Re: [802.3_4PPOE] Crossovers and straight-throughs (was diode bridges)

["Darshan, Yair" <YDarshan@xxxxxxxxxxxxx>]

Thanks, it helps a lot. It strengthen the working assumption that if something is working and nobody complains, don’t change it J.

Yair

 

From: Hugh Barrass (hbarrass) [mailto:hbarrass@xxxxxxxxx]
Sent: Wednesday, February 12, 2014 7:45 PM
To: STDS-802-3-4PPOE@xxxxxxxxxxxxxxxxx
Subject: [802.3_4PPOE] Crossovers and straight-throughs (was diode bridges)

 

All,

 

If anyone is interested in the reasoning (and history) behind this problem, please read on…

 

Telephony – a single pair is a communications channel, multiple communications channels may be carried in a cable. Therefore, the channel that is on (say) pair-A must always be connected to pair-A regardless of the connector type. What leaves one box on pair-A, arrives at the other box on pair-A – if they have the same connector then it will be the same pins. All cables are straight-through.

 

Multi-lane communications (think back to RS-232 if you like) – a cable is a communications channel with pairs carrying specified (often directional) functions. What leaves one box as “transmit” must arrive in another box as “receive.” If both boxes have the same connector, with the same pinout definition, then all cables with a single gender must be crossover (i.e. male-male or female-female must be crossover, male-female must be straight-through).

 

When Ethernet defined use of a twisted pair medium, it leverage cabling that was already defined for telephony even though Ethernet defined a multi-lane communication. This posed problems. For that reason, Ethernet defined two variants for the pinout – MDI was meant to be used for end-stations (including routers) and MDI-X was meant to be used by hubs (repeaters & switches). In the odd situations where someone wished to connect two switches together, or wished to connect an end-station directly to a router, then a crossover cable would be required. Eventually, the concept of auto MDI/X was added to the standard as an option but it has since become a basic requirement for any PHY design. There are still some systems in the field that do not support auto-MDI/X but, thankfully, no one ever tries to connect two of them together!

 

Because of this history, it is not easy to predict where straight-through or crossover connection may be. Almost all equipment works with either and so people do not bother to pay attention.

 

 

I hope this helps to put the discussion in context.

 

Hugh.

 

From: Dan Dove [mailto:dan.dove@xxxxxxxxxxxxxxxxxx]
Sent: Tuesday, February 11, 2014 4:35 PM
To: STDS-802-3-4PPOE@xxxxxxxxxxxxxxxxx
Subject: [802.3_4PPOE] Diode Bridges

 

Hi,

I have received a couple of requests from participants of this task force about Auto-MDIX and wiring questions in general related to polarity and pair swaps.

The PHY companies would be the ones to talk to, but in my experience many have bits to ID which pair is connected to which MDI PHY pair connection, and they also can detect/correct polarity flips and ID them as well. Of course, this means that some wiring defects/configurations are tolerated, and thus remain operational without user repair/replacements.

As I understand it, the question is whether or not pair/polarity swaps occur and at what rate, and if so, whether or not diode-bridges are required. My thought is that some relatively simple diagnostics could be done in the DC domain to ID the state of the link prior to applying power. I wouldn't want things to *POP* regardless of the statistical aspects of network wiring configurations.

Also, many MDIs either short their 4,5 and 7-8 pairs, or they terminate them to ground with 75 ohm resistors, or they terminate them to a virtual ground with 75 ohm resistors to a 50V cap to ground (150 ohms between wires). For Gigabit, most terminate to a virtual ground on all pairs with 75 ohms, a DC block between the powered pairs, and a 2Kv cap to ground. They may not provide DC blocking between the pairs that power is not anticipated on.

So this means you really have to think about a diagnostic process prior to power application regardless of the traditional wiring config. Most likely a high-impedance application of voltage to each wire/pair and measurement of the response on all other pairs would be required.

I hope this helps the group.

Regards,

--
Dan Dove
Chief Consultant
Dove Networking Solutions
530-906-DOVE - Mobile