In the rapidly developing world of Internet of Things (IoT), the radio frequency front-end (RFFE) of smart devices will have to handle higher data rates and access the full bandwidth of 4G/5G wireless technology. The reason for this, of course, is the growing demands of ubiquitous low latency data at higher operating frequencies required to accommodate enhanced data transmission capabilities and rapidly growing numbers of users.
RFFE is made up of a power amplifier, filter/duplexer, clock, switch and low noise amplifier. The system architecture for 5G RFFE will be extremely complex in a smaller footprint, requiring fitting 100+ RF filters into a smart phone. Cost, power efficiency, available space, and the ability to be manufactured in large quantities are the major factors that will drive future RFFE designs. The best of future RF filter designs will utilize the simplest possible design process, with smallest possible footprint, while maximizing performance. Key performance parameters for a RF band-pass filter are 1) low insertion loss to the desired signals in the passband and 2) enough attenuation of undesired interference in the stopband. Quality (Q) factor determines the filter insertion loss and filter roll-off sharpness. Coupling factor determines the sustainable filter bandwidth. Both the Q and coupling factor are directly dependent on the resonator design, which makes resonator design key to achieving the required filter performance. The trick to achieving well-designed filters, is to connect high-Q resonators of various impedances that add up to a desired bandwidth.
Current resonators used for RF filter applications are based on two methods of the excitation and efficient energy trapping of an acoustic wave (4000 to 1200 m/s) propagating in a piezoelectric material.
OnScale Simulations of SAW and BAW based resonators for RF Filter applications
Method 1: Is based on Surface Acoustic Waves (SAW), where the surface acoustic wave is generated on a piezoelectric substrate by metal interdigital transducers (IDT’s) on the surface.
Method 2: Is based on Bulk Acoustic Waves (BAW), where the bulk acoustic wave is excited by the application of electric field through electrodes above and below a thin piezoelectric plate.
Full 3D Finite Element Analysis (FEA) models are required to accurately capture acoustic wave propagation in a resonator. It is also required to characterize the effect of different non-idealities in the resonator structure based on the energy confinement of the main mode of interest to achieve high-Q resonator designs. Acoustic wavelength in these resonators are in the order of 1 um to 4 um, which requires several million degrees of freedom (DoF) to obtain accurate results. Legacy FEA software tools typically take 1,000+ hours and 600 GB RAM to solve full 3D simulation of such acoustic wave structures. So full 3D numerical models can prove challenging and computationally expensive when designing RFFEs. This forces engineers to make assumptions (simulating only 2D, for example) or to rely on analytical models. This can greatly limit the designer’s ability to explore multiple concepts or to mitigate the effect of non-idealities to achieve high-Q resonator design.
Leveraging OnScale for accurate and fast 3D simulations of RF MEMS Resonator
OnScale (our cloud-enabled FEA software) can be used for modeling viscoelastic dissipation, and thermal effects on the propagation of acoustic waves in these piezoelectric devices. OnScale allows users to specify anisotropic material properties and to rotate the crystalline cut angle for full stiffness, piezoelectric, and permittivity tensors along with accurate absorbing boundary conditions for the absorption of outgoing acoustic wave necessary for accurate piezoelectric resonator simulations. With OnScale, engineers can run thousands of simulations and set-up Design of Experiment (DoE). Large 3D models can be executed across thousands of HPC cores. The OnScale’s new CAD import kernel also allows import of GDSII layout files for better usability. These unique capabilities enable better high-Q piezoelectric resonator designs for RF MEMS filter applications eliminating expensive, iterative physical prototyping.
Dispersion, Impedance, Q-factor for FBAR Resonator results from an OnScale simulation
Key performance indicators (KPI’s) for resonator design are resonance frequency, Q-factor, coupling coefficient and impedance for RF filter applications. The above figure shows the dispersion curve, impedance, phase, and Q-factor for the thin film AlN FBAR. The dispersion curve identifies all the possible modes, energy associated with each mode, the cross-over between the main mode of interest and other plate modes when the FBAR resonator is excited. The impedance and phase curves quantify the non-idealities in the FBAR resonator design and the overall impact on the Q-values.
These results were obtained from a single time-domain simulation with OnScale. In contrast, legacy FEA software requires different modal, harmonic and analytical analysis. A direct runtime comparison between OnScale with MPI and the leading legacy solver was performed using the 3D square FBAR design. Table 1 shows comparison of OnScale to a commonly used legacy code, where both, the solve time and RAM requirements, were reduced by a factor of 99%.
+OnScale time domain solver from a single simulation following Nyquist criterion provides a fine frequency resolution of less than 0.2 MHz for an entire frequency spectrum from 1 GHz to 12 GHz which helps to accurately capture not only the fundamental mode but also the third and other overtone modes. Based on a frequency range between series frequency of 2050 MHz and parallel frequency of 2090 MHz this translates to 200 frequency points.
High-Q RF MEMS Resonator Optimization
Once a designer has decided on a SAW or BAW acoustic wave resonator, he/she must then optimize the resonator structure in such a way that the acoustic energy is trapped within the main resonator / IDT region to achieve high Q-factor values, at the same time suppressing the spurious modes. Five important choices in the design and development of high-Q acoustic wave resonator are the selection of:
- Resonator structure design
- Electrode and pad optimization
- Packaging effects
- Temperature compensation
OnScale provides integration of all the possible design parameters of a resonator parameter to perform fast parametric analyses for full design space exploration to optimize the resonator performance. Crystal orientation is one of the key design parameters for SAW resonator-based RF filter for achieving the desired Q-factor and coupling coefficient. Parametric analysis carried out for the cut angle varying from 0 to 180 degrees helps to obtain the KPI’s (impedance and frequency). Entire analysis for a 3D unit cell for 181 simulations using OnScale provides the KPI’s in less than 3 minutes on 8 cores per simulation. Similar optimization studies with respect to metallization ratio, power-flow, wave propagation angle, temperature compensation can be performed in OnScale in few minutes (as compared to days or weeks with legacy FEA software).
SMR resonator-based RF-filters require energy confinement in the piezoelectric layer based on the optimal selection of Bragg reflector layer, piezoelectric film and electrode thickness. Figure below shows the optimization study carried out based on Monte-Carlo Analysis to define the optimal tolerance on the piezoelectric (AlN) film and electrode (Mo) thickness. Simulation results help to determine the optimal tolerance requirements. The entire 2D-SMR simulation for 1000 possible combinations of AlN and Mo thickness using OnScale provided the optimal tolerances in less than 15 minutes and 30 MB RAM requirements.
Optimal Tolerance Limits for AlN and Mo Thickness for Yield Improvements using OnScale
OnScale empowers RF design engineers to perform optimization studies for different design parameters of acoustic wave resonator. Engineers can extract the key design parameters based on 3D simulations in less than few hours, with minimal memory requirements.