Network cables crossing electrical wires

[velocci]

Hi all, my house has cat5 cables running throughout it, going from each room down to the basement where my cable modem and router is. I tested the speed on my cat5 cables and got about 45MB/s transfer rate. I tested the one cat5e cable I had and got about the same thing. I should get more. all my devices are set to gigabit. I read of a situation once where this person was getting low transfer rates on his network and couldn't figure out why. The problem ended up being that his network cables were to close to power cables or running parralel to them. as was my case around the area where my modem and router were located. so shuffled the cables around so none of the network cables were near power lines. But i still got the same transfer rates. But at the ceiling of my basement, there are a few places where the network cables cross with electrical wires and coaxial cables for my TV. Anyone know if this would affect speed performace? Should I put soemthing between the network cables and the electrical wires? I have some pictures, but they are located on my hardrive and I don't know how to attach them.

 

[sturm]

Are your cables shielded? If not then I would re-run shielded cables where you are getting slow speeds.
And yes, anything that carries and electrical current or signal can cause EMF interference in network cables that are not shielded.

 

[thendershot]

how are you testing the transfer? a copy from 1 pc to another? via web server? smb? you can cross power cables but should NEVER run them parallel to each other. keep the number of times you cross power cables to a minimum. i do agree a shielded cable will be better then an unshielded.

 

[velocci]

this is how i tested it. i have a PC with windows 7 and a laptop with windows XP.
from the windows 7 box, i typed \\laptop_name\c$ to access the laptop. I then used window gui to transfer a file to the laptop (ie: copy and past). is there a better way to do it?

 

[thore]

Hello,

using Category 6 cables

If there is a rule "never to mix a weak current has a strong current"

 

[rand_79]

you most likely dont have a problem with your wiring.

try running some form of network benchmark.. not a copy paste

or you can run your test using a crossover cable.. or switch without using your inwall wiring to see what the speed is then.. i'd guess about the same as now.

 

[velocci]

is there a network benchmark software?

 

[palladin9479]

Ok first, check to see that each device is showing 1.0Gbps connection link inside windows, this will let you know if the NIC's were able to successfully negotiate a proper connection. If you are getting 45MBps then that should be about right for a TCP Samba file transfer from one windows system to another.

A 1Gbps connection can get the following,
1000 * .9 (10% loss due to wait times on the wire itself) = 900 / 8 = 112.5MBps pure data rate. TCP required that ACK packets get sent whenever a data-gram is received to verify that it reached its intended host. This kills network latency as the sending must wait to receive the ACK packet before sending the next set of packets (burst mode). Ontop of this SMB is very "chatty" with lots of overhead and connection / disconnection traffic. It is a very poor choice to benchmark file transfers.

BTW 45Mbps is about what I get doing a SMB file transfer from my server PC to my main working PC over a 1Gbps network, their both on the same switch inside the same room over Cat-6 cabling.

 

[someone19]

Eventually your network cables will have to pass closely to power cables, the best way to cross the power lines is at a 90 degree angle. Cross talk and interference is caused when the cables are run parallel, but not when run perpendicular. We used that technique to eliminate white noise from audio cables run through a theater.

 

[velocci]

Ok last night I ran a couple of tests on my network. I used something called Performance Test by Passmark Software. The first test was with my PC and laptop plugged into the router using cat5e and got 353Mbps. This morning I had the laptop plugged into the PC using a cat5e crossover cable and got 357Mbps. That's very disappointing. but it seems to me that the PC and/or laptop is slowing everything down. you think?

is this a good test for my network?

 

[someone19]

Keep in mind that you will never get 1000Mbit - ever. That's just the theoretical limit to transfer speed. I'd say that using a laptop would slow your transfer speed slightly, as the chip is designed to balance power use with performance.

 

[velocci]

I know i won't get to the limit. but i'm trying to go as high as I can. 353Mbps is not very high at all in my opinion. so what is a good test to use?

 

[ntrceptr]

I test my network with IxChariot and get near the theoretical limit of my GB network. Copy and paste of a file using windows shares is bad.
1 - The Protocal seems slow (at least to me)
2 - Your adding in the hard drive speeds which could be limiting your throughput on the test.

 

[velocci]

I went to IxChariot's site and you have to register to download the evaluation copy of this. do you know of a place where I can download it without registering?

I test my network with IxChariot and get near the theoretical limit of my GB network. Copy and paste of a file using windows shares is bad.
1 - The Protocal seems slow (at least to me)
2 - Your adding in the hard drive speeds which could be limiting your throughput on the test.

 

[palladin9479]

Another thing to remember, unless your using some pretty upper end NIC's then no PC will ever approach the 1.0Gbps limit. That limit is the limit on the wire, meaning any and all traffic going over that wire from multiple nodes would be included.

 

[velocci]

I have an update. this morning i ran the same test as yesterday (PC and laptop plugged into router using cat5e) and I got 426MB/s. yesterday it was 353MB/s. how did it get higher since yesterday? I also tried it with my PC plugged into the port using cat5 cable and get 419MB/s.

 

[palladin9479]

Ok either your mixing up your bits and bytes, or your system is glitching. 426 / 419 MB/s is not possible on a 1Gbps connection.

426 * 8 = 3408, or a 3.4 Gbps throughput.

 

[velocci]

sorry i meant Mbps.

I made a cat5 and cat6 cross over cable at work yesterday. I noticed that in the cat6 cable, there was a plastic piece in there that ran the entire length of the cable and its shaped like a + sign. it kept each of the twisted pairs seperate from each other. is this how all cat6 cables are?

so I used a cat5, cat5e and cat6 crossover cable and ran a bandwidth test between my PC and laptop and got roughly the same on average. i wish i had faster machines

 

[someone19]

What color code pattern did you use on the patch cables? If your not using the proper color sequence then you'll have crosstalk inside the cable (let alone from outside sources) and that will cause a very slow transfer speed.

http://www.ertyu.org/steven_nikkel/ethernetcables.html <- my favorite bookmark for when its been awhile since I've made a cable.

Sorry if its insulting, but nobody ever told me and I learned the hard way with lots of wasted cat5.

 

[velocci]

are you saying that there are multiple color codes to use to make a crossover cable?

 

[someone19]

Ethernet cables with RJ-45 plugs ( <- basically any network cables) have 8 wires, but only 4 are used. Pins 1,2,3,6 are used for data transmission, pins 4,5,7,8 are not used. Pins 1 and 2 need to be on one pair of wires, pins 3 and 6 need to be on the other pins used.

IF you wire a straight cable with the orange pair on pins 1,2 the green pair on 3,4 the blue pair on 5,6 and the brown pair on 7,8 then the data will travel on the orange pair, one wire in the green pair, and one wire on the blue pair. That's where the problem lies, by twisting the two wires of a pair they are able to have consistent interference, and the network cards are able to compensate for that. If the data is traveling on one wire of one pair, and another wire in one of the other pairs, the interference is not equal or correctable, resulting in a weak/slow connection.

The color doesn't trully matter per-se, but traditionally orange and blue are used for the data, and green and brown are not. The green / brown wires sometimes have less twists per inch as orange and blue and are not best for reducing interference.

A crossover cable simply takes pins 1,2 on one end of the cable and puts those colors to pins 3,6 on the other end, and vice-versa. Again, the specific colors don't matter as much as the same color needs to be on 1,2 and a different color needs to be on 3,6.

 

[velocci]

I used the color pattern found half way down on this page:
http://www.littlewhitedog.com/content-8.html

is that the right way?

 

[someone19]

Yup, that'll work just fine. If your making a straight cable, make both ends like the plug shown on the left of that diagram. For the cross over, make one end one way, and the other end the other way.

Also be sure to get the wires all the way up into the end of the plug. The hardest are on the end of the plug and they're the ones carrying your data. I like to pull the wires out from the outer cover slightly so I can get a really good push up into the plug, then push the cover into the plug before you crimp it.

Hope this helps. Drove me nuts trying to figure out slow connections until I found the correct color sequences.

 

[velocci]

but there is a difference in the color patterns that these sites used. what is the difference?
http://www.littlewhitedog.com/content-8.html

http://www.ertyu.org/steven_nikkel/ethernetcables.html

 

[someone19]

The crossover cables are the same on both sites - Err I just looked at them closer. The NOT USED wires are different, but as they are not used both color codes on the two sites are equally functional.