Preferred wireless communication between microcontrollers?
- Thread starter NeverDie
- Start date
NeverDie
Senior Member
Did you consider MQTT at all? Some people are utilizing MQTT brokers, which allows nodes to communicate via a publish/subscribe arrangement. However, although nodes can make their own decisions, I do get the impression that all the information flows through a single broker, and so if that were to fail, I presume it would take down the entire system (?). Nonetheless, it does seem to mean that the decisions themselves are not in a central controller.tattema said:Hi NeverDie
When I began building my HA system, I found that current designs relied on a central controller to do everything and to coordinate all node communication. At first I thought this was a good idea but then realised that I could have a more robust system if i didn't depend so much on the controller to manage each and ever event.
In the current design, nodes are intelligent agents that interact with each other rather than through the controller. In fact, the controller isn't needed for day to day operation of my HA system, just for maintenance and configuration purposes.
I'm really happy with the approach as it lets me segment the network so that local zones manage their own traffic without needing to funnel every message through the gateway/controller. This has made the network more robust, messages almost always are delivered and I don't need to rely on the internet for my HA system to function.
Here's my RF module based on the Semtech SX1232 radio. It's a 4 layer, .8mm thick board with castellated edges for SMD mounting. The PCB is impedance controlled and the output PI filter has been optimised to filter out stray harmonics.
IMAG1727.jpg
In this photo I have the radio mounted on my controller/gateway cloud interface so that devices can communicate with my phone through the cloud or over the local LAN. It runs a web server, CMSIS OS, IP6/IP4 and my mesh protocol.
IMAG1732.jpg
P.s sorry for the bodge wires... i forgot a couple of critical SDRAM connections :-(
Regards
Trevor
If not using MQTT, how is it that you propogate information in your network?
NeverDie
Senior Member
What I'm currently pursuing (only time will tell if it was a good decision or not) is the use of inexpensive NRF24L01+ modules for short range communication and higher power and pre-amplified versions of the same if long-range communication is needed. Not long ago this guy did an exhaustive comparison of different NRF24L01+ modules:tattema said:
https://www.youtube.com/watch?v=gtM832Z0ujE
Interestingly, he found that at short ranges, the relatively inexpensive modules with the PCB trace antennas easily outperform modules that can do well at long distances. For longer range communication, he recommends a module with a PA and LNA and which has metal shielding on it.
I went this route because I'm willing to trade-off some amount of performance that a fancier chip might provide in exchange for the easy setup and utilization that often comes with a more popular platform. However, that's me, and I certainly wouldn't blame someone if he preferred a different trade-off than the one I happened to choose.
I recently found out that many allegedly NRF24L01+ devices are most likely running on counterfeit chips.
http://forum.mysensors.org/topic/1664/which-are-the-best-nrf24l01-modules
Unfortunately, there often are at least some differences in how those chips perform, and so I view them as a possible source of unwelcome instability and/or poor performance. I'm not sure to what degree this is a problem with other radios too, so I'm just mentioning here as a warning so anyone reading this will be aware of it as a potential problem.
tattema
Member
NeverDie said:
Hi NeverDie
That's an excellent question. That's my understanding of MQTT too. My HA network uses a gateway/controller for configuration purposes in that rules are setup/configured in the controller and then deployed to participating nodes. Once deployed, the controller doesn't need to participate and can be switched off (not ideal because it's also an internet bridge) This way, operation is decoupled from the controller and pushed out to the nodes. Network joining and pairing requires the controller too as it maintains a dynamic database of local node addresses and other configuration details needed for a node to participate in a secure mesh.
I considered MQTT but moved on because I wanted nodes to communicate directly with each other and not through a centralised queue. Decisions in my HA network are not centrally coordinated so there is no need for messages to arrive and be queued centrally. MQTT is pub/sub but not point to point, rather, it's based on a star topology which for various reasons doesn't leverage nor take advantage of the benefits mesh technology can bring. I can see however that persons with different implementation requirements would find MQTT to be very useful, especially with it's guaranteed message delivery mechanisms.
Regards
Trevor
tattema
Member
NeverDie said:
I must admit that I'm very impressed by the performance of those devices. While the PCB antenna devices only provided 50m range, that limitation can be overcome by implementing a (bare) quarter wave whip antenna within the enclosure of advice effectively boosting range quite a bit.
I really like the compact PCB designs that can be achieved though using 2.4ghz technology like this chip.
What are your thoughts on the standard NRF24s with an without the external PA? Is there much of a cost difference at volume?
NeverDie
Senior Member
Hold on to that thought. I expect I'll be able to answer it much better in a week or two. By then some more hardware should have arrived, and I'll have better data. Shooting from the hip, though, I'd say that yes, at least for me there's a meaningful cost difference, but only because I'd like to establish a low cost wireless+arduino platform (around $2-$3 without PA+LNA seems achievable) that I could use to deploy a lot of devices. If it was just a few devices, then, no, it wouldn't matter.tattema said:
It's all relative though. If you compare it to the cost of typical z-wave devices, or to decent wireless weather station sensors, it's inexpensive with the PA and even more inexpensive without the PA. Either way, it's inexpensive. So, the answer somewhat depends on whether you're at the forest or tree level, if you know what I mean.
tattema
Member
Hi NeverDie,
I was curious as to why performance tests conducted in the youtube video (above) had significant signal drop-outs between 1000m and 1200m... puzzling because its' a huge dead zone. Upon closer examination, notice that he's right between two large antennas... who know's what they're pumping out but I'd wager that they are impacting reception somehow. 16:39 into the video you can see the chart. At 17:14 (1000m) you can see he's between the towers. In the HA space, radio means it's not reliable so that needs to be a fundamental consideration in any design.
Regards
Trevor
I was curious as to why performance tests conducted in the youtube video (above) had significant signal drop-outs between 1000m and 1200m... puzzling because its' a huge dead zone. Upon closer examination, notice that he's right between two large antennas... who know's what they're pumping out but I'd wager that they are impacting reception somehow. 16:39 into the video you can see the chart. At 17:14 (1000m) you can see he's between the towers. In the HA space, radio means it's not reliable so that needs to be a fundamental consideration in any design.
Regards
Trevor
NeverDie
Senior Member
Good catch. I think you're right, The size of that hole is much too large to be just a null zone. At time index 14:20 he refers to those two structures as carrying high voltage electric lines. Probably throwing off a lot of RF noise and effectively jamming the signals he'd be receiving on his mobile (bicycle) transceiver. They're at about 1000 meters from his measurement origen, so the shoe fits. I'd say you figured it out.tattema said:
You'd make an excellent Sherlock Holmes!
