Do you have a question on Embedded electronics and controls? Ask our Expert, Paul Clarke

Hi Paul,

I am designing a Duplex CAN bus architecture to be transmitted over a long cable (maximum 25 km.) with 1 conductor plus return path, which will also carry a DC voltage. What would be the best method for coupling and decoupling the data stream from the DC offset to minimise losses and maximise SNR?

Hello Raymond,

You don't quote the bit rate you need to support but I think you will have some problems:

CAN won't do 25km (at sensible bit rates) or mix with DC power using any standard physical layer.

You need a CAN modem at each end which will relay CAN traffic through the cable.

Without knowing a good bit more about the cable and bit/data rates you want to support it isn't possible to make rational suggestions about the design of the modems.

(It isn't just the CAN bit rate that matters but the actual amount of data to be transferred and acceptable latency or delay.)

(If you have deep(ish) pockets I can design this for you - it sounds quite fun ;-)

Thanks for the reply Michael. Alas, in these times of economic adversity, the "deep pocket" option isn't an option so I shall have to muddle along trawling forums and struggling through text books.

I haven't got any data on the reactance of the cable, but it has a characteristic capacitance of 0.6uF for 10000 ft. (I apologise I;m too lazy to figure that out in metric) with 100 ohms resistance (for 10000 ft.) on the forward path and 68 ohms (per 10000 ft.) on the return path.

I made an error in my original post, I meant 25000 ft. , not 25 km. Bit rate isn't enormously important, I need to multiplex 2 analog signals on one side and send a trigger from the other side. Delay could be an issue if it was significant. sub-millisecond resolution on the signals would be nice but whatever really, it's negotiable.

I suggested CAN bus as I know that this can be done, but OFDM or some other encoding would also work, it doesn't matter.

cheers.

Thanks for the reply Michael. Alas, in these times of economic adversity, the "deep pocket" option isn't an option so I shall have to muddle along trawling forums and struggling through text books.

I haven't got any data on the reactance of the cable, but it has a characteristic capacitance of 0.6uF for 10000 ft. (I apologise I;m too lazy to figure that out in metric) with 100 ohms resistance (for 10000 ft.) on the forward path and 68 ohms (per 10000 ft.) on the return path.

I made an error in my original post, I meant 25000 ft. , not 25 km. Bit rate isn't enormously important, I need to multiplex 2 analog signals on one side and send a trigger from the other side. Delay could be an issue if it was significant. sub-millisecond resolution on the signals would be nice but whatever really, it's negotiable.

I suggested CAN bus as I know that this can be done, but OFDM or some other encoding would also work, it doesn't matter.

cheers.

Hello Raymond,

Is this co-ax cable - if so then what matters is the characteristic impedance and the insertion loss at the frequency you operate at. To drive this length of cable you will need to properly impedance match the transmitter and receiver. You have the added complication of the DC bias. Don't forget that 25000 ft of your cable puts (168 * 2.5 = 420) ohms in series with your power supply which may limit how much power your remote end can use.

I wouldn't use CAN for this - you have a simple master and slave topology so CAN's networking virutes are wasted and its poor use of physical bandwidth will be a problem.

When you say you want to "multiplex 2 analog signals" do you literally mean couple one and then the other to the cable or do you plan to digitsie them and transmit data back. If so then what is the analogue bandwidth and signal/noise ratio you need.