For the past few years, insurance companies have been reporting an ever-increasing amount of incidents in the maritime world caused by lightning strikes. Here, we focus on how to protect the NMEA 2000 bus on a vessel against voltage surges caused by nearby lightning.

When reading up on how to protect against lightning strikes on a boat (e.g., Palstek, Blitzschutz auf Yachten by Michael Herrmann, VDE Merkblatt Blitzschutz auf Yachten), one generally finds that a lot of attention is paid to reducing the probability of a direct hit. And if hit, how to make sure the lightning strike has a perfect path to ground, avoiding human injuries and reducing damage to the vessel’s structure. Far less attention, though, is paid to protecting the electronic equipment on board, sometimes arguing this is only 3rd prio. However, at least in our experience, it is not very likely to have a direct lightning hit, but rather much more likely is a strike close by, perhaps within a perimeter of 50 metres. In 4 years we have had 5 such events. All equipment that was not powered on survived, but for the rest, every time we had some casualties.

When the lightning is very close, but not a direct hit, electronic equipment can be damaged by the large voltages induced on the electrical cables. The longer the cable, the more voltage gets induced. This means that two types of cabling are particularly affected, as they tend to be long: The 12V / 24V + and – cables to supply all devices with power, and the NMEA 2000 bus connecting all the devices. For the supply lines, some surge protectors exist, e.g., as offered by Dehn (German, English), and also discussed elsewhere (German), but only very few of these protectors work for DC and if so, they are specified for higher voltages. As to protecting the NMEA 2000 bus, virtually nothing is offered.

However, in my 5 lighting events so far, I could see consistently that the bulk of the damage was done via the NMEA 2000 bus, but only for those devices that were also connected to the 12V / 24V supply lines. The only exception is for wind, but it can be argued that it also has two inputs – the NMEA bus and the very long wire coming from the wind sensor. Usually, the NMEA 2000 chipset in devices like autopilot, chart plotter, or VHF radio got fried, whilst the device itself would still work as a standalone. (Well, with the autopilot there was no telling… 😉 ) However, devices that were only connected to the NMEA bus usually survived (except for the wind sensor).

In part, this may be due to my having somewhat reduced the negative impacts on the 12V / 24V cabling. There, given that I was far away from Europe at the time, I had not installed these professional surge protectors as offered by Dehn, but instead, I had opted for an easier and less invasive approach: Wherever I could access a 12V / 24V cable, (preferably + and – together), I added ferrite clips around them. Lots of them! These are not as efficient as proper surge protectors, granted, far from it, but they do damp sudden voltage changes and thereby give the protection circuitry in the devices more time to react.

But this approach obviously does not work for the NMEA 2000 bus, since these ferrites would smooth the data signals on the bus, making them unreadable in the extreme.

So, what to do instead?

A good friend of mine reminded me that on the physical level the NMEA 2000 bus is identical to the CAN bus used for many decades in the automotive industry and, more recently, also in Smart Home. Searching the internet we found two candidates that were offered as surge protectors for the CAN bus as used in Smart Home:

CAN-UES2

BlitzART

In essence, these devices offer multiple-stage protection with gas discharge, diodes, and ferrites. In the end, I decided to go for the CAN-UES2, mostly because in its technical circuit drawing I could identify all CAN bus cables. Here is a link for the NMEA 2000 cabling signals and colour codes. It is very simple, really: CAN_L is bLue, CAN-H is wHite, +12V is red, and GND is black.

Where do I need to install these surge protectors? Ideally, every device on the NMEA bus gets a surge protector in its drop line.

Unfortunately, it is a bit tedious to install these surge protectors, as they do not come with NMEA 2000 cabling. So, what one has to do is cut a drop line into two pieces and insert the surge protector in the middle. One has to make sure that the correct side of the surge protector is facing toward the bus, as it is not symmetrical. Also, one will want to make sure that the shielding metal sleeve of the NMEA cable is routed from one side to the other. And since the gas discharge will induce a lot of voltage in its neighbourhood when triggered, one will also want to have the surge protector in a metal case, so that nearby devices are less affected.

Additionally, I have split the NMEA bus into three segments, connected with galvanically isolated active ‘NMEA to NMEA converter’ bridges, where each segment is powered separately with a DC/DC converter, also with input and output galvanically isolated. This will reduce the induced voltage within each segment, simply because they are shorter. One power supply is located underneath the nav table where the majority of all the electric stuff is, another one next to the mast box, and the third one is located at the helm station. The Yacht Devices bridges were not quite straightforward to set up. One the plus side, I now have a reliable depth reading at the helm, which I often did not have — the helm is too far away from the depth sounder and the sounder may also be sending only weak signals on the bus, resulting in periods of time where the depth reading was not available at the helm — usually, when I needed it most. This had been a problem for years and now this is fixed, as the bridges repeat the signal and thus effectively boost it. That is good. What is not so good is that these bridges seem to introduce some fast packet errors on the bus. Initially, I also had problems that, e.g., the wind data would not show up on the Triton display unless the chart plotter was also on. In the end, this got resolved by temporarily connecting the segments 2a and 2b and performing a manual assignment of the sources on the Triton display. And finally, it should be noted that it seems to matter on which segment the master of a bridge is placed. The diagram below shows how it works for my network. Switching master and slave caused problems.

Should a bridge be killed in a lightning and I do not have a spare one with me (which I do 😉 ), I can always remove it and connect the two segments with a short patch cable.

Some cautionary notes: Do not have one DC/DC converter feed more than one segment, as this will kill the galvanic isolation offered by the bridge. Also, do not be tempted and deploy instead of the galvanically isolated active bridge a simple T-piece which only cuts the power, but lets the bus signals through. This would create an imbalance between the signal lines and the power lines on the bus, allowing large voltages to be induced between these two. Likely, this would mean that all those devices on the bus that only connect to the bus, but not to a separate power supply — which so far had been fine in my installation — would now become vulnerable. So, there is a risk of making things worse this way.

And finally, when applying these changes to the installation on my boat, I noticed that the input and output GND lines of the 24V/12V converters for the bus were actually connected with each other. That does not seem to make any sense, given that those are special, pricy DC/DC converters with isolated input and output — and why then go to all the trouble and expense, as well as accepting a reduced efficiency of these converters, only to reconnect the input and output in the end again? Does not make any sense to me, so I cut that connection. And whilst I was on this, I did the same for a couple of other DC/DC converters, like for the Fusion radio. The protection against lighting offered by galvanically isolated input/output of DC/DC converters is gone when connecting their GNDs on the input and output side…

All this will not help, anyway, when hit directly by lightning. But as said, this seems a much less likely scenario.

Incidentally, if you are interested in testing methodologies for CAN bus interfaces, have a look at this TI document. It talks about essentially the same protection elements as used in the surge protectors above. These should be built into the NMEA ports of our devices, but this does not always seem to be the case. For instance, my B&G autopilot features a separate domain for the NMEA bus circuitry, talking to the main board via opto-couplers, which is great. It also features thyristors (but only two, not three), but there are no gas-discharge tubes seen anywhere.

Now, this will take care of the NMEA bus, but as mentioned, the main risk seems to be for devices that have an NMEA bus connection as well as a separate power supply. So, we need something similar for the power supply of those devices.

It goes without saying that all these protections should be placed as close as possible to the device to be protected. Also, very important is that any surge-protection circuitry for the power supply of a device uses the same reference point for PE as the surge protector for its NMEA bus does. After all, it is relative voltages that matter.

The VHF radio as well as the stereo radio are both connected to 24V/12V converters with galvanically isolated input/output and thus may already be OK as far as protecting their power supply is concerned. They do need bus protection, though.

Finally, I have placed a Dehn surge protection unit identical to the one for the autopilot next to my main electrical distribution board where the master on/off switch is, and all the massive 100+ A fuses are. A much smaller variant of that is placed underneath the nav table where the distribution hub for all the electronics is and where the majority of all the DC/DC converters are placed.

Finally, a word on debugging methodology in case you do experience problems with your NMEA bus. If your bus is down, so you cannot see any devices from the chart plotter, say, then this usually means that there is at least one device on the bus that has its NMEA bus port damaged — for whatever reason. The best way to find this culprit is to split the NMEA bus into shorter segments, if possible (i.e., segments, which have their own power supply and are terminated at both ends with 50 Ω). In each segment, you will need to have a chart plotter or some other smart display from which you can inspect the bus status and get a list of all devices found on the bus. If you have only one power supply to the bus, then you will need to tackle the entire NMEA bus in one go. In the next step, you disconnect about half the devices on this bus segment and then power up the segment again. Have a look at the device list on the chart plotter and see whether it is still empty. If yes, disconnect another half of the remaining connected devices and check again. Repeat this procedure until you can see the device list populated with the names of the devices still connected. The bus is working now and the culprit must be one of the disconnected devices. So, start connecting some of those last devices again, just a few, and check whether the device list is still populated. Continue doing this until the bus is off again and no devices are seen anymore. The culprit device is among the devices added in the last step. So, disconnect half of them again and check again whether the bus is working or not. By trying out different devices of this last batch you can work out which device is the culprit that makes the bus go belly-up. Once you have found it, keep it disconnected and continue adding the remaining devices. Check every so often whether the bus is still working, as it may well be that there is more than one culprit device. In the end, you will have all culprits disconnected and the bus is still up, so you know which parts to order as replacement. Be mindful, though, that lighting-induced damage may take months to show up in a device. Initially, the device may still be healthy, but it will gradually degrade and eventually fail. So, in case of lighting damage, it could be that the above procedure needs to be repeated after half a year or so.