Thanks for helping @seeers at Github.

Still no communication issues to report, so just two additional remarks on the MCP2551 usage following some short tests on that:
They also seem to work at 9600 Baud and also using the MySensors standard (AltSoftSerial at PINs 8 and 9) is possible.

So for now, I'd recommend everybody thinking about new nodes to give this type of hardware a try. The only disadvantages may be

• the limitation in payload - but that seems not to be any practical issue in the use in an MySensors environment
• they might be more sensitive in crossing BusHigh and BusLow connection, so make sure all highs are on the same wire and also all lows

My favorites are kind of violet modules. Didn't come across them earlier, so I'd see all PINs on one side as a modest point for improvement. As there are also no resistors at all, just pads for soldering, there's also not the need of starting with desoldering unnecessary parts - just make sure, you have the right stuff to add .

As everything still seems to work as expected some futher remarks on my findings/hypothesis and todays setup:

General remark: As a lot of things have changed over time and some of my tests turned out to be contraproductive, it's hard to sort out THE root cause now. But as others had some issues with colliding messages too, I'd bet on that and really appreciate if the patch #1142 would find it's way to everyones codebase (@seeers seems to have some issues at github, can one of the mods help him out of that, please?) .

Back to my setup:

• Nodes are powered now through a central 5V DC supply, only 3.3V conversion remains locally, GW through USB
• GND of central 5V is not connected to GW GND (in case of trouble, I'd add a "resistored" connection)
• All nodes use Hardware-Serial, Baudrate is @19200
• No debugging messages on nodes activated (if someone wants/needs it: swap debugging output to altsoftserial...)
• Transceiver used: MCP2551, most of them on modded TJA1050-Boards (don't forget to dissolder the R120) => no DE-Pins necessary
• Gateway is a Pro Micro
• Termination resistors: 2kOhm at the last node in line (CANhi->5V and CANlo->GND), 120Ohm (A-B) only at last node and GW

Remarks:

• My long-time GW was a regular MAX485/MAX487-Nano using AltSoftSerial. IMO this over time had been one of the most reliable components in my setup. Conclusion: Most likely this lib is NOT to blame for any trouble I ran into.
• Nevertheless hanging nodes with regular RS485 trasceivers caused problems to some extend. So I really like the CAN logic to switch the transceiver off when it's MC seems not to work properly to avoid infection of communication from other nodes. What I didn't test yet: Using CAN transceivers together with altSoftSerial. So expect some additional info on that later, as this might be helpful for debugging over USB as on any other node
• To some extend all of the transceivers seem to be able even to transmitt in case A-B voltage level is above the "critical value" - depends a little on the strength of the power source. So I'd see this just as some type of "most likely" indicator of a bus problem now.

Hope someone might find that summary helpful and once again:

Thanks a lot to all those people here and at FHEM-forum trying to help me out of that never ending mys(t)ery!

@jonathan-pucel Have a look at the code in this post.

Found this some days ago: https://github.com/mysensors/MySensors/pull/1142

Adding these changes to a 2.3.0-alfa base seem to make a big difference in reliability .

So if there's other users having similar problems: please make also some testing of this patch.

You will get further updates and some more info on my recent setup, so far: Thanks a lot for all the ideas and hints to improve things!

@ehome said in Looking for reliable PIR human detection sensors:

@rejoe2 Have you already made some experience or used these microwave sensors as motion sensor?

I use several of them on a couple of RS485 MySensors nodes and also know some others in the FHEM forum using them with ESPEasy. So as far as I know the most common problems with the RCWL's are power-related (ESP8266 is a 3.3V-device).

As I have some problems with my RS485 setup the following is somehow preliminary (seems the 12V to 5V conversion is root cause; this also affects the RCWL's):
The ones I use myself are in my cellar and in the garage (two rooms in the later). In the cellar, detection is indirect (PIR on same wall as entrance door): Works very well, but also detects the opening of the door from corridor towards a second cellar room - there's a small hole in the wall between these two rooms near the PIR's position... So motion is also detected when just passing the cellar.
In the bigger garage room, it works perfectly - when there's enough power for the entire node - the converter on the pro mini seems not to supply enough power to get the RS485 chip working sufficiently in all situations. There it's placed on a side wall - no detection, when passing outside, but opening the front door is detected pretty reliable, in some cases, one has to do one step into the garage, detection angel is 80+ degrees .
The third RCWL I only recently installed, so there's not too much enough experience to talk about, the others are up and running for months now. But at least the radio fields of both RCWL's in the neighbouring rooms in the garage seem to not influence one another I had some doubts on that point as I first tried to use 3 of the RCWL's on the cellar node to monitor the two cellar rooms and the corridor; that didn't work - as I now think due to powering issues.

I did some experiments wrt the resistor placement, but by now, all are just without one (but I can also state: one can see a difference)

So keep us updated once the parts you ordered are up and running!

The radiowave based ones are imo better wrt. that, so why not give them a try?

According to my experience results depend much on the environment, especially wrt. their ability to "look" through walls. If it's bricks without holes, you may get also satisfying results.

Besides all other things a stable power sourcing seems to be essential . I use 5V, but some report also satisfying results when sourcing the IC directly with 3.3V.

There's the option to reduce responsiveness also by adding a resistor, more details can be found here.

@qobouky First: using more than one SPI device on one MCU is possible - this requires just some more preparation e.g. in before() in case of using the same SPI interface (initialising other SPI devices prior to nRF will be addressed through the MySensors libs in presentaion() and setup()).
If you plan to use a LCD screen, you may need a more powerfull MCU like the ESP or an STM32 (th later offering also more hardware SPI interfaces). There are some examples available you may use as a reference:
https://www.mysensors.org/build/dollhouse#tv-node
https://www.mysensors.org/build/scene_controller

@qobouky If I got you right, you question whether it is possible to receive the signals sent from your nRF24-Nodes via the wireless transceiver of an ESP8266.
The short answer is no.
Both may use 2.4GHz as a physical transport layer, but the encondeing/decoding of the information is done completely different. You may compare that to two humans speaking, but one chinese and the other english - they may hear the other one producing sound, but will not be able to understand the message.

@btmerz If the controller is not present, the nodes won't enter loop(). To avoid this, use (3000 may be changed)

#define MY_TRANSPORT_WAIT_READY_MS 3000

@nofox Did you measure voltages on the bus?

Some additional info: I did some futher experiments using also MCP2551 as transceiver chip. As they are PIN-compatible to the TJA's, it was just a modification of the modules. I can confirme thess also to work.
General "restriction" is one has to use HW-serial. For me, that's ok, but if you want serial output for debugging, you also have to use altSoftSerial and use it for debug output.

This morning, I replaced the end-of-line module with a resistorless one (so also NO termination resistor is used on this side, only at the GW!).

At least, the Bus is working for some minutes now @56700 with just GW (MAX487) and one node for each of the tranceiver chips (MAX487, MCP2551 and TJA1050).

If that works more reliable, I'll switch all nodes to MCP2551 and lower the transmission rate, most likely to 19200 Baud. Will take again some time, but I'll keep you updated. For this I plan to use a STM32F103 as GW, if possible, I'll try to expose also the other two Serial Interfaces to the OS, so one could attach up to two other GW's (for other physical transport layers) to the STM using just one USB connection. But that's another project on another planet and if there's someone out there with more experience on how to do that: Thank's a lot...