I recently bought an RF explorer http://rfexplorer.com/ handheld spectrum analyser to get a better look at the normal EU LoRaWAN channel lineup and to measure the noise floor and potential interference during site surveys at gateway locations.
I’m very impressed, both with the handheld unit and with the PC software that can access the unit via USB. I’m also more and more impressed with the very low-cost Semtech technology.
This is a screenshot of the PC application after about 48 hours of running in my workshop with a number of devices and a couple of gateways outside. The lines are min and max-hold.
The markers are at the centre frequencies of the channels. I have no idea what the intermittent peaks around 869 Mhz and the constant peak at 870 MHz are. The noise floor is sitting at about -108 dBm as expected.
This is a screenshot looking at the channel separation between channel 3 at 867.1 MHz and channel 4 at 867.3 MHz. I’m impressed to see that the channel separation at 867.2 MHz is about 30dB.
So, a good cheap tool on a good cheap radio system… bodes well for 2020!
A noise floor of around -105dBm to -110dBm is typical, and does not seem to vary much between cities and hilltops in the countryside. I have an RFExplorer too.
And yes, the observed sensitivity limit for a LoRa signal is the noise floor - the SNR for the spreading factor. So a max ‘Real World’ sensitivity of -125dBm to -130dBm.
Certainly when I have done hilltop to hilltop long range testing with known transmiiter powers and antennas (so you can calculate the signal level you receiving) the failure level is very close to noise floor minus the SNR for the spreading factor, and as much as 15-20dBm outside the datasheet sensitivity figures.
Hi @TonySmith, most of my work is in LoRaWAN on farms in Scotland (lots of hills) and I use an RX sensitivity of -120 dBm for SF11 and an RX sensitivity of -115 dBm for SF9 radio prediction. We use SF9 planning to create coverage maps for customers with a Service Level Agreement (SLA). Farmers are very accustomed to patchy mobile network coverage so we provide the predicted LoRaWAN coverage data to them as Google Earth KMZ files and they generally become very supportive.
We have processed the radio metadata from a very large number of LoRaWAN uplinks RX at gateways and very rarely see a reported gateway RSSI below -120 dBm.
As reported by @LoRaTracker, most noise floor measurements show -105 to -110 dBm - except in heavy industrial environments where it’s higher, often around -100 dBm and can be very intermittent as machinery starts/stops and can temporarily rise as high as -90 dBm, which reduces range very significantly.
That’s exactly what I see. Graphing the gateway received RSSI for quite a few nodes over several years data and they are strikingly similar. All flat bottomed at -120dBm, rarely see anything less than this.
My installations are on relatively flat ground, with a number of nodes at the extreme of coverage.
With an inexpensive instrument, I’d want to try to make sure I was actually measuring environmental noise and not the limits of the instrument or its own noise figure.
Also important to note that the actual units of environmental noise are dBm per Hertz - the amount of power that implies depends on what bandwidth you integrate over. Radio engineering usually assumes physics imposes an ultimate noise floor of - 174 dBm/Hz, if you integrate that over 125 KHz bandwidth that’s a factor of 10^5.1 so it becomes -174 + 51 or -123 dBm of noise power in a typical LoRa uplink channel. Before any comparison can be made, it’s important to be using the same units.
Of course the chirp coding then further narrows that mathematically, which is why we can receive with a negative SNR relative to the noise in the channel’s simplistic bandwidth.