Bypassed ethernet barrier

Thread Starter

butthead

Joined Oct 21, 2018
36
I am looking at some PCBA's with ethernet RJ45 connector regarding surge test susceptibility. Common legal requirement are 1.5kV isolation.

The first issue that I see is PCBA's with RJ45 connector metal housing grounded to logical 0V (ground plane). Like Raspberry PI for example. Look at the bottom side of it. A surge impulse of 1,5kV on the shield of a STP cable will see no isolatation at all. It has direct galvanic connection to 0V. Surge impulse has lot of energy and as I see it will destroy the electronics at the first impulse.

The second issue I see is PCBA's with good isolation (4mm) between RJ45 connector and logical 0V, but with just 0.5mm isolation between the ethernet status LEDs (routed directed from the PHY) and RJ45 housing. This makes the 4mm barrier to have a weak spot of just 0.5mm, which definately should destroy the electronics too when STP ethernet cables are used.

What are your thoughts and experiences?
 

ronsimpson

Joined Oct 7, 2019
2,985
I think you have the idea backwards.
The cable has no shield and the metal housing does not effect the signal. In normal conditions the housing does not conduct current.
The shell needs to connect to ground. There should be no isolation. My products are tested with a spark machine that injects high voltage to all connectors to see if the product will fail. The spark will hopefully jump to the shell and save the electronics. It would be very bad if the spark jumped to one of the data traces on the board.
1573395452395.png
The LED wires are shielded from static (sparks) by the metal housing. They hide in safety.
If the metal housing was not connect to ground (very well) the housing would build up voltage and conduct the high voltage inside the box and kill many things.
 

Thread Starter

butthead

Joined Oct 21, 2018
36
The cable has no shield?
STP stands for shielded twisted pair.

Normal conditions?
Surge is not a normal condition. When applying surge test pulse on the shield of a STP ethernet cable, the RJ45 housing (like in your picture, or what's populated on a RPi) have galvanic contact with the STP shield. That pulse will if the housing is connected to logical 0V (like Rpi), pass on to the PHY ground and destroy the PHY and other electronics, in probably 10 times of 10.

There should be no isolation?
Ethernet has a legal requirement to 1.5kV minimum.

Static sparks?
You mean ESD? A surge (what this case is about) is a high energy transient, while an ESD is a low energy transient.
No matter if the LEDs are hide in a isolated housing, if the PCB CAD has a low isolation (0.5mm) to RJ45 housing, the 1.5kV isolation will not work, even if the rest has 4mm isolation.
 

ronsimpson

Joined Oct 7, 2019
2,985
Maybe we are talking about different things.
All my connectors connect directly to ground. (0V) Like on the Raspberry Pi. I do not have ESD failures. I usually have a ring of ground along the edges of the PCB to pick up ESD and carry it off to 0V.
I do not understand why you need the Ethernet shield isolated. That will kill the product for certain.
The PC I am typing on has the Ethernet shields & the USB shields not only connected to 0V but also directly connected to the metal case. There are fingers of metal that make the shield to case connection.
1573397634766.png
I am not talking about the data lines. They can be isolated or not. We sometimes run "power over ethernet". I should not have brought that up but power and ground is sent over the same cable along with the data.
--edited--
I see you are using Cat-7 or shielded cables. Then I think there can be no "isolation" or the shielding does not work.
1573398700485.png
 
Last edited:

Thread Starter

butthead

Joined Oct 21, 2018
36
Surge. Not ESD. They are two different things. Surge transients have much more energy than ESD. ESD is not the topic here.

I think the RPi is wrong with grounding the RJ45 housing to logical 0V (not same as chassis ground), because the 1.5kV surge transient on the shield will go directly to the 0V plane on the PCB. Ethernet has regulatory requirement for isolation of 1.5kV. If the shield are connected on the PCB to logical 0V, then there is no isolation anymore.

You probably don't have failure from surge transients, because you likely don't have any surge transients.

STP shield (through RJ45) housing should connect to chassis, but not logical 0V. (If there is an electrical conducting chassis).

Data lines must have 1.5kV isolation to logical 0V. That's why ethernet transformers are tested for 1.5kV isolation and bob-smith capacitor are rated to minimum 1.5kV. If you see no magnetics or barrier on the PCBA, then the magnetics are built-in to the transformer.

PoE still have to have 1.5kV isolation.


edit: No, not necessarily a CAT7. An STP cable. Can be Cat 5 or 6, but the STP type.
 

djsfantasi

Joined Apr 11, 2010
9,156
Surge. Not ESD. They are two different things. Surge transients have much more energy than ESD. ESD is not the topic here.

I think the RPi is wrong with grounding the RJ45 housing to logical 0V (not same as chassis ground), because the 1.5kV surge transient on the shield will go directly to the 0V plane on the PCB. Ethernet has regulatory requirement for isolation of 1.5kV. If the shield are connected on the PCB to logical 0V, then there is no isolation anymore.

You probably don't have failure from surge transients, because you likely don't have any surge transients.

STP shield (through RJ45) housing should connect to chassis, but not logical 0V. (If there is an electrical conducting chassis).

Data lines must have 1.5kV isolation to logical 0V. That's why ethernet transformers are tested for 1.5kV isolation and bob-smith capacitor are rated to minimum 1.5kV. If you see no magnetics or barrier on the PCBA, then the magnetics are built-in to the transformer.

PoE still have to have 1.5kV isolation.


edit: No, not necessarily a CAT7. An STP cable. Can be Cat 5 or 6, but the STP type.
If you aren’t using CAT7, then the shielding is not connected to ground on either device. Hence, the metal case is isolated from the cable.
 

Thread Starter

butthead

Joined Oct 21, 2018
36
Either you think 1.5kV isolation is not required or that the product has 1.5kV isolation when the RJ45 is connected to logical ground?
 

SteveSh

Joined Nov 5, 2019
109
In a more general sense, the outermost shield of a cable should be connected to chassis (ground) at both ends. I agree with nsaspook where he said that "An improperly grounded STP is worse than UTP". When we build shielded cables for running between boxes, the outer shield is always connected to chassis on both ends, through a metal backshell that makes contact with the shield 360 degrees around the cable. The metal backshell mates with the metal housing of the chassis connector, which is intimately connected to the chassis. A kiss of death from an EMC compliance standpoint is to connect the outermost shield to chassis through a pigtail.
 

Thread Starter

butthead

Joined Oct 21, 2018
36
Take a look at this article f.x. at figure 11, in the link below. If you connect the cable shield to logical ground 0V, then the isolation barrier is lost. According to surge testing EN61000-4-5, for STP at both ends,, the test pulse is applied to the chassis. If the chassis is non conducting, the test pulse is applied to the STP shield. With chassis ground connected to logical 0V, the expected result is failure to the electronics when the surge pulse is applied.

https://incompliancemag.com/article/designing-ethernet-cable-ports-to-withstand-lightning-surges/
 

djsfantasi

Joined Apr 11, 2010
9,156
STP cables have shield connected to RJ45 connector. Even if they are Cat5 or 6. Measure and you will see.
Sorry, I disagree! Perhaps that applies to the cables that you own, but it is not required.

There are many here with networking experience. I have extensive experience, having been responsible for many networks, for 24x7x365 e-commerce sites with multiple redundant networking and computing resources, that had 99.999% uptime.

CAT7 is the only standard that requires the shields of a cable to be grounded through a device. It appears to me (and I could be wrong) that you need either additional practical training or more research.
 

Thread Starter

butthead

Joined Oct 21, 2018
36
We are a bit off topic, as the case (still unresolved as I see it) was isolation barrier and chassis connection to logical 0V (actually how it is done in the PCB CAD). But that's OK. This is interresting too. Do you mean that there is no requirement for a mass produced Cat5 or 6 STP cables to be crimped to a shielded RJ45 plug, if it is going to be sold as a STP cable? I probably misunderstand you.
 

djsfantasi

Joined Apr 11, 2010
9,156
We are a bit off topic, as the case (still unresolved as I see it) was isolation barrier and chassis connection to logical 0V (actually how it is done in the PCB CAD). But that's OK. This is interresting too. Do you mean that there is no requirement for a mass produced Cat5 or 6 STP cables to be crimped to a shielded RJ45 plug, if it is going to be sold as a STP cable? I probably misunderstand you.
There is no requirement for a mass produced Cat5 or 6 STP cables to be crimped to a shielded RJ45 plug, if it is going to be sold as an STP cable!

Yep, that is what I mean.

We are on-topic as without my supporting information, I doubted that you’d believe me.
 

Thread Starter

butthead

Joined Oct 21, 2018
36
So the shield of the cable is just left unconnected at both ends? Can you provide a link to such cable?

The topic as I see it is RJ45 receptable chassis grounded to logical 0V would not comply to the 1.5kV isolation requitement. No matter of which cable being used.
 

djsfantasi

Joined Apr 11, 2010
9,156
In
So the shield of the cable is just left unconnected at both ends? Can you provide a link to such cable?

The topic as I see it is RJ45 receptable chassis grounded to logical 0V would not comply to the 1.5kV isolation requitement. No matter of which cable being used.
Nope, I think that’s a good exercise for you. It’s not hard, I found one in 5 seconds. And maybe it’s true that it wouldn’t comply to that standard. It’s not my fault or the cables fault. It’s the designers fault who specified a cable type where the shield is not grounded. As I’ve said more than once, CAT7 is the only shield specification where shielding is grounded to the connector case.
 

ronsimpson

Joined Oct 7, 2019
2,985
Hours ago I was reading a standard for building wiring. I don't remember where I was reading. They said that for connected shielding, the entire building must have connected shielding. There can not be a mix of shielded and unshielded cables.

I have wired up schools with unshielded cables. My house(s) and lab are all unshielded.

You probably don't have failure from surge transients, because you likely don't have any surge transients.
What we build must pass UL, VDE, CSA etc. testing. I hate watching them test. They try to get the spark to enter the box at every opening. The connectors are a point where they test many times. We do not want to see the CPU reset or anything.
I don't see how to protect the electronics with out passing the surge energy to ground.
 
Top