3
\$\begingroup\$

I am looking for suggestions on protecting a gigabit ethernet port from ESD between the jack and the magnetics.

Due to the nature of the equipment design, the magnetics cannot be placed next to the non-standard (ie not RJ-45) circular connector. The transformer is about 3 inches away after the signals traverse a flex and onto a PWB. Currently using a Semtech Rclamp0524 device on the flex right at the connector for the high-speed Low-C ESD protection that doesn't affect the eye diagram of the PAM-5 differential signals.

However we have found that when connected to some network routers/switches with PoE (which tends to be turned on by default on those pieces of equipment) the TVS diodes get destroyed (sometimes short -> port quits working, or sometimes open -> port works but now unprotected). The theory is that the V/I characteristics on these during the PoE probes makes the port look like there is a PD controller on it and the PoE runs the voltage up and destroys the diodes.

We are going to move these TVS diode arrays to between the transformer and the MDI/PHY port to protect the phy. However I also need to protect the other circuitry on the PWB (near the flex interface) from seeing the ESD transients (every other interface port has ESD protection right at the back of the mil-circ connectors to keep it off the main PWB). The concern is that any TVS capable of withstanding PoE voltage has too much capacitance and will degrade the eye diagram to the point that 1Gbs operation won't work.

Does anyone have recommendations for parts and/or an alternate solution that could be used to kill ESD on the flex/connector circuit?

Circuit Layout: enter image description here

Simplified Schematic: enter image description here

\$\endgroup\$
10
  • \$\begingroup\$ You need to definitely show a schematic. You also should be considering EN pr IEC 61000-4-5 indirect lightning surge protection levels and not piddling ESD events. 61000-4-5 is standard for normal ethernet interfaces. \$\endgroup\$
    – Andy aka
    Jun 29, 2020 at 17:37
  • \$\begingroup\$ I will draw up a diagram. IEC specs are not applicable in this case. Testing is to MIL-STD-461G CS118. CS117 is not applicable in the vehicular enclosures this equipment gets installed into. \$\endgroup\$
    – BrianB
    Jun 29, 2020 at 18:06
  • \$\begingroup\$ Please draw a diagram, but yes, Rclamp0524 has four signal pins, and if you are using that at the connector side of POE ethernet, that's just plain shorting the POE voltage out if it is between positive supply pair and negative supply pair. \$\endgroup\$
    – Justme
    Jun 29, 2020 at 18:42
  • \$\begingroup\$ @Justme - I don't think so. It has to go through the reverse breakdown of the zener diode. There is no forward-bias path between pin pairs. \$\endgroup\$
    – BrianB
    Jun 29, 2020 at 18:59
  • \$\begingroup\$ Justme - further, the 2-pair (4 wires) on the Rclamp are connected to Pairs B+/- and C+/- with A+/- and D+/- connected on another Rclamp part. So for both Modes A & B of the PoE spec, DC+ and DC- are never on the same Rclamp device. \$\endgroup\$
    – BrianB
    Jun 29, 2020 at 19:51

1 Answer 1

Reset to default
1
\$\begingroup\$

The Rclamp0524 devices are incorrectly connected.

Usually Ethernet lines don't have ground-referenced ESD protection on the cable side of the transformer, you only see something like a TVS connected to both pins of a pair, if anything, and the Rclamp device would be on the PHY side of the transformer. On the cable side simple TVS between wires of a pair can be used, so there will be four of them, and they do not have to withstand the PoE voltage. Of course a special ESD device meant for protecting PoE enabled devices can be used.

Just for sake of simplicity, it says in the Rclamp datasheet that any pin to ground has a clamping voltage of 15V at 1A.

Notice that you have the Rclamp ground pins of all Ethernet pairs connected together, let's not call it ground for now, but common reference.

So due to the protection device, there can only have 15V from PoE+ voltage to common pin, and 15V from common pin to PoE- pin. So whenever PoE voltage of even 30V is applied, you would have clamping current of 1A via Rclamp devices.

In practice, as each diode has forward voltage of about 0.6V and the TVS has reverse breakdown at about 9.36V, it will start conducting at somewhere near 12V already.

So you must not reference or protect the ethernet pairs to ground or to each other, and with the Rclamp device, it does not withstand PoE voltages.

\$\endgroup\$
5
  • \$\begingroup\$ The PoE providing equipment is supposed to probe the line for a PD device as part of negotiation. Our belief is that the reverse breakdown V/I characteristics of the Rclamp device makes it look like a 19 to 26 KOhm impedance and the router then ramps up to full voltage. \$\endgroup\$
    – BrianB
    Jun 30, 2020 at 13:15
  • 1
    \$\begingroup\$ As to the grounding, I don't quite follow - if the TVS is line to line, where does the transient energy flow during an ESD event? You can't assume that a cable is connected when the energy reaches the pin (in fact MIL-STD-461G CS118 calls for contact discharge on the individual pins with return through chassis) and I stated that the energy cannot be allowed to flow down the flex to the other circuit card. The only place to shunt that energy is directly to the chassis ground. \$\endgroup\$
    – BrianB
    Jun 30, 2020 at 13:22
  • \$\begingroup\$ It is more common for Ethernet ESD application notes to not have line-to-ground protection on the cable side, only line-to-line protection. If you have an ESD event on one pin, it travels into transformer and out from the center tap, via 75R resistor and the capacitor to ground. As the transformer might convert the pulse to the PHY side, the PHY side protections take care of that. If PoE is passive, you need to support more than 48V between pairs, it won't negotiate, and two Rclamps will clamp that to below 30V. \$\endgroup\$
    – Justme
    Jun 30, 2020 at 14:27
  • \$\begingroup\$ @Justme I sincerely hope there aren't any switches acting as passive PoE sources by default. Enabling passive PoE without a compatible device on the other end is equivalent to adding an injector blindly, and will likely fry the other device. \$\endgroup\$
    – TooTea
    Jul 25, 2021 at 19:41
  • \$\begingroup\$ @Brian I'm collecting information whether or not the various ESD standards require contact discharges into the individual connector pins. You've mentioned that the MIL-STD-461G prescribes contact discharge on the individual pins. I've read the section CS118, and I only found reference to discharge into shells of connectors (5.16.3.3 item b). Am I missing something? When your device was actually tested, were there contact discharges into pins? (Discharges into pins are a worthwhile engineering test in one's own lab. I'm asking about formal qualification tests only.) \$\endgroup\$
    – Nick Alexeev
    Apr 17 at 22:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.