Hi everyone! Recently I have gotten interested in communication protocols and have been looking for one (with little success, hence my question) which would satisfy the following requirements (in prioritized order):

• All devices should be able to initiate communication (not your typical master slave with polling)

• Devices should be able to address messages directly to one another without a single one permanently taking on some host/master/relay role

• Devices should be hot-pluggable (A new device may be connected to the network or removed from it without problems even when the network is up and running)

• The protocol should be stable and immune enough to noise so that it can be sent through wires up to at least 5 meters, ideally even up to 10-50 meters.

• It should be simple, implementable even on low-perfomance microcontrollers.

• It should have its maximum transfer rate in the kilobytes per second range or preferably even megabytes per second.

I have looked at I2C, which for the most part would be fine, but as far as I understand it was not meant to be sent through a wire and it would be quite vulnerable to interference. (It also is by default a master-slave protocol, but it also has a multi-master mode, so maybe making every device a slave and a multi-master master at the same time could work?)

The rest of the protocols I looked at were all master-slave, which, for my purpose, is unacceptable.

That being said, I would be grateful if you could give me some pointers as to what protocol may satisfy these requirements and perhaps even on what sort of wires and signals it would require (twisted pairs? Differential signal? Shielded twisted pairs? Could plain simple wire suffice? etc.)

    • surepancakes@discuss.tchncs.deOP
      link
      fedilink
      English
      arrow-up
      3
      ·
      1 year ago

      Thank you for your suggestions. Rs-485 seems to be what I was looking for. CAN could also work, but is it reasonably implementable? So far everyone who I spoke with about CAN said that it is a rather complex protocol with a lots of finicky details and an extra long standard. Perhaps I have just talked to the wrong people. I will look into it more thoroughly, and thanks again.

    • surepancakes@discuss.tchncs.deOP
      link
      fedilink
      English
      arrow-up
      3
      ·
      1 year ago

      Hmmm… It does tick a lot of the boxes, but the ethernet protocol is way too complex with all of its layers and not reasonably implementable on a low-power microcontroller. Also it requires separate hubs for connecting multiple devices together unlike i2c, which is daisy chainable.

      • wildbus8979@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        5
        arrow-down
        1
        ·
        edit-2
        1 year ago

        A lot of microcontrollers, ESP32 or STM32F4 to give two examples, have the PHY built in.

        But if that’s too much for your taste, then CAN bus might be what you’re looking for.

    • surepancakes@discuss.tchncs.deOP
      link
      fedilink
      English
      arrow-up
      2
      ·
      1 year ago

      Thank you for your suggestion. Is it really that easy to implement? So far the people I talked with irl told me otherwise, but I will look into it and judge myself whether I am up to the task.

      • oldfart@lemm.ee
        link
        fedilink
        English
        arrow-up
        2
        ·
        1 year ago

        Exactly, hard/easy depends on your background. It’s been almost 20 years since I worked with it, I’m sure there are ready made libraries or chips implementing 99% of the protocol these days.

    • surepancakes@discuss.tchncs.deOP
      link
      fedilink
      English
      arrow-up
      2
      ·
      1 year ago

      This seems more than interesting. Will check it out in detail, since I haven’t heard about it until now. Thanks for the suggestion!

    • azdle@news.idlestate.org
      link
      fedilink
      English
      arrow-up
      2
      arrow-down
      1
      ·
      1 year ago

      MQTT isn’t peer-to-peer, it requires a centralized broker that all devices connect to.

      If OP were looking at network protocols, CoAP would be the right option, but it looks like they don’t even want to bother with IP.

      • surepancakes@discuss.tchncs.deOP
        link
        fedilink
        English
        arrow-up
        3
        ·
        1 year ago

        Indeed, I was looking for a simpler and lower level protocol. Thanks for mentioning the network protocol anyway, because I have never heard of CoAP until now. It may come in handy in the future (or maybe for someone else reading the post).