LowPowerLab Forum

Hardware support => RF - Range - Antennas - RFM69 library => Topic started by: tve on February 12, 2017, 03:37:58 PM

Title: how to choose among equivalent LoRa modulations
Post by: tve on February 12, 2017, 03:37:58 PM
I've been playing with sx1276 radios in the 433Mhz band for the past couple of weeks and one question I haven't really been able to answer is how to choose among the equivalent modulation schemes. For example, 125khz bw, SF12 results in 30 symbols/sec and -137dBm Rx sensitivity while 10.4khz bw, SF8 results in 40 symbols/sec and -138dBm Rx sensitivity. I'm kind'a wondering what the tradeoffs really are.

The datasheet says that the freq tolerance is the lesser of 25% of bandwidth and 50/100/200ppm for sf12/11/10 respectively. I can certainly attest to the fact that a 2500Hz frequency offset at 10.4khz bandwidth kills communication and I have the nodes dynamically adjust frequency and bit rate. The ppm tolerances seem much wider, if I understand correctly (50ppm @ 433Mhz ~ 22khz, right?).

On the other hand, according to the datasheet switching to narrower bw and thereby using lower SF for the same data rates gives a little edge in terms of Rx sensitivity. For example, @125khz bw there are 3dB steps going from SF7 to SF10, and then SF10-11 gains 2dB, and only one extra dB to SF12.

Obviously, narrower bandwidths use less spectrum allowing more channels to be packed in, if that's the objective.

I noticed that LoRaWAN sticks to 125khz bandwidth from SF7 to SF12 (plus one 250khz combo a the high end) and I have to wonder why they didn't go to 32khz/SF10 which, interpolating the datasheet, would probably gain another 2dB. Maybe keeping it simple was more important. It's also hard to tell how the sensitivity really scales at the long range end. Isn't it weird that a long-range radio datasheet has most of the stats at the short range end?

Anyone have thoughts or real-world experience? I did my first long distance tour yesterday using 10.4khz/SF9/CR6 and easily got to 8.5mi where there was reasonable line of sight. This was using a simple dipole at one end and a simple l/4 car roof magmount at the other and only 17dBm Tx at the car end. I was sending 22-byte GPS packets and getting an ACK back. The attached map shows the track with the RSSI level superimposed.
Title: Re: how to choose among equivalent LoRa modulations
Post by: tve on February 18, 2017, 12:45:39 PM
Apparently not much LoRa stuff going on here, but for the sake of the next person I'll answer my own question having received some info on the semtech forum.

I just experimentally validated why LoRaWAN stays at 125khz bw. I had picked 10khz/sf7 to get 766ms/20-byte pkt @-136dBm sensitivity in some of my experiments. LoRaWAN uses 125khz/sf11 to get a similar 660ms/20-byte pkt @-136dBm sensitivity. I have two radios that have approx 4khz xtal difference, which @433Mhz is ~9-10ppm. Using 10khz bw they don't connect, the data sheet says they have to be within 25% of bw, which they're not. Using 125khz bw they do connect. I also tried 125khz/sf12 and they still connect as expected given that the data sheet only requires 50ppm. Done.