Background InformationLast September captcha designed, tested, and optimized designs for dipole PCB antennas in order to allow anyone to simply order low cost and high performance antennas for their RF projects such as Moteino. He made these designs available on OSH Park here:
https://oshpark.com/profiles/captcha. Many forums users have used these designs and found them to perform better than the 1/4 monopole antenna which ship with the Moteino giving either a longer maximum range before signal dropout or a stronger signal at the same distance. Since then, captcha has done even more characterization of various antennas and shared his methodology and results here:
https://lowpowerlab.com/2017/08/17/simple-monopole-antenna-testing/. There is excellent information in both the original PCB antenna thread as well as the antenna shootout post and I do recommend that anyone interested check them out.
Motivation of Current WorkBuilding on captcha's original work, I wanted to offer smaller and much lower cost versions of the same PCB dipole antennas. The price from OSH Park probably isn't an issue for those of us who live in rich countries, but not everyone on these forums lives in a rich country. Plus since captcha's original work, we have become aware of lower cost PCB fabs which produce great work. At the same time, I shrank everything down as much as possible around the resonators to try to get the antenna to fit within the smallest (and cheapest) board area possible. There was also some discussion in the original antenna thread about wide resonators and the fact that they allow for wideband antennas and may also allow for a reduction in resonator length (remains to be tested at the time of writing).
An edge-mount RP-SMA connector has five prongs arranged in two rows. The center prong is the signal from the radio and the other four prongs are GND and are connected to the outside barrel of the connector itself. To keep things small but with wide resonators, I tried sneaking the resonator connected to ANT underneath one of these GND prongs. All this required was removing the top pad that this GND pin would connect to and Felix suggested adding some silkscreen to further separate the two. The final result of the connection from the RP-SMA to the two resonators looks like this:
Schematic of ANT Resonator Sneaking Under a GND PinClose Up Showing Unconnected GND PinThere is no electrical connection between the rightmost prong (GND) and the copper rectangle beneath it (ANT) even though they are only separated by the solder resist and silkscreen layers. From there, I just had to shrink the board around the resonators right up to the 15-mil offset that PCBs.io requires between copper and the board edge. The final result of this optimization is shown below for 915MHz.
Compact Design Compared to Baseline DesignSome savings in length but the real board area savings comes from compacting everything into the minimum allowable board width of 0.25".
Validation of RF PerformanceFinally, to ensure that this optimization didn't come at the expense of performance, I tested the average RSSI between a gateway equipped with the PCB antenna under test and a stock Moteino using a 1/4 wave monopole 50' away broadcasting at 300kbps at -18dbm (power level 0 on the RFM69CW) with some walls between them. Each packet sent out by the node has a packet ID in the payload so that the gateway knows if there was a missed packet and 20 packets were received by the gateway for each test. The RSSI from these 20 packets was averaged to allow comparison between a known good design and my experimental version. The results obtained were (larger or less negative values are better):
Captcha's reference standard: -88.90dbm and 1 missed packet
ChemE's experimental design: -85.45dbm and 0 missed packets
Perhaps it is a reduction in PCB around the resonators or perhaps it is due to the widening of the resonators, but this small test would suggest that the new design is no worse than the standard and perhaps it is more sensitive.
CostMy designs are shared on PCBs.io and the direct links and costs are below.
915MHz -
https://PCBs.io/share/4XRDg - $1.20 each vs. $4.37 at OSH Park
868MHz -
https://PCBs.io/share/46L3w - $1.28 each vs. $4.65 at OSH Park
433MHz -
https://PCBs.io/share/rJ36o - $2.56 each vs. $9.02 at OSH Park
A cost reduction of around 72% without any apparent sacrifice in performance. I will attach the eagle files to this post in case anyone would like to modify my designs. PCBs.io ships free to any country in the world (how is this possible?) so these should be usable and affordable to just about everyone.