ESD protection on an isolated circuit

Question

I'm currently working on an RS485 DMX circuit and am thinking about proper protective measures.

I've oriented myself on this question. I understand that galvanic isolation does not equal ESD protection, so I need some TVS diodes on the incoming signal lines which clamp to chassis ground, not isolated ground (because the isolation devices only work up to 600V/1000V.)

This raises some further questions:

1. Do I need to filter the input of the DCDC as well (because of switching noise?) If yes, how much? I don't think I need a full CM-filter though, as the signal is relatively slow at 250kbps.
2. Although the circuits are isolated, they are not allowed to drift apart too much if I use TVS diodes, otherwise they'd trigger randomly. As I use ESD-protected drivers, I don't have to choose tightly specified diodes. I only need to protect the isolation devices, not the driver. But I still need to keep chassis ground and iso ground close enough that the TVS doesn't trigger. Should I just put 10Mohm between the grounds to couple them, or am I overthinking this? (see Edit)
3. The former point adresses data line ESD events. What about ESD on the ground plane? Usually you don't protect GND from ESD events, but if I don't then the ground has +8kV in respect to the signal lines, which can't be good either - and I'll violate the 600V isolation to the main circuit. So can I just put a TVS on isolated ground to chassis ground? How does this mesh with the ground coupling?

Maybe it's an XY question. What I want to do: Protect my IO from ESD and misconfigured devices down the line. Is this appropriate or am I overthinking stuff?

Edit: I've read ANSI E1.11 (the DMX Standard), which says that galvanic Isolation for transmitters is not necessary, but can be done (4.2 & Annex A1). I'll go with isolation, because the ports can also act as an receiver which should be isolated. Also, I'll need to connect iso gnd to chassis gnd with >22MOhms.

So question 2 has answered itself. Question 1 & 3 still remain; does the R between iso gnd and chassis gnd pose problems to ESD events? I haven't found a reference implementation for ESD protection.

Answer 1

250kbps is quite fast for RS485. How far apart is your nodes?

How big your EMI problem will be (and how strong protection is required) is significantly tied to location of your infrastructure. If they are all contained in the same building a tiny TVS diode, such as the SM712 will suffice. However, if you are outside, with a cable that is some hundred meters long you need to "beef up" with a GDT or MOV, and appropriate supporting components.

Boruns has some great articles on this topic: Link

Answer 2

ESD & Overvoltages: question 3

High level

1. As far as I know ESD does not transmit along cables and only accessible parts on the equipment itself must be protected: if you can touch the pins of a connector, then you must protect them; a display must be protected.

2. Different are overvoltages due to indirect lightning effects and induction from power cables.

3. Different also is polarity reversal and accidental mains voltage connected to the lines.

4. Field ground is exposed too, as cable shields bring in phenomena of the (2) and (3) type.

Bits of answers

1. If you use ESD-protected drivers, you are okay with fingers touching pins.

2. The level of insulation of the "dirty ground" connected to the field protects against overvoltages: if drivers and DC/DC are insulated for 2.5 kVrms 1 min., the they withstand approx. twice as surge voltage (slightly more than that, but there are different types of surge waveforms and equivalency is not well documented; see old IEC 60950-1).

3. If you want to protect the "dirty ground" for overvoltages, you should understand to which other ground you should terminate the SPD (Surge Protective Device) you put in. If you have an external ground for safety and no other part of your equipment is connected to it directly, you can discharge the "dirty ground" to it when you approach the level of rating of your galvanic insulation (let's say 5kV surge voltage). I did it once with a Bourns 2093. But this is for really exposed cables, in my case running along a metro track on a viaduct.

4. One consideration of the isolation rating: you speak of "600V". You mean 600Vrms for 1 minute I think: isn't that low for the devices available nowadays, e.g. from Maxim and Texas Instruments? I think 2.5 kVrms is viable, and 3.75-4 kVrms are also quite available. Then -- as said -- multiply by approx. two for surge peak voltage.

5. If you connect ISO ground (dirty ground) and chassis with a resistor, that resistor must be rated for the insulation level, so you need a chain of resistors, and even the PCB where you could mount the chip resistors is at stake. Thus you need resistors with lead wires.

   Take into account also that the coupling capacitance between dirty ground and chassis is likely to shunt the ESD discharge (in case there is one on such dirty ground). After estimating it, you could help it with a tiny high-voltage capacitor (e.g. some pF), that would close the circuit during the ESD. Y1/Y2 types are acceptable for safety-relevant applications.

---

DC/DC filtering: question 1

Your DC/DC is not that big source of disturbance and I do not think it will propagate farther to be a concern for EMC certification. In any case, you can put a couple of chip inductors and a transversal capacitor between pos and neg, that works as a diff mode filter, but with the low-ed inductor (the one on the neg) you balance the impedance and you get some common-mode filtering, just in case there is really need.

This if the DC/DC output really disturbs the driver.

Common values may be some hundreds µH and some µF. It should cover all possible switching frequencies, from some tens of kHz and above.