Ethernet over non RJ45 connectors?

[sddsfgadasd]

Hi

I'm working on a design where it needs to use only circular DIN connectors (for instance https://www.amazon.co.uk/PIN-Solder-Connector-Audio-Video/dp/B008MI9G34) to connect to a board which has a generic ethernet PHY IC (not yet selected). From a signal integrity perspective, this sounds to be a challenge because there may be an impedance mismatch at the connector end because it is not a standard ethernet connector. I'm thinking it might be ok because the ethernet throughput should be under 10Mbps (only using 10 or maybe 100BASE-T) and I could use twisted pair cable for the other end of the connector cable.

Any issues from an EMI perspective that can be addressed? Additionally, should I still use a discrete ethernet isolation transformer (magnetics) on the board?

[jonpaul]

Bonjour

Ethernet has a RS-485 PHY, and 110 Ohm balanced on UTP, TP.

The TX, RX, transceivers are balanced and transformer coupled. 
Thus some CM noise immunity and tolerant of overshoot, etc.

Broadcast and pro audio industry uses a similar protocol, AES/EBU at baud rates 6.144....24.576 Mhz
We use all types of CAT5,6,7 and many connectors besides RJ45, eg XLR, DB25, DIN, with no issues.
The small Zo discontinuity in connectors   different  from 110 ohm is not signifiant if the cable is 110 ohm balanced TP, UTP.
See AES/EBU std, and Ethernet PHY, also Belden Cable for digital audio cables and connectors

RE EMI,
"Any issues from an EMI perspective that can be addressed? Additionally, should I still use a discrete ethernet isolation transformer (magnetics) on the board?"
A transformer at both ends is required for CMRR and isolation between devices.
SMD and TH parts are available from Pulse Engineering, Halo, Coilcraft.
https://www.haloelectronics.com/products/lan/ethernet/discrete-transformers/10-100base-tx/
https://www.mouser.com/datasheet/2/336/-349690.pdf

Bon journée

Jon

[jkostb]

No problem. I have done this for an industrial project and it passed without problems all EMI tests. You still need to use an ethernet transformer and can use another type of connector. Wiring should be shielded and if possible twisted pair. For speeds till 100Mbit/s this will work.  The max distance will drop if wiring is not twisted pair

[Siwastaja]

Wiring should be shielded and if possible twisted pair.

I think it's the other way around: not "if possible", but must be twisted pair; and when it is twisted pair, shielding is optional and usually not necessary, because twisted pair itself rejects interference by its mechanical construction by cancellation, as you probably are aware of.

In other words, bog-standard Ethernet cable: CAT5, CAT5e or CAT6. Wrong type of connector does not change rest of the advice, normal Ethernet stuff applies.

I'm pretty sure if you tried speaker cable of any non-trivial distance (> 1 meter or so), both communications and EMC would fail.

But the connector represents such short length that the connector causing impedance discontinuity does not matter much. Just keep the transformers very close to the connector, and PHY close to the transformers, and it's fine. Of course I'm assuming the connector is not insanely large. With 100M ethernet and some 1-2cm of wrong type of connector, you'll likely pass EMC; at very least it will work "in practice".

[T3sl4co1l]

8P8C is an abomination unto electronics; a DIN connector is perfectly fine.

Plain vanilla "multi-conductor cable" is probably fine too, certainly for 10BASE over modest lengths (10s of m, probably less than the full 100m though).  You may want to inspect it before applying connectors, to ensure you're using adjacent wires (they won't really be pairs).  Shielding will probably help in that case.

With twisted-pair cable, as said, unshielded is fine, both connectors and cable of course.

On that note, if you do use shielding, note it's meaningless unless connected at both ends, at least at RF.  It doesn't need to be galvanically tied to ground, but enough bypass caps in parallel are needed to tie it in to keep noise out.  Typically a connector will be assembled with foil tape joining the screen/braid to the shell; a single wire (gathered-up braid or drain wire) is worse, and the longer the unshielded length is, the worse still it gets until it might as well be unshielded.  The length dimension being on par with the frequencies of concern, so, RFI and ESD are a problem for more than a few inches unshielded length.

Tim

[jonpaul]

bonjour cher T3sla4ciol

Re 8P8C Ethernet connectors "abominationn..."

Au contraire !! in defense of the RJ type telecom connector,

RJ45 was designed by Western Electric and BTL, to serve specific requirements of the telecommunications network, ( same as 2 and 4 pin RJ11 for POTS).

Reliability of contacts, ease of termination, low cost. The gold plated wire spring contacts are self cleaning. Rated from dry Signal to max V and I of POTS and PoE Ethernet.

I have not seen an intact RJ11, RJ45 or RJ13X fail.  Neutrik and high quality professional connector manufacturers sell panel mount, waterproof, and sealable versions.

Just the ramblings of an old retired EE !

Bon Soirée

Jon

[voltsandjolts]

Ethernet is remarkably robust really.

Here's a 100Mbps ethernet connector used on 100+ metre jumper cables, deployed at 4000m water depth and plugged in by ROV.

DIN plug won't a problem

[bingo600]

I've seen Ethernet interfaced via 9-Pole serial's w screws, that was due to the "Railway Vibration approval"

/Bingo

[langwadt]

https://youtu.be/LUYSfEL9E8I

[T3sl4co1l]

bonjour cher T3sla4ciol

Re 8P8C Ethernet connectors "abominationn..."

Au contraire !! in defense of the RJ type telecom connector,

Hm, I should be more specific.  Allow me to correct "electronics" to "electromagnetism".

Tim

[jonpaul]

Rebonjour Cher T3sla4ciol

Bare RJ45 were never intended to be EMI Shielded

The PHY Ethernet is balanced and controlled rise time

A wide range of Shielded RJ45 cable and panel mount connectors are available

Many Ethernet modules have RJ45 connection with built-in CM or DM EMI chokes and 1:1 transformers.

Bonsoir

Jon

[ajb]

From a signal integrity perspective, this sounds to be a challenge because there may be an impedance mismatch at the connector end because it is not a standard ethernet connector. I'm thinking it might be ok because the ethernet throughput should be under 10Mbps (only using 10 or maybe 100BASE-T) and I could use twisted pair cable for the other end of the connector cable.

There's always something of an impedance mismatch with standard 8P8C connectors due to the pinout and basic connector structure.  Also consider the average lowest-bidder cabling install, where who knows how far back from the jack the pairs will have been unwound to make the termination easier.  So, it's good to consider best practices in your designs, but the whole system is designed to tolerate a certain amount of non-ideal wiring/termination, and is relatively forgiving as long as there aren't too many sins on a given link. 

Any issues from an EMI perspective that can be addressed? Additionally, should I still use a discrete ethernet isolation transformer (magnetics) on the board?

Yes, you should always use a transformer on your board to provide isolation from the outside world, and Eth transformers often include common-mode chokes as well that will help improve signal integrity.  Technically you can use capacitive isolation instead of a transformer, but this is generally only used between PHYs on the same PCB or possibly in a backplane system.

[Swake]

Late 80's and start of 90's it was relatively common to find cat3 level wires running 10baseT nearby mains wires under the floor in open space buildings. Search for images of patch panels of that time and you'll have no doubt your solution will work. The patching was generally done with flat cable.

[tooki]

8P8C is an abomination unto electronics; a DIN connector is perfectly fine.

Plain vanilla "multi-conductor cable" is probably fine too, certainly for 10BASE over modest lengths (10s of m, probably less than the full 100m though). 

I can assure you this is incorrect: back in the days when many small businesses were just installing their first Ethernet LANs, I fixed more than one ill-performing network where the problem was that they’d used “silver satin” 8P8C cords (typically intended for ISDN) instead of twisted pair. More than a few feet of it cause significant packet loss on 10Base-T.

TCP/IP is quite good at masking said packet loss (at the expense of total throughput), but on pickier protocols like AppleTalk (which many of my clients were still using; I was a Mac consultant back then, right when everything was transitioning to TCP/IP) the packet loss caused significant problems.

[redkitedesign]

8P8C is an abomination unto electronics; a DIN connector is perfectly fine.

Plain vanilla "multi-conductor cable" is probably fine too, certainly for 10BASE over modest lengths (10s of m, probably less than the full 100m though). 

I can assure you this is incorrect: back in the days when many small businesses were just installing their first Ethernet LANs, I fixed more than one ill-performing network where the problem was that they’d used “silver satin” 8P8C cords (typically intended for ISDN) instead of twisted pair. More than a few feet of it cause significant packet loss on 10Base-T.

However, nobody would still use the 10Base-T PHY's your customers were using (this would have been like 30 years ago?).
Modern equipment will use PHY's capable of 1000Base-T. They have much better signal quality on the output, and are much more sensitive at the inputs, especially when only doing 10Base-T

Of course, this is definitely not robust or standards-compliant, but you might get away with it.

[tszaboo]

M12 D coding is made for Ethernet, which is a similar size circular connector, with readily available stuff for it.

[T3sl4co1l]

I can assure you this is incorrect: back in the days when many small businesses were just installing their first Ethernet LANs, I fixed more than one ill-performing network where the problem was that they’d used “silver satin” 8P8C cords (typically intended for ISDN) instead of twisted pair. More than a few feet of it cause significant packet loss on 10Base-T.

Hah, good to know!

Twisted pair it is then.

Tim

[sddsfgadasd]

Thanks all for the suggestions For EMC and ESD, is it ok to follow the discrete magnetics design as in Figure 3 from https://ww1.microchip.com/downloads/en/Appnotes/VPPD-01740.pdf?

Since there is no RJ45 chassis, does a Bob Smith termination just connect to the casing of the DIN plug (which is connected to board ground)?

[ebastler]

I'm working on a design where it needs to use only circular DIN connectors [...]

Umm, but why?? Even if you can get good signal integrity for Ethernet over a DIN connector, won't the users hate you forever since you force them to use (buy? make?) custom cables? What is the compelling reason against using a standard RJ45 connector?

[T3sl4co1l]

I'm working on a design where it needs to use only circular DIN connectors [...]

Umm, but why?? Even if you can get good signal integrity for Ethernet over a DIN connector, won't the users hate you forever since you force them to use (buy? make?) custom cables? What is the compelling reason against using a standard RJ45 connector?

Seems not unheard of...
https://www.globalsources.com/DIN-cable/Mini-DIN-8pin-to-RJ45-cable-1141079898p.htm
See a few for MIDI, which I guess makes sense; maybe some phones have used it too, not sure?  RS-232?

To be sure, M12 and such are far more commonly used for Ethernet.

I guess there's power (not PoE) and other signals, which wouldn't exactly be compatible with standard.

Since there is no RJ45 chassis, does a Bob Smith termination just connect to the casing of the DIN plug (which is connected to board ground)?

RJ45 chassis..?

Chassis chassis.  Is this in a metal box?  (If not, maybe it should be, or at least a shield/plane to tie together all the connector grounds?)  Presumably the shell/shield also connects there, or at least with some RF bypassing (to afford some galvanic isolation -- alleviate ground loop).

If it's completely and utterly floating, like, a battery powered box (or line powered, for that matter), no other connections, plastic case, single connector -- BS would return to the shell, if anything at all.

Note that, because your other wires are in the same cable, they should be terminated the same way, for the same reason... this may not be easily arranged depending on what the "signals" are, and the "power" may need a CMC to introduce that impedance at high frequency.

Tim

[berke]

I'm working on a design where it needs to use only circular DIN connectors [...]

Umm, but why?? Even if you can get good signal integrity for Ethernet over a DIN connector, won't the users hate you forever since you force them to use (buy? make?) custom cables? What is the compelling reason against using a standard RJ45 connector?

Seems not unheard of...
https://www.globalsources.com/DIN-cable/Mini-DIN-8pin-to-RJ45-cable-1141079898p.htm
See a few for MIDI, which I guess makes sense; maybe some phones have used it too, not sure?  RS-232?

Unlike USB3, Ethernet is fairly robust electrically, even at gigabit speed because it's relatively slow per pair as it runs at 125 Mbaud.  For a couple meters, as long as you use twisted pairs, the connectors won't impact things much, you can use .1" pin headers (I tested this for a few weeks myself because I had ordered the wrong size of plugs) or probably even screw terminals and get away with it (if you don't care about emissions).

Maybe it's unfair but RJ45-style connectors are seen as flimsy and do not inspire much confidence in high-vibration environments.  The plug plastic is definitely unwelcome in space because it is likely to off-gas and condense on delicate optics.  As a result for mil/aero RJ45 is typically replaced with those eye-poppingly expensive circular connectors (hundreds of bucks per connector; add zeroes for the tooling) or micro-D connectors.  This is why there is a market for non-standard, industrial Ethernet connectors, see for example Glenair.

[sddsfgadasd]

Is this in a metal box?  (If not, maybe it should be, or at least a shield/plane to tie together all the connector grounds?) 

Yes, I can use a metal box which would be shorted (at DC) to the shell of the connector. Would you need to then connect the chassis to the PCB ground using a small capacitor (100nF) to prevent ground loops as you mention or is wiring it directly to PCB ground using a low DC impedance connection better?

Thanks 

[T3sl4co1l]

Unlike USB3, Ethernet is fairly robust electrically, even at gigabit speed because it's relatively slow per pair as it runs at 125 Mbaud.  For a couple meters, as long as you use twisted pairs, the connectors won't impact things much, you can use .1" pin headers (I tested this for a few weeks myself because I had ordered the wrong size of plugs) or probably even screw terminals and get away with it (if you don't care about emissions).

Headers are quite good, actually.  The diameter and spacing gives somewhere in the vicinity of 100 ohms diff, so they can be about as long as you care to use and it doesn't matter.  Best of all, you can choose your own pinout and not split a pair across three fuckin' pins..  They fall apart at higher frequencies (gigs) where the mismatch and stub length of the PCB pads and surrounding plane cause significant mismatch (these are small/short structures, hence the much higher frequency of relevance), and where the non-twisted and unshielded geometry makes for significant emissions.  By then (i.e. PCIe etc.) you're going to be using fine pitch, high speed connectors anyway (usually with well defined impedance, SMT pads, and shields with lots of ground pads).

Is this in a metal box?  (If not, maybe it should be, or at least a shield/plane to tie together all the connector grounds?) 

Yes, I can use a metal box which would be shorted (at DC) to the shell of the connector. Would you need to then connect the chassis to the PCB ground using a small capacitor (100nF) to prevent ground loops as you mention or is wiring it directly to PCB ground using a low DC impedance connection better?

Depends.  It's certainly an option though.

Tim

[westfw]

You might want to look into the new "Single Pair Ethernet" standards.

[jonpaul]

broadcast industry uses métal shielded RJ45 eg Neutrik.

Many pairs of Ethernet are connected via a DB connector

Jon