Ethernet CAT7 disaster

Posted by: Gajdzin on 11 April 2014

Today was a big day of hacking the walls and putting in new cables. I got a dedicated power supply spur for the Naim system and a new ethernet cable - STP CAT7 in place of the 10 years old CAT5. Here's the problem: the new cable, although it checks correct on cable tester, dropped the network speed across the house from 1000mbps to 100mbps. In fact I had 2 cables run and they both have the same effect.

 

The installer has no idea why it's happening, I searched some internet fora, found some similar cases when CAT5 was providing gigabit ethernet and when replaced with CAT7 the speed dropped to 100mbps, but no explanation why that could be happening... The cable was run through the house by an electrician, not a network cabling company, so his knowledge is limited, too.

 

Any ideas...?

Posted on: 13 April 2014 by Simon-in-Suffolk

Well I am a solutions  architect, so those traits are part of my professional world, it's interesting you picked up on it and thanks      My professional  engineering 'specialisms' are VoIP systems, and audio and data convergence. However I do have routing and switching  experts in my teams, some of whom hold CCIE certifications and their engineering ability with data comms  is pretty impressive..

Anyway let us know how you get on with the analyzer...

Simon

 

Posted on: 14 April 2014 by engjoo

It is strange the CAT7 do not require anything more than "snap on connectors". From what I know, even CAT6 requires proper termination tool and connectors.

Posted on: 14 April 2014 by Aleg
Originally Posted by engjoo:

It is strange the CAT7 do not require anything more than "snap on connectors". From what I know, even CAT6 requires proper termination tool and connectors.

These "snap on connectors" are sophisticated shielded field-mountable professional grade connectors.

A bit of creme-de-la-creme under the RJ45-connectors.

 

cheers

 

Aleg

Posted on: 14 April 2014 by Gajdzin
Originally Posted by engjoo:
It is strange the CAT7 do not require anything more than "snap on connectors". From what I know, even CAT6 requires proper termination tool and connectors.
There are several models of RJ45 plugs that do not require a crimping tool. One example of such plug can be seen on the ends of AudioQuest Vodka and Diamond cables. They look like this:
That's what I bought, 4 of them for 2 cables. Quite expensive - ca. 10 Euro per plug. And, judging from comments on several fora, I may not be the first one who has problems with those...
Posted on: 14 April 2014 by Gajdzin
Originally Posted by KRM:
Audioquest upgraded the Vodka cables after I bought them. They very kindly replaced my old ones free of charge. Again, I believe the change in design was related to earthing (according to the dealer).

Keith, do you know what they did to AQ Vodka shield? If I remember correctly, when I bought my Vodka cable I checked it with a multimeter and the shield was connected at both ends...

Although Simon is encouraging me to let the shield float at both ends, I am now leaning towards leaving them connected because: to disconnect the shield is quick and easy: a female-to-female coupler + 50cm patch cable with unshielded RJ45s. But if they are disconnected at the main cable level, it's not easy to connect them back without re-crimping the plugs...

So if I decide on shields connected at both ends, using a couple of couplers and patch cables I can actually compare both options.
Posted on: 14 April 2014 by Simon-in-Suffolk

Gajdzin, if you are looking at definitely using the shields - and in some high noise commercial/industrial environments or where 10Gbps is required it is required,  but as you realise careful preparation of the termination is required and earth / star earth bonding is usually required.

 

http://www.panduit.com/heiler/...20041307%20FINAL.pdf

 

You can see why I stick with UTP or float the shield or if I have to use STP professionally I  use professional structured wiring specialists.

Simon

 

Posted on: 14 April 2014 by Gajdzin
Originally Posted by Simon-in-Suffolk:

Gajdzin, if you are looking at definitely using the shields - and in some high noise commercial/industrial environments or where 10Gbps is required it is required,  but as you realise careful preparation of the termination is required and earth / star earth bonding is usually required.

 

http://www.panduit.com/heiler/...20041307%20FINAL.pdf

 

You can see why I stick with UTP or float the shield or if I have to use STP professionally I  use professional structured wiring specialists.

Simon

Yes, I understand - all I'm saying is that if connect the shield, I can easily disconnect it with a couple of couplers and CAT5 patch cords, and then I have a choice to compare But if I ask the installers to disconnect them, I have only 1 option.

Posted on: 14 April 2014 by KRM

Hi Gajdzin,

 

I don't know what they did, I'm afraid. I do know it took them months, though. They definitely changed the design and then had production issues. The new ones are fatter than the old ones.

 

The dealer was aware that there is a potential issue with earthing and you have to be careful with plugging shielded cables into metal sockets as they can act as ariels. I have metal at both ends (naim components and Netgear Gigabit switch) so I'm ok. I was told all this before they changed the design so it may not affect the new cables.

 

They have a customer service helpline and email service so they can probably advise on earthing issues.

 

That's the extent of my knowledge.

 

Keith

 

 

 

 

Posted on: 16 April 2014 by Fred Mulder
Originally Posted by Gajdzin:
Originally Posted by engjoo:
It is strange the CAT7 do not require anything more than "snap on connectors". From what I know, even CAT6 requires proper termination tool and connectors.
There are several models of RJ45 plugs that do not require a crimping tool. One example of such plug can be seen on the ends of AudioQuest Vodka and Diamond cables. They look like this:
That's what I bought, 4 of them for 2 cables. Quite expensive - ca. 10 Euro per plug. And, judging from comments on several fora, I may not be the first one who has problems with those...

 

Hi, is supose you are using the telegartner MFP8 T568? Is your problem solved?

 

I'm using those plugs without any issues what so ever (connected to Ecolan Cat.7a  S/FTP kabel 1000MHz, 4x2xAWG23/1 PIMF, FRNC-C, 100 Ohm). I really hate pulling cables thru the house/pipes.. some a 20+ metre with corners... Aaargh! Hopefully they can do their work for a long time.

 

Cheers, Fred

Posted on: 16 April 2014 by Gajdzin
Originally Posted by Fred Mulder:
Hi, is suppose you are using the telegartner MFP8 T568? Is your problem solved?

Yes, I think it's Telegartner plug. The problem is partially solved - I cut off both plugs on the main line and crimped ordinary RJ45s myself. A grounded one on the side of the switch and a non-grounded one on the side of the router. The cable performs fine now.
The other cable in the wall I leave with the plugs because I'm waiting for someone to lend me a Fluke CableIQ cable qualification tester. A simple cable meter shows correct pin-out, but the able refuses to work over 100mbps. I want to understand the root cause of the problem, that's why I'm not touching it and waiting for the Fluke
Posted on: 16 April 2014 by Adrian F.

AFAIK: As long as you mix unshielded and shielded in the same network, you will always have problems with shielding and grounding. That will never work properly...

 

Decide for one system or the other for your whole network topology.

Unshielded will be cheaper and easier to install. That may (but doesn't have to ) work with do-it-yourself.

 

When you have to work with shielded cables, get a pro to install and measure the whole network (incl. all gear like switches and patch-boards, which all have to support shielding and must be correctly installed = grounded to work properly).

Posted on: 16 April 2014 by Gajdzin
Originally Posted by Adrian F.:

AFAIK: As long as you mix unshielded and shielded in the same network, you will always have problems with shielding and grounding. That will never work properly...

 

Decide for one system or the other for your whole network topology.

Unshielded will be cheaper and easier to install. That may (but doesn't have to ) work with do-it-yourself.

 

When you have to work with shielded cables, get a pro to install and measure the whole network (incl. all gear like switches and patch-boards, which all have to support shielding and must be correctly installed = grounded to work properly).

OK, that's another bit of wisdom I didn't possess when embarking on this project... Oh, well, I'm still searching high and low for someone with a cable quality tester, the commercial installers have them but are not interested in coming to a residential place even if I pay them, the home installers don't have them. But when I finally find one, we'll know exactly what hurts this installation.

Posted on: 16 April 2014 by Simon-in-Suffolk

Adrian

 

>AFAIK: As long as you mix unshielded and shielded in the same network, you will always have problems with shielding and grounding. That will never work properly...

 

It will make no difference. You typically will only shield specific links that need it - either through specific link speed or protection against EMF. The consideration with shielding is usually about minimising circulating earth currents in the shield and that means star grounding and/or bonding the earths together (or letting them float either at both or one end) . However that is for a physical electrical link between two points, such as between two switches or a switch and an edge host. It has nothing to do with the network per se.

 

Simon

 

Posted on: 18 April 2014 by Simon-in-Suffolk

Gajdzin, any update?

Posted on: 18 April 2014 by Gajdzin
Originally Posted by Simon-in-Suffolk:

Gajdzin, any update?

My company's cabling contractor is coming with a Fluke CableIQ next Wednesday. Then we'll know

 

In the meantime - Happy Easter!

Posted on: 18 April 2014 by Simon-in-Suffolk

Indeed Happy Easter

Posted on: 23 April 2014 by Gajdzin

OK, the gentlemen with Fluke CableIQ meter have just left and here's what I learned:

 

1. The cable make and quality matters. A cheap CAT7 cable ordered from the internet (mine is a "Maxcable") very often doesn't hold the specs. I'd be better off with a quality brand CAT6 shielded cable.

 

2. When using solid core cables it's essential to use special plugs for solid core cables.

 

3. I should not have terminated the cable in plugs, but in sockets, with a short, flexible patch cord from the socket to the router. When terminated with sockets, it's enough to twist the cable coming from the router a bit and the connection is compromised. Also, when using short patch cords at the ends I'll have full flexibility to ground the shield at both ends, at one end or not at all, depending on which patch cables (shielded or not) I use.

 

Now I'll have to wait another week for them to come down and replace the plugs with proper sockets for this type of cable and measure again. They said it cures 90% of such problems. But if I still get very bad measurements, like now, I'll have to replace the entire 25m of cable in the walls.

 

Lesson learned: when going above CAT5, leave the entire installation, crimping, etc. to specialists... And let them use the brand and model of cable they know and trust.

 

If only learning on one's mistakes wasn't so bloody expensive...

Posted on: 23 April 2014 by GraemeH

Sorry to hear of your plight. I hate to say it but I have 20m Cat6 running under my floor. A few quid on Amazon and works perfectly.

 

G

Posted on: 23 April 2014 by Gajdzin
Originally Posted by GraemeH:
Sorry to hear of your plight. I hate to say it but I have 20m Cat6 running under my floor. A few quid on Amazon and works perfectly.

Just because you get the signal on the other end doesn't mean it works perfectly That's what I learned today from the gentlemen with the Fluke CableIQ meter. I do get my gigabit signal now (after changing the RJ45 plugs last week), but the cable has a million signal quality problems that can only be seen on such cable quality meter.
Posted on: 23 April 2014 by Simon-in-Suffolk

Gajdzin, the key thing is the signal quality as far the switch is concerned.

This is my NDX port on my Cisco 2960 switch. This feeds into a long run of flat Cat5e and then connected to a short 4m run of shielded Cat 6a. This is over a few months and not one error! The keen eyed amongst you will see I grabbed this from the switch whilst streaming 16bit/44.1kHz WAV.

 

5 minute input rate 40000 bits/sec, 80 packets/sec
5 minute output rate 1475000 bits/sec, 125 packets/sec
34873986 packets input, 2694230351 bytes, 0 no buffer
Received 537093 broadcasts (533557 multicasts)
0 runts, 0 giants, 0 throttles
0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
0 watchdog, 533557 multicast, 0 pause input
0 input packets with dribble condition detected
63556820 packets output, 79155362395 bytes, 0 underruns
0 output errors, 0 collisions, 3 interface resets
0 babbles, 0 late collision, 0 deferred
0 lost carrier, 0 no carrier, 0 PAUSE output
0 output buffer failures, 0 output buffers swapped out

 

Simon

Posted on: 23 April 2014 by Gajdzin
Originally Posted by Simon-in-Suffolk:
Gajdzin, the key thing is the signal quality as far the switch is concerned.
This is my NDX port on my Cisco 2960 switch. This feeds into a long run of flat Cat5e and then connected to a short 4m run of shielded Cat 6a. This is over a few months and not one error! The keen eyed amongst you will see I grabbed this from the switch whilst streaming 16bit/44.1kHz WAV.

Unfortunately my switches are of the dumb variety and don't generate any logs...

Posted on: 23 April 2014 by DaveBk

I've been running 10m of fibre into 15m of CAT6 solid core punched into sockets with 1m patch leads at both ends for about a month - also zero errors on my Cisco SG-300 switch. When ethernet works, it just works...

 

I hadn't realised you had put RJ45s straight onto the end of solid core CAT7 - I guess the crimp and twist geometry is sufficiently off spec to cause all sorts of crosstalk and interference. Hope the cable guy can sort it all out for you.

 

Dave.

Posted on: 23 April 2014 by Gajdzin
Originally Posted by DaveBk:

I've been running 10m of fibre into 15m of CAT6 solid core punched into sockets with 1m patch leads at both ends for about a month - also zero errors on my Cisco SG-300 switch. When ethernet works, it just works...

 

I hadn't realised you had put RJ45s straight onto the end of solid core CAT7 - I guess the crimp and twist geometry is sufficiently off spec to cause all sorts of crosstalk and interference. Hope the cable guy can sort it all out for you.

 

Dave.

Well, my internet works, I get 1000mbps, but the CableIQ showed crosstalk, standing waves, noise, you name it - it's bad. The cable guys also didn't like it that I wired my cable as cross. After the initial problems when I could only get 100mbps on this cable I read up on the internet and came to the conclusion (incorrect, as it turns out) that the run between a router and a switch should be a cross cable even if the switch is auto sensing and can work with both types of cable. The guys yesterday said CAT7 shielding is designed for straight through connections, and it was probably my re-crimping of RJ45s and not moving from straight to cross that improved the performance until the switch managed to negotiate a 1000mbps connection with the router.

 

Nothing to do but wait for next Monday...

Posted on: 24 April 2014 by Simon-in-Suffolk

Its like being back in school

 

Ethernet Pass Through Cable:

Switch to Host

Switch to Router

Switch to Switch

Switch to WAP

 

Crossover  Cable:

Router to Router

Router to Host (Excluding switch ports)

Host to Host

 

Most devices - and nearly all consumer devices - auto detect send/receive pairs using Auto MDI-X .

Therefore the need for the correct cable type is largely removed.

 

The grounds are bonded and are not related to the send / receive pairs.

The negotiation of 1000Mbps (by identifying the 4 pairs as opposed to 2 pairs) has little bearing on the cable quality - unless really awful!

 

To get a real world view of data integrity use a managed switch and look at the counters (which is why they are there). CRC errors usually indicate a flipped bit. This can get worse if the cable gets pinched or compressed..

Posted on: 29 April 2014 by Gajdzin
Originally Posted by Simon-in-Suffolk:

The negotiation of 1000Mbps (by identifying the 4 pairs as opposed to 2 pairs) has little bearing on the cable quality - unless really awful!

Well, mine was really awful then. Dear Naimers: do NOT buy CHEAP twisted pair cable from the internet! I won't write here which make it was, but if anyone's interested, shoot me an email.

 

Yesterday was 2nd visit by the ethernet guys with their trusty FlukeIQ meter and 50m of very handsome, purple-clad CAT6 STP cable. First they replaced the plugs on my original CAT7 cable with some shielded, female sockets they know and trust. No improvement. Then they inspected every bend in the cable to see if it's not too sharp - didn't help. Finally, there was nothing else to do but replace the cable with the purple monster.

 

Three hours later it was in place and - zero errors. Immediate 1000mbps negotiation between devices. With one exception - when we used shielded patch cable on both ends of the main cable. My switch could only do 100mbps then. This is where Simon-in-Suffolk says: "I TOLD you so!!!" So I settled on shielded cable on the switch end (different room than Naim gear) and unshielded on router end (near Naim gear) - the cable is now grounded on one end.

 

I swear next time if I have to rewire my house I'm going optical

 

Many thanks to all in this thread who helped me throughout this misadventure.