I am designing a Power Line Communication Network which has to power several LED lights, which will be consuming top 5A of current, and send their control information, RGB values for each light. The data rate does not need to be very high, 5kbps is enough. The modulation scheme is not designed yet, I am first evaluating the electrical technological feasibility. Therefore, I am planning on designing just specifying a frequency range where the data will be transmitted. The higher the frequency, the easier it is to distinguish the supply (Low frequency) from the data (high frequency), so I set a minimum frequency the data spectrum will occupy.

The application requirements are:

  • 12V DC PLC or 24V AC PLC
  • Max DC current: 5A
  • Data minimum frequency: Flexible, but I'd like it to be around 100kHz (or less).
  • Min Data Rate: 100bps
  • As cheap as possible

The power supply side will be something like this, with a microcontroller generating the modulated data to send:

enter image description here

(Could also be an AC supply)

The main design challenge is the electrical summing operation to form the Supply+data PLC signal. Initially I wanted to use the MAX20340 IC, but it can't handle the current this application needs.

As a first approach, I am following this answer, the DC supply goes to a LPF, the data through a HPF and I get the superposition. I simulated that in LTSpice, being the filters a simple inductor and capacitor:

LTSpice Simulation

I think something like this could work, but I am not sure. For example, the gain peak caused by the LC resonation may cause problems. I'm also not familiarized with PLC-specific implementation problems. I thought about removing the inductor, what specs of the voltage source should I check to see if this is possible? I only thought about it's output impedance, which in this case I'd like it to be non-zero, but with the currents I'm handling I don't think this is really an option.

Another option I saw is using a transformer, but not in the PLC context. I'm not sure if I can isolate the data generation block's GND from the power supply GND.

I also saw PLC is used in Power Over Ethernet (POE), and that there they also use transformers, but using 4 instead of 2 wires: enter image description here Source: https://kintronics.com/how-power-over-ethernet-works/

To sum up, I've thrown everything I've seen so far, but:

  1. I'm pretty lost when it comes to deciding if what I'm seeing is a viable solution or not.
  2. What aspects of the solution I choose should I analyze to know I won't have problems when implementing this.

As a side-comment, I read this answer and freaked out, that's why I am really doubtful about the solution I choose working in real life.

Edit1: I added the possibility of using AC supply and communication instead of DC, but I haven't researched it yet.

Edit2: I lowered the bit rate requirement to 100bps, I realized I could manage to work with this really low speed. This allows me to consider for example the X-10 protocol.

  • \$\begingroup\$ see yamar.com \$\endgroup\$
    – Axis
    Commented Oct 20, 2021 at 19:30
  • \$\begingroup\$ I checked their ICs but I think they can't handle this amount of current. For example, the SIG100 mentions in the pinout description (Table 1) the output current and the TXO's output current is maximum 66mA. \$\endgroup\$ Commented Oct 21, 2021 at 11:34
  • 1
    \$\begingroup\$ It seems to me you confuse the power supply output power and the communications output power. The data signal is superimposed on the power line and does not require very high power. If your lines aren't too long ofcourse. For instance in your PoE example, the datalines are modulated (and recovered) on the data wires with use of transformers. On the other side, DC power is recovered with a DC-DC converter which is usually also isolated to conform to IEEE spec. \$\endgroup\$
    – N. Berg
    Commented Oct 22, 2021 at 12:46
  • \$\begingroup\$ Thanks for your comment! I understand the difference but I may have been unclear when writing the question. The current the IC (MAX20340 or YAMR SIGXX) handles is indeed enough for the data, but the problem is that, according to my understanding, its output is directly the powerline, so all the power should be handled by the IC, right?. Or are you suggesting somehow adding the IC's output to a 12V line? In that case I think I would be in the situation presented in the first block diagram, replacing the MCU block with a dedicated IC, but I'm not sure that's how they are supposed to be used. \$\endgroup\$ Commented Oct 24, 2021 at 17:17
  • \$\begingroup\$ "its output is directly the powerline, so all the power should be handled by the IC, right?" - no. The PLC chips only modulate data onto an existing power line. Look at figure 1 on the first page of the datasheet yamar.com/datasheet/DS-SIG100.pdf A compressor and air-conditioner surely don't run on 66mA... \$\endgroup\$
    – Axis
    Commented Oct 24, 2021 at 21:38

2 Answers 2


I think a much easier way to do this that meets your requirements would be to just invert the polarity of the DC current to send serial bits- current in one direction is a mark while current in the other direction is a space. Like an RS-485 current loop, but with 5A current and a full reversal between states. Very, very simple and extremely robust.

To drive the pair of power lines, you can use a cheap and readily available H-bridge chip or board (they have these on Amazon for around $15), or roll your own with MOSFETs. Connect this to any serial data source.

On each LED node, you'd have a full bridge rectifier to provide the power to the LEDs and a small (i.e. $0.25 ATTINY) microcontroller. Add a smoothing cap across the rectified DC. A $0.10 5V voltage regulator powers the microcontroller. To recover the data, you could do something as simple as feeding one of the AC conductors into one of the IO pins on the microcontroller through a large current limiting resistor. Since almost all microcontrollers (i.e. ATTINY) have built-in protection diodes on the IO pins, you could simply read the IO pin to read the data bit off the line. You could even potentially connect the resistor directly to the microcontroller's hardware serial RX pin and have it decode the serial data for you.

Since you will already have a microcontroller in each node, mind as well have it also control the RGB LEDs. Connect three of the PWM pins from the microcontroller each to a MOSFET (possibly with a current limiting resistor) to turn on/off each of the 3 RGB LEDs. Give each node an address, have firmware in each node look for its address followed by RGB values in the serial data stream, latch those values into the RGB output PWMs... and now you have a working system - potentially for less than $1 in hardware per node and less than $20 for the transmitter.

enter image description here Very rough sketch. There should be at least 3 LEDs connected to the MCU (R,G,B) and there can be any number of additional nodes hanging on the lines to the left...

(Note there lots of details and optimizations not mentioned here, this is just an overview!)

  • \$\begingroup\$ Thank you very much for your answer, you really thought about all the implementation details! It seems to be a really good option. The transmitter may be more expensive than with the YAMAR SIG100 option (the IC costs around $6), but the receivers are reaaally simple and cheap. I see two drawbacks but I may be able to live with both of them. 1. Because of the diode rectification voltage drop, I wouldn't have 12V on the receiver side. 2. The H-bridge's switching losses due to switching the whole voltage and not just a little AC signal. I'll analyze it further to decide what to do, thanks again! \$\endgroup\$ Commented Oct 27, 2021 at 15:13
  • \$\begingroup\$ Just the SIG100 chip is >$5, to build a full board likely many $10's of dollars even at quantity. If you are price sensitive on the transmitter, you can build a 12Vx5A H-Bridge out of MOSFETS likely for <$5. Power losses in the transmitter at 100bps are likely negligible compared to power used even by a single LED. \$\endgroup\$
    – bigjosh
    Commented Oct 27, 2021 at 19:23
  • \$\begingroup\$ If you really care about diode forward voltage drop, you can rectify with MOSFETs... but I doubt you care since no LEDs need 12V and so you are likely going to be using a current limiting resistor in series with the LED anyway so just use a slightly smaller resistor. \$\endgroup\$
    – bigjosh
    Commented Oct 27, 2021 at 19:25
  • 1
    \$\begingroup\$ @frsc Valid concerns, but at the OP's required data rate of 100bps the switching losses and EMC should be negligible. \$\endgroup\$
    – bigjosh
    Commented Jun 6, 2022 at 14:49
  • 1
    \$\begingroup\$ @bigjosh Right, especially if data transmission is only needed at rare intervals. Still, if you want to avoid some rather large H-bridge, you could use the OP's original approach of modulating the data onto the supply voltage (something like described in electronicdesign.com/technologies/communications/article/…). This would even allow for bidirectional transmission if required. I'm currently thinking about which way to go for a very similar application as described by the OP. \$\endgroup\$
    – frsc
    Commented Jun 7, 2022 at 18:05

I'll use the YAMAR SIG100 IC, as suggested by N. Berg and Axis.

The IC's data is coupled to the powerline through a capacitor, and on the receiver side another capacitor filters the DC component. Therefore, the only component that must handle all the current is the optional inductor connected to the main power supply.

enter image description here

It was chosen because it is probably the option that minimizes the development time, a requirement I didn't specify but that is always present.

Another option that met all my requirements was given in bigjosh's answer. It is cheaper but requires more development time.


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.