Ethernet, Over DC Power

We’re used to extending our network connections and being no longer constrained in our use of Ethernet by proximity to a switch or hub. Our houses routinely contain wireless networks, and of course powerline-Ethernet units passing data over our mains wiring. [Peter Franck] had a similar problem but without the mains power, for a distributed sprinkler system he needed to send Ethernet over DC cables.

The solution is a surprisingly simple one, taking one of those powerline Ethernet units and converting it by removing its mains power section. These devices contain the Ethernet and powerline modem chip with its associated circuitry, and a small switch-mode power supply. He’s removed the power supply and put in a capacitive coupling to the DC cabling, resulting in a relatively inexpensive DC powerline network device.

Powerline Ethernet devices are not without their own issues, for instance they are not popular with radio amateurs due to their effect on the RF noise floor. We’d therefore be curious to see what the RF emissions are like for this hack, but we still think it’s a useful weapon in the armoury as well as something to do with all those surplus powerline Ethernet bricks.

63 thoughts on “Ethernet, Over DC Power

    1. Where did you get the 30m from?

      To quote the source: “The purpose of this hack is powering and connecting multiple Raspberry Pi computers controlling a 24V DC solar irrigation system in a remote place with no AC or network line in reach and just 3 wires in the ground.”

      Possible reasons I see:
      1. W-LAN and water from an irrigation system isn’t the best combination.
      2. keep the air free from unnecessary low-usage networks. different networks for irrigation system (red), yourself (green) and guests (yellow/orange).
      3. better reliability
      4. PXE/etherboot/netboot

      1. To quote the source: “transmits 100 MBit/s over a thirty meters spool of telephone cable”

        Second point under “some data:” on Peter Franck’s site.

        Now obviously we don’t know if any/all of his runs were 30 meters but the point still stands that given those distances WiFi could have reached the devices. On the other hand there may not have been a good way to implement wifi.

        1. Well, the location where this is going to be deployed has topological difficulties that would render a WiFi system very costly. Having said that, there is a WiFi uplink to the internet from one of the nodes.

        2. BTW: the telephone cable was just meant to be an example lab setup. Actually, the deployment scenario is various lengths of 2.5 square mm power cables. 100 MBit/s is absolutely not necessary for the application. It just leaves a comfortable margin for signal attenuation in the field. I can add some more data when it will finally have been deployed.

          1. Please share more once you have details. I am quite curious how this will work. I would normally think of using RS485 for this, but then it gets complicated when you actually want internet access instead of just a serial link.

    1. Could only have been about 5 years before doing it all in 10/100 base T came down below total cost. Running 10 base 2 through the whole house was pretty decent even until media files got large.

      For giggles sometime I’m gonna try making baluns for 50 ohm to 300 ohm and try connecting vintage PCs with coax cards through Cat 5 runs.

      1. 10Mbit Ethernet ran just fine on phone line, and I don’t mean the twisted cat3 kind. Was it noisy? Yes. Did it allow 100 meter distance? Dunno. Never tried more than ~30, but guessing 100 would have been a stretch. 100Mbit over Cat3 was fine too in a pinch.

  1. It’s a nice idea, but for a sprinkler system does he really need 100MBit/s? I would guess some old RS232 would be enough for some command and status. IC is less than a dollar i guess, EMI is much better and it will work for hundreds of meters (depending on speed).

      1. I’ve seen this done before on HaD (maybe without the power injection) but what about the throughput? AFAIU WLAN and powerlan are both half-duplex but WLAN uses either 2,4 or 5GHz while HomeplugAV is between 2 and 68 MHz(!).
        The later seems far better suited for normal coax cabling (the ones used for DVB-T/S2).

        1. You just need decent enough cable, and it doesn’t even need to be THAT good. Losses over coax are linear with distance, they’re quadratic for free space propagation. Cabling always wins.

          Over 100 ft you can use pretty much anything you want RG-6 with F connectors is still only going to be like, 20-30 dB loss, which is the equivalent of free-space with 20 dBi gain antennas on both sides.

    1. Yes, that’s absolutely possible. Homeplug AV does not necessarily need power on the power wires. It has been proven by this hack.
      You can easily “UPS” the adapters by supplying 4.5 to 17 Volts to the modem circuitry but take extreme care when hacking them to stay on AC mains to avoid safety issues.

    2. A good idea, I wonder if the powerline devices would perform any better in this case too, with the noise reduced. Heck they might perform better without a blackout just by being powered from a more stable DC source..

      1. If you remove the switching power supply as described in the original posting it would in deed reduce the noise on your medium. However, the effect may be negligible since the original circuit provides inductors L2 and L4 as a low-pass filter between its internal power supply and the medium.

  2. This type of tech is actually being standardized as single pair Ethernet in IEEE and will be making it’s way into cars and industrial plants. Allegedly early tests even got it to work over two spools of barbed wire!

    1. Ok that’s just amazing and needs to be everywhere right now. I’d love to see 2.1mm plugs with this replace USB for Arduinos. If this ever makes it into an Espressif product it will be so useful!

  3. I did this once (happened as a side effect really). Hot a pair of AV200 tenda units, supposedly broken. Did some digging, supplied external 3V3. In the process, i figured i should test the connection without AC power between the units. Sure enough, it worked at full 100Mbit.

    Pretty nifty hack. Too bad there aren’t any consumer devices, that would use existing powerline chipsets but allow you to use any type of compatabile wiring you want.

    1. I chose the ones with a 100 MBit Ethernet interface for power consumption reasons. They were supposed to be the least power-hungry devices by spec. The modem is probably potentially faster than 100 MBit.
      You could choose the ones with a GBit Ethernet interface where power consumption is not an issue.

        1. Let me quote https://www.tp-link.com/us/home-networking/powerline/tl-pa4010/

          “Interface: 1*10/100Mbps Ethernet Port”

          So the limiting component is the Ethernet interface of this specific HP/AV adapters. Since I don’t even need 10 Mbps in my application I happily traded power consumption for speed.

          If you have a need for speed and you don’t care about an extra Watt, TL-PA7017 would be your best bet. Though I haven’t tried it out personally I would expect a very similar design except for the Ethernet interface part. At least from the outside they look very very similar.

          see https://www.tp-link.com/us/home-networking/powerline/tl-pa7017-kit/ for specifications

  4. If you used these on the common-mode center taps of PoE, could you do EoPoE? I’m actually curious if you could get gigabit speeds on the ethernet itself (despite the added common-mode noise) plus whatever speed you get from the powerline adapters, to increase total throughput.

    1. I would rather try multi-GBit interfaces to increase throughput in a scenario where structured cabling is already installed. I strongly doubt PLC will increase overall bandwidth on top of GBit ETH, with or without PoE, though I haven’t actually tried it.

    2. Your idea inspired this nefarious possibility:
      Could it be possible to establish a hidden/dark net through a PoE switch to jump VLANs?

      Basically connecting two compromised devices equipped with powline-LAN adapters on the common-mode center taps to two ports of a PoE switch on different VLANs.

      Depending on how the switch “does” its PoE it should be possible, shouldn’t it?

  5. Hi everyone, I am trying to implement this hack with a pair of TP-Link AV600 powerline adapters. However, I was thinking of setting the transmission line voltage to 48V instead of 12V. Basically, I will be using 12V for powering the signaling circuit and 48V for transmission. Do I need to make the ground common? Or both the grounds should be isolated?

    Any help is appreciated.

    1. Hi Abhilash Ingale,

      Your Application should be functioning just fine. The DC voltage on the transmission line doesn’t matter as long as it is lower than mains Vpp (roughly 330 VDC).

      The different grounds should be isolated.

      The Modem circuit is isolated from the line circuit by design.

      Peter

      1. Thanks Peter.

        I have a follow-up question.

        I read that these decoupling caps are for filtering high frequency noise from the input. I do not have a10nF ceramic cap.

        Could I keep the existing caps if I am okay with a little noise / loss in data transfer speed? Will it be good enough to demonstrate its working?

        Please help.

        1. The capacitor is not meant for noise reduction because in this case the “high frequency noise” is the actual modem signal ;-). The series capacitor’s function is a high pass filter that couples the high frequency signal of the modem into the medium.

          Long story short: the original capacitor is just fine!

          I replaced it only due do its bulky form factor. You definitely don’t need to do that.

          1. Thank you for the insight, Peter!

            One more request –

            I need your assistance in removing components from the AV600 PLC receiver (please refer the link for images). I am confused on which components need to be removed in the red bounding box area? Which of the capacitors should I keep and which of them should I desolder ?

            I am a newbie and not experienced enough to read multi-layer PCBs. Besides, this powerline adapter is not readily available in my country and hence, I don’t want to make rookie mistakes. I hope you would be able to help me out here.

            ( Drive link – https://drive.google.com/open?id=1aTOFlh7yqFL8tstmnm1XYTQA_-bwurfl )

          2. I think you need to keep C1, R5, D6 and the maroon colored brick (a fuse that was black in my device).

            Check if one terminal of C1 connects directly to the ferrite transformer. If yes, that is the right component to keep.

            Anyway, mark the capacitors in order to remember which one was where, so if you removed the wrong one, you can just put it back in and all will be fine. Don’t worry. You won’t break it.

          3. Hi Peter,

            Thank you so much for the time that you are putting in. However, I did not understand which transformer you were referring to when you said ferrite transformer. Is it the ‘big one’ or the ‘small one’?

          4. Oh, I’m very sorry for that.

            The component I tried to refer to is marked “T4” in your circuit.
            The dark square thing with 4 terminals, a few turns of tan wire on a core, a little white dab of white paint on top and the PCB slit cut underneath.

      2. Hi Peter,

        So I wanted to share some developments. I desoldered the unnecessary components on both the adapters and have put the pins for powering the signaling circuit. Could you please check the polarity on the receiver? Please take a look at the image. The polarity of connections seems to be reverse of the transmitter adapter; I checked the continuity of one terminal with negative pin of electrolytic cap.

        Also, how should I connect the transmission cables? Will polarity matter there? If yes, could you please point out which ones should be positive and which ones should be negative (consider the orientation in the attached photograph) ?

        Soon after this is resolved, I would wiring it up and sharing thee updates!

        1. Hi Abhilash,

          I cannot reliably tell from the pictures the correct polarity of the power input.
          As i remember it there should be a series diode to protect you from damaging anything by reversing the polarity. You are absolutely right that the negative power input shall be connected to the negative terminal of the large electrolytic capacitor.

          The power input polarity must match the capacitor polarity.

          The polarity of the line terminals is irrelevant. Either way will do. For a testing setup i would suggest you to connect the devices by just two wires on your bench. Applying voltage to the line is not necessary for them to function.

          If you plan to continue hacking SMT stuff you should possibly get in reach of a hot air gun eventually ;-)

          1. Hi Peter,

            Thank you again for clearing the doubt. I have wired up the system and as you suggested, will try to test them without voltage first. Will update you soon about the outcome.

            And yes, I will try to get some stuff that would ease handling SMT stuff and keep things a lot cleaner. Thanks for the suggestion!

          2. Hi Peter,

            I could successfully implement the hack as per your directions.

            I also conducted a few tests on different line voltages (0V,12V,48V) and I am getting at least 50Mbps download speed (the fluctuations could be because of my breadboard connections or internet service provider as well). In practical terms, I could stream 1080p 60fps YouTube videos without buffering.

            Overall, this hack works like a charm! Thank you for all the assistance with the build.

  6. Hi Abhilash,

    you are absolutely right in that respect that the negative supply input is the one that must connect to the negative terminal of the large electrolytic capacitor. I can’t see it reliably in the picture so consult your multimeter to check.

    Input polarity must match the capacitor’s polarity!

    The polarity of the transmission line does *not* matter. For testing purposes i would suggest you to connect just two wires between the devices. There is no need to apply any power to the line in order to transmit data.

    If you continue to hack SMT stuff you should probably try to get in reach of a hot air gun eventually ;-)

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.