Typically, Ethernet uses a star topology. Which means that if you were to use a protocol based on TCP/IP —like MQTT— for wired devices, you would end up with a (number of) central node(s) and one cable for each device. That's a lot of cable.

I was researching alternatives for wired devices and the ones with best support or quality (in Europe) seem to be KNX and Loxone Tree, but they are proprietary and expensive. I'd much rather stick to open protocols.

Is it possible to have a bus topology for home automation using Ethernet? Ideally, I would throw one cable and connect devices (mainly smart actuators with ethernet connection) in serial, and be able to address them individually.

The idea is to complement a wireless installation of non-critical sensors and actuators which is using Zigbee and MQTT, with a wired installation of more critical devices, like dimmers, doors, blinds, etc.
Ethernet is already running through the system (for LAN or cameras), hence the question. A different approach is to have a bus installation that can "talk" to MQTT without needing extra gateways.

  • 13
    Frankly, if cabling is that much of an issue, go wireless. That allows for the star topology with currently available hardware and using it as designed instead of ancient (10base-2 and 10base-5) hardware or hacking which leaves you totally on your own. I'm very much a proponent of wired, but if you're complaining about too much wire...
    – FreeMan
    May 23, 2022 at 21:39
  • 4
    Realistically, no. Look at wireless and/or switches distributed over the coverage area. The cost of ethernet cable is nothing compared to the time and aggravation you're looking at with cobbling something together.
    – gnicko
    May 23, 2022 at 22:27
  • 2
    What exactly are you trying to do via home automation that requires ethernet? What are you trying to do that can't already be done by existing home automation platforms that are wireless and/or send their data via the powerline? If you want to build your own economically, I'd look into 1-wire. Can you give us more specifics about what you're trying to do and why you think ethernet, or TCP/IP for that matter, are necessary?
    – mikem
    May 24, 2022 at 5:44
  • 1
    You may consider alternative protocols like RS-485
    – user253751
    May 24, 2022 at 8:17
  • 6
    Could it be that you are a bit cofused by "star", singular? You have a central switch, running cables to each room. Not "each device". I mean, you can. But you can also, once there in a room, add another switch. Stars, plural.
    – DevSolar
    May 24, 2022 at 16:25

13 Answers 13


10Base-5 (thicknet) and 10Base-2 (thinnet) physical layers are bus topology, if you want ethernet on a bus topology. At 10 Mbit and 100 Mbit top speeds, of course.

Might be a bit hard to find hardware for them other than scrapyards, these days, but I can't say I've looked for any recently, so perhaps it's still being made by some crazy fool...and there's plenty in scrapyards.

The more typical solution to "that's a lot of cable" for modern ethernet at modern speeds is to put switches nearer to devices (so you have multiple stars which connect to each other, rather than one star that connects to everything.) And use higher speed links between the stars if speed is an issue.

Or use some serial bus technology (like RS-422 or 1-Wire) rather than ethernet if what you want is better served by a slow serial bus. (I seem to have misremembered and not read throughly before putting that in - RS-422 is more point to point, while RS-485 as mentioned by Greg Hill is better suited to bus use and probably what I was thinking of.)

You can probably get plenty of cheap or free obsolete 10Base-T or 100Base-T switches (or hubs, egad!) if you want to do the "multi-star thing" on modern cable without breaking the bank, at moderate-to-slow speeds with the option to upgrade if your use of the wiring changes. If you are not familiar with old tech, you might have to learn about "uplink ports" or "crossover cables" to use some of that old stuff without auto-sensing.

  • 3
    Just a note that with 10Base-5/2 dealing with the physical coaxial cable and associated taps and signal loss was a major hassle.
    – Armand
    May 23, 2022 at 20:26
  • 12
    No, it was a Major hassle with 10Base-5 (I have installed vampire taps...) - it was much less of a hassle (hardly even a 2nd Lieutenant) with 10Base-2. Pull a terminator, add some cable and a Tee, put the terminator back on - or disconnect a cable, put in another Tee and a cable - easy-peasy. But they are both basically on the scrapheap of history for valid reasons.
    – Ecnerwal
    May 23, 2022 at 20:32
  • 1
    LOL Have no idea what you just said... and not even embarrassed about it. :-)
    – JACK
    May 23, 2022 at 20:57
  • 2
    You just drill a hole in the active network coax (at a designated mark on the cable jacket) and jam a spike into the hole. Fun times... en.wikipedia.org/wiki/Vampire_tap The thicknet cable was massively unmanageable and stiff, to boot.
    – Ecnerwal
    May 23, 2022 at 21:03
  • 2
    @Ecnerwal - showing your grey hairs. I worked in the group at DEC that developed that stuff.
    – DaveM
    May 24, 2022 at 0:14

So far as I'm aware there isn't any off-the-shelf Ethernet solution that uses twisted pair wiring and allows bus (as opposed to point-to-point) topology.

That said, if you can tolerate the latency, you could put an Ethernet switch at each device location and daisy chain them. It would be possible that a device could be made with a built-in Ethernet switch, but considering this kind of device is often very cost-constrained, it's unlikely such would exist.

If you were into hardware and software hacking it might be possible to tweak an Ethernet phy chip to do half-duplex and then wire a multi-drop bus with many nodes.

You might do better do pick a bus that is designed for multi-drop and long distance (RS-485 and CAN come to mind)... or just pay the little bit of extra money for cable and one or two Ethernet switches with high port count.

  • 3
    +1 for daisy chaining. You can get $10 2-port switches these days, eg. ebay.com/itm/313992370614, which you could use one per device along a bus; (chain in, device, chain out).
    – dandavis
    May 23, 2022 at 20:36
  • 4
    If the switch also gets power from PoE and forwards that POE then you don't even need a separate power run for either the IOT or the switches. May 24, 2022 at 9:42
  • 2
    @dandavis: I'm not sure that specific device will be of much use - as far as I can tell only 2 of its ports will work at once, and you'll need 3 working for the daisy chain in/out + device...
    – psmears
    May 24, 2022 at 10:11
  • 4
    @dandavis that one is an odd duck -- it's an A/B or SPDT switch for LAN! Not useful for daisy chaining unfortunately.
    – Greg Hill
    May 24, 2022 at 18:30
  • 2
    3-port switches do exist however: geizhals.eu/… But other options may be cheaper, I have multiple 5- and 8-port switches I got for free (used). One could also use old routers (just disable the DHCP servers and set unique IP addresses for the web interfaces).
    – Martin
    May 25, 2022 at 12:09

Solutions exist for both automotive Ethernet and industrial Ethernet applications.

The automotive PHY standard for multi-drop on a single twisted pair is 10BASE-T1S.

Industrial Ethernet fakes a multi-drop topology by placing an "extender" at each device. An extender is basically an Ethernet switch and has two ports for the multi-drop links both supporting PoE in and PoE out, and one to four local ports. As well as allowing a multi-drop cable run topology the extenders allow network segments many kilometres in length.


If you are really into Ethernet with the TCP/IP protocol you could consider the possibility to deploy some "power line Ethernet adapters". These can link together several local network segments using the AC powerlines as a kind of a "bus structure".

You could even connect a small hub at the power line adapter to branch out to several endpoints locally.

My personal experience with AC power line adapters for Ethernet is that they work but performance is less than advertised.

  • 3
    AIUI, Ethernet over Powerline will work across the whole house, but it works best when the outlets are on the same circuit breaker.
    – FreeMan
    May 23, 2022 at 22:57
  • 4
    @FreeMan they also can work between houses too, an interesting "boundary" condition especially when the earlier units are default-configured with no encryption.
    – Criggie
    May 24, 2022 at 4:28
  • 2
    LOL. I hadn't thought about that, @Criggie, but it's an excellent point!
    – FreeMan
    May 24, 2022 at 11:50

Home automation is very low bandwidth, excluding anything that does video streams like doorbells or security cameras.

As such, ethernet is overkill AND you may have a lot of home automation sensors in places that make no sense for an RJ45 ethernet jack.

Many sensors support low-power wireless protocols like Z-Wave or Zigbee, which use one base station to connect multiple sensors to ethernet. Those sensors can run off a couple of watch batteries for up to a year.

https://en.wikipedia.org/wiki/Z-Wave and https://en.wikipedia.org/wiki/Zigbee


There are a bunch of different ways to do this: 1 8 wire ethernet cable can be split into two logical 4 wire ethernet cables.

You can also use single pair ethernet.

But the easiest thing is to consider a snowflake topology -- multiple mini switches with an ethernet backbone; these switches can be powered via POE.

  • 1
    This is it right here - OP should be deploying switches where suitable. I might have said "hubs" but switches are cheap enough nowadays. May 23, 2022 at 22:39

The other consideration is find a better cable source - 8 wire/4 pair stranded cat6 UTP is not too expensive at around $170 USD/1000 foot box. If you can locate your patch panel centrally that helps a lot with minimising total run.

Also see if you can install ethernet jacks in a way that make them useful in the future for other purposes.

Ethernet based on cat5 might save a few dollars, but the cost is in the labour not the parts, so don't cheap-out and shorten total installed life.

Note some suppliers have odd lengths, with amazon showing listings for 250 foot, 500 foot and 1000 foot boxes. Make sure you're comparing similar products.

For longer life, consider 22AWG rather than the cheaper and thinner 23 and 24AWG. Thicker wires do cost more, but they carry POE better and further with less loss.


Disclaimer: I work in IT and I also do lots of home automation.

Do not try to invent something new because you will end up with unexpected behaviour you will have a hard time debugging. Stick to well-known technologies.

If your problem is "lots of cables", go wireless (this is going to be the case anyway with your IoT devices that discuss with the MQTT broker). We are talking about insignificant bandwidths.

There are industrial solutions for your problems but they will be expensive and mounting them will be a problem (the person that will do the work will have no idea). Then you need to make sure they fit to your device, or buy a converter.

Ethernet is indeed a lot of volume. But does this really matter? Do you care about the volume of your plumbing or wiring? This is a home so you can more or less predict the right places and the right amount of sockets (1) so go for something reasonably future-proof and add one more socket on the diagonal from the places others will land. I would also plan for a power socket nearby, just in case.

(1) ha ha ha ha ha ... oh god ... ha ha ha. sorry.


EtherCAT is a automation technology based on daisy chain and hub+spoke , as well as drop-line topology using standard ethernet technology (100BASE-TX, 1000BASE-T, and 100BASE-FX + others) .

There are a variety of PLCs , relays, I/O, bus couplers, and sensors that are available to use. Because of the topology it greatly reduces wiring requirements for automation networks. Effectively, your entire wiring topology is a single run of suitable cable snaking its way between all the devices (daisy) . For redundancy this can be doubled or split into several branches each for resiliency.

Since each segment is a bona-fide ethernet link, in a daisy chain, up to 100m/300' are allowed between nodes (and several km for fiber), so nodes beyond 100m/300' can be reached with a single cable run, assuming there are some nodes closer than 100m/300' away from each other to make the network.

Downside, as a primarily industrial technology it carries a price premium. There is a limited amount of free tools/software to work with the protocol, but it is based on standard ethernet frames.

enter image description here

EtherCAT Topology, source: beckhoff

enter image description here

Visual depiction of the topology, linking 3 nodes together (motion controllers) source: beckhoff

COI disclosure: I don't work for the non-profit ethercat technology group (ETG) or Beckhoff GMBH which develops the technology, but I have contributed to the specifications published by ETG and do have some vested interest in its long term success.

  • 3
    +1 because this is serially wired as the OP asked for. However while it is Ethernet, it's not TCP/IP and you will not find any off-the-shelf home automation gear supporting it. Only industrial equipment that you'd have to "roll your own" home automation with.
    – IceGlasses
    May 27, 2022 at 4:45

I have been there myself. Because of nearby transmitters etc I could not use WiFi so I elected to go to a bus topology as it appears you did. I looked at several and built both a small CAN and RS485 networks. The end result, for me, CAN was the best for the physical layer. It also has many other benefits in that since I build my own nodes I can use a lot of different controllers. I ended up selecting Arduino (mostly Nano) and a Pi for the master.

This is replacing a system that has been in operation for over 30 years. A typical node would have 16 outputs which I have a custom relay driver built and operating as well as a few other items. The existing system has 126 24V DC relays switching the mains installed in various locations throughout the home.

The existing interface to the PC has 128 inputs, the 128 outputs from the PC are HCMOS signals. The interface also has a RTC, CAN, and a few other items including a display all with an arduino Nano.


Is it possible to have a bus topology for home automation using ethernet?

Basically, no.

Ethernet gives you flexibility, high throughput, low latency, and galvanic isolation. But that comes at a price: each device needs a relatively smart microcontroller (usually with an IP stack), a MAC+PHY, magnetics, jack, etc. All that is too expensive for simple home automation stuff like actuators and smart switches. In addition, each ethernet port uses a non-negligible amount of power, something like 0.5W idle. If you plan on using many devices, that's a real problem.

That's why no-one uses it for this purpose. Also, it is not possible to daisy-chain.

These days, the simplest solution would be wireless, something like ZigBee or nRF24L01. Some people even run mesh networks on nRF24L01.

Wireless is nice because you don't have wires, but the things still need power supplies. If you have lots of things, that means a lot of power supplies.

20 years ago the solution would have been CAN bus on twisted pair, that does exactly what you want: a bus that you can tap anywhere to add a device. However these days there's a lot of existing designs, either DIY or commercial, using these inexpensive wireless chips. If you don't want to reinvent the wheel, that's the way to go.


Twisted pair base10-T and base100-T can use a shared conductor bus-type physical layer, this is built into the general ethernet protocol.(link protocol is layer 2 or 3, physical is layer 1) So the nodes can all be on one unswitched conductor pair permanently soldered together if desired, and they listen for packet collisions. Similar to a conversation in a group where people listen for someone to use their name(mac address) to receive a message, and they wait for an opening to inject their own message packet. If they accidentally start talking at the same time they both stop, wait a short random time, and try again.

Typical cat5/6 cable contains 4 pairs. Only 2 pair are needed for base10-T and base100-T; 1000M uses all 4 to avoid high signal frequencies. 10-T can use cat3 (or possibly less for short segments), and if building devices yourself you can use the smaller 4p4c telephone connectors rather than the fat 8p8c.(rj45 is a specific wiring of 8p8c)

There are new 10BASE-T1S and 10BASE-T1L physical layer standards that use only a single twisted pair pair and a new style connector, I believe T1S is shielded(automotive) and T1L is long distance.

As for topology a bus-hub or switched star can easily have sub-hubs/switches, stars of stars. A hub is a simple copper bus no moving parts no brains, all members are openly sharing the same electrical signals. Hubs were cheap and small, mostly replaced as switches(layer 2) came down in price and bandwidth needs grew. Switches avoid congestion between high numbers of nodes by directing packets only to specific cables. That cable could have a hub and multiple nodes but only those local nodes will see each other instead of getting chatter from the whole subnet, the only exception is a broadcast address packet [ie xxx.xxx.255.255] which switches will distribute to all cables in that whole subnet. Broadcast is used for things like requesting a dynamic ip address and network info from an unknown server at startup and to learn the mac addresses of other nodes in the subnet.


I will probably be blasted for this but a possible solution is to wire a star configuration to appear as a bus. I have not tried this but I think it it will work. CAN requires twisted pair which ethernet cable is. It has 4 pairs, you will only need two, so far so good. as an example use the orange pair as the 'input' for each node and the blue pair as the 'output' for each node. The blue and orange would be connected at each node. Between nodes you would connect the blue from the preceding to the orange pair to the next unit. In short the bus is run from node to node where the two pairs are also connected at the node. I am assuming the runs are brought to a common point. Electrically it makes one single bus which it runs besides itself in the stubs. At 500kbs a 100 meter (328 feet) should work. More at a lower baud.

  • Signal reflections will limit the speed. That’s an issue for any multi-drop network even if it isn’t physically a star topology and that’s why hardly anyone does this.
    – Navin
    Jun 11, 2022 at 23:39
  • If the bus is properly terminated, signal reflections will not be (much of?) a problem.
    – Milind R
    Aug 3, 2023 at 13:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.