Is it possible to make a device that can detect but not interfere with an ethernet signal?

How would I do that?

3/17/2015 3:30:15 PM

remotely or by being in-line?

3/17/2015 3:55:22 PM

http://en.wikipedia.org/wiki/Port_mirroring

Not hard to find an inexpensive smart/managed switch with mirroring.

3/17/2015 5:06:00 PM

^^Being inline.

^I'm trying to build a device that senses the presence of ethernet and does some stuff with the signal. buying a switch with port mirroring kinda defeats the purpose.

3/17/2015 6:15:36 PM

hax teh planet: http://ossmann.blogspot.com/2011/02/throwing-star-lan-tap.html

3/17/2015 8:50:33 PM

^That's on the right track, but I need my circuit to actively sense the presence of the right type of signal (10baseT) and then provide some out bits.

How do?

3/18/2015 9:18:09 AM

Do you wan't it to detect every possible ethernet signal or a particular one..

Have to be precise on those kind of things.

3/18/2015 9:58:14 AM

I'd start by looking at autonegotiation:

http://en.m.wikipedia.org/wiki/Autonegotiation

What are the bits you want to output and where are they going?

Quote :

"Have to be precise on those kind of things."

This would help enormously.

3/18/2015 10:56:19 AM

Sounds like he's trying to do something sketchy...

3/18/2015 11:00:03 AM

I'm not trying to do anything sketchy.

I'm trying to detect Ethernet carrier signal - the connection pulses that are described in the Electrrical Signals section of the wikipedia page. Let's say for the sake of a proof of concept that I'm only interested in 100Mb BASE-T and below. I explicitly DO NOT want to actually read or parse the data that's being transmitted, only that a device is connected to the wire and is reaching out to detect the presence of another device connected to the wire.

circuit would put out a constant ~5V (whatever a typical arduino might expect) as long as there is a signal, and 0V if there's not one.


I'm kinda feeling through this blind... So if there's more information you need let me know what it is and I'll provide it. I appreciate the help.

3/18/2015 8:44:11 PM

since it uses differential signaling (in most forms), you could build a circuit that would detect a voltage from pin 1 to pin 2.

if data is being sent you should always have a voltage, should be able to detect it.

or you could look at a non contact method.

http://makezine.com/projects/non-contact-voltage-detector/

3/18/2015 10:55:01 PM

Tap into it w a buffer opamp (high impedance input won't interfere w the signal) and run it to a rising edge trigger interrupt. In the ISR set a flag and do something in software or set a digital output high.

3/18/2015 11:30:28 PM

But why are you trying to do this... it would be more effective and easier, if you're trying to create an external device to make a device that reads the status lights of most switches (these often indicate connection rate).

Or make something in-line like ^^ suggests.

Trying to measure this inductively has a lot of external factors that affect the design, and is probably a bit more complicated than you're trying to get.

But to do what you'd want to do, you're better off looping some wires around an ethernet cable, hooking it up to an oscilloscope, look at what the electrical signal is doing, then see if there are any thresholds or timings you can pick out with a µcontroller to report the status you want.

3/18/2015 11:36:14 PM

Are you effectively trying to make a Link light?

Stop being so cryptic about what you are trying to accomplish.

3/19/2015 8:55:56 AM

I don't know how a link light works, so I'm not sure whether that's what I'm trying to make.

The end goal is for this circuit to be used as part of an automatic pre-switching physical crossbar. An arduino will cycle through several different physical cables and will set relays to either connect this physical cable to another physical cable, or disconnect from the physical cable based on whether a device trying to speak ethernet is detected on the line. If you're familiar with patch panels, the idea is to put in a simple and cheap device that automates the process of connecting and disconnecting lines because it's boring and time consuming.

I have the arduino switching mechanism working, I just need the part that senses the presence of the ethernet link from the host across the line to know whether to connect or disconnect.

3/19/2015 8:45:14 PM

I could be completely missing the point here, but why not get a big enough switch so every wire is always patched in?

3/20/2015 1:02:46 AM

^This

3/20/2015 6:23:38 AM

The short answer is "money."

As small and medium businesses with their own IT infrastructure grow, they outgrow patch panels by several connections. As long as only $patch_panel_ports connections are being used at any time, a business can save money by just going and switching out unused wires in the panel with wires that are in use. Switches are expensive, they have software that makes them really expensive, and it's also costly in terms of time and manpower to send someone to switch out the patch cables constantly.

If someone made a much cheaper solution than a switch to the patch panel problem, they could alleviate growth pains for smaller companies by pushing out IT costs and decreasing manpower required to service the infrastructure. our solution is this brain dead patch switching arduino.

Is it guaranteed to work as a business? No. We're going to try it because we think it brings value.

We have a huge chunk of the work done. What I need is helping determining the best way to non-disruptively detect signal on the line.

3/20/2015 9:38:28 AM

^I do analog and digital work. Contact me for work.

3/20/2015 10:47:22 AM

Why bother switching on signal? Your typical use case is probably going to work with scheduling. Allow the user to program a schedule for switching.

3/20/2015 10:53:24 AM

Quote :

"Your typical use case is probably going to work with scheduling."

^That's not true.

3/20/2015 12:39:37 PM

Okay. I don't believe you.

3/20/2015 1:53:12 PM

If the business is big enough that they have their own office space (read: not regus or the like), the cost of switching and patch panels is a pittance. If the company is technically mature enough that they need a segmented, managed network, it's not expensive to expand the network. Finally, if the company is technically unsophisticated AND they own their own space, they'll go wireless before they fuck with this. I'm all for the entrepreneurial spirit, but this is a complete waste of time and energy for folks that seem otherwise talented enough to even embark on this.

3/20/2015 2:24:01 PM

You're entitled to your opinion, but I honestly don't give a shit what you think.

I need someone who can help me design a circuit that does one thing: sense ethernet. If you can help me with that, please do. Otherwise, STFU.

I'm not in here to get business advice, I'm in here to get technical advice.

3/20/2015 2:48:29 PM

ha so you're trying to sense when there's no signal on a patched connection, and the switch OUT the patched connection with a random other cable?

Or you trying to sense when there is signal on an unpatched connection? I am not familiar with the kind of patch panels you're using, but I don't think you can easily sense a signal on an unconnected ethernet.

3/20/2015 3:07:06 PM

Quote :

"Or you trying to sense when there is signal on an unpatched connection?"

This one.


Then we'll need to know when the signal ends as well.

3/21/2015 8:58:53 AM

I could be miss understanding what you're trying to do, but a NIC is constantly sending out link pulses waiting for a response, which then triggers the auto negotiation process which then brings the connection up on higher layers of the OSI.

To my limited knowledge, the link pulses don't change whether a computer is trying to reach the Internet or not. So if you have an unconnected etherent cable, there's no way to know that it wants a connection vs sitting idle.

It may be possible to create an application that runs on the host that overrides this behavior (I think windows puts pings on the wire regardless of connection state but not sure) , but even then it might be difficult to detect these on an open circuit twisted pair.

3/22/2015 12:17:58 AM

Have you actually priced out the cost to do this the correct way, with more patch panels and switches?

How are you going to implement your home grown fix? You would at least have to buy all the jacks and whatever multiplexed switching setup you are going to use, plus circuit boards, power supply, etc....

I just don't see how you can even fathom doing a hack like this instead of just expanding their network capacity.

For $100, you can go the cheap route, skip the patch panels and terminate the cables directly into a 24 port gigabit dumb switch: http://www.amazon.com/TRENDnet-24-Port-Unmanaged-GREENnet-TEG-S24Dg/dp/B0044GH27U/ref=sr_1_2?ie=UTF8&qid=1427030605&sr=8-2&keywords=gigabit+switch+24

Stack 10 of them up if you have to.

Stop wasting time on a solution that isn't going to save any time or money.

3/22/2015 9:25:09 AM

the difference between needing to patch in a cable or not is the difference between having a host plug in to the corresponding jack or not.

As an oversimplified example, consider the following:

There's a school library with 50 jacks available for kids to plug their laptops in to, but the library only has a 24 port switch connecting the library infrastructure to the school's backbone. If someone needs internet and plugs in a cable to an unpatched connection, an IT guy has to walk over to the patch panel and switch that cable out with another one.

There won't be any pulse signals on a line with no host connected unless it's been auto-patched into the switch - at that point the line might be receiving the link pulses from the switch. In which case I may need to be able to distinguish between link pulses sent from a particular direction (coming from the switch not connectedd to a host vs coming from the host not connected to a switch.) Not sure how to solve that problem.

3/22/2015 2:07:59 PM

let's follow your oversimplified example a bit further. what school (or any building) is built with the foresight to add ethernet jacks in some 50 locations, but without enough switching/patch panels to accommodate them? no one does one without the other. if they do, they're stupid, and the solution is what Wolfmarsh mentions, or they do one even simpler and more versatile, and add wifi. you can't possibly sit here and think an arduino and a limited volume board (that probly cost on the order of $75 together) are some how more effective than a dumb 24 port switch for the same price. hell in an environment like this, 10/100 switches would work fine and those are $15 for quality old stock.

[Edited on March 22, 2015 at 4:21 PM. Reason : im trying to be nice and pragmatically explain how terrible this idea is]

3/22/2015 4:13:19 PM

I need help designing a circuit, not understanding enterprise IT.

That example was illustrative, and as I wrote in the post, overly simplified.

following it any further is retarded, because it's not a real life scenario.

3/22/2015 5:20:13 PM

If I understand this correctly, you basically want to build some sort of X:Y ethernet mux where you have more X than Y?

Quote :

"Tap into it w a buffer opamp (high impedance input won't interfere w the signal) and run it to a rising edge trigger interrupt. In the ISR set a flag and do something in software or set a digital output high."

This seems like a start.

If this is just a matter of monitoring for nothing and then something then I think as long as your device can respond to the pulse widths of an ethernet line then you could just constantly monitor for activity as denoted by a line that is occasionally changing from nothing to something.

Only issue I see is most high speed digital stuff wants impedance matching so even a high impedance op amp hanging on the line might cause reflections. If you have access to a good enough scope you could check to see what happens when you do this. You could some real quick and dirty proof of concept here without spending too much scratch.

How are you implementing the "switching mechanism"?

This sounds a lot more like you're trying to design something to automate testing.

[Edited on March 23, 2015 at 9:58 PM. Reason : .]

3/23/2015 9:57:09 PM

Quote :

"This sounds a lot more like you're trying to design something to automate testing."

Sounds like he is trying to solve a problem that doesn't exist, with a complicated solution that nobody in the future will be able to support because it's a stupid hack.

I could put this circuit together in an hour, but I'm not about to help this guy anymore.

[Edited on March 24, 2015 at 9:24 AM. Reason : .]

3/24/2015 9:23:29 AM

Jeez that's a little harsh... Sometimes the journey is more important than the end. Even if this was a completely pointless product, what's the harm in just exploring an arduino prototype for this 1 applicantion?

3/24/2015 12:11:55 PM

It seems to me OP is looking for someone to provide a final answer rather than actual help developing his own. He's more interested in you taking the journey than himself. Not a good way to get help, particularly when it's something he seems interested in commercializing.

(IMHO)

3/24/2015 1:05:27 PM

Is there not an answer already? I'll admit, I didn't come in here thinking I'd need to invent a way to measure this sort of thing, has no one done it before? Is it not clear how they'd do it?

I'm not an electrical engineer, my knowledge is working at best.

3/24/2015 2:54:51 PM

http://www.digikey.com/short/7rnffn

3/24/2015 6:40:53 PM

Quote :

"I could put this circuit together in an hour, but I'm not about to help this guy anymore."

Really? If your real wang was as large as your e-wang you'd spend a measly 1 hour to prove it. I bet you can't get this to work in a week of trying.

3/26/2015 10:39:18 PM

Quote :

"Really? If your real wang was as large as your e-wang you'd spend a measly 1 hour to prove it. I bet you can't get this to work in a week of trying."

You aren't going to goad me into doing it.

It may seem harsh, but I'm not going to waste my time on something I think is being done the wrong way.

3/28/2015 9:30:46 AM

if you honestly are looking at that number of devices that are going to be off, you could throw hubs in there and avoid the hassle.

your talking about re-creating a switchboard from the 1940s

3/29/2015 2:04:32 AM

If you literally could do this in an hour you'd do it regardless of whether there is some better way and with what information has been posted here you clearly can't know what that could be.

3/29/2015 4:24:02 PM

whatever happened with this dumbass idea?

7/15/2019 3:05:49 PM

I'd be interested to know as well. I've been in IT now for almost 15 years, and I just don't see the issues out there that this solution is trying to solve.

7/16/2019 9:05:56 AM

A Rube Goldburg robot that constantly cycles around checking to see if someone has plugged an Ethernet cable in somewhere, goes and grabs a free cable and plugs it in for you, all without getting the cables tangled. Nevermind detecting when an Ethernet connection is in use, this robot would be insanely complicated and expensive. Because once it plugs in an Ethernet cable, it needs to keep accessing all the other available jacks. And however simple the circuit is to monitor an Ethernet port for someone plugging in, until you're making a million of them it'll always be cheaper than just buying a cheap mass produced Ethernet switch to provide actual Ethernet to that port, nevermind the cost of the friggin' robot.

Another Ethernet switch port and a patch cable will ALWAYS be FAR cheaper than a circuit to monitor insertion on the port plus the railing for the robot to crawl along to reach all the ports. As such, this never becomes cheaper, even if you have something crazy like thousands of possible Ethernet ports with only, say, 24 of them in use at any one time, and some reason WiFi can't be a thing (top secret?).

That said, if you got rid of the monitoring circuit, say people used their phones to ping the robot over WiFi to have a particular secure jack patched in, then a patch running robot becomes an actually good solution in this ridiculous scenario.

7/20/2019 9:07:03 AM