15

I am designing a small sensor which fits into a custom aluminum enclosure. The sensor uses CAN bus and a long cable to connect it to the host and power. The shield of the cable will ground the enclosure through the connector.

Sensor

While developing the sensor I realized how important it is to ground the enclosure, otherwise my ADC readings are more noisy.

Now my dilemma is that if my costumers forget to ground the shield or use a shielded cable (which they will) the performance will be poor, as the enclosure will be left floating.

Connecting the power ground to the enclosure (shield) inside also seems like a bad idea as this creates a ground loop.

Is there any way I can get around this problem? An AC coupled shield?

W5VO
  • 18,303
  • 7
  • 63
  • 94
SensorMan
  • 211
  • 3
  • 5
  • The AC shield coupling sounds problematic as it will cause an AC ground loop. I suppose you cannot integrate a separate inner "shield," or a cheap version thereof, such as guard traces and extra (quiet) ground planes? –  Nov 20 '13 at 06:21
  • An inner shield cannot be realized in the current design. I'm not quite sure why I have this problem, but putting it inside the enclosure makes everything worse. The AC coupling was something I saw used in USB devices. They have a parallel 1Meg resistor and 4.7nF cap between ground and shield at the device. – SensorMan Nov 20 '13 at 08:09
  • Which means that your USB device, if their shield is additionally grounded externally, will have a ground loop with built-in impedance of 1Meg || 4.7nF. Much better than a practically zero resistance, but not perfect, either. Have you tried yet what the empirical impact of this approach is on your prototype? –  Nov 20 '13 at 11:14
  • Hi pyramids, no I have not. I will test the 3 different approaches and let you know :) – SensorMan Nov 22 '13 at 08:10

2 Answers2

2

Two solutions:

  • Include a shielded cable with the sensor. Put a sticker on the sensor saying: shielded cable is required.

  • Spend enough time to understand why your ADC readings are sensitive to the missing shield connection. Maybe you can improve this by changing the filtering on the signals on the inner conductors?

For a product like this I have seen good results from connecting the board ground to the enclosure (not sure if you have done this, but I see no reason not to).

Also I have seen this type of product with a full metal enclosure and only one shielded cable connecting to the outside being very immune to ESD and burst tests (CE) as well as very silent on EMI. So unless there is something I haven't picked up, you should worry if that is not the case.

Good luck.

Rolf Ostergaard
  • 4,448
  • 18
  • 24
  • Thank you for your answer Rolf. I have the option to connect the board ground to the enclosure, as shown in the picture, but I just worry about ground loops forming, as the shield is then also connected to the board ground through the connector. I'm not sure if I understand your answer correctly: You have seen both types of grounding? 1) Board ground connected to enclosure. 2) only the shielded cable connected to the enclosure, and the board is not connected. Is this correct? Which one would you recommend then? Thanks again. – SensorMan Nov 20 '13 at 09:50
  • 4
    I see no reason not to ground the board to the enclosure. And you need to explain where the "ground loop ghost" is and what harm it does so I can understand if I would worry. – Rolf Ostergaard Nov 20 '13 at 10:41
  • I think he meant the loop formed by shielding and ground line from the cable. Could you please explain on that? – richieqianle Nov 26 '15 at 11:11
1

Connect your circuit ground to the enclosure. This alone will shield your circuit. Ground loop happens when you have two earth points with different potentials. If you connect everything to the same enclosure, with a thick enough cable, you wouldn't experience any ground loop.

By the way, try measuring the noise on your ADC input when no ground is available. I am pretty sure it is a 50-60Hz power line induced noise.

Ali Alavi
  • 325
  • 2
  • 10