Long Range Embedded Systems
Very recently we’ve had a proliferation of small cpus with Long Range transceivers that fit into a chip socket. I need one for an application so wanted to compare their RF and other capabilities.
What is Long Range Communication
Besides being a generic for communication that goes a long way it often refers to using the unlicensed 868MHz (EU) or 915MHz(US) frequency bands. They have limitations on how much data you can send and how much power you can use but they are otherwise pretty open.
Europe seems to be setting up LoRa on towers for cloud networks of IOT data. Pretty neat. You may read more anywhere (Wikipedia).
Here’s what I can reliably test to include in a comparison. In order:
- Receiver sensitivity. How much noise is in the receiver? How well does it block out nearby signals? How well matched to a 50ohm load is the antenna circuit? How well does it shrug off temperature changes?
- Transmitter performance. Does it hit the maximum legal limit? How well matched is the feed line for the antenna to 50 ohms? Does it generate additional garbage in the spectrum? Does it overheat?
- Antenna match. Are there commercial antennas it’s well matched to that give it good-in-class performance?
How far will this go?
Well, that’s the question. We can measure the answer deterministically.
When we send from one board to another the system looks like this ->
The example here shows a loss in the SPACE circle of 48dB. The actual loss from each foot of distance in space depends on what’s in that space, but it will be the same no matter how good your receiver is.
When the receiver calculates an RSSI value, that’s correlated to the power level it’s receiving.
So, if we know, for example:
- The transmit power is 20dBm (max)
- The receiver can receive down to -140 dBm
This means that we have a link budget of 160dB. That’s how much loss space and cabling can produce before we drop the signal.
Math: double the distance from the transmitter adds 6dB of loss, all things being equal.
Antennas are really complicated. If you’re an audiophile — an antenna is like a loudspeaker as a complex load. I care because I have to match boards to antennas.
Vendors like to claim that an antenna is a 50 ohm resistive load across the frequency band. They also claim the board receivers and transmitters are tuned for 50 ohm resistive at the center frequency. Sure.
Antennas have three characteristics we can measure.
An ideal (different from optimal) antenna will send all of the input power out in a spherical shape. This is a gain of 1. Antennas get gain>1 by being directional — putting the available power into a smaller area. I’m ok with some directionality for increased gain but generally that’s confusing for a customer (point this side of the box at that other box a mile away you can’t see).
Gain is measured in dB (logarithmic) so zero equals ideal sphere.
The RF transmitter on the board has a nominal impedance of 50 ohms. Envision a source with a 50 ohm resistor in series. You get the most power transferred when the antenna looks like a 50 ohm resistor.
So, a badly mismatched high-gain antenna is a toothpick that will overheat the transmitter with reflected voltage.
Antennas can be wide or narrow band. We want the antenna to be full gain through the frequency range then drop off as fast as possible to avoid receiving/sending out-of-band stuff.
Until test equipment arrives I’m doing impromptu testing. Now, it’s comparing the antennas using a single device. I’m using the Feather M0 Lora because it has the most options and is simple to use. I soldered a U.FL connector on the bottom of it.
In one corner I’ve got a Feather M0 Lora with a simple 1/4 wave wire antenna. In my informal testing this works incredibly well, as you’d expect with zero connectors or coax in the middle.
In the other corner I’ve got another Feather. This one has a U.FL connector. I also have these antennas (antennae?). This set includes a wire, 2 flat antennas and 2 whip antennas.
The wire is about 3" long. The usual formula is about 232 / F (MHz) feet long. This produces a quarter wave antenna that requires a ground plane but isn’t very fussy.
Testing the connector
To ensure the connector was reasonable I kept the wire antenna on the second Feather, measured the response, then put a U.FL->SMA piece of coax on the U.FL connector and stuck the wire into the center of the SMA. In fact I got very similar results, minus some small loss for coax adapter — so the U.FL was installed well.
Here’s a comparison table of the various antennas and how they worked.
- Straight Wire: Best result -49dB. Uniform.
- Molex ISM 105262 Flexible Antenna: Best result -49dB when pointed straight at it. Orientation is rotated. Directional.
- Siretta Quad-band PCB Antenna: Best result -51dB. Very uniform.
- SMA+Wire Antenna: best result -49dB.
- Short Stub Antenna: best result -49dB if straight coax and hung straight. Moderately uniform at -55dB.
- Long 915MHz Dipole Whip: best result -49 when hanging straight down. Directional.
Surprisingly, if I carefully pointed each antenna then they ended up within a couple of dB of each other. In my opinion the straight wire is the best. A thin wire produces a moderately narrow band antenna and if you tune it carefully it works very well.
The worst issue with a wire antenna is that it needs a ground plane and it radiates along the ground, which is not optimal for LoRa. A dipole is a better design since it does not require a ground plane and radiates less along the ground.
… in progress… comments?
If you’ve gotten this far: this project may proceed a little now but the actual test equipment doesn’t arrive until, theoretically, end of March. So,… temporarily on hold.