PJON (pronounced pigeon as I have just learned) looks exciting but it seems like it’s trying to provide globally addressable devices, as well as local ones?
In the case of global addressing, don’t we already have an open and free protocol in the form of IPv6? Some quick googling found LWIP which provides an IPv6 stack in “tens of kB” which seems comparable to PJON’s “4kB to 8kB” but maybe that size difference really is important?
"The PJON protocol v3.1 in local mode supports connectivity for up to 254 devices, in shared mode supports connectivity for up to 4.294.967.295 buses (groups of devices) and up to 1.090.921.692.930 devices."
That means 32 bits for the bus id, and 8 bits for the device id.
Another interesting side-effect of the address-space scheme and the dynamic meta-data inclusion in the packet is the overhead:
"It supports interoperability between systems that use a different configuration and provides with high efficiency including only the protocol's features used and the overhead effectively required (5-22 bytes)"
In terms of minimum requirements PJON is uncomparable to IPv4 or IPV6 needing only in the most limited configuration: 1 IO pin, 16MHz clock frequency, 4144 bytes of program memory and 46 bytes of RAM
Thanks for the information, and this release! The minimal memory requirements are impressive.
With 32 bits of global addressing space, is that for routing, or more to avoid address collisions when two networks are joined in an ad-how way?
I noticed the PJON bus-ID assignment process and it seems a bit like MAC vendor ID assignment, so I’m guessing the latter?
I don’t have much experience in this space, but it feels like it would be useful if the bus-addressing could be optionally extended to incorporate IP addressing and routing infrastructure. Perhaps IP to PJON bridging is already very easy?
Although not using PJON addresses as IP addresses, PJON has a few strategies that allow a PJON bus to span a Ethernet+WiFi LAN and/or WAN.
I use this in my PJON-based home automation system using clusters of SoftwareBitBang based devices (a bus on a single wire plus ground), linking these clusters together and with the master with routers using the DualUDP strategy over my LAN. PJON is great for cases like this.
It could easily cross the Internet as well but in that case I would recommend using site-to-site VPN commections for security as the PJON traffic is not encrypted (encryption is a little too heavy when running on limited devices like Arduino).
Ciao gongoiler, the 32 bits address identify a group of devices that share the same bus id, so yes, many groups of devices (that may have the same 8 bits device id) can coexist without addressing collisions.
I have never thought about the bridging strategy you propose, yes effectively it may not be hugely complicated :) if you would like to share your idea with other users feel free to open an issue here: https://github.com/gioblu/PJON/issues
I would be curious to see to what extent it supports error recovery. Perhaps I missed it, but I didn't see anything that specifically addressed that.
This seems like it would be very compelling, especially for small scale stuff. Being able to wire something as small as an ATiny85 into a computer and communicate over a relatively high-level protocol would be very useful for interfacing with sensors and other peripherals.
I wonder how difficult it would be to adapt something like USBtiny[0] build an adapter. For potentially a few $ worth of parts, you could connect to all kinds of nifty embedded peripherals, albeit at relatively slow speeds. One obvious use case that comes to mind would be controlling lights / sensors / fans etc. in custom computer builds.
There's an optional 8-bit CRC and sequence number, with a single bit ACK.[1] This is even weaker than ARCnet, circa 1977, one of the earliest protocols for talking to low-end devices.
This is for talking to things like lightbulbs with some US$0.50 CPU inside. It's for really low end devices. It's best for idempotent operations that can be requested repeatedly without harm and are timeout-tolerant.
Small correction: Using CRC is not optional. Setting the crc flag makes it use CRC32 even for very small packets. The default is to use CRC8 for packets of a few bytes and then upgrade to CRC32 over a small size threshold. This is to minimize the size of otherwise tiny packets.
Also, the ACK implementation depends on the strategy (physical media type), and it is not a single bit but a single byte, and it seems to me to be robust. I am sure the creator, gioblu, has some more comments about this.
Great work on PJON! Where can I find documentation/more info on how to build this for STM32 microcontrollers, like the STM32F091CCU7 using SerialAsync (over PLC)?
Happy to see somebody from my city dedicated to such an ambitious project. I wish you all the best Giovanni! I see you are based in Via Larga, do you guys have an office there?
I work in embedded and I’m still not 100% clear on what this does after reading the README.md. I find that for a lot of projects actually. Is this an MQTT alternative? I find use-cases or working examples to be key, even better if they compare with pros/cons to alternatives.
PJON is based on a concept of selectable "strategies" to support using the protocol with different physical media. Some of these, like the one named SoftwareBitBang which allows a bus on a wire connected directly to an I/O pin, are very compact.
There are also strategies for TCP and UDP based communication, essentially using the same packet structure but sending them as TCP or UDP over a LAN or WAN. These strategies need the Ethernet library, and use more memory.
Still, the memory requirement is small enough to run a router with two PJON objects with the strategies SoftwareBitBang and DualUDP on an Arduino Nano with a W5500 Ethernet shield. This allows easy traversal between different physical media, and IP-based communication, also with PJON-based software running on a Windows or Linux computer.
Nice one, never heard of pigeon (PJON) before but it has potential. I understand that it’s mainly for low end devices, but I’d like to see it combined with https://balena.io
Very hard to read in mobile github read me text. Is this require special hardware? Or usable even if you just have a python interpreter say. Or is it a replacement of ip stack. Wonder.
In the case of global addressing, don’t we already have an open and free protocol in the form of IPv6? Some quick googling found LWIP which provides an IPv6 stack in “tens of kB” which seems comparable to PJON’s “4kB to 8kB” but maybe that size difference really is important?
LWIP project: https://savannah.nongnu.org/projects/lwip/