11
\$\begingroup\$

At a job site we typically have 10 vehicles connected to a single vehicle (star layout) by a 100m 5 conductor cable carrying power(unused), ground, shield, CANH, and CANL. The shield is grounded on both ends, but the ground is effectively the battery ground, which as far as I know isn't tied to any other external reference.

What (if anything) keeps the shield from acting as giant antenna? My understanding is the shield is meant to block EMI from entering the can bus, and may also be to limit external emissions from the bus, but I thought that to prevent radiation both ends needed to be well grounded; I'm not sure how (or if) a floating battery based ground effects it.

Disclosure: Not a EE/CE major.

\$\endgroup\$
  • 2
    \$\begingroup\$ Very Vaguely as I never went deep into these things. The shield does not stop anything- it actually picks up all the noise and get carried over to the enpoints for noise cancellation.So that the same noise on your BUS gets removed and your left with a nice clean signal. Now the big problem is that you have floating grounds and you could have a ground-loop problem.and this actually affects the Signal and not the shield! A major problem experienced in CarSound Enthusiasts. A ground-loop isolator is the only thing that comes to mind. \$\endgroup\$ – Piotr Kula Jun 9 '11 at 18:31
  • \$\begingroup\$ I agree with @ppumkin, I think you may need better isolation since each system is a floating ground from each other. Also typically the shield is only grounded at one end, say also as the star center. \$\endgroup\$ – kenny Jun 9 '11 at 18:37
  • \$\begingroup\$ If a signal output is capacitively coupled to a floating shield, and a signal input is capacitively coupled to the shield, then the shield will conduct the signal from one to the other. If the shield is grounded relative to these signals, then they will be capacitively coupled to ground instead of to each other. \$\endgroup\$ – endolith Jun 9 '11 at 19:17
  • \$\begingroup\$ Yea- thats why i never went deep into those things.. i know it makes sense @endolith .. but you know :) Hopefully it helps the OP \$\endgroup\$ – Piotr Kula Jun 9 '11 at 19:26
9
\$\begingroup\$

In an ideal system, the shield will not radiate (or pick up) and EMI. As soon as I see an ideal system we can break out the Campaign, as I haven't seen one yet in 25 years in this business.

The floating (battery) ground has no real influence over the problem-- well, at least no more influence than anything else.

An important thing about shields is that there should normally be no current flowing on that shield. Any current flowing will likely be radiated as EMI. The mistake that many people make is that they connect SHIELD directly to a GND pin of the same connector. That's not correct. SHIELD of the cable should be connected to the the metal chassis of the box, and GND should be normally connected to the ground inside of the PCB it's talking to.

I'll also note that EMI is a hugely difficult subject that is often more art than science. This answer just barely touches the surface, as we don't have the time or space to give this a comprehensive answer. But I suspect that you might not want the more detailed answer anyway.

\$\endgroup\$
5
\$\begingroup\$

The shield is meant to keep outside signals from getting onto the CAN lines, and signals on the CAN lines from radiating. Radiation from the CAN lines is not likely too much of a problem since CAN is a differential signal. One line goes up as the other goes down. If these are carried on a good twisted pair of the right impedance, there should be little radiation of the CAN signals.

However, there is other inevitable noise in your system that is not just the differential CAN signal. This noise could easily get on the shield, especially if the shield is hooked up so that it might carry some ground return currents. Yes, the shield is a long thin antenna wire.

I would connect all the shields at the center of the star, and leave them disconnected at each endpoint. Don't connect anything to it, just leave it floating. Use only the ground line inside the shield as the ground reference of the CAN bus and to tie the grounds of all the units together. Make sure that all the endpoints are not othewise grounded. Only the center node, if any, should be explicitly grounded.

\$\endgroup\$
  • \$\begingroup\$ Would it not be a good idea to take all the floating points and deal with them in the center somehow? \$\endgroup\$ – Piotr Kula Jun 9 '11 at 19:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.