Log in

How to Optimize 5G SAW Filters Using OnScale

May 21, 2019


OnScale is the first CAE package to bring enterprise level simulation to the cloud. Once you upload your problem, you immediately gain access to each PC level computation and are charged on a pay-as-you-go basis.

Every additional core-hour you use for your problem only costs $10 per core-hour – so our is extremely customized to your exact project needs with the goal of helping you cut costs on unnecessary license fees. In addition, our scalable HPC platform can run billion degrees of freedom problems, in parallel!  Our simulation capability along with our flexible licencing system delivered by a SaaS model allows you to accelerate R&D processes in our very secure and high performing cloud through

What is a Core-Hour?

A Core-Hour is a measurement of computational time by the processor core. In OnScale, if you run one CPU for one hour, that is one core-hour. If you run 1000 CPUs for 1 hour, then that is 1000 core-hours. Simple!

In this blog post we will discuss OnScale’s capabilities compared with other legacy CAE solutions, the types of simulations we can do with SAW filters, and our suggested techniques to exploring expansive design spaces using OnScale.

Register to Watch Full Webinar on this Topic

OnScale Capabilities

Mechanical Wave Propagation Problem:
3D SAW Filter Device
Piezoelectric Problem:
Guided Wave Propagation Through CFRP Laminate Structure


First example above is a basic mechanical wave propagation problem. The second is a fully coupled piezoelectric simulation. RF Filter models often incorporate elements of both of these simulation types. If you take a close look at the “runtime” for both problems above, and compare OnScale with a known legacy multiphysics code – you’ll see that OnScale is able to process problems 100-1000x faster than other legacy software, bringing a new approach to RF filters simulation.

In the case of the legacy CAE solution, we’re very restricted by simulation time and capability as we can only analyze one simulation at a time. This means that we can only focus on small simulations and will probably only explore a few designs before moving forward to a prototyping stage. This carries a lot of risk, cost and time. With OnScale, we can evaluate thousands of potential designs in parallel on the cloud which allows us to completely understand a large design space. Once we’ve selected a candidate design, we can evaluate it in full 3D, allowing us to reduce risk cost and time to market.

Unit Cell Getting Started (minutes)

Full 3D Performance verification (hours)

Cross Section /3D Finger Exploring design options (10s of minutes)

We can tackle the simulation of SAW devices on a number of levels. The first is a simple unit cell model, which can be used to begin to flush out a design. It runs in less than a minute. We can then move on to larger models looking at cross sections or full 3D models of soft fingers which allow more subtle aspects of the design to be analyzed. These simulations run in a mere tens of minutes. Running many of these simulations in parallel, allow us to cover a very wide design space in a short amount of time. Finally, by moving to a full 3D design, we can get real performance verification for our design before we get the final result of the design process. We can track our basic KPIs such as insertion loss, queue, and resonant frequency all the way through, to device and even system level effects.


Here’s how OnScale’s capabilities can be applied to wafer bonded TC SAWS. This simple unit cell model of a single port Surface Acoustic Wave (SAW) filter allows rapid simulation of device performance. This allows a very wide range of designs to be explored before moving to more complex full 3D simulations. The model comprises a pair aluminum electrodes on a piezoelectric substrate. The substrate can either be Lithium Tantalate (LiTaO3) or Lithium Niobate (LiNbO3). The model is set up to simulate a Y-cut, which can be rotated to the desired angle. The base design generates an SH mode at 1.5 GHz using a finger pitch of 1.3 µm.In this case, we use an example of a 36 degree Y-cup LiTaO3 piezo layer bonded to a silicon substrate. The unit cell model gives us a very simple way to model an idealized SAW structure by applying periodic boundaries in both lateral directions. We create an infinite structure while minimizing the model size.

Model Geometry


Register to Watch Full Webinar on this Topic

This particular input file can be downloaded at the end of our SAW 3D Unit Cell tutorial. The file can then be found in File Explorer. By opening in the IVE this allows you to view the code behind the template. Here you will see there are a number of design variables, allowing you to sweep a very large number of designs. OnScale allows you to run a preview of your model beforehand so that you can see the model geometry and the effects of any changes in variables that you’ve made. Before running your simulation, you can run a quick estimate to see how long the job will take, the RAM required and how many core hours are required.You can then run your simulation on the cloud to begin the process of uploading your model to OnScale.

This particular simulation will run for around 2000 cycles to allow the device to ring down to a steady state. When the job is completed, you will be presented with a screen which shows you all of your results which are currently stored on the cloud. You can select the job which you just ran, click on the simulation and right click to download all of the results from that particular run. Once the results have downloaded, you can move to push processing mode by clicking at the top right of the screen. Once again, use the file explorer to open the relevant files. This time select a time data file and a data output file. By selecting the voltage, time history and clicking impedance, this will generate an impedance spectrum for your run. Then plot it in a log scale, allowing you to view it in the window. You can create multiple views which can be used to visualize the mode shape at resonance alongside the impedance.

You can enter custom scaling, and by running the video for multiple loops you can see how the device is behaving. Once finished with post-processing, the work space can be cleared and you can move back to the ID. When revisiting the design parameters, you will see that two have been marked with the SIM X command, which means that they can be varied in cloud sweeps. If you reopen the cloud schedule you will see that the two variables selected with SMI X are available for either thing. Now this allows you to perform a sweep. As an example, you can sweep the electrode thickness from 150 to 250 nanometers in 10 steps and sweep the PS effectiveness from 0.2 microns to one micron in 10 steps. This is going to create a simulation sweep with a total number of a hundred different designs. If needed, you can change the number of cores that are allocated to this job and view the estimated amount of time that is required. This time when you click run,  you will see that a hundred jobs are being spooled up. The jobs will all progress at slightly different speeds depending on if the thickness of a piezo layer is varied. The thicker the piezo layer, the more time required for calculation. Once the job is complete, you will see the cloud simulations closed down and all of your simulations which can all be downloaded.

As well as having links to Matlab, OnScale has some local options for bulk processing of large scripts, which can produce a very similar view.

Register to Watch Full Webinar on this Topic

Full 3D SAW Simulation

Above are our example results from a design with a piezo thickness of 0.6 microns and an electrode thickness of 200 nano-meters and simulated in full 3D. To do this, we’ve defined the device with a hundred pairs of IVTs along with 20 grating fingers at each end. The grating fingers are connected to electrical ground and the total device covers an area of 315 by 58 microns. The simulation requires just over 34 million degrees of freedom and runs just over four hours on a 16 core node. As you can see from the results, a number of things have changed since moving from the basic unit cell model. The first is  an expected reduction in Q as we now have more mechanisms on edges where we’re losing energy. The second is that we have some spurs evident in the impedance plot. We can also see some additional strong resonance modes above two gigahertz.


This was only a short list of the examples we have which are purely focused on SAW design, but we have a range of other topics including SMRS and FBARS covered for any of these models. On our examples page there are many more that you are able to download and run.

Why not download the SAW 3D Unit Cell model we used in this blog post and try it for yourself! Post your results and don’t forget to tag us!


Download OnScale for free today!

Your Comment

Your email address will not be published.