NMEA 2000 Cable: The Cathedral, the Bazaar, and the Evil Snake on Your Boat

Everything on your boat talks to everything else, until one day it doesn’t. A display goes blank for four seconds. A depth reading freezes. An autopilot drops its heading source and finds it again before you have finished swearing. You blame the instrument, then the brand, then the man who installed it. You almost never blame the cable — because nobody has ever told you what an NMEA 2000 cable actually is. There is a reason for that, and it is not a technical one.

What We All Know, More or Less

Most of us can give a rough account of NMEA 2000, and most of us are roughly right.

It is not a marine invention. Underneath, it is CAN — Controller Area Network — the bus developed for the car industry for “real-time engine and transmission control, anti-skid breaking systems, and to replace wiring body components.”^[1] On top of CAN sits the protocol that everyone half-remembers as “the SAE J-something”: SAE J1939, the CAN-based network the Society of Automotive Engineers developed for trucks, buses and trailers, later adopted by the International Organization for Standardization as ISO 11783 for agricultural machinery.^[1] NMEA 2000 is built on J1939 and defines its own messages.^[2] It runs at 250 kbit/s, which buys a backbone of up to 200 metres.^[1][2]

The name suggests the year, and the name is very nearly honest. The NMEA 2000 Standards Committee began work in 1994; after five years of development the standard went through an eighteen-month beta test; and the National Marine Electronics Association “completed and released the NMEA 2000 Standard for Serial-Data Networking of Marine Electronic Devices during October 2001.”^[1] It was later standardised internationally as IEC 61162-3.^[2]

Even the plug is borrowed. The committee did not design a marine connector; it went shopping. “The Open Devicenet Vendor Association’s physical layer cable and connector specification was chosen since it was a proven and robust specification,” the association’s own paper records — “a standard for cables and connectors for automotive, robotic and other terribly rough environments.”^[1] The smaller of the two sizes, “Micro”, is the M12 five-pin barrel connector specified in IEC 61076-2-101^[2] — the same M12 that bolts sensors onto factory floors in every industrial estate on earth.

So: an automotive bus, wearing an industrial-automation plug, adopted by the marine trade in 2001. That is the modern way of wiring a boat, and there is nothing wrong with any of it. It is a good standard. Everything that follows is a story about who is allowed to read it.

The devil is in the detail. And the detail is lying inside your boat at this moment, behind the panelling, under the sole, running from one end of her to the other, in the form of a black snake — one that, in all probability, has nothing to do with NMEA 2000.

The Certified, and Everybody Else

On paper the rule could not be firmer. Certification is not a badge you may collect if you feel like it; the association’s own position is that “any device providing a NMEA 2000 interface must pass the self-certification test before that product may claim that it offers NMEA 2000,”^[1] and, more bluntly still, “All products that read and or transmit NMEA 2000 information must be certified by NMEA.”^[3]

Cable has its own approval list. On 4 September 2026 the association’s certified-products database, filtered to NMEA 2000 Approved Cabling, returned ten manufacturers across thirteen listings:^[4]

• Actisense
• Amphenol LTW Technology
• Garmin
• KUS Americas
• Maretron
• Oceanic Systems UK
• Outdoor Solutions Electronics
• Phoenix Contact
• Shenzhen Element Automation
• Turck Inc. USA

Ten. For every boat in the world.

Now notice what that list does not tell you. The cable in the bilge of most cruising boats came in a box carrying one of the big instrument brands, or no brand at all — and the list cannot tell you who actually made it. A branded cable may well have been manufactured by one of those ten and sold under somebody else’s name; that is ordinary practice across the whole of manufacturing, and it is invisible from outside the box. What the list gives you is ten names you can check. What it cannot give you is any way to trace the cable in your hand back to one of them.

And there is something stranger still. Most of the instrument suppliers do not use the standard cable at all. They ship their own: Raymarine sells SeaTalk ng and then sells you an adaptor to get from it to “NMEA 2000 (DeviceNet)”;^[13] Simrad sells SimNet and then sells you a SimNet-to-Micro-C cable to get back out of it.^[14] Underneath, it is the same bus. On top, it is a different plug, and a small industry of adaptors to undo it — usually very expensive ones.

The easy explanation is the commercial one, and everybody reaches for it first: own the connector, own the accessory sales. But it is not the whole explanation, and it may not even be the main one. There is a good reason a manufacturer might not want your network held together by whatever five-pin cordset you found online — and it is the same reason this article exists. Keep reading; it will be obvious by the end.

Nor is this new, and nor is it confined to cable. When Panbo audited the certification database in 2011, Ben Ellison found that of Raymarine’s range “no C Wides, no autopilots, no instruments” appeared; that for Simrad and Lowrance “neither the NSE/NSO/NSS nor the HDS Series are certified”; and that Furuno listed six certified devices in total — “three of the Big Four NMEA manufacturers.” His conclusion is the one worth keeping: “Certification is not a guarantee of interoperability.”^[5]

And he is right, which is precisely why none of this looks like a problem from the cockpit. The gear works. You plug the yellow thing into the blue thing and the depth appears on the screen — for years, on thousands of boats, with hardware that never went near a certification tool. The standard is good enough that violating it politely still mostly works.

Mostly. And Then the Snake

Sometimes it goes bad, and nobody can say why.

The fault is never the clean kind. It is the four-second dropout at 0300. It is the wind instrument that disappears only when the engine is running, or only when the shore-power charger is on, or only on port tack, which cannot possibly be a real electrical category and yet there it is. It is the fault that goes away when you wiggle something, which is the worst kind of all, because a fault that answers to wiggling has already told you it is physical and you will still spend the winter buying electronics.

So we point at the instruments. Then at the brand — their gear never plays nicely. Then at the installer. We take the display out. We send it away. It comes back with a note saying no fault found, which we resent, and which is true.

Sometimes the devil has the shape of a snake.

And a cable was a snake to begin with. It lies coiled in a locker until you need it. It goes in behind the panelling, through the bulkheads, under the sole, and then it runs the whole length of the boat in the dark — touching everything, visible nowhere. Once it is in, nobody looks at it again for ten years. Of all the things aboard that could be quietly wrong, it is the only one you have deliberately hidden.

And the temptation it offered was the oldest one there is, and it was not even elaborate. It looked exactly like the thing it was pretending to be. It fitted. It was a quarter of the price. Nobody in the transaction told a single lie — the seller because he was never asked, the buyer because he did not know there was a question. You fed it through the bulkhead with your own hands and felt clever about the money.

Then it lies there, being a cable, doing nothing wrong that anyone can point to. And when the bite comes it leaves no mark: no burnt smell, no blown fuse, no error code, nothing to photograph and nothing to send back. Only a screen that goes blank for four seconds at 0300 and is perfectly innocent again by the time you have found your glasses.

The Cathedral: A Standard You Are Not Allowed to Read

Here is the part nobody tells you, and here is why nobody tells you.

The NMEA 2000 standard is eight documents — a main document plus appendices covering the application layer, the message database, the certification criteria and test methods, application notes, and the ISO and CAN layers underneath.^[1] You cannot download them. You cannot even buy them the ordinary way: “Due to copyright and licensing, the NMEA 2000 Standard is not sold via an online store,” the association states, and “all Marine Industry NMEA 2000 products must be certified based on the NMEA 2000 License Agreement.”^[3] The association claims copyright over the standard and its full contents are not publicly available.^[2]

This is where NMEA 2000 parts company with almost every other standard you will ever meet. ISO, IEC, ASTM and the rest all run web shops. You pick the document, you pay — often a great deal — and you download a PDF. The price is the barrier, and once you are past it the knowledge is yours: you may read it, build from it, teach it, quote it, argue with it in public. That is what a paywall is. It rations access; it does not impose silence.

There is a particularly odd consequence here. IEC 61162-3 — the international adoption of NMEA 2000 itself — is sold the ordinary way, from the IEC web shop, to anybody with a credit card: 355 Swiss francs — about $440 — for the 54-page consolidated version.^[21] If you build marine electronics for a living, that is an obvious purchase and you make it without thinking. If you own a boat, or run a chandlery, it is not — and nobody would expect you to. Which is exactly the point: the two people holding the cable at the moment the decision gets made are precisely the two for whom buying the document makes no sense at all. The same technology, wearing an international number, behaves like every other standard in the world. It is only the NMEA original that cannot be bought online, and only the NMEA licence that asks you to keep quiet about what you have read.

What you buy is a licence, and the licence comes with a vow. As Peter Hayden set out in 2016, quoting the agreement itself, the standard “is confidential and not accessible to anyone who does not hold a license,” and “the licensee of N2K is obliged to hold the standard and all associated documentation confidential.” The licensee may not “transfer, sublicense, lease, lend, rent or otherwise distribute the NMEA 2000® Standard to any third party,” nor “make the functionality of the NMEA 2000® Standard available to third parties.” His summary is hard to improve on: it works “like a secret society. Nobody knows what goes on inside, and as a contingency for being let in on the secret, you have to agree to maintain the secret.”^[6]

Let us be careful here, because the easy conclusion is the wrong one. The standard is good. It has held a whole industry together for twenty-five years on a bus designed for lorries, and it works. The certification programme is serious. And the confidentiality is not a racket either: the priest is the custodian of the Verb, and he keeps it in the confessional, as he should. Standards bodies have run this way for a century, and the man guarding the text is doing precisely the job he was given.

The trouble is not that the Verb is kept. It is that it cannot be preached by anyone who belongs to the church.

What fills a vacuum like that is a bazaar. Not a conspiracy — a bazaar: a large, cheerful, entirely rational market in things that fit. The connector is an industrial standard that predates the marine use and is manufactured by the tens of millions for factories.^[1][2] Anyone may build an M12 five-pin cordset. Anyone may photograph it against a blue background and write “for NMEA 2000” underneath. And because no buyer can read the standard, no buyer can contradict it.

So we buy whatever is put in front of us. It looks like an NMEA 2000 cable. Often, in the ways that matter, it is not.

Why We Can Tell You This

A fair question, and it has a short answer: Galvanic Works is not — not yet — an NMEA member. No licence agreement binds us to hold anything confidential. Not yet, at any rate — we are still weighing up whether joining is the right thing to do.

That is an expensive position, and we should be straightforward about what it costs us. Some people will not buy our products because they are not NMEA 2000 certified. That is an entirely fair position for them to take. A certificate is a real thing, it means somebody checked, and we do not have one.

What we would put beside it is this: not being inside the cathedral has never meant building to a lower standard — if anything it obliges us to the opposite, because we have no certificate to hide behind. So we do whatever is needed to build to the highest standards we can hold ourselves to: galvanic isolation, high-voltage protection, the best electronic components we can source — parts specified to the most stringent standards their makers publish, rather than the cheapest that would pass — and the most demanding industrial norms we can find to design against. The boat does not care whose logo is on the paperwork. It cares whether the hardware survives what a boat does to hardware.

And where a standard can be read, we read it. We have read IEC 61162-3 — NMEA 2000 as the IEC publishes it — in full. It sits on the shelf with the others we design against: IEC 60945 for marine equipment, IEC 62923 for alert management, EN 55032 and EN 55035 for emissions and immunity, the IEC 61000-4 series and EN 61000-6 for the electromagnetic environment, EN 62368-1 for safety, IEC 60664-1 for insulation coordination and clearances, IEC 60529 for ingress protection, EN 18031 for radio-equipment cyber security, EN 62311, EN 62479 and EN 50663 for RF exposure, IEC 62133 for cells, IEC 60073 for indication and colour, ISO 2631 for the motion a hull puts through a human being, IPC-2221 and IPC-6012 for the boards, and AEC-Q101 for the ESD protection parts. Every one of them can be bought and read by anyone who wants to check our homework. We elect to hold ourselves to the most demanding of them we can find.

And the position is liberating in other respects. It means we work the way everyone outside the cathedral works: from the public standards underneath (CAN, J1939, ISO 11783, IEC 61076-2-101, IEC 61162-3), from manufacturers’ published datasheets, from installation guides written for professionals, and from listening to real boats and reverse-engineering the PGNs and messages that cross them. Everything in the rest of this article is drawn from documents anybody can obtain, and every one of them is cited at the bottom so you can go and check us.

The Catch-22 That Kept the Sermon Unpreached

Now look at the trap that sets, because it is the real reason you have never read this anywhere.

Everyone who knows the standard in detail has undertaken not to describe it. So the only person free to explain what an NMEA 2000 cable actually is, is somebody who is not inside the standard — and being outside the standard means being uncertified.

Which makes the explaining an act against the explainer’s own interest. Everything that follows in this article argues for the standard: that it is a good one, that its certification is real, and that the part which carries it is the part to buy. We make that argument holding no certificate ourselves. And by raising the subject of non-compliance at all, we invite the obvious question straight back at us. Does the Galvanic Voice have compliance gaps of its own, of the same kind we are about to describe in these cables?

It may. It may not. We are not certified, and we are not going to pretend the question stops politely at the chandlery door. It applies to us exactly as it applies to anything else on the bus that has not been through the certification process.

We are not the first to arrive at this, and it is worth saying so. CANboat is the best-known open-source NMEA 2000 project in the world, and its README states the position plainly: the database “is copyrighted by the NMEA… Access is restricted to members and parties that pay for it. If they do so they are not able to divulge the content of the database, thus making it impossible for open source developers to get access to it. For this reason we have reverse engineered the NMEA 2000 database by network observation and assembling data from public sources.”^[19]

Its author, Kees Verruijt, put the trap in one sentence in a Free Software Foundation interview in 2016: “I could join the NMEA and buy the NMEA 2000 standard… but then I couldn’t actually release my program as free software. So I figured that the only way out was to reverse engineer the protocol from scratch.” He was careful to add what we would also say: “I’m not trying to subvert their business model here, just offering an alternative.”^[20]

So the people who could speak are not allowed to, and the people who are allowed to speak indict themselves by speaking. Silence is the comfortable choice for everyone in the industry, which is precisely why the Verb has stayed so well kept for twenty-five years.

We owe you honesty, and this is it.

Paradoxically, a tribute to the standard we do not adhere to.

But because we do not adhere to it, we can tell you.

The Details: Micro-C, and Its Twin

To see how close the two worlds are, put the construction of the two side by side: an M12 five-pin cordset that carries NMEA 2000 approval, and an M12 five-pin cordset built for general industrial fieldbus use that does not. Both descriptions below are taken from published manufacturers’ datasheets.^[7][8]

It seems identical. Every row a buyer could reasonably check says the same thing on both sides, and the only line that differs is a line about paperwork. Both of these are serious products from serious manufacturers, sold at broadly similar prices; nothing here is a bargain and nothing here is a fake. They are simply the same recipe, cooked for two different kitchens.

And it goes on seeming identical after you have bought it. It looks the same in your hand. It fits the same. You wire it in, the network comes up, every device answers, and nothing anywhere tells you a thing.

It is not the same cable. It is physically and electrically different.

Which ought to be a comfort, and is not. If the copper is the same, the difference that matters cannot be in the copper. It is in something the datasheet does not show you at all, and we will come to it.

Now the part that decides what actually ends up on your boat, and it is not that comparison at all. Go online and buy “an NMEA 2000 cable” and you will not be offered either of those two products. You will be offered a third thing entirely — and the sellers do not even hide what it is. Here are two of their own product titles, verbatim: “NMEA 2000 N2K Backbone Drop Cable IP67 Micro-Change M12 A Code 5 Pin Male to Female Extension Cable For Garmin Lowrance Simrad B&G Navico DeviceNet CANOpen Can Bus Actuator Sensor”, and “NMEA2000 Backbone Drop Cable… IP67 Waterproof — Certified Universal Compatibility”.^[18] One cable, advertised to the marine market and the factory-automation market in a single breath, with the word “Certified” used as decoration.

And at the factory gate, that is where the money is. Ask a supplier to build the cable properly — to the construction the standard describes — and the quote comes back at roughly four times the price of the lookalike sitting beside it in the same catalogue.^[12] Not four times an approved cordset from a serious manufacturer, which costs what serious cable costs. Four times what the thing that looks like it costs.

Because the same ingredients do not make the same recipe if you do things differently. Everything in the table above is a list of ingredients, and the ingredients really are the same. What a list of ingredients cannot tell you — and what nobody tells you — is how the pins are connected.

That is the good news and the trap in a single line. The good news is that the industrial part is a serious product built to a serious specification, which is why the bazaar mostly works. The trap is that once “an M12 five-pin cordset with a drain” is the whole definition in the buyer’s head, everything with five pins qualifies — including the parts where the drain does something else entirely.

And the subtle differences matter. That is the entire reason they are in the standard. A shield is not decoration and a pin assignment is not a convention — they are the difference between a network that shrugs off your alternator and one that does not, and you will not find out which you own on a calm afternoon at the dock.

Which hands you a test you can apply without reading a single datasheet, and without anybody’s permission: the price. At a factor of four, the real thing is not something you buy by accident, and it is certainly not something you pay for without noticing. So if you have bought a cable labelled NMEA 2000 and the price did not raise your hair, you were almost certainly not buying an NMEA 2000 cable. Most of us never have. Which means most of us have something else lying in the bilge, doing something else with the shield — and that difference is anything but cosmetic, as the next three sections are about to show.

Pin 1: The Pin That Should Not Connect to an Ordinary Wire

Five pins, and four of them are ordinary wires. The fifth is not supposed to be. The assignment is fixed and it is the same on Micro-C and Mini-C:^[1][7][9]

Two rows differ. The connector, and pin 1. And they are the same difference, seen from two ends.

In the approved cable, pin 1 is the drain wire. The shield runs down it, connector to connector, tee to tee, the whole length of the boat, to the one point where it is grounded.

In the other cable, pin 1 is just a wire. It joins pin 1 at one end to pin 1 at the other and it carries the ground, connector to connector, tee to tee, down the whole network, exactly as it appears to. What it does not carry is the shield. The foil has been bonded to the connector shell instead — and a shell grounds a shield only if the shell itself is grounded, which on a network whose tees are moulded plastic it never is.^[7][11]

So the ground runs the length of the boat, and the foil lying a millimetre away from it never touches it. The shield floats.

Pin 1 carries neither data nor power, which is exactly why it is so easy to ignore. It is the pin the drain wire lands on, and the drain wire is what ties the foil to a reference. The foil is what shields; pin 1 is how the foil gets its connection to ground. Maretron’s installation guide states the job plainly: “The drain wire shields the signal, power, and ground wires from external Radio Frequency Interference (RFI) and helps reduce RFI emission from the cable.”^[9] So the foil shields only if that connection exists: continuous along the whole run, and made at exactly one place. The association’s own barrier-strip diagram labels the SHIELD terminal at the backbone termination with three words: “No Other Connections.”^[1] Maretron puts it in capitals: “The NMEA 2000® network should be grounded at ONE location… In addition to the ground wire, connect the drain or SHIELD wire at the supply ground location and NO other place.”^[9]

Why the Ground Is the Whole Point

It is worth being exact about what a shield is, because “shielded cable” sounds like a property of the cable, and it is not. It is a property of the installation.

Grounded, the foil is a screen. Interference arriving from outside — the alternator, the inverter, the radar cable somebody ran alongside it in 2013 — lands on the foil and is carried away to ground instead of arriving on the signal pair. The foil takes the blow and puts it somewhere. That is the entire mechanism, and it needs somewhere to put it: a ground, at one point, once.

Ungrounded, that same foil does not become neutral. It becomes a long conductor lying hard against your data wires with nowhere to send what it collects. It floats. It picks up whatever is going past and holds it, and at the frequencies its own length happens to favour, it resonates — at which point it is no longer screening the signal pair. It is broadcasting into it, from a distance of half a millimetre.

Every sailor already knows this shape. A sheeted sail is a wing. Trimmed, made fast at both ends, it takes the wind and turns it into something useful and quiet. Let the sheet go and the cloth has not become harmless — it has become a flogging sail: the same material, doing the exact opposite of its job, shaking the whole rig, audible from the next boat, beating itself to pieces.

An ungrounded shield is a flogging sail lying against your signal wires. It is not a shield that has stopped working. It is a shield working in reverse.

Which brings us to the thing that is genuinely, physically invisible. An M12 shielded cordset can terminate its shield in two completely different ways, and industrial catalogues sell both as ordinary stock — the distinction is written into the product names themselves, “shielded to coupling nut” against “not shielded to coupling nut.”^[10] One lands the drain on pin 1. The other grounds the shield to the outside of the connector — the shell and its coupling nut — and leaves pin 1 — wired from end to end, exactly as it always is — bonded to nothing. Outside grounding. That is the difference the datasheets do not show you, that no amount of comparing foils will reveal, and that the whole rest of this article is about.

To be precise about what does not differ: in both cables the bare drain lies against the shield along its whole length, and that contact is the connection. Neither construction asks for a separate bonding step, and no installation instruction tells you to make one. The whole difference is where that shield is terminated at the connector: on pin 1, or on the shell.

So here is the secret, and it is one sentence long. The main difference is that in one case the shield is attached to pin 1 — and in the other case it is not. Whether that is the only difference, we are in no position to tell you: we have not read the standard, and we are not permitted to. It is certainly the one that matters, and everything else in this article follows from which of the two you happen to have bought.

From the outside, with the cable in your hand, the two are indistinguishable. And I will admit it plainly, because there is no point pretending otherwise: it is hard even for me to tell which one I am holding without stripping the insulation back and going at it with a multimeter — put one probe on the shield and find out whether it answers at pin 1 or at the connector. There is no marking on the jacket that will tell you. There is no way to see it. You either take the cable apart, or you take somebody’s word for it.

Metal or Plastic: The Wrong Argument

We all tend to think that a metallic connector must be better shielded than a plastic one. It is a perfectly reasonable instinct — metal shields, plastic does not — and almost everywhere else in electronics it would be the right one. You pick up the heavy one, you feel the machined nut, and you conclude you are holding the serious cable.

Here it is wrong, and it is wrong for two reasons nobody tells you: the tee it screws into is plastic, and the standard carries the ground through pin 1, not through the connector.

Sometimes the connector is plastic. Sometimes it is metallic. It settles nothing either way — because in both kinds of cable, pin 1 runs through from one end of the cordset to the other. Put a meter on pin 1 at one end and pin 1 at the other and you will get continuity whatever you bought. That test tells you nothing at all. The question was never whether pin 1 is connected end to end. The question is whether pin 1 is connected to the shield.

And the tees — the parts that actually join the network together — are plastic through and through, including the approved ones. Maretron’s Micro Tee: moulded body thermoplastic PUR, contact carrier PA 6 nylon, contacts brass/gold, coupling nut brass/nickel.^[7] Actisense’s four-way connector: housing TPU, flame retardant to UL94V-0 and halogen free, contact carriers PA66 with 30% glass fill.^[11] Walk any boat and you will find the same thing: the trunk is held together by plastic.

Which settles the argument, and not in the direction anyone expects. In a tee, the connector is never attached to ground and never carries the shield across to the other cables. There is exactly one continuous electrical path for a shield from one end of that backbone to the other, and it is pin 1.

So a cable that grounds its shield to the connector cannot transport that ground to the next segment. It stops at the first tee it meets. And since a tee is what every single device on the boat hangs from, the shield is broken at every device — every tee, every drop, every one of them a break. Not one long shield running the length of the boat to a single ground, which is what the standard asks for, but a row of disconnected stubs, each one ending at a piece of plastic.

None of this shows up on a multimeter check of the network. Power is there. NET-H and NET-L are there. The bus talks. You have simply been running an unshielded network past your alternator, your inverter and your radar cable, and you will find out on the night it matters.

Why It Works Anyway, Nearly Always

We should stop here and give NMEA 2000 its due, because an article about one failure mode can leave the impression that the thing is delicate. It is the opposite. The bus on your boat is among the most robust engineering aboard her, and the reason it is robust is that it is gloriously, deliberately slow.

It runs at 250 kbit/s.^[1][2] Set that beside the rest of the boat. Fast Ethernet is four hundred times quicker. The gigabit socket on your laptop is four thousand times quicker. Even the ageing Wi-Fi 5 access point in the saloon — the ordinary two-stream sort that shipped in every router for a decade — runs at 867 Mbit/s, some three and a half thousand times quicker.^[15] In an industry that sells speed, the marine standard went slower than anything else on board — and that was not an oversight, it was the entire design. Actisense puts the reasoning in a sentence: NMEA chose CAN over “something like Ethernet… which has a much wider bandwidth” because “CAN has the ability to guarantee message delivery even when the network is 100% loaded” — “an important feature when you consider that the network is likely to contain important GPS information.”^[17]

Slowness buys immunity, and it buys a great deal of it. At 250 kbit/s a single bit lasts four microseconds; on a 100 Mbit/s link a bit lasts ten nanoseconds. To corrupt a bit here, interference must hold its disturbance four hundred times longer than it would need to on the office network — and most electrical noise on a boat is a spike, not a siege. The standard’s own numbers show the bargain being struck: at 1 Mbit/s the network may be 25 m long, at 250 kbit/s it may be 200 m.^[1] Speed was traded away for length and for noise margin, knowingly.

Then it is traded away twice more. The signal is differential — NET-H and NET-L carry the same message in opposite senses, so interference that strikes both wires equally is subtracted out at the far end and never becomes data. And CAN underneath supplies frames with, in the association’s words, “robust error checking, confirmed frame delivery, and… deterministic transmission times.”^[1] A frame that is damaged is detected and sent again, and nobody in the cockpit ever learns it happened.

That is why the bazaar works. A whole network built from the wrong cable will run for years and give every appearance of being perfectly fine, because a bus designed to survive a lorry engine bay has margin to burn, and most boats never get round to burning it. Anyone who tells you the cheap cable “doesn’t work” is wrong, and the sailor who has run one for a decade without trouble is not a fool. He is a man with margin.

And Then the Boat Starts Making Noise

Margin is finite, and a boat is not the quiet place it looks like at anchor. Everything below is ordinary equipment, fitted for good reasons, and every one of it is a source the shield exists to deal with:

• the alternator, ripple rising and falling with engine speed, running the length of the boat on heavy cable
• the inverter and the battery charger, switching hard, all day, often on the same side of the boat as your instrument runs
• the bow thruster, the windlass and the electric winches — hundreds of amps for a few seconds, exactly when the helm most needs the depth and the wind to be right
• the autopilot drive motor, working continuously on passage, generally within a metre of the very network that steers it
• the fridge compressor and the pumps, cycling all night while everyone is asleep and the anchor watch is the only thing running
• MPPT solar controllers and a wind generator, chopping current continuously in daylight and breeze
• LED lighting with whatever switch-mode driver was cheapest that season
• the VHF or SSB transmitting, and the radar turning — deliberate, powerful radio energy, radiated on purpose, a few metres from a cable whose job is to ignore it
• shore power, and everything on the pontoon that shares it with you

Each of these on its own is nothing much, and a properly shielded network shrugs at all of them together. That is what the shield is for. But a network that has quietly spent its margin before the season starts — because the drain is grounded nowhere — meets that list with nothing in reserve. It will still work most of the time. Most of the time was never the problem. The problem is the ten minutes a year when the windlass is running, the engine is charging hard, the radar is on and you are coming into somewhere unfamiliar in poor visibility, which is precisely the ten minutes you bought the instruments for.

The Same Boat, Wired Twice

Put the two cables into a real installation and the argument stops being abstract. Same backbone, same tees, same devices, same terminators at each end. The only variable is which cordset came out of the box.

Built with cable that lands the shield on pin 1, the drain is one continuous conductor from one end of the boat to the other, grounded at a single point at the supply — and grounded once and once only, because “grounding at more than one location may produce ground loops.”^[9] That is a screen. It works.

Built with cable that grounds the shield to the socket, nothing changes that you can see, and everything changes that matters. The shield stops at the first plastic tee. Then at the next. What you have is not a screen at all but a row of isolated stubs, one per device, each of them ungrounded, each of them a flogging sail lying against the signal pair — and every one of them installed by somebody doing exactly what he was told.

This is the fault that has no name in any menu. Nothing is broken. Nothing is loose. Every connector is tight, every device answers, and the boat has been quietly unshielded since the day it was wired — waiting for the alternator to sing. Then a dropout at 0300, and a winter of blaming the display.

Nobody Missed This

It would be comforting to think the exposure was an oversight. It cannot have been. It falls straight out of two decisions that were both, on their own, plainly right.

The first was to take an existing industrial connector system rather than invent a marine one. The committee went looking for “a specification that would more than meet the rugged marine requirement for durability and safety” and adopted the ODVA physical layer because it was “a proven and robust specification… for automotive, robotic and other terribly rough environments.”^[1] That is how you get a connector made by the million, at industrial prices, already tested in places far less forgiving than a bilge.

The second was the topology: a bus that crosses the whole vessel in segments, joined by tees, so that any device can be hung off it anywhere, and any device can be unplugged “without affecting any other device.”^[9] That is what makes the boat extensible for thirty years instead of finished on launch day.

Each is a virtue. Put them together and you get this, necessarily: a safety-relevant network assembled from many separately bought pieces, using a connector the rest of the world also uses for something else. There was no third option. And there is nothing wrong with the outcome — done correctly, it works completely, which is exactly what the fleet demonstrates every day.

The catch is where the responsibility comes to rest. The yard wires the boat once, properly, out of one supplier’s box. Then the boat is yours. And from that day, every instrument you add, every device you move to a better position, every refit and every repair means one more cable to buy — and the choosing is now yours, or the yard’s, or the workshop’s, or the travelling electrician’s who fits your new transducer in a marina in August. None of them is inside NMEA either. None of them has read the standard, and none of them has any way of checking the cable in the van. The standard cannot make that choice for you. The tee will not check it. The network will not complain, on the day or for years afterwards.

If the cable is wrong, the mistake is yours: made in good faith, in a chandlery or a browser tab, in the name of not overpaying for something that looked identical.

“So How Do I Know If My Cable Is Good?”

Which is the question anybody sensible asks at this point, standing in front of a boat already full of cable that came from somewhere.

First, the weak indicator. If the connector is plastic, you have a better chance — because there is no metal shell there for the foil to be bonded to, so the shield has a better reason to be where it belongs. A better chance is not a certainty, and it is emphatically not a test. It is a hint, and the hint runs opposite to everybody’s instinct, which is the reason it is worth stating at all.

Then the only method we know of that actually answers the question. Take a blade to the outer jacket — carefully, somewhere in the middle of the run where it does not matter — and cut through the outer insulation until you reach the shield underneath: a foil, or a fine metallic mesh. Make a proper contact with it; foil is thin and it is easy to think you have it when you have not. Then test continuity between that shield and pin 1.

Your cable is now slightly damaged, and you have the answer. There is no version of this where you get the second without the first. That is the whole difficulty in one sentence: the one fact that determines whether your network is shielded is a fact you can only obtain by cutting the network open.

And there is a third answer, which is not a test at all. Buy into a system whose connector you cannot get wrong. Raymarine’s SeaTalk ng and Simrad’s SimNet do not use the NMEA 2000 connector^[13][14] — and there is no marketplace full of five-pin lookalikes for them, because there is no lookalike to make. Whatever cable is in one of those networks came from the manufacturer, built to the manufacturer’s specification. The mistake this whole article is about is simply not available to you.

Which puts the earlier question in a different light. We said the obvious explanation for a proprietary connector is commercial — own the connector, own the accessory sales — and that it might not be the whole of it. Here is the rest of it. A closed connector system removes, at a stroke, the one failure the standard cannot defend itself against. It is a crude answer to the problem, and an expensive one for the owner, who pays for every adaptor. It is also, by design or by accident, an effective one — and a manufacturer who has watched enough warranty returns come back from networks built out of whatever fitted might reasonably have concluded that crude and effective beats open and unverifiable.

The Solution, Which Is Disappointingly Simple

After all that, the remedy is one line long and there is nothing clever about it.

That is the honest end of the argument, and notice where it lands: on the side of the standard. The certification exists, it means something, and the cable that carries it does exactly what it claims. The trouble was never that NMEA 2000 asks too much of you. It is that almost nobody is in a position to tell you what it is asking. Not even the person selling you the cable: with the shield hidden inside the jacket and the connector moulded shut, he has no way of knowing which of the two he is holding either. He has the label, and the label is the whole of his evidence — exactly as it is the whole of yours. So a decision that should have taken ten seconds at the counter has instead been made, silently and wrongly, on a great many boats.

Buy the approved cable, ground the shield once, and the whole subject disappears for the life of the boat. That is not a bad deal against the alternative that merely fits just as well.

The Verb, Undelivered

There is no villain here, which is the frustrating part. The standard is good. The priest is honest and keeps his vow. The disciples respect the rules and stay silent. The certification programme is serious. The industrial connector is a fine piece of engineering. The bazaar sells competent parts at honest prices. And still the sailor stands in the middle of it holding two cordsets that are identical in every respect he is able to observe, and one of them will quietly unshield his network for the next ten years. Nobody put the snake in the garden. It simply grew there, in the space where the sermon should have been.

He is in that position for one reason only: the document that would tell him the difference is one he is not allowed to read, and the people who have read it have promised not to say.

A Verb kept faithfully in the confessional is still a Verb nobody outside has heard. And a congregation that never hears the sermon does not stop believing — it goes and buys its candles in the square, from whoever happens to be selling them, at the price that does not raise anybody’s hair.

A Confession, Since I Have Been Handing Out Sermons

What follows I write as a sailor and a customer, not as Galvanic Works. Everything above is the company speaking; this part is me, standing in the same chandlery as you, with the same wallet.

I did not know any of this. I should say so plainly, because it would be very easy to write an article like this one from the comfortable position of having always known, and that is not where I am writing from.

I have bought all the wrong cables. All of them, over years, with exactly one exception — and the exception was not judgement, it was luck. Every other length of cable I have ever bought for a boat I bought the way this article says everybody buys them: it had five pins, it fitted, the price did not raise my hair, and I was pleased with myself for not overpaying. And it works. That is the part I keep coming back to — it even works, most of the time.

And the person who finally educated me was not an engineer, not a standards body, not a manual, and certainly not the cathedral. It was a Chinese sales representative on Alibaba, who explained to me, patiently and without being asked, why he was in a position to sell me either the cheap cable or the expensive one — and what, precisely, he would be selling me in each case. He was under no obligation to tell me anything at all. He simply knew, and he thought I ought to know too.

And the reason one costs more, he explained, is not margin. It is work. The expensive cable has a real shield, and a drain wire that has to be found, dressed and landed on pin 1 — at both ends, on every single cordset. That is a different job from pushing five identical wires into five identical holes, and it is a different machine, a different operator and a different reject rate. There are further reasons besides, which we will not go into here.

So the sermon did reach me in the end. It reached me in the square, from the man selling the candles.

And the bazaar, today, is global. The square is not the chandlery down the pontoon any more; it is a browser tab, and the man in it is on the other side of the world. That turned out to be no obstacle whatsoever to his knowing more about how your boat is wired than anyone you can reach by walking. It is a strange arrangement — the cathedral silent, the square well informed and open all night — but it is the one we have.

So we do the only thing left to do with what we have learned: we take it out into the square and shout it. Not from the steps of the cathedral — we are not allowed up there, and the people who are cannot say a word. From the middle of the market, where the cable is actually bought.

Mind the snake in your bilge. It had five pins, it fitted, and it worked most of the time.

Pin 1 is always connected. Sometimes it is protecting the other wires. Sometimes it is not. Nobody tells you which — and your $5,000 autopilot, on your $1,000,000 boat, can be misbehaving for a $10 reason.

References

1. Spitzer, S. (Technical Director, NMEA), Luft, L. (USCG Research and Development Center) & Morschhauser, D. (Mystic Valley Communications). NMEA 2000® Past, Present and Future. RTCM 2009 Annual Assembly Meeting and Conference, St. Petersburg, Florida, May 2009. National Marine Electronics Association. (CAN and SAE J1939 origins; ODVA physical layer; Micro-C and Mini-C pin assignment; topology, terminators and powering rules; the eight documents of the standard; certification programme; October 2001 release and 1994 committee; standards-committee membership lists.)
2. “NMEA 2000.” Wikipedia. Retrieved 4 September 2026. (SAE J1939 basis; 250 kbit/s; IEC 61162-3; M12 five-pin connector per IEC 61076-2-101; copyright claimed over the standard and full contents not publicly available.)
3. National Marine Electronics Association. “NMEA 2000® Standards.” nmea.org/nmea-2000.html, retrieved 4 September 2026.
4. National Marine Electronics Association, certified-products database, filtered to NMEA 2000 Approved Cabling. web.nmea.org, retrieved 4 September 2026: ten manufacturers across thirteen listings.
5. Ellison, B. “NMEA 2000 certification, the elephants in the room.” Panbo, 23 June 2011. panbo.com
6. Hayden, P. “How ‘Proprietary’ is NMEA 2000?” Adventures of Tanglewood, 3 June 2016, quoting the NMEA 2000 License Agreement. mvtanglewood.com
7. Maretron. Micro Cordsets, Tee and Powertap Datasheets. maretron.com (PDF) (Micro double-ended cordset: thermoplastic PUR body and contact carrier, brass/gold contacts, brass/nickel coupling nut, 4×22 AWG plus 22 AWG drain wire, three-level foil shielding, HDPE data-pair and SRPVC power-pair insulation, 4.0 A, 250 V, IP68, NMEA 2000® Approved, IEC 61162-3; Micro Tee CM-CF-CF: thermoplastic PUR moulded body, PA 6 nylon contact carrier, brass/nickel coupling nut, 4.0 A, 60 V, IP67.)
8. Turck. RSC RKC 572-2M — Cable for DeviceNet and CANopen, Extension Cable (ID U0323), datasheet revision 21 February 2025. turck.us (PDF)
9. Maretron. NMEA 2000® Network Installation Guide. maretron.com (PDF) (Five-wire cable construction and the drain wire’s function; wire colour/name/usage table; maximum cable distances, drop lengths and cumulative drop length; 50-device limit; Powertap, end-powered and middle-powered networks; per-segment current limits; the 1.5 V rule; the single-point grounding requirement; common fault list.)
10. Mencom Corporation, DeviceNet and Micro DC (MDC) cordset catalogue, retrieved 4 September 2026: product families are named for their shield termination — e.g. “DeviceNet Drop, M12, Cordset, Shielded Cable, Not shielded to coupling nut, 5 Pole…” against “MDC, Cordset, Shielded to Coupling Nut, 5 Pole…” mencom.com
11. Actisense. A2K-4WT NMEA 2000 4-Way Connector. actisense.com (Housing TPU, flame retardant to UL94V-0 and halogen free; contact carriers PA66, 30% glass filled.)
12. Galvanic Works supplier sourcing, 2026: cable built to the NMEA 2000 construction is quoted at approximately four times the price of visually equivalent five-pin M12 cable offered for sale as NMEA 2000 cable.
13. Raymarine. NMEA 2000 (DeviceNet) (Male) to SeaTalk ng Spur (Female) Adaptor Cables, product page, retrieved 4 September 2026: “Adapts Raymarine SeaTalk ng instruments and displays to work on a NMEA 2000 DeviceNet network.” raymarine.com
14. Simrad. SimNet to Micro-C (male) cable and Micro-C (female) to SimNet cable, product pages, retrieved 4 September 2026. simrad-yachting.com
15. Bit-rate comparisons: NMEA 2000 at 250 kbit/s;^[1] Fast Ethernet (IEEE 802.3u) 100 Mbit/s; Gigabit Ethernet (IEEE 802.3ab) 1 Gbit/s; Wi-Fi 5 (IEEE 802.11ac) 866.7 Mbit/s for the common two-stream, 80 MHz configuration. Ratios are of nominal bit rates and are given for scale, not as a measure of delivered throughput.
16. Alicat Scientific. DeviceNet Standard Pinouts, DOC-PINOUTS Rev. 5, July 2023: 5-pin M12 — 1 Drain, 2 Bus (+) 11–28 Vdc, 3 Bus (−) common, 4 CAN_H data signal, 5 CAN_L data signal. alicat.com (PDF)
17. Actisense. The Ultimate Guide to NMEA 2000® Networking, Revision 4, 2021. actisense.com (PDF) (Adoption of the DeviceNet hardware standard; the choice of CAN over Ethernet for guaranteed delivery at full load; the distinction between “NMEA 2000 Compatible” / “CANbus approved” and “NMEA 2000 Approved”; wire colours and network diagnostics.)
18. Marketplace product titles for generic five-pin M12 cable sold as NMEA 2000, retrieved 5 September 2026 — quoted verbatim from the sellers’ own listings. The same cable is advertised simultaneously for marine (Garmin, Lowrance, Simrad, B&G, Navico) and industrial (DeviceNet, CANopen, CAN bus) use.
19. CANboat. CAN Boat — NMEA 2000 and NMEA 0183 utilities, project README, Apache License 2.0. github.com/canboat/canboat
20. Free Software Foundation, Licensing and Compliance Lab. Interview with Kees Verruijt of CANboat, 18 May 2016. fsf.org
21. IEC Webstore, IEC 61162-3 Consolidated version — Maritime navigation and radiocommunication equipment and systems, digital interfaces: 355 CHF (approximately USD 440 at the rate of 4 September 2026), 54 pages. webstore.iec.ch
22. Hero photograph: Câble NMEA 2000 dénudé by Loxyger, Wikimedia Commons, licensed CC BY-SA 4.0; cropped. commons.wikimedia.org