Optocoupler for Ethernet connection

Question

First of all sorry about my English, it's not great. Second of all the question:

I want to create a unidirectional Ethernet connection between my laptop and desktop! For that I created a circut like this (Link):

SideA           SideA           HUB PORT
-------         -------         -------
x x r r         x x r r         r r x x
6 3 1 2         6 3 2 1         1 2 3 6
| | | |         | | | |             | |
| +-/ |         | | | \-------------/ |
| |   |    OR   | | \-----------------/
+-|---/         | |             
| |             | |
| |             | |
| |             | |
6 3 2 1         6 3 2 1
r r x x         r r x x
-------         -------
SideB           SideB

It works good but because of some security problems I need a way to isolate SideA and SideB from eachother in physical layer. To do that i think optocouplers are good choices but there are some problem that I face them :

1. Fast ethernet (100baseTX) uses signal level of -1v 0v and +1v and optocouplers can't cross negative voltages.
2. Optocouplers bandwidth are limited to maximum 90 Mbps (I saw datasheets and the max was 90 Mbps) and I need 100Mbps. Let me know if higher baudrate is available.

Please share with me any idea can solve this problem. Just consider that main problem is to isolate SideA from SideB in phisical layer to be sure data flow is just form SideA to SideB and not viseversa. selection of optocouplers is just for this; i mean optocouplers are unidirectional because they use light and light just can flow unidirectionally(i'm not sure used a correct word or not!!!)

Answer 1

The transformers used in the ethernet interface already isolate the signals galvanically.

There are probably no opto-isolators available for the required speed.

Why do you think you need opto-isolators?

Answer 2

The simplest way to optically isolate two copper ethernet ports is to use a media converter to change to fiber-optic ethernet.

In a way this is an "XY Problem": You're trying to make sure that the directionality of an ethernet device can't change against your will, but you're asking about optocouplers because you assumed that's the way to solve your problem.

As it turns out, changing to fiber optic ethernet also solves the directionality concern: A fiber ethernet connection uses two strands: a transmit and a receive. Using only one strand guarantees unidirectionality. Not all hardware will support this.

Answer 3

You need to understand how ethernet works. It's not a simple piece of wire.

Opto isolation provides galvanic separation but is not the solution to filtering data.

On the physical level, as well as data frames there are synchronisation signals (called FLP, fast link pulse, IIRC) going both ways all the time, that is how the link light and speed detection works. If they are removed the link breaks and the datalink goes down. It is true that you can break the link in one direction and it might work (been there, done that) but it's not the correct solution and will break some chips. (some types of ethernet are half duplex so data travels in both directions on the same wires)

To filter data (which is transferred in frames, completely separate from the synchronisation) you need a Managed Ethernet Switch. I find the cheap ones from Mikrotik very good and easy to use.

Locate a managed switch between the devices that require data isolation. That will allow you to only allow Ethernet Frames in one direction through the management console.

Answer 4

Ethernet requires bi-directional communication to handle handshakes, packet acknowledgement, etc.

Figure 1. Bi-directional TCP traffic. Image source: Geeks for Geeks.

The linked article explains that

TCP provides reliable communication with something called Positive Acknowledgement with Re-transmission(PAR).

This requires the bi-directional communication so your proposal will not work.

Answer 5

I'd like to toss my 2 cents in for those reading this at a later date. It's now March 2023.

1.) There are two basic protocols used in modern network communication.

• TCP and UDP.
• TCP uses a rigorous firmware method to insure that data is received correctly and without error in the order it was transmitted.
• While UDP Transmits the data and and only deals with basic data integrity. Data can be out of sequence and if a packet goes missing, no errors are created.

2.) Any time you have a network to hardware connection, such as a file server or even a dedicated data storage device, things still need flow control. For this you require a bi-directional connection also called Full Duplex. This is achieved by using two twisted pairs (or more) of wire, one for transmit, the other for receive. There is a fully recognized Half Duplex Ethernet connection that is, to the best of my knowledge, seldom used, but... read on...

3.) Ethernet started on Half Duplex as it was a wireless UHF connection between computer installations. Like a CB radio, the sender had to initiate a transmit, it could be done using a single twisted pair of wires.

4.) Today - well not today, some time ago actually - a method of using different voltage levels was used to transmit/receive Full Duplex data over a single twisted pair of wires (just 2 wires). It is commonly used in machine control and the automotive industry. It ads a new character to the 100BASE-T format and is called 100BASE-T1.

5.) In a typical 10BASE-T or 100BASE-T configuration, if a user were to convert the twisted pair of wires normally used to transmit data, into a one way optical connection, there is no network protocol or device control instruction that could reverse this to read the data. The interfaces might receive and execute the function, but the optical conversion would still be the same, one way.

But, this would create a series of issues that would eventually prove fatal to accurate and intact data storage.

The most basic issue would be flow control. At the receiving end, you'd need a fault tolerant buffer at least the size of a few hours of required data storage. And, the other issue would be data transmission fault tolerance.

How do you get the system to respond to a lost or corrupted package? You don't, it's gone, possibly skewing/corrupting all the following data. What if your hard drive crashes? All data is destroyed and new data is lost.

========== Other options include a hardware firewall at the receive end, with stateful packet inspection and full administration control where you could setup a channel as receive only (the channel, not the drive). This would prevent any data from being read from the WAN side, but still allow full use of the TCP or UDP protocols. The data would still be available on the LAN side.