Can we improve LoRa device performance?

(LoRaTracker) #1

Someone asked a while back as to whether a LoRa device\module could be improved by adding bandpass filters and\or an LNA.

I eventually got around to writing up some tests I had done, initially at 434Mhz, BW125000 and SF7.

By using software that counts packets sent at different powers you can work out at which point (transmit power) the link fails. You can then make a change, add a bandpass filter, LNA, antenna then run the test again and measure the change in dB.

In summary, adding a bandpass filter made the link worse, due likely to the filter insertion loss.

Adding a high specification LNA, SP7000 @ £500, improved the link by +5dBm
Adding a more modest LNA (G4DDK @ £55), improved the link by +4dBm

Both the LNAs are high specification units, yet the improvement in link budget is modest. It is conjecture as to how much gain a much simpler LNA added to a typical LoRa module might add.

The full details of the test, how it was done, and the graph of results will be found here;

I will be testing SF12 shortly, and I suspect that since this mode operates much further below noise level than the SF7 used above, the performance improvement will be less.


so its definitely not easy to improve for example rn2483A hardware… leave us to add more gateways (if possible) ? better antenna’s ?

(LoRaTracker) #3

Does not look like it.

With a gateway that has a remote antenna, then a masthead LNA would reduce cable loss but that introduces problems in separating TX and RX paths.

Better antennas is one improvement of course but the other obvious one is to just increase spreading factor or use a TXCO and reduce bandwidth, but both of those options increase TX time and have implications for the fair use policy.

I note the ‘new generation LoRa’ devices have no performance improvements apart from 2dBm extra in possible power, which is of no consequence to most applications where TX power is already restricted.

(Will3509111) #4

Great work! I’m also doing some test around my gateway:

I’ve try adding MAX2640 and later SKY67150_396LF with better NF (both with SAW filter & attach under antenna with tee bias), tested with multiple high altitude ballooning. but LNA doesn’t help that much in my case. LNA’s NF might have some impact on link budget, since in real case noise floor is higher. I can see RSSI going up, but SNR still perform the same.
I’m trying right now is by using high-end LDO(TPS7A470) to power SX1255/1257 and moved to 4-layer design to see if I can improve sensitivity…

(LoRaTracker) #5

Testing modules A versus B to see if ‘noise’ from a linear regulator is causing issues should be straight forward to simulate. Power the receiver from a 3V standard LDO, then a couple of 1.5V batteries and see if there is any difference.

(LoRaTracker) #6

SF12 Tests

I repeated the same tests at SF12, and the results were a surprise.

Since SF12 operates at up to –12.5dB bellow noise over SF7 (the previous tests described above) I had expected the LNAs to have a lesser effect, the reverse proves to be the case.

The SP7000 LNA added around 7dB signal gain over the Simple link and the G4DDK added 3.5dB.

The extra 7dB of link margin at SF12 from a good LNA would approximately double the distance the SF12 packets would reach. This could have a significant benefit in some applications.

As for producing a modified LoRa module with an LNA, maybe you might get circa 3dB link gain at the high spreading factors.


Very interesting indeed
Did you verify what this would bring in practice? (I assume so given your profile name :slight_smile: )
This is probably hard as you need LoS conditions to verify (which is probably hard for these distances :wink: )

(LoRaTracker) #8

I dont need to verify it in practice, it was already a practical test and there is no particular need to ‘verify’ using LOS tests.

In effect the SF12 tests I did showed that distance ‘x’ needed 10dBm TX power for the link but with the LNA in place you needed only 3dBm, that is what the testing was actually measuring, how much TX power do we need for the link.

Now if you were using LoRa to say track a high altitude balloon and the reception limit at 10dBm TX power (from the tracker) was 250km, then if you added the specified LNA to the receiver reception distance should increase from 250km to 560km, this is based on standard link calculations, to cover twice the distance you need four times the power.


I fully agree and understand that :slight_smile: I was just curious how the effect of an LNA would work out in practice (which mostly isn’t LoS at larger distances) so if indeed:

also works out like that in practice. So I mentioned LoS because the statement about doubling the distance is only valid for LoS I guess?

(LoRaTracker) #10

The link improvement is valid for other environments also.

Problem is that in other situations it can be difficult to predict ‘exactly’ how much further you will get particularly in urban areas as they vary so much in density.

(LoRaTracker) #11

My particular interest has been the potential use for LoRa in small satellites, either direct node-node or lately via TTN\LoRaWAN.

It is unclear if LoRaWAN would be permitted in space, but node-node should be possible under Amateur radio licensing and here LoRa has great potential for small satellite communications.

I note there is a ‘LoRaWAN in Space’ topic at the upcoming TTN conference, I hope to be there.


For the planning of mobile networks (which use a frequency band quite close to 868 MHz), they use a decay of the signal strenght in urban area’s (non-LOS) proportionate to a factor R^4.


How about using a diversity antenna to eliminate the effect of multipath fading? Has anyone tried this?


I don’t know, but there are several types of diversity receive techniques - see wiki
for eliminating the fading effect you should think of spatial diversity imho where the 2 (or more) antennas have a certain phyisical distance which is difficult to achieve in an endnode.

(LoRaTracker) #15

Not tested diversity myself.

But I have done a fair bit of signal strength testing on foot in the urban area around me. It is noticable that signal strength can vary quite a bit at shoulder height, 5dB or more over say 3 or 5 metres. But that is at the end node where spacing antennas by that much would be difficult.

With the gateway antenna in a good position, elevated and clear of obstructions, I wonder if there would be any benefit in antenna diversity at that end.

(LoRaTracker) #16

On the topic of what governs LoRa performance, I collected my thoughts into this post;

My problem with receiver sensitivity and link budgets