Potential Solution For Auto-Negotiating Ethernet Issues

[Paralel]

I found a potential simple, cheap solution to the auto-negotiating ethernet issue that creeps up from time to time with the old 10M ethernet adapters/transceivers. All you need to do is use an ethernet cable with RJ-45 plugs wired for the old Cat. 3 standard.

I've been having trouble getting my PB 520 to talk to any of my modern networking hardware. I bought a NOS Farallon transeiver and it came with the original Cat 3 cable that was only wired for two pairs to the T568B standard. I tried the old Cat 3 cable just for the hell of it and it is now talking to everything correctly. I guess only being wired for two pairs the hardware I was connecting automatically knew that it was supposed to default to 10M.

It's pretty easy to turn a Cat 5 cable into a cat 3 cable just by snipping a few wires, so I figured I'd share this and see if anyone else has had a similar experience.

 

[nvdeynde]

I use old 3Com Officeconnect Hubs, pure 10 Mbps ( model Hub 8C ), they even have a coax connection too and can be bought very cheap.

Then I uplink them to a gigabit switch, however all my vintage Apple computers are connected to the 10 Mbps hubs. I never had any problems.

 

[bse5150]

I use an old 10 Mbit NetGear hub to connect my two 68k Macs. Well, I used to. Now I have a 100 Mbit NIC in my Quadra 700. So it's just my Quadra 605 that does the NetGear shuffle now.

 

[NJRoadfan]

Something is up with those Ethernet cards if removing the unused pairs "fixes" the problem. Both 10BaseT and 100BaseTX use the same pairs of wires in CAT3 and CAT5 cable. the only difference in the cable catagory is the quality rating and shielding. Both have 8 wires (4 pairs). I think the unused pins of the RJ-45 connector on Ethernet devices are left floating and not connected to anything beyond the jack. I wouldn't be surprised if these picky cards do something silly like grounding those pins, which is really bad practice considering those lines could be carrying Power Over Ethernet.

That being said, I have never run into this problem with 10BaseT NICs not interfacing with auto-negotiating hubs. That includes cards based on the SONIC series chip used in 68k Macs, the Fastpath 4, and Nubus cards (its used in just about every 3rd party Macintosh NIC since its register compatible with Apple's built-in NIC drivers). All my patch cables have all 8 wires crimped in the RJ-45 jacks.

 

[Paralel]

I wonder if Farallon used a strange wiring standard for their Cat 3. 10M cables. The cable uses the blue and orange pairs (pairs 1 & 2 for the T568B wiring standard), compared to the typical green and orange pairs (pairs 2 & 3 for the T568B wiring standard) for Fast Ethernet.

 

[NJRoadfan]

Wire color doesn't matter, its all copper. Just as long as they are wired to the right pins on both ends it should work fine. I did run across a network installation by a jerk that didn't follow the colors in the TIA/EIA wiring standards (10BaseT uses the orange and green pairs too). I wasn't too happy with that person as it likely was intentional on his part so someone else couldn't fix his handiwork.

 

[James1095]

Wire color doesn't matter, its all copper. Just as long as they are wired to the right pins on both ends it should work fine. I did run across a network installation by a jerk that didn't follow the colors in the TIA/EIA wiring standards (10BaseT uses the orange and green pairs too). I wasn't too happy with that person as it likely was intentional on his part so someone else couldn't fix his handiwork.

It does matter, well not so much the color itself, but each differential signal has to use a twisted pair of wires. You can't use one wire from one of the four pairs and one wire from one of the other pairs for the same signal. Differential signals are transmitted as the difference in potential between two wires rather than the absolute level relative to ground. When sent over a pair of wires that are twisted together, both wires in the pair are exposed exactly the same to any interference and the difference between the two remains constant. If you cross the pairs, you lose the shielding effect.

Years ago when 100Mb ethernet was just becoming common, I spent hours tracking down why one of my PCs would connect fine set to 10Mb but not 100Mb. Finally figured out that when my friend made the cable, he just wired it in an arbitrary manner with the same color order on each end. It tested fine with an ohm meter and it was good enough for 10Mb but it didn't work at all for 100Mb.

 

[insaneboy]

I'll have to give that a try...

my Asante mini EN/SC(well not mine any more) as well as the Dayna Etherprint-T both will not work with 10/100 *switches* on a 10/100 hub they both worked just fine. Ethernet card in the SE/30 is fine on a 10/100/1000 switch.

 

[Paralel]

Worst comes to worst, I'll send you my other Cat-3 10M cable that came with an Asante AAUI transceiver and you can give that a shot.

 

[insaneboy]

Thanks for the offer, but I have the parts here to make one. Once I have time to play around. Would be nice to ditch all the 10base-T hubs I have kicking around (three, two can do 10base-2 as well).