23 May 2011

Observing the electrical behaviour of a NMEA 2000 network

Earlier I wrote that I was going to investigate why Simrad/Navico is now advocating adding a Simnet Noise Filter to more complicated networks.

First I measured a small bench network using Simnet cable at home.

A screenshot of a simple network (one sender, one Actisense NGT-1 and about 5 m of cable in total, with no terminators. I was expecting swinging of the observed signal when it returns the bus to a non-driven state, and indeed the digital scope proves it worth and shows this to good effect. As you can see, the rising edge when the sender drives the bus the wave is pretty neat. The swinging happens when the sender stops driving the CAN output driver and the voltage returns to zero, with quite large oscillations around 0 V.

Adding a 68 Ohm resistor (which is the nearest equivalent of 2 x 120 Ohm terminators) results in the picture shown here at the right. As you can see the ringing is a lot less now.

Then I added 40 m of Simnet cable:
As was expected, this added a bit of capacitance to the cable: the rising flank of the pulses aren't straight up anymore, especially once the voltage reaches 2,5 V, and the ringing is back. I didn't expect that.

And finally I switched to a 39 Ohm terminator, which is what you end up when you install a 'Simnet Noise Filter' at the mast base which is what Simrad is telling me to do. We now get a bit of overshoot on the rising flank as well, and you can see the slower and reduced ringing because of the changed impedance.

Last weekend I finally had the chance to measure my own boat's onboard network, which has a what I'd call decent mid-level network with 13 active senders and about 60 m of cable. The added load of the devices and longer cable results in a cleaner waveform. Nothing really wrong here, but still there is a bit of overshoot on the rising signal.

The last image is of course with the filter installed. The effects seen in the test network were seen again. No overshoot or ringing; this is now completely gone. The rise and fall time is still completely acceptable.

The noise filter also has as an added benefit that the network will work more reliably when the cable going up the mast to the wind sensor is removed, either because accidentally or on purpose.


The Simnet Noise Filter seems to work and improve the signal behavior of the two networks that I tested it on, and I intend to keep it installed. It is recommended for large(r) Simnet networks.

It would be interesting to see the effect of such a network on other brands of NMEA 2000 cable.

1 comment:

  1. hello, how do you aquire the signal? I see that you are using one channel, you connect the can H to center of the probe and the canL to ground of the probe? I've seen people measuring the differential lines using 2 channels, one normal and the other on inverted and using the function add between the two channels... also I've seen a rigol with can decoder integrated have you got the chance to see if decodes nmea 2000?