5

I am designing an electronic product which will be housed in a plastic (non-conductive) enclosure. There are a number of, mostly shielded, I/O cables (Copper 10/100/1000Mb/s Ethernet, USB 2.0, RS232, low/audio-frequency analogue inputs). The product will be installed inside a metal cabinet in an industrial (noisy) environment, and I am concerned about ESD and EMC issues.

At schematic level, there are a number of TVS diodes, ESD current limiting resistors, filter capacitors on the I/O lines. Now it is time to do PCB layout and routing and I am wondering how best to implement the ground system.

I have turned to (the guru) Henry Ott's Electromagnetic Compatibility Engineering, and have found two figures in particular and the corresponding text, very useful:

  1. For ESD reasons: Fig 15-17 (image below) shows how the use of a metal plate in the enclosure is used to divert ESD currents away from the PCB, via its stray capacitance, to (earth) ground. The key is to make a low impedance connection between I/O cables/screens and metal plate.
    This is intuitive to me and is also suggested as a solution to ESD in this electronics.stackexchange answer. Ott Fig 15-17. Use of a metal ESD ground plate to divert ESD current from the PCB.

  2. For EMC reasons: Fig D-13 (image below) shows how the use of a metal plate in the enclosure is used to create an intentional reference plane with the cable so as to form a monopole antenna which can be controlled, so as not to radiate, by bonding the cable and plate together. Again, the key is to make a low impedance connection between cable/screen and metal plate. Ott Fig D-13. Use of a metal ground plate as monopole reference plane..

My question is: can the ESD ground plate/reference plane be implemented, instead of using a metal plate, by using an internal PCB copper layer? And if so what are the benefits/drawbacks?
It would make the implementation far simpler (no mechanical design, easier to make low impedance connections between plane and IO cable/shield) and more cost effective. I cannot find anything in Ott (or elsewhere) that explicitly states this is not a good idea, but I cannot recall ever seeing an entire plane being used this way; sometimes chassis GND or IO planes are used, but normally these are in regions near the IO area not an entire plane (I would still do an IO GND plane in addition to the reference plane).

Follow up question: for the EMC case when an intentional/controlled monopole antenna is formed with the reference plane; whether using a metal plate or PCB plane, what stops the IO cable (now bonded to the plate/plane) from forming an antenna with other power/signal GND layers on the PCB?

Gman
  • 51
  • 4
  • 2
    I think this is an *excellent* question. Henry Ott's book on EMC is great, however today's designers have to deal with many difficult non-ideal situations that go far beyond what the book covers. For example, Ott emphasized that a cable shield is an extension of chassis shield, thus it should always be 360-degree terminated at the chassis at both sides (for RF), then to the circuit ground. Yes, it's 100% ideal and correct. Yet, it's often simply not an option if your digital board has a tiny USB-C connector in a plastic enclosure. The question is what to do in these "pathological" situations. – 比尔盖子 Dec 17 '21 at 16:57
  • 1
    When you're willing to make compromises, some solutions certainly exist. For example, for a digital board with a USB connector, the industry has generally accepted that you should connect the connector shield to the solid circuit ground plane with strong, low-impedance connection, same for ESD protectors. It's so well-accepted that it became a requirement in the USB-C specification. It's not ideal, but better than nothing, often works satisfactorily enough. I think it's actually a great opportunity for a subject expert to write a new book to systematically explain these design decisions. – 比尔盖子 Dec 17 '21 at 17:05
  • What kind of cables attach to your device? Like for USB it’s usually already shielded at the USB host. Like already mentioned I think I would use ESD at every point where the encasing (printed? Already made?) is breached by SD slots, buttons and connectors. The controlled monopole could also be a receiving antenna! I would be careful, what kind of EMI do you suspect? From the board itself or from outside? – RemyHx Jul 23 '22 at 06:49

2 Answers2

2

If a device has no earthed metal chassis (or metal foil in a plastic enclosure), we have to consider two cases:

  1. The device is fully floating, e.g. a mobile handheld device with no cables. In this case ESD is rather uncritical, because the device will only charge its self capacitance. Metal parts should be tied to th local ground plane to distribute the charge efficiently. EMI is addressed by good layout with minimum loops.

  2. The device has IO cables which bring an earth reference but no real earthing conductor, e.g. mobile phone connected via AUX to mains supplied speakers. ESD will now flow down these cables. So you have to make sure that this current can reach it in a benign way, e.g. through a copper plane where the high current only causes some mVs of voltage offset. That means each lane of the IO cables must have either a direct bond or ESD diodes to the copper plane, not necessarily because ESD enters through these ports, but because they provide the exit. EMI has to be solved again by good routing. But there is one more complication: large common-mode currents in the cable to turn it into an antenna. This cannot be influenced much from the floating device end, but a ferrite bead on the cable should be recommended to damp these currents.

Tim Williams
  • 22,874
  • 1
  • 20
  • 71
tobalt
  • 18,646
  • 16
  • 73
1

Based on what you have described, your device isn't internally referenced to earth, so the last thing you want to do is have the ESD event come in a connector wander around your PCB for a bit before exiting through another connector on its way to being sent to earth somewhere else. You're just making trouble for yourself. Put ESD protection on all signal/power lines as close as possible to the connectors, and routed to the connector grounds and you shouldn't have any further ESD concerns.

As for the EMC side of thing, the metal reference plane you have shown appears to be there to deal with a residual common mode signal that comes through the cable. It is usually far better to try to eliminate such common mode signals first, before you start worrying about mitigation options. If you design your communications interfaces, connectors and cabling correctly in the first place, then you probably won't need to do anything else. With EMC prevention tends to be the better option if you have the luxury.

Jon
  • 4,894
  • 13
  • 20
  • re: ESD; The device will be DC referenced to earth, but this will be via a connector+wire to AC safety earth or an earth electrode. This is a high impedance path which the high-frequency components of ESD will not want to flow through. Without the dedicated plate/plane ESD may prefer to jump onto one of my power/GND planes and through it's parasitic capacitance to earth. The ESD can only go back out through the connector (as you suggest) if that connector has the lowest impedance path to earth, but then why would it have entered my circuit via the connector in the first place? – Gman Feb 22 '19 at 03:33
  • 1
    re: EMC; most cables will have common mode chokes and shielding (in attempts to block common mode signals coupling onto cables or radiating from them), however because this is a plastic enclosure the shields will not be 360 degree terminated to a chassis in the classical way - so where do I terminate them? If terminated to signal/circuit ground then they will pollute my circuit, but if on a separate baseplate, they can harmlessly flow through the baseplate's parasitic capacitance to earth. – Gman Feb 22 '19 at 04:25