Fixing 10base-T on more ethernet cards

[Glenn Anderson]

Here is what I have done to fix an Asante MCiLC.

The card I have looks like it was designed for both TP and coax, but then a version of it was made with the coax components removed. Unfortunately that version still has the media detection and switching circuitry present, which suffers from incompatibility with ethernet switches that have Auto MDI-X.

The fix I have come up with is to disconnect the link detection input, and force it to output a constant link detected. This involves removing 3 components, C1, C2, and R10, then linking 2 points. I've attached an annotated image of my modified card.

An alternative would be to cut the AUI/TPI trace between U2 and U1, then link the AUI/TPI input to U1 to GND.



I'm also working on the Asante MCiNB, and the DaynaPort E/II-3.

 

[Glenn Anderson]

Here is a picture of an unmodified DaynaPORT E/II-3, and a schematic I've made. The link detection circuit seems similar to the Asante MC3NB, with R17 setting the timing for link detection. I've ordered some surface mount resistors, when they arrive I will try changing R17 to 220K.



There are also possibilities around the unpopulated R27 and JP3 where the card could be hard wired for TP, or maybe a resistor and removable jumper installed for manual selection of media.

 

[Paralel]

I had no idea that MDI-X was incompatible with the switching protocols for cards that supported 10BASE-T as well as 10BASE2. That might explain a few of the issues I've run over the years of trying to get old system onto a modern network.

Strangely, I'm actually looking to make a new 10BASE2 installation in my parents house using NOS equipment since they have an existing installation of unused RG-58 that runs throughout the house (I'm not too familiar with this type of cabling, before my time, so I'm not sure what they would have used it for at one point in a house) and need to have hardwired hardware in a few different rooms for various things where 10 Mbit/s wouldn't be an impediment. Never thought I'd see myself using 10BASE2 in the 21st century.

 

[Glenn Anderson]

I had no idea that MDI-X was incompatible with the switching protocols for cards that supported 10BASE-T as well as 10BASE2. That might explain a few of the issues I've run over the years of trying to get old system onto a modern network.

This is by no means the general case, there are cards like the Asante MacCon+IIET that have automatic media detection and don't have this problem.

The problem cards I've seen all use the DP83902, which doesn't have built in support for automatic media selection, and then try to add on 10base-T detection with an external circuit. Technically speaking I don't think these problem cards are actually compliant with the 10base-T specifications, as they don't output link pulses until they detect the other end of the link. This makes link detection very reliant on the other end of the link outputting link pulses, which breaks with Auto MDI-X as the ethernet switch isn't constantly outputting link pulses.

Because of this, connecting two of these cards together with a crossover cable doesn't work, as both cards end up waiting for the other end of the link to start outputting link pulses.

Cards that use the DP83902 with manual media selection work just fine.

Strangely, I'm actually looking to make a new 10BASE2 installation in my parents house using NOS equipment since they have an existing installation of unused RG-58 that runs throughout the house (I'm not too familiar with this type of cabling, before my time, so I'm not sure what they would have used it for at one point in a house) and need to have hardwired hardware in a few different rooms for various things where 10 Mbit/s wouldn't be an impediment. Never thought I'd see myself using 10BASE2 in the 21st century.

I'm trying to think what RG-58 might have been used for other than 10BASE2, maybe amateur radio?

 

[Paralel]

I'm trying to think what RG-58 might have been used for other than 10BASE2, maybe amateur radio?

That would make sense regarding the previous owner. The RG-58 traced back to near where the mast for the house used to be, before I took it down due to the structure almost being rusted through and being a risk to the house (it's too bad it was in such poor shape, I had all sorts of ideas for Wi-fi experiments)

 

[Glenn Anderson]

I replaced R17 on two DaynaPORT E/II-3 cards, changing it from 100K to 200K, and they both now work with my TP-LINK TL-SG1016DE ethernet switch. Image with the location of R17 highlighted attached. Larger resistor values like 220K, 270K, or higher, should also work.

 

[Glenn Anderson]

Here is a schematic for the relevant parts of the MCNB/MCiNB, picture also attached. Still working out what to do with this one. The input to the 26LS32 could be hard wired, like with the MCiLC, however that would completely disable AUI. The alternative would be something like cutting the AUI/TPI input to the DP83902, and connecting that to a new jumper.

 

[Glenn Anderson]

Here is a mod for the MCNB/MCiNB that hardwires it for 10base-T. It involves cutting a trace, then soldering a link between two points. Pictures attached.

This mod is cutting the automatic AUI/TPI link between U1 and UA5, then tying the AUI/TPI input on UA5 to GND. C8 and C9 can also optionally be removed to disconnect the automatic link detection circuitry from the 10base-T input.

 

[GondolinOnMyMind]

Here is a picture of an unmodified DaynaPORT E/II-3, and a schematic I've made. The link detection circuit seems similar to the Asante MC3NB, with R17 setting the timing for link detection. I've ordered some surface mount resistors, when they arrive I will try changing R17 to 220K.

View attachment 10213View attachment 10214

There are also possibilities around the unpopulated R27 and JP3 where the card could be hard wired for TP, or maybe a resistor and removable jumper installed for manual selection of media.

This is interesting. I have been struggling with this adapter a couple of evenings now. The only way to get it to get the link up was to to force the switch port to half duplex 10Mbps. And even then it takes ages on the switch end. Dayna tests, both hardware and software, claim the nic to be good.
I still can´t get L3 to work, though, but I guess that is due to my total lack of knowledge of networking these old beauties.
Anybody knows what are the capabilities of this nic? Does it do better than 10 half duplex?
Quadra 700, macOS 7.5.

 

[Glenn Anderson]

This is interesting. I have been struggling with this adapter a couple of evenings now. The only way to get it to get the link up was to to force the switch port to half duplex 10Mbps. And even then it takes ages on the switch end. Dayna tests, both hardware and software, claim the nic to be good.
I still can´t get L3 to work, though, but I guess that is due to my total lack of knowledge of networking these old beauties.
Anybody knows what are the capabilities of this nic? Does it do better than 10 half duplex?
Quadra 700, macOS 7.5.

These cards are only 10Mb half duplex. If your ethernet switch has a way to turn off Auto MDI-X, that should help.

If you are using MacTCP, it doesn't support DHCP. You would need to have BOOTP support on your network, or manually configure the IP address.

With a Quadra 700, you might want to consider getting an AAUI adapter for the built in ethernet. The Quadra 700 built in ethernet is still 10Mb half duplex, but is a newer 32bit ethernet chipset with DMA, connected to the main system bus.

 

[GondolinOnMyMind]

Thanks @Glenn Anderson

I have bid on an AAUI adapter. Fingers crossed.

I'll drop my Cisco config here, in case it helps somebody.
interface GigabitEthernet0/24 description "OLD MACINTOSH" switchport access vlan 700 switchport mode access switchport nonegotiate duplex half speed 10 no mdix auto spanning-tree portfast end

 

[Kai Robinson]

Sticking this thread!