1
\$\begingroup\$

Topology I'm designing a network of sensors to be placed in an outdoor environment, see the attached diagram. The network will consist of 30 data acquisition boards (very low power consumption), divided in 10 zones ("Silo bags" in the image). Each zone will be separated by 5m and the Main node (in charge of the communication with a remote server, power distribution and configuration) installed in the middle, so the horizontal wires will have a maximum length of 25m. Each zone have a length of up to 75m (vertical wires) with 3 data acquisition boards in each zone.

I was planning on using UTP CAT5 or CAT6 cables, with 2 conductors for data (CAN bus) and the other 6 for power distribution (3 for GND and 3 for Vcc). The idea is to design the nodes with 2 RJ45 connectors and connect each data acquisition board in a chain. I would also need to design a board with 3 RJ45 connectors as a T-connector for the UTP cables (UTP splitter in the image).

  • What do you think about this scheme?
  • I know that CANbus is very immune against noise and the length are not near the limits, but can it bring me trouble having so many connectors?
  • What about the DC power distribution? considering that the whole system will be powered by a battery placed near the main board (I will make intense use of the low-power modes of the MCUs). What about the power losses?.
\$\endgroup\$
  • 1
    \$\begingroup\$ You may be able to get away with a simpler wiring scheme. \$\endgroup\$ – Jeroen3 Dec 7 '19 at 21:13
  • \$\begingroup\$ @Jeroen3 While power over data lines is clever idea, I would not call it "simpler" in this case. And introducing AC coupling into already stretched topology seems to be a recipe for disaster IMHO. \$\endgroup\$ – Maple Dec 7 '19 at 21:20
  • \$\begingroup\$ Why would you use RJ45 for an outdoors environment application? What part of this is actually outdoors? Somewhere along the way, there will be a need for IP rated connectors. Also, you don't want 50ohm impedance. \$\endgroup\$ – Lundin Dec 9 '19 at 11:49
3
\$\begingroup\$

What do you think about this scheme?

From the OP it seems you describe mixed topology with very long runs. The CAN bus is not particularly suitable for this. It is designed for single long run with multiple very short stubs, i.e. the opposite of what you propose. You might be able to make it work but only at the extremely low speed.

My recommendation would be to use CAN bus repeaters in place of those "UTP splitters" in the picture. You can use commercial modules, like this one or make your own based on any MCU with built-in dual CAN controllers. Note, that 75m long runs will still be a limiting factor in the maximum speed you can get from this. 1 Mbps is definitely out of question.

I know that CANbus is very immune against noise and the length are not near the limits, but it can bring me trouble having so many connectors?

Connectors should not be a problem if you design your boards properly. Texas Instruments has many papers with recommendations that you should follow, like this one or this.

Correct termination is the key. CAT5/CAT6 cables you are planning to use have 100 Ohm impedance, which is different from 120 Ohm CAN specs. You need to match this with 100 Ohm termination resistors and then make sure 50 Ohm total is within the acceptable load parameter of your transceivers. Here is useful TI forum topic on this subject.

Another thing to consider is using isolated transceivers and dedicate 1 or 2 conductors for "CAN ground", instead of using negative power line for it. This is because 75m cabling in outdoor environment might result in common mode voltage going beyond CAN specification.

What about the DC power distribution? considering that the whole system will be powered by a battery placed near the main board (I will make intense use of the low-power modes of the MCUs). What about the power losses?

The right way to deal with power losses is to increase bus voltage and use DC-DC POL modules on your data acquisition boards.

In our modular wheelchair electronics we are using CAT5 cables and dual RJ45 connectors on each board. 4 out of 8 conductors carry 24V power and each board has tiny 24-to-3.3V DC-DC converter, powering MCU and all attached sensors. All boards are daisy-chained and extra jacks at the two ends have plugs with terminating resistors.

Note, that we did not use all available conductors for power. Instead, we have "enable" signal line that controls those DC-DC converters. This allows us to effectively power down entire network by the command from master node, further increasing power savings.

\$\endgroup\$
  • \$\begingroup\$ Thanks for the complete answer. I don't have hard speed requirements, I need to send a few measures per day from every node, and sometimes deliver an alert when the vibration sensors exceed a threshold. \$\endgroup\$ – Gaspar Santamarina Dec 7 '19 at 19:25
  • \$\begingroup\$ The DC-DC converters are buck converters right?. Why would you need the control signals? Why not cut down the power from the main board? Or you use the control signals to power down each node separately?. \$\endgroup\$ – Gaspar Santamarina Dec 7 '19 at 19:44
  • \$\begingroup\$ Yes, they are step-downs with 95% efficiency. Cutting down the power from main board requires additional components while control signal can be directly connected to enable pins of the converters. Besides, as you've guessed it, some of the nodes have to be permanently active, so we have dip switches on them connecting converters either to control signal or the pull-up, depending on node function. \$\endgroup\$ – Maple Dec 7 '19 at 20:19
  • \$\begingroup\$ Forgot to mention that control signal is also routed to INT pin on MCU of permanently powered boards, so that software can further improve power savings or change the behavior depending on network status. I actually discussed the design on this forum. You can see the discussion and preliminary schematics here. It is different from what I described above but that is due to our specific requirements not pertinent to your case. \$\endgroup\$ – Maple Dec 7 '19 at 21:09
  • \$\begingroup\$ Great, gonna take a look. Thanks for all the info! \$\endgroup\$ – Gaspar Santamarina Dec 7 '19 at 22:47

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.