[Bit of a Rant] Is the NMEA Corrupt or Incompetent?

[moseriw]

Ah, and NMEA 2000 is just an addon to the common automotive used CAN Bus.

[goboatingnow]

Originally Posted by moseriw

Ah, and NMEA 2000 is just an addon to the common automotive used CAN Bus.

It’s not an addon , the protocol borrowed certain aspects of J1939 but that’s about it.

[Jammer]

Originally Posted by BrightSEALAB

1- In 1985, the NMEA released it's first standard aimed at unifying communications between multiple brands of marine appliances. At that time a robust and well tested system of networking computers and peripherals together was already in existence in the form of 10Base2/5 Ethernet.

Cabling an Ethernet system was remarkably easy. Using inexpensive flooded coax wiring anyone could lay wire at custom lengths and adding nodes/peripherals along the backbone was as simple as cutting a wire and compressing a couple of CATV fittings.

I worked with 10base2 systems back in the day and I can assure you that they were not an improvement in reliability or ease of use over NMEA2000. It used BNC connectors, not CATV fittings. You may be thinking of 10broad36, which used 75 ohm coax and CATV fittings, but which never caught on because of the high component cost for each device.

2- By 1995 Ethernet had evolved into an even faster, more robust, easier and cheaper network physical layer. 10BaseTX could transmit 100,000,000 bits per second between anything connected to a switch or router. New devices just needed a home run to the hub. Weatherproof cabling was common and waterproof connectors were available and inexpensive. Cabling a small business with dozens of devices had a cabling cost of $100 or so.

I was working with this stuff in 1995. Let's get the history of it right:
* It's 100baseTX, not 10baseTX. There were competing standards such as 100baseT4 and various proprietary alternatives. It was a new standard and it wasn't clear whether it was going to be another Betamax.
* We typically paid an installed cost of around $80 per drop for cabling (including labor), plus another $30 for a pair of patch cables (one for each end).
* Weatherproof cabling was available but was expensive and difficult to terminate. Then and now, there are waterproof connectors but they are not fully compatible with 8-position modular ("RJ45") plugs and sockets. The retail price today is $30 for each plug or socket; at my day job we buy thousands and pay a little more than half that. Most systems (then and now) use cable glands which are bulky, and which are not truly IP68 as installed in most cases.
* In 1995 the higher speed 100baseTX components were expensive driving the cost of compatible equipment and hubs up much as is the situation with 10GbaseT today
* This was before power-over-ethernet and so the ability of NMEA2000 to provide power was not present.

So what does NMEA do, but take the worst aspect of 10Base2 and make it the central 'feature' of NMEA 2000. And pile on top of that, NMEA decides to go with a backbone and drop topology using multi conductor wires that cannot be built to size on the site by the owner installer. One must assemble their network infrastructure from expensive premade 'certified' little bundles of wire and connectors... $18 for a 20" drop wire. $20 for a t-fitting. And even then, most of the device makers have put their own special connector on their stuff and likely aren't letting 100% of other manufacturer's data translate.

I can field terminate and splice NMEA2000. Doesn't mean you can. Most boaters can't field terminate something to a compression terminal with a captive screw, so the NMEA2000 system was designed with the idea that connectors were permanently attached in a controlled, production environment. I don't think that's a bad choice, because it makes for a more reliable system.


I would agree that the standard connectors are poorly chosen because they are too large to run through holes of reasonable size. This is the main reason manufacturers have chosen alternative connectors.

3- Now the NMEA rolls out NMEA OneNet, and surprise!, It's based on the ethernet standard that's been available for 38 YEARS!

NMEA OneNet uses M12 X-coded connectors, power over ethernet, IPv6, and 1GbaseT, none of which were available 38 years ago.


The components required to implement all this are, at this point, very expensive. The connector, the magnetics, the PoE, and the 1GbaseT transceiver together cost over $50 a set in quantity (1000s). A 3' long weatherproof patch cable in M12X is going to set you back about $40 in the same quantities.

[wholybee]

I think they did a good job. NMEA0183 electrically *is* rs422. Then the protocol (sentences) were specified to use with it. RS422 was, and still is, widely used industrially. It is very reliable, works with a wide range of cabling, in noisy environments at long distances. Because it is so widely used, the engineering for manufactures to implement was is done, and ICs readily available. And anyone in the electronics industry or computer industry understood how it works so it was easy to install.

I was using 10Base2 in the early 90's. That was far more expensive, and while the wiring was rather simple, overall it was very complex compared to RS422. It would have been a nightmare to implement and make work. Remember how many networking companies there were in the 90s and none of them worked with each other, even though they were all 10Base2?

NMEA2000 is based on CAN, again with the benefit that the industrial world has been using it, the bugs are mostly worked out, and people understand how it works. It makes even better sense because it offers connectivity to other equipment that finds it's way onto a boat, like engines and chargers.

I get paid to (among other things) fix networking issues. Ethernet is not simple, nor is it any better understood by the layperson than RS422 or CAN. At least I can explain RS422 to a layperson to a reasonable understanding in an hour or less. Most people can't make 2 computers in their own home share files to each other.

[muttnik]

Like many folk coming to marine electronics from a computing background, I had a certain "WTF?" to some of the approaches and especially the whole "pay to play" standards of the NMEA but I don't think the OP's hyperbole is helpful. "Corrupt" implies illegality or at least conscious immorality and is wholly unwarranted.

10base2 was inappropriate for the 1980s marine data comms use case. Consider the cost of transceivers, consider the cost of implementing drivers on constrained devices. More importantly...what protocol would you have run on top of that? An IP stack? For a 1980s boat network? Why is OneNet a decade late? A primary reason is the use of IPv6. From a computing standpoint that was completely the right choice but it ignored the skillsets of people in the marine industry and the availability of IPv6 network stacks for the sorts of devices the marine industry wanted to use.

As to why NMEA-2000 is persisting with the advent of OneNet, there's definitely a discussion to be had there but a sensible argument needs to address the "real time" rationale for why it will persist (at least for the time being) alongside OneNet.

Are there even any OneNet devices yet?

If "how good" a protocol is is measured in its longevity, NMEA-0183 was pretty good...

[CarinaPDX]

Agreed that "corrupt" is far too harsh of an accusation. And I see less of an issue with the NMEA standards than the individual manufacturers' choices to subvert them by using proprietary sentences, PGNs, connectors and extensions (e.g. RM's original SeaTalk is based on an early 0183 standard but wasn't upgraded AFAIK). By offering new features of interconnection within a product line with proprietary sentences or PGNs (e.g. selecting a target on the AIS or MFD that then makes a DSC VHF call) a manufacturer can get a lead on its competitors without giving away its idea early by having to jointly agree to them. And the high cost serves to discourage smaller companies (and even some larger ones) from getting NMEA-certified product status - hence NMEA-compliant products.

I am rather surprised that anyone has fond memories of the old ("thick-net" and "thin-net) ethernet coax wiring. There were so many failures resulting from IT "professionals" thinking they could install a connector correctly that troubleshooting a network meant checking the connectors first. It was a friggin' nightmare. By contrast Maretron's field-installable N2K connectors (proper DeviceNet spec) are about as simple and robust as one could ever imagine, with the wire connections color-coded to match the cable spec. It is just wrong that one has to use pre-made cables and run the cable with the connector on the end. Most of my connections use pre-made cables, but longer runs I make up in place. Of course it you must use the proprietary offerings then suffer away... Hint: Furuno has always used DeviceNet connections, and Maretron offers great DevideNet interconnect solutions.

Furuno has been using Ethernet connections between their radar scanners and fish finders and MFDs for at least 12 years (my MFD with radar scanner has it). They encapsulated N2K in the data stream so that things like GPS can be connected via N2K at the scanner. Unfortunately they did modify the ethernet standard a bit so care must be taken if using a non-Furuno ethernet hub. With an NMEA ethernet-based standard this shouldn't be a problem in the future, if the manufacturers actually drop their proprietary implementations and go with the OneNet standard. BTW I see N2K and OneNet coexisting quite happily as they really are doing different things.

Greg

[billgewater]

I am the troglodyte. My instrumentation is all independent. Nothing is shared except the power supply. OK it is clumsy but it's worked for me for over 40 years. Nearly all components have been replaced at different times except for the auto pilot installed about 1983. I use Open CPN via a small laptop with the CM 93 charts but I can't feed it to the autopilot so I lay chart lines on the screen and then adjust the autopilot to follow that direction. It's not perfect but it's OK.

[Lasseman]

Dont think they are corrupt or incompetent. But its time to move on for sure.

A new standard owner is needed, industries today want "real" standard organisations like IEC, IEE etc. Standard organisations under mandate from United Nations and not a "members by payment fees and regulations under nda only". NMEA might work today but its not future.

I have IT com background and i dont think IP based networks is future for marine needs. NMEA organisation IP based proposed OneWire standard will never take off. Its to messy, to big and no one is queing up to implent it.

I think marine electronics future is Ethernet! But Ethernet without IP. This is what industries is now implementing for similar use cases as marine needs. The standard is called TSN/AVB networking. Car industry has under IEC standardised 1GB one pair cabling togheter with TSN. Two (small) wires with 1GB troughput, sub 2us sync.

Audio/video industry is jumping all aboard on TSN, concerts running TSN Ethernet mixing audio with whatever in same fibers. Metallica, is one ambassador for this. Video guys also, superbowl was largest implentation.

Latest ASEA robots is using TSN, medical labs is using TSN networking.

Replacing CAN networks with TSN is already happening and it is going fast. NMEA next gen should follow same path.

Benefit of TSN, (replacing CAN networks)

Unified network infrastructure: same network cables are used to pass traffic of all types (legacy IP and TSN). IP traffic will be automatically rate-limited before it can ever disturb the TSN traffic.

Plug-and-play: no network management ever needed, no matter the size of your network. (No IP address management, No DNS, no routing rules)

Multicast transmissions: any number of devices can listen to the same digital TSN stream.

Low latency: minuscle latency is optimized automatically according to the network topology. Meaning the provision for the maximum time an TSN packet takes to go through the network from the talker to the farthest listening device. In the simple case where all your devices are connected to one single switch, latency can be reduced down to 250 microseconds. If 6 cascaded switches separate the talker from the farthest listener, the recived latency can be set to 1ms.

Automatic overload prevention: whenever a segment of the network is approaching its physical bandwidth capacity, TSN will automatically prevent any new stream to be added on this segment.

Determinism: performances of TSN are guaranteed, they are not the result of statistical phenomenon (IP address management/routing/DNS lookups); TSN either always works, or it never works. “It works” means TSN traffic are going through the network and reaching their destination on time. “It doesn’t work” means that you have no TSN traffic at all.

Synchronization: all devices listening to the same talker are able to use the received TSN data at the exact same time within +/-1us.

Think about cameras, radars plus traditional sensors beeing in within +/-1us with its data communication, over a oldschool Ethernet cable togheter with whatever other IP traffic you have, all in the same one (1) cable. That is the future.

There is however a future for NMEA to jump on this train. Becouse an marine touch is still needed. Someone needs to marinify TSN and common decisions needs to be made like:

• Common data format for the traffic. Mapping N2K group messages into TSN first gen1, while working on redoing/unifying N2K/0183 mess into an gen2 data packet application level groups.
• Plug-and-play synchronization of tsn clocks through the network itself.
• Inter-operable seamless redundancy.
• Unified control of TSN devices central controller software design.

My 2cent

[continuouswave]

NMEA is an organization that serves its members. Recreational boats are very unlikely to become members of NMEA.

There is a boater-developed communication protocol: SIGNAL K.

SIGNAL K solves all the complaints about NMEA and its legal and financial barriers to use and information about the protocol. Read more at

https://signalk.org/overview.html

However, I expect that SIGNAL K will be useful to only a very small fraction of recreational boaters, principally whose boaters with hardware and software experience who like to fiddle with electronics on their boats.

[Cabo Jim]

So I consulted with a friend of mine who happens to be have expertise with communications. Here is his reply:
So generally he is just wrong. He's switching between a bunch of layers on the OSI model.

We often move NMEA over Ethernet and fiber etc. Years ago it was a bigger deal to have a whole protocol stack and Ethernet processing just to pass a bit of ASCII info at a low speed. A serial connection with simple serial processing (native to microprocessors since 70's) has been a reasonable transport for a long time. And ya Ethernet is better, simple, cheaper - now - and ya we use it now.
So - all over the place about bitching about NEMA plugs and transport, it's neither of these it's a protocol. And it's 35 years old and evolving.(slowly)

[Hoodsail]

After reading this thread I am convinced the Golden Globe Racers have it right with their electronics for solo, unassisted circumnavigation being limited to:

- A stand-alone satellite tracking system (the skippers cannot see) for web tracking updates.
- A two-way satellite short text paging unit. (to race headquarters only) for twice daily 100-character text reports.
- Two hand held satellite phones (for important calls to Race headquarters only) for one a week safety check in only.
- A sealed box with a portable GPS chart plotter (for emergency use only)

Don't believe NMEA ???? is an issue...

[muttnik]

Originally Posted by Cabo Jim

So I consulted with a friend of mine who happens to be have expertise with communications. Here is his reply:
So generally he is just wrong. He's switching between a bunch of layers on the OSI model.

We often move NMEA over Ethernet and fiber etc.

Your friend may be a comms expert but I'll guess that he's not familiar with the NMEA standards: they're "full stack" standards and cover everything from the physical connectors up to the application layer so criticising choice of connectors and datalink is not invalid, even if I disagree with the OP.

If "NMEA" in the quote above is short for "NMEA-0183", I'll suggest that it's incorrect to say that we often move NMEA-0183 over ethernet because NMEA-0183 is by definition a serial protocol. To re-reference the OSI model, we may move the application layer of NMEA-0183 over (IP over) ethernet, but that's either a different standard (e.g. IEC 61162-450, another "pay to play" standard I'm not so familiar with) or the non-standard which is used by OpenCPN and other applications which consume "NMEA-over-IP".

The OSI model isn't necessarily a great fit for the marine data comms use case as more than half the layers may be redundant. Lasseman's post above raised the question of whether a network layer was necessary as it isn't in many automation/control applications. I'll argue that if nothing else, consumer demand for data over wifi in this part of the 21st century demonstrates that it is.

Does anyone know if any of the big manufacturers have OneNet products planned? Last boat show I went to (Boot 2020) none of the reps on the stands had even heard of it.

[loneshark64]

Originally Posted by wholybee

NMEA2000 is based on CAN, again with the benefit that the industrial world has been using it, the bugs are mostly worked out, and people understand how it works. It makes even better sense because it offers connectivity to other equipment that finds it's way onto a boat, like engines and chargers.

I get paid to (among other things) fix networking issues. Ethernet is not simple, nor is it any better understood by the layperson than RS422 or CAN. At least I can explain RS422 to a layperson to a reasonable understanding in an hour or less. Most people can't make 2 computers in their own home share files to each other.

I agree, I am an old ex IT guy as well, and have no problem with NMEA as a standard. In concept a bus architecture proven in industrial automation applied to lower bandwidth comms and specifying marine grade elements end to end makes sense. The target user is a guy that owns a boat or a dude working for a marina. Plug and play is good.

The problem is vendor implementation. A lot of us have old boats. I felt like an idiot trying to get my network set up. I had an old Raymarine autopilot (0183) a Raymarine Axiom plotter (Seatalk NG) a VHF with DSC (0183 or NMEA 2K) a Link 2K (0183) and Raymarine Quantum radar. I tried to start with the first 3 only and failed.

Autopilot required 0183 to NMEA2K converter bus and cables ($100)

Plotter required SeaTalk NG backbone and cables PLUS another interface cable because the Axiom is Seatalk NG but does NOT have a Seatalk NG plug! It has a fricking Devicenet plug! ($150).

VHF required a drop cable but had to go from NMEA to Seatalk NG plugs. I tried various things. ($20 each). Ultimately I found that Standard Horizon did not correctly implement NMEA2K on the VHF and they gave me a bunch of try this/try that BS and I eventually gave up and bought a standalone GPS antenna for the GPS on the radio to make the DSC work.

I sold all the NMEA cables and converters, busses and other garbage on EBay and gave up.

Moral of the story is Raymarine makes good gear but have totally screwed the pooch on connectivity across generations because they wanted to make money on cables. And Standard Horizon is Chinese garbage. Shopping for new boat, buying Garmin.

[rslifkin]

Originally Posted by loneshark64

And Standard Horizon is Chinese garbage.

Their NMEA implementations aren't great, but the actual radios are quite good. They're made by Yaesu who is right up there with Icom in terms of quality (and reputation for radios on non marine applications).

[gchead]

Why use cable at all?

These days each instrument should come with a wifi module (5 USD at any electronics store) so all devices can talk to each other wirelessly using NMEA over IP.

OpenCPN already supports this technology.