I just want to report that I’ve had some limited success modifying the wixel antenna as follows. The zig-zag portion of the PC trace antenna (the side that has no connection at the board edge) was severed with a quick grind of a dremmel tool. The other half of the PC trace, that part that makes one loop and terminates at a circuit ground was left intact. The middle trace that formerly branched off in two directions was scraped to make soldering possible, and a 1/2 wavelength piece of insulated #24 wire was soldered at that point. I also coiled this short piece of wire (62.5MM) around a pencil, which resulted in about 3 turns, and then stretched it out, over a distance of about 30MM. This is similar to what tbabun did, except that I left the ground loop part of the trace intact, and used 1/2 wavelength instead of full (A full wavelength of 2.4Ghz is 125mm, so i/2 that is 62.5mm.), and I also had experimented with consolidating the wire (coiling), as described.
My tests were conducted making the above modification to a transmit only application, in which a short packet (20 bytes plus overhead) was sent 10x per second. The receive side (receive only) application simply gathered these packets, and maintained statistics based on the average LQ (radioLqi() checked after each good or bad packet receipt), the total number of packets received over the course of 2 seconds, and the total number of errant (bad checksum) errors. A baseline test was performed with the unmodified wixels separated by an ordinary house wall (plate over wood frame), about 2 meters total. This was a threshold in at which some occasional errors were expected, so I’d more easily see the positive or negative effect of each modification tested. A second test would then be done with the transmitting wixel placed 30 feet away, with two walls in-between. This was a threshold where I would begin to see as high as a 50% error rate with the unmodified wixels. My statistics were monitored by using the USB on the receive side as a virtual serial port, using a terminal program. Every two seconds, a report would be prepared and displayed.
To gauge overall quality of communication, I have to say that I have always found the LQ numbers to be all over the map. Other than statistically higher average numbers when the distance and wixel orientation is more ideal, it does not reliably correlate with the more important statistics of (a) how many packets are received within a given time frame when compared to the expected number, and (b) how many of the packets received were “good” (meaning, they did not result in an internal checksum error ( radioCrcPassed() returned 0).
So all that said, after trying several combinations of wavelength, with or without the ground loop severed, and with or without the added antenna wire coiled, and moving the transmitter antenna to a few different orientations, I found that with my 1/2 wave addition as described, there was a marked increase in the overall number of error free packets received when compared to the unmodified transmitting wixel, though I could NOT correlate this to a significantly higher an overall higher LQ statistic. This antenna combination also dropped my 50% error rate in the second test (30 feet apart, two walls between) almost half… so I’ll say about a 25% error rate. Further, the coiling of the antenna extension seemed to cause the improved receiver statistics to be less affected by changes in the transmitting antenna orientation.
Interestingly, some other variations were not as promising…
With the ground loop severed, neither a full wave nor a 1/2 wave antenna addition, coiled or not, seemed to be an improvement over the unmodified wixel. I assume this is because the loss of the PC trace ground loop put the overall antenna impedance way out of range.
I’m not an RF engineer, and so I have no clue why this would be (perhaps impedance again), but even with the ground loop trace intact, the half wave antenna noticeably outperformed the full wave addition.
Coiling the wire was an afterthought, to see if any added antenna was still worthwhile if it had to be crammed into a smaller space. As I mentioned, coiling the antenna did not seem to hamper the improvement at all, but it did seem to make the antenna orientation less significant.