Comparing a TTN-gateway with iC880A to one with RFM95W

TL:DR; We’ve built a full-blown multi-channel gateway with the iC880A chip and a professional antenna and it performed *magnitudes* better than our previous single-channel gateway with the RFM95W chip. Read below to find out about our results.

A few weeks ago we tested a single-channel TTN-gateway built with an ESP8266. Surprised by the results, we decided to fund a full-blown gateway and put it to a test in the same scenario. No sooner said than done!

The gateway, an iC880A connected to a Raspberry Pi 3 Model B and paired with an Aurel 868 ground plane antenna, has been deployed in the exact same location of the last test (on top of a chimney, around 8m height). Again we locked the nodes to spreading factor 12 to create a comparable setup and walked around for about six hours in our sub-urban town (about 20km) and recorded the results with TTN-Mapper. We’ve paid attention to share as many measurement points as possible with our previous test. Note: with the RFM95W, we used a small helical 1/4 lambda antenna, this time the Aurel 868.

Results and findings:
First off, here you can find an interactive result set of the new gateway. For comparison reasons: here is the dataset of our RFM95W-test. We always publish our raw result sets so you can independently evaluate them.


Image 1: Screenshot of the iC880A’s range test results (enlarge to see location markers)

Below you can find some locations in comparision:

LocationDistance to gateway (m)
RFM95W (dB)iC880A + Aurel GP 868
Change
Point A80
-122 dB-67 dB
increase of 55 dB
Point B192-115 dB-90 dB
increase of 25 dB
Point C270-120 dB-99 dB
increase of 21 dB
Point D442-123 dB-98 dB
increase of 25 dB
Point E985-124 dB*-113 dB
increase of 11 dB
Point F920-120 dB-88 dB
increase of 32 dB
Point G916-123 dB-117 dBincrease of 6 dB
Point H1080-124 dB*-118 dBincrease of 6 dB
Point I620-124 dB*-105 dBincrease of 19 dB

* We haven’t seen the RFM95W to report anything lower than -124dB and believe this to be its lowest reportable RSSI.  It’s likely to have had a RSSI below -124dB.

You can see that most measurements were significantly better with the iC880A, especially in the sub 1km range. 21db means that the received signal was about 128x stronger! The long range improved too, not to mention that we had coverage where we did not even expected it.

When seeing this unexpected result to solidify, we decided to skip some close area measurements and went as far as we could get by foot. Due to the topography of our landscape, this involved going beyond the hilltops.

Surprisingly, we were able to receive packets from the nodes for a pretty long time after crossing the hilltops. Only after we went a few hundred meters behind the hills, we had no coverage.

This was unexpected because in theory, the soil and rocks should totally attenuate the signal. At 2500 meters, the radius of the first fresnel zone is about 15m. When looking at the topography involved with the packets received e.g. from the north-eastern test area, we can see that the fresnel zone is nowhere to be even partially free.


Image 2: Screenshot of airlink.ubnt.com, showing the totally blocked 1st fresnel zone (here: on 900 MHz)

What can we learn from this? Either the attenuation is not as big as anticipated, or reflexions came into play.  As there were no big, plane or metal objects in sight, such a reflexion must come from above. While we’ve heard of reflexions in the atmosphere, this might have been our first time to experience it ourselves.

What do you think? What is more likely? Let us know what you think!

Critics
We cannot rule out that one of the two chips reports wrong RSSI (while we experienced a solid increase in range with the iC880A!). Also, tests took place on two different days (Weather seemed comparable to us).
Both gateways were carefully soldered, but of course we can only test a two samples here. We’d love to see the results confirmed by independent groups, with their own gateways.

Conclusions
It’s worth building gateways with the IMST iC880A and the Aurel GP 868 antenna. The drastically higher RSSI will result in higher bandwidth / lower air time. Not to mention that this setup is proven to work with hundreds of nodes around, all sending on different channels and spreading factors. 

Future / prospect
Now that we’ve sucessfully tested the gateway in a suburban area, we would like to evaluate it in a dense, urban area and help spreading The Things Network in Germany. This will likely be the city of Bonn, in Germany.

Stay tuned for the results.

Credits
Thanks to everybody involved in funding this gateway.
Thanks to @jpmeijers for mapping service (ttnmapper.org)
Thanks to @galagaking for providing instructions on GitHub
Gateway built by Janek Thomaschewski and Manuel Schmidt.
Tests made by Ramdan Hamdan and Manuel Schmidt on 10.09.2017 in 53894 Mechernich (Germany)