Do Gateways forward all LoRa messages?


I really did not find an answer to this seeminly simple question yet (seems its time to read the complete specifications :slight_smile: ):
What exactly does a gateway do with messages that do not belong to its own network?

  • Does it simply forward any LoRa packets it receives and leaves it up to the network server to decide that the packet does not belong to this network (so a rather dumb gateway forwarding everything it receives)

  • Or is there some intelligence on the gateway, meaning it knows which DevAddr’s belong to its own network and only forwards these messages?

Thanks & best regards


There are some gateways with some filtering mechanisms on them, but that’s not a normal situation.


Just to be sure: gateways should not forward packets for which the CRC failed. So, to be nitpicking: not any packet per se.

1 Like

OK, bit existential, but I am descartes after all, so, is a packet a LoRa packet if it fails CRC or is it just noise?

And if a developer codes in the woods, can he leave out the comments?

1 Like

Hmmm, I guess that CRC failures indeed indicate noise. :slight_smile:

However, LoRa does not require CRC to be enabled. But LoRaWAN does, for uplinks (just like it requires specific other LoRa-settings, such as a coding rate of 4/5.)

So, CRC-less LoRa exists as well. Now, would a gateway recognize/reject a CRC-less LoRa packet…? Maybe that’s covered in Phantom Packets, which I remember being a nice read anyhow. A teaser quote and image:

If that is the case how can it be that the CRC is valid for about 70% of the phantom packets?


It would be, but Stuarts shut down his old website and hasn’t moved all his articles over to the new one yet.

But it’s beer o’clock and the Air Con has overflowed again, so we can stop now!

Ah, thanks for the heads up. I found the new locations, and updated @LoRaTracker’s forum post.

I remember the Phantom Packets postings - as yuo say it was a good read and good/interesting investigation by Mr. R. CRC invalid packets can be thought of as equavalent to noise for the front end…and definetly as interferers (by channel & SF), and given one has to partially process to determine if CRC failed then also represent a processing load on the system so worse than noise in impact in that it takes up a decoding/dsp/processing resourse… do CRC fails in the woods look like an unintentional DOS attack?! :rofl:

Nick’s right its beer oclock and the heat is getting to me (yes I know unusual for the UK!)…

Along with a prodigious memory for previous posts, you should get a Sainthood badge for sorting these sorts of things out

1 Like

Stuart’s moved all his posts over since I last looked for something the other day on CAD - thanks Stuart, it is a very useful resource.