Soitec Smart-Cut™ technology which allows for high uniformity layers and high-quality volume manufacturing.

This structure guides the acoustic wave at the surface of the substrate and confines its energy in the top thin lithium tantalate layer with very low losses (see Figure 1b). With this engineered substrate, filter designers have access to a substrate material with a better coupling factor (k2) and a lower thermal expansion coefficient. This enables them to design resonators with a high-quality factor at higher frequencies and target larger bandwidth filters with low temperature sensitivity. It also provides the capability to integrate multiple filters on the same die.

The POI substrate is composed of three layers, a piezoelectric material, a buried oxide and a silicon layer. The thin and highly uniform piezoelectric layer confines the energy of the guided wave, enabling high performance acoustic characteristics. The buried oxide selects and guides only high velocity waves and constrains the piezo material, reducing thermal expansion and in-turn temperature sensitivity. This structure allows for high signal selectivity, as well as frequency stability when temperature changes. Therefore, it also simplifies the manufacturing process compared to TC-SAW since filter device manufacturers do not need to add a thick layer on top to constrain the piezo material, thus improving the coupling efficiency.

Table 1

Figure 2

Figure 2 SAW resonator design.

The very low insertion losses achievable by a SAW filter designed on a POI substrate allows device manufacturers to efficiently manage energy consumption. Compared to existing solutions, SAW filters on POI have a high Q-factor, high coupling for large bandwidth filters, extremely low TCF and efficient integration of filters on the same die (see Table 1).



In addition, it should be noted that designing filters on POI substrates requires very similar skills to those required for designing SAW filters on bulk piezo wafers and manufacturing devices on POI substrates is straightforward (standard metal layer deposition for the main part) using a small number of manufacturing process steps.

SAW Resonator and Filter Designs on POI

Measured performance of SAW resonators built on lithium tantalate wafers and SAW resonators built on thin film POI were prepared and characterized. The results demonstrate the performance improvement of the POI substrate. For this experiment, a single port resonator of dipoles using 120 finger pairs and 20 electrodes on each side acting as mirrors was manufactured. The acoustic aperture was set at 40 λ and the distance between fingers and electrodes set at 1.2 μm with a ratio metal/spacing of 0.5 (see Figure 2).

A 1.6 GHz central frequency was targeted for those resonators and used tip probing to measure their characteristics. The POI substrates used had the following characteristics: 600 nm thick (YX)/42° LiTaO3 on a 500 nm thick SiO2 on a high resistivity Si (100 crystal).

Figure 3

Figure 3 Resonator k2 measurement of bulk and POI substrates.

Coupling k2

Figure 4

Figure 4 POI-based resonator response - max. Q at anti-resonance.

Figure 5

Figure 5 Sensitivity of velocity vs. temperature.

The coupling k2 of the POI reached 8.13 percent when the bulk LiTaO3 wafers used for conventional TC-SAW devices was limited to 5.98 percent (see Figure 3). k2 is calculated as 1-fr2/fa2 (fr is the resonance frequency and fa is the anti-resonance frequency). The benefits of the higher k2 provided by the POI substrate enable design of larger bandwidth filters to address some of the new 5G bands (up to 6 percent bandwidth of the center frequency).

Q-factor

Another significant performance improvement of the POI substrate appears on the Bode Q factor at anti-resonance. Under the same conditions, the Q-factor of the bulk LiTaO3 reached 935 compared to 2200 on the POI substrate (see Figure 4). This value should enable SAW filters to compete against BAW filters in the L- and C-Bands.

Temperature Compensated Factor (TCF)

The TCF factor (cubic polynomial fit) on the POI substrate is also significantly reduced. We are able to achieve much less than 20 ppm/K (typically below 10 ppm/K) while the bulk LiTaO3 would be around 40 ppm/K. Figure 5 shows the quasi-compensation of temperature effect on a 1.4 GHz resonator - second order effect is notable (TCF1 = −1.93 ppm/K, TCF2 = 403.5 ppb/K)

Based on our characterization work on the resonators, a ladder architecture was designed and simulated for a SAW filter at 2 GHz. The filter resonantors were implemented on POI wafers and performance measured. There was no target frequency band, but rather the filter was designed to take advantage of the modes of propagation that the POI substrate can enable.

The resulting extrapolated filter had 80 MHz bandwidth (1 dB band), less than 2 dB insertion loss, rejection greater than 40 dB and a group delay variation of about 50 ns or better. The absolute TCF was under 10 ppm/K over the entire operating range. Bandpass characteristics could be further optimized with filter design, but these results illustrate what can be achieved on this new platform (see Figure 6).

Figure 6

Figure 6 Transmission and group delay of the 2 GHz SAW filter fabricated on POI: wideband (a) and passband (b) performance.

Summary

The 5G roll-out is bringing a set of new challenges to front-end module devices, including form factor, thermal dissipation and performance. Improved performance is needed in order to achieve a successful roll-out. Filters are playing a key role since their numbers are increasing dramatically to support the new bands requirements in front-end modules.

Soitec has developed a new type of engineered substrate consisting of a very thin and uniform single-crystal lithium tantalate layer on a thin silicon oxide layer on high resistivity handle substrate using its Smart Cut™technology. The proposed solution provides resonators and filters with figures of merit in line with 5G filter requirements, particularly regarding the quality and coupling factors that are both greatly improved compared to standard SAWs on bulk piezoelectric material.

Designing, integrating and manufacturing filters on POI wafers remains straightforward as it relies on similar techniques used for SAW devices fabrication. SAW filters on POI can also compete with BAW filters for required frequencies in L- and S-Bands as they also bring the required performance.

The piezoelectric material expertise associated with Smart Cut™ technology allows Soitec to manufacture large volumes of uniform engineered substrates in their dedicated production line and is available to support the stringent filter requirements of 5G.