2

I'm currently designing a board that will have:

  • a primary power supply of 24VDC from an external converter
  • a shielded Ethernet connector
  • two BNC connectors (for two +/-10V signal, 14 bits, <100Hz)

The housing is in full plastic (screw headers also) but the front panel (and back panel) is metal.

The Ethernet shield is connected to the front panel (metal contacts on the connector).

The BNC are isolated from the front panel.

Power supply connector is isolated from the front panel.

Nothing is connected to the the back panel.

On the 24V power supply rail, I have one isolated DC/DC converter for the logic supply, and another isolated DC/DC converter for the analog supply. Output grounds of digital and analog are connected together in the regulators zone.

The power supply and the acquisition system that is connected to the BNC outputs, will be in the same room. The acquisition system inputs are also isolated. The board will be connected to a single remote Ethernet peripheral (direct connection), but with about 40m of length.

I'm questioning about the isolation or connections between the following references:

  • the "chassis ground", i.e. the front panel plate and the Ethernet shield reference
  • the "power ground", i.e. the input power supply negative reference
  • the "electronic ground", i.e. the isolated converters output ground

My objectives :

  • be sure that remote noise / perturbation incoming from Ethernet shield will not damage or perturbate the board.
  • be sure that ESD discharges on the front panel or the BNC metal part won't damage the board

This is what I was expecting to do: (but after reading and reading documentation, I feel never happy with what I do)

  • on the input power rail, I already have a TVS (SMCJ48CA-TR) and a common mode filter (744272251).
  • on each BNC, there is also another TVS (SMA6J13CA-TR) between central pin and ground pin.
  • connect "power ground after common mode filter" to "electronic ground" on each DC/DC converter with a 100pF/3kV capacitor
  • connect "power ground before common mode filter" to "chassis ground" through a 100pF/3kV capacitor (power and Ethernet connectors are close)
  • connect "electronic ground" to "chassis ground" or "power ground" (don't know) through a 100pF/3KV capacitor close to the BNC connectors pins.

I agree, these choices are result of "let's try and see" thinking. This may be shocking, morally or electrically, I don't know :)

I'm afraid of ESD discharges, I guess that's not with 100pF that I will get rid of them. And if a discharge happen on the BNC connectors, where should it go safely?

What should I add or modify or suppress?

Also, not sure where to make the connections before or after the common mode filter (that presents a certain inductance!)

Thank you.

SamGibson
  • 17,231
  • 5
  • 37
  • 58
aurelienr
  • 21
  • 1
  • Not what you were asking about, but if you get too much noise from the dc-dc converters you can start with a higher voltage and follow it with a linear regulator. Or just first try CLC filtering. – Rudy Dec 18 '18 at 01:38
  • 1
    About the ESD discharges: a capacitance between the "shocked" terminal and the reference potential sometimes helps very much, despite the low value required. This is due to the fact that the ESD is not a voltage or current disturbance, but it is a *charge transfer* disturbance so a somewhat higher capacitance in many cases lowers the shock voltage below dangerous levels, so don't be afraid of using small values of capacitance. – Daniele Tampieri Dec 18 '18 at 05:38
  • Ground is defined as your 0V reference but is also the return path to signals and power. Therefore EMI susceptibility needs a **profound understanding* of dielectric breakdown and RF current paths, such as ESD < 1ns rise time or >300MHz BW. Look at expert instrument design. – Tony Stewart EE75 Dec 18 '18 at 14:58

0 Answers0