Ethernet and DIN 41612

[homebrew]

I want to use these standard DIN 41612 backplane connectors (2x32 pins in A-C configuration) together with a custom backplane in a standard subrack enclosure.
All signals are absolutely uncritical, except for for a single Gbit ethernet link (one on each module). The corresponding switch would be implemented on the backplane.
I certainly know that these connectors might not be specified for these speeds. On the other hand Gbit ethernet is fairly robust...

Would you give it a try or do you have any other suggestion?

Best,
Homebrew

[TheUnnamedNewbie]

If the links are for internal communications within a system it should be fine. Guaranteeing performance over all circumstances is another matter tho.

But I have seen people do absolutely horrible things to ethernet over twister pair (like paying no attention to actually using the pairs at all) and somehow it always seems to work, so I think you'll be fine, esp. over short links.

[Benta]

Probably an overkill, but you can get 41612/60603-2 connectors with coax inserts:

http://www.te.com/usa-en/products/connectors/pcb-connectors/backplane-connectors/eurocard-connectors.html?tab=pgp-story

[tecman]

Running the ethernet through a pairs of adjacent pins should work fine.  Typical connectors, like RG45s, are not constant impedance connectors and rarely cause issues.  If you keep your wiring clean and short to the connector, and on the pcb go to a differential set of runs recommended for ethernet by the PHY manufacturer you should not experience issues.

Paul

[homebrew]

Running the ethernet through a pairs of adjacent pins should work fine.  Typical connectors, like RG45s, are not constant impedance connectors and rarely cause issues.  If you keep your wiring clean and short to the connector, and on the pcb go to a differential set of runs recommended for ethernet by the PHY manufacturer you should not experience issues.

Paul

That is very good news! Thank you all for the replies.

Of course digging deeper into the topic, a new question surfaced:

On the boards I'll probably have an USB->Ethernet bridge. Normally a transformer is used to couple the signals. Hence I would need a transformer on each board and one for each slot on the backplane implementing the switch. The question is whether these can be skipped in favor of capacitive coupling. I found a promising application note:

http://www.ti.com/lit/an/snla088a/snla088a.pdf

They even describe the backplane scenario using only capacitors between phys. However, I have my doubts that this can be generalized. The datasheets of the specific components I looked into do not cover these applications. Does anybody have experience with that?

USB-Ethernet-Bridge
http://www.newark.com/microchip/lan7500i-abzj/usb-2-0-to-ethernet-controller/dp/69W9448

Backplane Switch
https://www.digikey.com/product-detail/en/microchip-technology/KSZ9897RTXI/KSZ9897RTXI-ND/6623477

Any Ideas?

[dmills]

You can get switches that have RGMII interfaces for external PHYs on some or all ports, and it is entirely possible to run this to your cards rather then ethernet (This part has ports 7 and 8 set up this way, so if you can live with just two slots having ethernet then you can run RGMII or such (4 bits each way, 125MHz DDR, which should be fine).

That way you only need the magnetics on the external ports, not on the ports connected to the on card MACs.

There are ways to use capacitive coupling instead of the transformers, but setting up the DC bias can be a pain depending on the details of the pin drivers.

73 Dan.

[tszaboo]

I've sucessfully run 100Mbit ethernet through that DIN connector, 4  layer cards. Magnetics on the backplane, phy on the card. Gigabit should not be too much issue, the bandwidth of the signals is not 10 times as high. Just use common routing rules, and common sense.