In a little experiment with a 95 meters high node last week, SF7 packets were received almost 50 km away in line-of-sight.
I was expecting that to happen for SF12 packets (which were even received 73 km away). No problem for SF12: if capacity needs to be increased, more near-by gateways can be deployed and the network servers can use adaptive data rate to tell the nodes to use more efficient spreading factors, and hence limit their reach. Or, so I thought.
But now, seeing long distances happen for SF7 as well, I wonder if a 100 meters high gateway might receive too many packets from (far-away) nodes that happen to have (near) line-of-sight, eating its capacity and causing collisions for nodes that rely on that gateway? Even when using directional antennas, a 50+ km long area would still be huge.
So: could 100 meters be too high if the number of nodes increases…?
(Note that I am not an expert at all!)
The gateway position depends a lot on the type of antenna used.
The Taoglass antenna has a 5 dBi gain and there are other antennas up to 9 dBi gain.
Now we need to ask how these antennas work. A normal 1/4 wave antenna has a donut pattern:
The antenna sits in the middle of the donut so it radiates (and picks up waves) below and above the antenna.
For the Taoglass, and other antennas with amplification, this donut is flattened.
Back to gateway height.
If the gateway is placed high and the antenna is a +9dBi antenna, there is no reception on ground level near to the antenna.
So in the end it all depends what your goal is and it also depends on the environment. In the middle of an open fiels, 5-10 meters can be high enough. As strange as it sound, this is also the perfect height in an area with only high rising buildings. Placing the antenna on top of a high building will reach the area around that building but most likely not the building itself.
Line of sight is not absolutely necessary but it will affect the distance. My office in Eindhoven is located around 3 km from a TTN gateway and 4 km from a KPN gateway. In both directions there are high buildings that block the direct view but reception is still good (RSSI of -114dB for the TTN gateway and -96dB for KPN)
In the (near) future there will be more gateways. When gateways are cheap enough you’ll see gateways every 1 km or less from each other. Most of these gateways will be placed in homes, just like wifi routers, and these will have a short range. Only few gateways will be placed outdoors at high places for long range use. These will make sure that we indeed have a 100% coverage of a given area.
This mix in gateways makes the network better. Remember that gateways have the same duty cycle rules as nodes so a dens gateway network with lots of gateways makes for a network that can handle large amounts of sensors requesting for an ACK or even data sent to them.
That’s what I wonder about: if many nodes happen to reach a highly positioned gateway (regardless if these nodes also reach other, close-by, gateways), then the highly positioned gateways would just get too many messages? (Or: collisions.)
Same problem there: if a highly positioned gateway sends a downlink, it reaches a large area, increasing chances for collisions?
That’s the big trick of a good LoRaWAN network. The ADR (Adaptive Data Rate) feature of LoRaWAN will make sure that nodes switch down to a lower Spreading Factor (higher datarate) and lower power such that it will still reach nearby gateways but not disturb gateway traffic from gateways that are further away.
In a controller environment, the network operator will monitor the gateway traffic and flag an issue. But within TTN I think we still have to use our own common sense. Within the Nijmegen area we are now looking at the people who have, or have planned, to place a gateway and determine antenna locations in such a way that we have a good coverage of the area. 4-6 gateways (with antennas on high locations) will be needed to cover the whole Nijmegen, Wijchen, Malden, Molenhoek area. More people will add gateways, resulting in a dense network of small cells accepting traffic from more nodes.
That mechanism works fine for stationary nodes, for moving nodes ADR does not work as both the network and the node would have to adapt constantly to changing conditions.
Also, if the node is already using SF7 (like it was in my tests) then there’s nothing lower than that? (LoRa defines SF6, but LoRaWAN does not use that. Or maybe ADR could make it use FSK, but even then there’s not a lot more to be adapted?)
@Batilan was wondering if packets with a much lower SNR might simply not affect packets with a much better SNR, rendering my observations moot? (Yes, please!)
Indeed, moving nodes will use a fixed SF/power. But there is a fat chance that moving nodes will not use an optional antenna placement and thus have a very limited range.
Moving nodes are often ‘labels’ attached to objects like crates, trolleys or containers to track their position. It is expected that Spreading Factor and Transmit Power are set such that the highest possible battery life is reached while still allowing for a good enough range for tracking.
Do you also know the power setting used?
The LinkADRReq message in LoRaWAN (see section 5.2 of the LoRaWAN specification) will request the node to switch not only the SF but also the TXPower setting (+20 … +2 dBm for EU)
Yes. As soon as there are more nodes communicating on the same frequency the noise floor will raise and signals with a much lower SNR will just drown in the noise.
But then, I still think the 95m high tower is a bit too high for this geographic location, apart from the fact that you will create a dead spot around the antenna location. Placing an antenna like a Taoglass on top of the 7 story high building (I guess about 30 meters) is a very good position.
In short, the “best position” always depends on the local situation. You always want a location to be covered by multiple gateways - this reduces the amount of lost messages due to interference, signal loss or gateway outage. Having 20 gateways placed on lower building in a city is preferable above just one placed on top of the church in a village.
Site surveys are an important method of determining where gateways can/need to be placed.
Very well possible.
You just need a good antenna cable and some connectors.
Good antenna cable that is easy to manage is something like Aircell 7. The gateway will nost likely have a normal SMA connector, there you place an SMA connector on the cable and on the antenna side an N-connector.
Now either use an N-female/female chassis that you mount on a bracket with a N-male/SMA-female connector to mount your standard antenna or just connect the cable directly to a Taoglass antenna using the N-connector mounted.
In all cases, make sure you have some self vulcanizing tape to make all connections water tight. Also keep the antenna cable as short as possible. It’s much better to have a 20m network cable to the gateway with 5m antenna cable than 25m antenna cable with a short network cable.
To quote myself…
In Slack, @Thomas mentioned that the reach of nodes at ground level will be much less, as the earth’s surface will be in the first Fresnell-zone, limiting the maximum distance. (I hope I paraphrased/translated this correctly.)
Why the height of >6 meters you ask?
This has to do with the Frenell Zone: https://en.wikipedia.org/wiki/Fresnel_zone
Calculate here: http://www.wirelessconnections.net/calcs/FresnelZone.asp
Not my cup of tea, but it sounds convincing.
What should be the minimum distance or dB separation between the Gateways.
Can the gateways be too close, will it cause any interference ?
Please share your feedback on this Question.
The rule of thumb is 2 meters away from everything in about the same frequency band, so including cell phones.
Thank you, request you to advise if 2 meters distance is not maintained, What kind of issues will raise.
Will there be any effect on Packet loss, as mostly gateways are Receive mode.
it is said they may slowly cook each other. This is because when they are sending this would be a stronger signal than the other gateway could accept. Having said that I have never heard of any gateway that was damaged that way, with the remark that it may not be that obvious if it would be slow deterioration.
Not sure what to think about though.
As the uplink and Downlink are IQ inverted, will the packets collide which are at same SF and same centre frequency.
thank you for your clarifications.
It is clear that Gateway will not be affected by end node and GSM. But i would like to understand the affect of gateways when they are nearby 2m.
Why is it recommended in this thread to place the gateway indoors with a long antenna cable connecting it to an outdoor antenna, rather than the gateway in a waterproof enclosure on the roof directly attached to an antenna? Is the latter approach also reasonable? Will TheThingsNetwork gateways have PoE capacity?
It’s recommended to place the antenna outdoors if you want wide area coverage outside. Placing the gateway indoors and connecting the outdoor antenna through a cable is just something people do, but placing the gateway outside works just as well. This is my preferred method because it saves you from having to buy expensive coax cable (or bad cheap cable).
You do have to keep some things in mind though, like temperature variations (freezing, overheating in the sun), water ingress and moisture. But in general, just putting it in a white IP-x6 enclosure with some cable glands on the bottom side for antenna, power and network works just fine. I have such enclosures containing WRT-54GL routers on top of some roofs for more than six years now, and they still work fine.