Gigabit Switches on a Long Ethernet Run

[tbutler@ofb.biz]

Howdy, folks,
I've run a brand new Cat 6 cable through an empty conduit between buildings. The length is a bit longer than I thought -- approximately 260 ft. (the underground conduit isn't as direct as I hoped). In addition, on my cable run, I left about 20 ft of slack -- 10 ft on one end, 7 on the other -- "just in case."

I have been trying to put Trendnet 8-port unmanaged gigabit "GreenNet" switches between the different runs. They were cheap ($25) and seemed to be well liked. So, I have one of those in the church office; the cable then runs 280 ft to the next switch. When I plug the two together, the lights do not light up on either switch (both of which work, otherwise). However, I plugged in a lowly Linksys WRT54G router instead and was able to establish a 10/100 connection successfully.

Thus, I'm wondering if the Trendnet switch's "GreenNet" functionality is causing me trouble. Apparently, it automatically reduces power to the ports to try to lower energy consumption. I'm wondering if it simply is trying to be too green and is thus unable to accomplish what a very basic -- non-green -- Linksys router can do.

Question: Does anyone know of an affordable (preferably sub-$50) gigabit switch that might emit a signal with more power to edge me closer to the 100 meter limit on Ethernet? I'd rather not stick to using the Linksys WRT54G as my switch in this particular location...

Thanks!

 

[tbutler@ofb.biz]

Here's an update: I tried replacing one of the Trendnet switches with a 5-port Netgear ProSafe Gigabit switch. The connection was shaky between the 5-port and the Trendnet 8-port on the other side of the 280 ft run, but unlike the Trendnet switch, the Netgear managed to alternate between grabbing a 10/100 and gigabit connection (the Trendnet, as I mentioned yesterday, established no connection at all).

Meanwhile, decided to locate a switch on the one end of the line right where the line comes in the building, which allowed me to clip about 30-35 ft off that run. With that, I was able to maintain a reliable gigabit connection between the Trendnet and the Netgear. Hopefully with two Netgears and the shorter run, I'll have enough wiggle room that I can hook up my APC ethernet surge suppressors without killing the link.

Progress! I thought someone might find this intriguing. I've typically treated unmanaged switches as essentially generic, swappable items. But, here's a very good, real world sign that they are not.

 

[thetoad30]

Unfortunately what you're seeing is a result of many people never utilizing the 100m (330ft) distance that Ethernet allows. Because of this, and the green movement becoming more popular, people are wondering why a switch needs so much power when sending most signals over, at most, 20 ft of cabling in the home. More people use around 5 ft per cable, so the savings could be enormous. I would contact Trendnet and find out what the problem is - but I am guessing they have problems figuring out to send max power over such a long connection. I'm also curious if they were intended to actually be used together on one link. As you found out, a previous gen switch worked just fine.

For my switches, I like NetGear. I did just upgrade to some Cisco switches that were managed, but they have unmanaged switches as well. Cisco is good hardware. Linksys, not so much these days. NetGear is solid, and if you get any of the regular switches from TP-Link, TrendNet, and the like, you should probably be okay.

In short, unless you're running the thing off of solar power or are paying for fuel for electricity (generator), then just buy a regular switch. The power savings is seriously miniscule (you're talking a couple of watts) for the extra headache you get, in my opinion.

 

[tbutler@ofb.biz]

In short, unless you're running the thing off of solar power or are paying for fuel for electricity (generator), then just buy a regular switch. The power savings is seriously miniscule (you're talking a couple of watts) for the extra headache you get, in my opinion.

Thanks for the information. The one problem I've been running into is that it seems like most switches -- especially on the lower end -- are labeled as green these days. The Netgear switch I was testing with is also green, although perhaps it does a better job of calculating the needed power than the Trendnet one. I'll have to try your suggestion and talk to Trendnet to find out more, although since the Netgear switch seems to be working pretty well, I think I will swap some of these switches out, since I'm within my return period for them. The Trendnet switches do seem to work OK on shorter runs, which makes sense if I'm running into a problem with an overly green switch.

Thanks again for your help!

Tim

 

[thetoad30]

Sure thing. Glad I could help. I do tend to stay with Linksys and NetGear when it comes to just switches. Stay away from the Linksys routers though. I just don't find them to be worth it anymore - too cheap, and too clunky in the firmwares and specs.

 

[tbutler@ofb.biz]

OK, here's an update: with Netgear switches, everything works beautifully, until I put the APC surge protectors back on. I have (or, rather, want to have) an APC PNET1GB on each end of the outdoor ethernet run before the ends go into the respective switches. When those are on, the signal degrades enough that I'm still losing probably 80-90% of the packets sent in a ping test and everything else is pretty well unusable.

Is there some other brand or model of surge protector that would degrade the connection less, perhaps? I'm starting to wonder if I need to investigate doing a fiber run instead...

Thanks!

Tim

 

[thetoad30]

Why are you worried about surges on a Cat6 cable?

If you get enough of a jolt through your own run cables, you have been hit. Chances are more than just your switches are destroyed, and insurance should cover it.

Now, external devices direct into yours are a different story. I have cable coming into my modem, but don't have it shielded. I figure if my modem fries, I'll get my ISP to replace it. Same with my satellite provider (but that hits my house so see the first paragraph).

Have you made sure to ground the two protectors? I'm also curious if by grounding these on both ends you create a ground loop, where there is potential between the two grounds causing your interference. Try using only one at a time and see if it works; if it does, you have a ground loop.

 

[tbutler@ofb.biz]

I haven't grounded the surge protectors yet -- I figured if they weren't working, it wasn't worth getting them grounded. Although perhaps the lack of ground could cause a problem?

As to why, I've read in various places that any ethernet cable that goes outdoors (even in a conduit) should have surge protection to protect against lightning strikes. Is this overly cautious? I do think the conduit has issues as well -- there seems to be some water in it -- so, I suspect that would make it more likely to conduct electricity from lightning strikes...

 

[thetoad30]

I'm not an electrical engineer, so I'm not entirely sure how a surge protector on a LAN cable works. I would say try grounding to see if it works? I don't know.

And consider this: If lightning hits anywhere near either building, you are going to have electricity coming through the ground, neutral, and any "underground" wires. That LAN cable is the least of your problem, honestly. Even a wall surge protector cannot defeat a true strike. They're meant to shunt spikes coming through the wires from a far-off strike to that wire. At least that's what I've always read and interpreted.

 

[tbutler@ofb.biz]

Thanks. You may have a point. I tried grounding the surge protectors, but to no avail. The network still goes up and down constantly when using the surge protector (but not at all when the lines go straight into a switch). The same model surge protectors work fine on a somewhat shorter run between the building at the long end of the 260ft run and a third building.

I wish I could figure out a good way to diagnose the problem. I could take them off, but I've heard horror stories of surges coming through ethernet wires buried underground. And while mine is in a conduit, the conduit seems to have holes in it, so I don't want to count on it isolating the wire from whatever else is in the surrounding dirt...

 

[thiggins]

My guess is that the components in the surge protector along with the capacitance and inductance of the long cable are forming a filter that is knocking down the signal to a level where it is producing a lot of packet loss.

Run 100 pings or so and see what you get for packet loss.

Given your conduit is buried, I'd skip 'em. TheToad is right, if you get a strike strong enough to put a surge in your buried line, it's gonna get to your equipment via plenty of other paths.

 

[thetoad30]

I understand your point.

Here's something else I just thought of: These "green" switches do something to detect the length of the cable. I'm not sure if they send power down it, or send different frequencies down it, or how they work. But, my bet is that those surge protectors are preventing that signal from going down the line.

I'd call up APC or research this a bit on google - I did try a quick google search and didn't find anything right away or I would have posted it.

 

[tipstir]

Green switches just lower power usage. I have two of those here from Trendnet. I felt they're slow even with the 1.25MB and 2MB for packet buffering. I did open one up to see what's in there must bare bones using very old controllers.

 

[tbutler@ofb.biz]

Thanks, TheToad, Tim and Tipster. On each side, the first switch isn't attached to anything of great worth. So I could, if nothing else, put the switch "out in the open" and then put the surge protector between it and the main network...

The green switches still bother me a bit, even though the Netgear ones seem a lot better. I'm curious what would happen if I found a good, non-green switch. Maybe I should look into that, too.

Tim, as to ping, that was how I was measuring packet loss. The last time I tried it, after getting grounding hooked up on both ends, it wasn't quite as bad -- maybe 50% loss. I could actually navigate around the web... but the amount of loss definitely took a toll on the system and made it practically unusable (albeit, somewhat rewarding feeling that it at least barely worked).

 

[westom]

The last time I tried it, after getting grounding hooked up on both ends, it wasn't quite as bad -- maybe 50% loss.

Even 90% packet loss is a completely defective circuit. If not using an oscilloscope, then just close your eyes to shoot a gun at any direction hoping to hit a target. Without actual signal levels (numbers), then everything is only wild speculation.

Same reason why dB numbers are so critically important in solving ADSL problems.

Now, the protector. A lightning strike to one building is a direct strike to everything on the network in the other building. Any wire inside any outgoing cable must connect low impedance to each building's single point earth ground. Not wall receptacle safety ground. Earth ground. 'Single point' also has serious engineering significance.

Well, connecting any of eight Ethernet wires directly to earth ground means no signals. So we connect all eight to earth via a protector. If that protector does not make a low impedance (ie 'less than 10 foot') connection to earth ground, then protection has been compromised. Then lightning does not find earth destructively via 'other building' electronics. Instead, it is earthed at the 'struck building'. No damage.

Generally, those Ethernet protectors do not list a critically important number - capacitance. Too much capacitance (especially on a wire that so much diminishes signals) means an even lower signal to noise ratio. To say more using these described concepts means an oscilloscope or some other signal strength measuring device. An answer without numbers, well, close your eyes and fire the gun wildly at some target.

Relevant to your problem are two concepts. The first is based in a concept called the telegrapher's equation. Second is the time required for a signal to reach the far end and reflect back. For example, when your router sends a signal out on that 300 foot cable, when is the output voltage first seen at its cable connection? After the signal has gone out to the far end and then is reflected back to the transmitter.

Other parameters (that should be irrelevant to you) are gauge of that CAT6 wire, any connections (that wire should be 300 feet continuous), and impedance balancing loads inside each router. Knowledge necessary to solve your problem means you also know why those basic concepts are relevant. A wire (for some reason) only 2 AWG thinner can have electronic consequences.

Best is to start by testing using a more 'controlled' sample. Take a 300 foot (or maybe longer) roll of Ethernet cable. Connect both routers to that roll. How do each perform with 300 or 400 feet not inside those many 'signal distortion' devices (ie conduit) and not challenged by other variations? If symptoms are different, then we can discuss factors that make the interbuilding connection problematic. The number of suspects is, well, you don't have enough fingers and toes. Only useful answers must come with numbers.

 

[matthelm]

I haven't grounded the surge protectors yet -- I figured if they weren't working, it wasn't worth getting them grounded. Although perhaps the lack of ground could cause a problem?

The surge protectors act like capacitors (as does the long wires), which at higher frequency acts like a short. Add in a weak signal (because of the length), and the inductance of the longer wires (length again), and you are just asking for issues.

As to why, I've read in various places that any ethernet cable that goes outdoors (even in a conduit) should have surge protection to protect against lightning strikes. Is this overly cautious? I do think the conduit has issues as well -- there seems to be some water in it -- so, I suspect that would make it more likely to conduct electricity from lightning strikes...

Yes, you need these, or at least on one end, but without the ground connection, they will not protect the circuit.


Per your post about fiber, I think if you can do fiber with a max adder of $200, I go that way for sure. MUCH safer. Now if it was in a single building ...