I ran into some troubles trying to design a wired bus which will connect in about 64 testing devices. My first choice of mind was CAN bus, known of it's reliability and low EMC interference. However, during my research, I have discovered that there is limitation of minimum distance between CAN nodes, related to load capacitance and media capacitance: http://www.ti.com/lit/an/slla270/slla270.pdf

enter image description here

The nodes are connected through PCB traces and connectors, I have estimated media capacitance of 0.62 pF/cm (single 1 mm width trace to ground plane through 2.1 mm FR-4, is it right way to calculate?) and TJA1040 transceiver has input capacitance of max 10 pF.

My bus layout is:

enter image description here

Node to node distance is about 60 mm. From the graph, minimum distance is about 180 mm.

I also have some problems understanding bus characteristic impedance. In my design, I can manipulate CAN_L and CAN_H traces width and spacing, minimum width and spacing is 0.5 mm. With 2.1 FR-4 thickness, trace width of 1 mm and spacing of 0.5 mm (possible to manufacture for me) I can achieve 120 Ohm differential impedance, which is.. required, yes?

CAN Bus is a differential bus. Each differential pair of wire is a transmission line.Basically, the terminating resistor should match with the Characteristic Impedance of the transmission line to avoid reflection. CAN bus have a nominal characteristic line impedance of 120Ω. Due to that we are using typical terminating resistor value of 120Ω at each end of the bus.

I am asking this question because I wonder how it's (differential impedance) related to media capacitance. If I could manipulate media capacitance, I could lower minimum distance and maybe it would work with CAN.

How I calculated media capacitance: enter image description here

Differential capacitance: enter image description here

If there's no possibility to make it work with CAN, what serial interface would you suggest instead of? Would RS-485 be better choice?

  • \$\begingroup\$ Are you sure your Conductor Height (H) value is correct? 2.1mm seems high. Is your board two or four layers? A lower H value will increase the media distributed capacitance, which will improve your minimum distance. I've run several CAN nodes at much closer distances than what that graph suggests (granted, they weren't at the highest data rates either). \$\endgroup\$
    – youtooth
    Commented Sep 25, 2017 at 18:59
  • \$\begingroup\$ @youtooth Yes, the data is right. This board is 3.2 mm thick, 4L. Inner insulation is 2.1, copper thickness is 70 um. 2.1 mm is measured from TOP (layer 1) to layer 3. \$\endgroup\$ Commented Sep 25, 2017 at 19:04
  • \$\begingroup\$ I could make these traces on Top layer (1) and layer 2, insulation thickness between them is something around 0.2, but then I had to make trace width like 0.1 mm to maintain differental impedance of 120 Ohm, which is not possible. \$\endgroup\$ Commented Sep 25, 2017 at 19:13
  • \$\begingroup\$ Any reason that layer 2 isn't used as a ground plane here (even if locally for these signals)? Either way, whatever connectors and cables you're using will add additional capacitance onto the 62pF/m you've calculated. So it's likely that the your total distributed capacitance is closer to 100 pF/m (just a guess). These constraints also apply to all serial interfaces, so RS-485 will still have similar concerns. \$\endgroup\$
    – youtooth
    Commented Sep 25, 2017 at 19:13
  • \$\begingroup\$ As I said, it's a lot easier to get 120 Ohm impedance with such spacing. May I ask, you said that you ran several CAN nodes, this means less than 60 mm? I thought that my case is extremaly special. How it was performing? \$\endgroup\$ Commented Sep 25, 2017 at 19:17


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.