How to tune antenna efficiency based on simulation

Every connected device needs an antenna to send and receive data wirelessly. A functional antenna can be a simple strip of copper which converts a conducted electromagnetic field on a printed circuit board to a radiated field which propagates through the air. The efficiency of this conversion influences the connection range as well as the power budget of the device. The simplest way to increase efficiency is the antenna tuning, also called antenna matching. At the same time this optimization can be achieved without an additional hardware loop. This article describes how the antenna can be tuned very accurate with a simulation-based approach.

The goal behind the antenna tuning is to convert your antenna impedance to the impedance of your transceiver, usually 50 ohms, for highest power efficiency. The antenna can be tuned with passive components such as inductors and capacitors. There are multiple white papers available which show how this impedance matching can be achieved (1). These white papers describe the theoretical concept very well. In practice, however, one is often faced with a totally different result. The next sections show the reason for that and how a simulation-based approach can provide more accurate antenna tuning results. The workflow will be demonstrated using increasingly complex and more accurate simulation models.

Let’s analyze an unmatched BLE antenna s-parameter file on a smith-chart, which has been measured by a vector network analyzer or short VNA (Figure 1). This is the baseline for our simulation. At the Bluetooth center frequency of 2.4GHz, the antenna impedance is 3.19ohms + j13.5ohms, far away from the desired 50 ohms. The Smith-Chart Software (2) is a nice tool to evaluate possible antenna matching topologies to tune the antenna. In this example a L-network was used to convert the impedance from 3.19ohms +j13.5ohms to the desired 50 ohms (Figure 2).

Once we simulate this network over the frequency range from 2.0 to 3.0 GHz, we have a very well-matched antenna (Figure 3). A return-loss of < -9dB at the band-edge means that at least 88% of incident power is received at the antenna, while the remaining 12% are reflected by the antenna. Component losses are ignored in this simulation so far.

If this component network is now soldered on the printed circuit board and measured with a VNA, the result will be different. The reason is that the simulation so far only used ideal-components and ignores any stray capacitance, inductance caused by the printed circuit board.

The following three simulations demonstrate iteratively the influence of idealizing components and of the mentioned stray capacitance and inductance caused by the printed circuit board. To that end, the s-parameter models of the used inductors are imported, the component distances are measured and the via inductances are approximated to be 0.2nH.

The return-loss graph (Figure 5) clearly shows the shift in resonance frequency of the antenna caused by non-ideal passive components, physical distance between the components and vias to the reference ground plane. This means that the antenna radiates at a frequency other than 2.4GHz, which results in efficiency degradation.

In the example above the trace lengths and additional via inductance were approximated. To eliminate the last bit of guesswork we must create a realistic simulation model of our pi-matching topology. In our workflow we create an electromagnetic model using a 3D high frequency simulation software called Ansys HFSS. The resulting s-parameter file of the electromagnetic model can then be included in antenna matching circuit simulation.

Once the simulation model is created, different component configurations can be simulated using circuit synthesis and optimization software, such as Optenni Lab for maximum performance of the antenna.

The simulation-based approach has multiple advantages. Whereas with a vector network analyzer usually only the return-loss (S11) is measured, the forward transmission coefficient (S21) is also available in the simulation. This information is needed to know how much of the incident power is received at the antenna port and thus also taking component losses into account. A further advantage is the sensitivity analysis of the passive components, as the component tolerance can have an influence on the return-loss and shift the resonance frequency of the antenna. This can be demonstrated on a different antenna matching topology which looks very good at first consideration (Figure 7).

The return loss with maximum of -10dB between 2.4 and 2.483GHz (Bluetooth/WiFi frequency band) is very good. The disadvantage of this topology is the inductance L1. The component value of L1 is very low with 0.56nH and its tolerance is very large (almost 20%). A tolerance analysis (Monte-Carlo simulation) shows that this small inductor leads to a large sensitivity in return-loss. As this tolerance might be too large, depending on the return-loss target, one solution could be to use tighter component tolerances or switch to another matching topology.

Once the final network is chosen, the components with the correct value will be soldered on the printed circuit board and measured again with the VNA to verify the simulation. Figure 9 shows the comparison of the simulated return-loss and the actual measurement captured with a VNA. This comparison shows how accurately the simulation matches reality.

Tuning the antenna with the help of a simulation has become the standard at Zuehlke Engineering, as it is very efficient, and the maximum performance can be extracted. In addition, a simulation can already be performed on the first prototype to know the maximum possible efficiency of a given design at an early stage of a project.

(1) https://abracon.com/uploads/resources/Abracon-White-Paper-Antenna-Impedance-Matching.pdf

(2) https://www.fritz.dellsperger.net/smith.html