This morning, some colleagues helped me to test the rf performance of one of my sigfox tracker. I’m really not an RF expert, I try to improve but there still be a long way to go…
In this design I’m using a TD1204 and a Wurth smd 868 antenna ; I have a 100 ohm impedance net between both (sure it is bad) and LC circuit (that sound not to be the good one – btw).
The process to to the test was to use a reference emitter using an antenna with 0db loss / gain and an emission power known ; measure the received power from this to another one with a spectrum analyzer. This gives a power loss reference due to the environment. This as been done in a specific room avoiding wave reflection. Then change the reference emitter by the device to be tested and see the difference in the power reception. Turn the device and test in different directions.
In the best case, the power loss with this antenna is -10db classing my device in the U2 device class. (not really good, but sound acceptable regarding the size of the object) The results vary a lot depending on the axes of the device regarding the antenna. This means with a such antenna you have to expect to have multiple antenna at different location to receive the message to ensure you get it !
So the antenna worked as expected even if the impedance of the circuit is not really the good one and if the ground plan was really short due to the sizing of my prototype. Interesting to see.
On the side it was interesting to see the sigfox channel hoping and the different power pic coming from this channel hoping and multiple transmission. Making sigfox protocol concrete on a scanner screen ! As much as we have seen the used frequency is around 868.1xx.
Now, I need to understand how to create a good matching circuit … seems to be a really more complicated question … any post around this could be interesting to read… please post link.
The new center frequency is 868.13MHz (old one was 868.2MHz) is the intermediate frequency base stations are listening and a 192KHz span. Considering 100Hz wide each channel, you have arround 1920 channels, but not all are usable with all transceivers in the market and channels in the middle, near IF frequency are not recommended because less sensitivity due to I/Q missmatches.
At the modem, you can know AT$IF? or set AT$IF= the center frequency used. The channel used is random around this center frequency. For tests you can use AT$ST to establish a fixed channel and use always the same in your tests.
I am also not good at RF skills, try to search a würth application note about the layout needed and how to adapt it to 50 ohm needed for the module.
Thank you for these precisions, really great !
First of all, most of Nestor’s comments are correct, except that base stations are not less sensitive near the “zero” because this is compensated (without it, as a matter of fact, you have usually degradation depending on gain distribution). This being said, in normal mode, a modem is “choosing” its frequency following a secrete and unique (programming) pseudo randomization linked to its unique id (synthesizer programming). This frequency is then physically transmitted with a natural error due to crystal error, and temp, adding some “randomization” to the process, and helping for capacity at the end (which is anti-intuitive, but works).
From matching stand point, the best way would be to measure real impedance of the antenna with a compensated cable (antenna disconnected from modem, but in its practical environment (ground, packaging, battery…Etc…)). By “compensated” I mean that the vector analyzer calibration must be done at the end of the last cable in the arrangement (or that all piece of cable not comprised during VNA calibration must be entered in length and velocity factor in the VNA menu). The antenna impedance can be then read, and the proper matching to the 50 Ohms “wanted” by the modem can be calculated, with a computer Smith program for instance (Smith, Smith Chart, MIMP…Etc…).
Without VNA, the old TFMS (Tune For Max Smoke) method can be used. It is longer and much more painful as it is a dichotomy process (5 to 6 path minimum, with relative measurements between trials).
Do not expect to have more than 3 or 4 dB of improvement, those small antennas having a maximum -6/-8 dB gain in the best direction despite what datasheets are saying.
Much better antennas can be obtained, even for miniaturized environment, through dedicated design using wires or PCB, taking complete environment into account…But it takes more effort and time indeed. Good luck for further improvement. Hope to see you soon on SIGFOX network ! Best Regards.