How to solve the problem that is SeaTalkNG

A forum for discussing boat or trailer repairs or modifications that you have made or are considering.
User avatar
mastreb
Admiral
Posts: 3927
Joined: Wed Feb 09, 2011 9:00 am
Sailboat: MacGregor 26M
Location: Cardiff by the Sea, CA ETEC-60 "Luna Sea"
Contact:

Re: How to solve the problem that is SeaTalkNG

Post by mastreb »

Hi Robert,

You're supposed to put the power supply NMEA 2000 cable in the middle of your network (by physical cable length), and only supply power at a single location due to the possibility of introducing ground loops with multiple power supply points. This also makes it easy to switch all the NMEA-2000 equipment on and off. If you have to supply at multiple points due to draw, it gets a bit complicated--they want you to make two networks and bridge them with opto-isolated NMEA-2000 bridges. Of course, that's just a best practice, likely you could try powering and just see if everything works.

Matt
User avatar
Crikey
Admiral
Posts: 1833
Joined: Sun Apr 17, 2011 12:43 pm
Sailboat: MacGregor 26M
Location: Washago, Muskoka, Ontario, Canada, Earth, Singularity.Suzuki DF60A. Boat name: Crikey!

Re: How to solve the problem that is SeaTalkNG

Post by Crikey »

What happens to the stability of the backbone when device proprietary code is piggy-backed into industry standard NMEA2000 sentences being emitted from one component? Does it just garble the accuracy of some numerical readouts elsewhere in the chain, or does everything fail?
Too bad all feeds cant pass through a PC somewhere using configurable software translation and control.
R.
User avatar
mastreb
Admiral
Posts: 3927
Joined: Wed Feb 09, 2011 9:00 am
Sailboat: MacGregor 26M
Location: Cardiff by the Sea, CA ETEC-60 "Luna Sea"
Contact:

Re: How to solve the problem that is SeaTalkNG

Post by mastreb »

Nothing happens, they're simply ignored by devices that don't know how to interpret them. NMEA 2000 specification allows for manufacture specific codes.

The wireline protocol is all the same, typical CAN-BUS and actually the same as the J1939 specification for semi-trucks. It uses Carrier-detect Multiple Access with Collision Avoidance to ensure that devices don't talk over one another; essentially, devices wait until the bus is silent, and monitor their own transmission as it goes out on the bus. If it gets garbled, they stop talking, back-off, and try again at a random time later. It's the same mechanism used on Ethernet computer networks.

There's also an extension where "dumb senders" are allowed to just blurt out their data on the bus a few times a second, and all the smart devices will back off and wait, and other dumb devices just get stepped on. This is used by tank senders, GPS units, rudder position indicators, etc. where if an update is missed its not critical. The bus is limited to 64 devices total, so its unlikely that there would be so many devices as to cause constant contention.

So there's nothing to fear from proprietary codes except incompatibility.

The thing that bugs me most is that there's no standard specification for how to configure devices over the bus, which means you have to buy a MareTron display to program maretron devices, a garmin display to program garmin devices, etc.
User avatar
RobertB
Admiral
Posts: 1863
Joined: Sun Jun 05, 2011 7:42 pm
Sailboat: MacGregor 26M
Location: Clarksville, MD

Re: How to solve the problem that is SeaTalkNG

Post by RobertB »

The power issue aside (and much easier to deal with): I called Raymarine to ask about using the P70 control head as the connection to the NMEA 2000 DeviceNet components (the Garmin chartplotter and the ETEC engine feed). They stated I should use a connection off the Sea TalkNG backbone, that the information could get messed up. OK, great, no real useful information here - just asked me to spend another $100 in cables and connectors. Kind of makes me want to take the P70 apart and see if the two network ports are directly wired together.

I am curious if this is more directly related to power (the other two wires in a NORMAL NMEA 2000 bus). Curious if the only components that require power supplied by the bus in the whole setup are the direction sensor and the control head. The Autopilot computer, the chartplotter, and the engine all have their own power feeds and may not draw power from the NMEA 2000 bus.

So, do I hook up power to the sensor side - a total bus/spur run of 4 meters to the sensor, 6 meters to the control head, and 11 meters to the engine (5 meter backbone, 1 meter power, 3 meter spur cable to the sensor, 1 meter to the control head, and 5 meter engine cable). The computer is hooked in at the same 5-way block the sensor and power is.
Or do I hook into power in the pedestal close to the control head - 9 meters to the sensor, 2 meters to the control head, and only 6 meters to the engine.
The later may reduce the average run but the first case I am pulling power off a relay near the engine while the pedestal power is at the end of a long run. I also suspect the most critical component for power is the sensor (and that the engine probably does not use power off the bus).

Any logic to this rant?
User avatar
mastreb
Admiral
Posts: 3927
Joined: Wed Feb 09, 2011 9:00 am
Sailboat: MacGregor 26M
Location: Cardiff by the Sea, CA ETEC-60 "Luna Sea"
Contact:

Re: How to solve the problem that is SeaTalkNG

Post by mastreb »

Ultimately you want the shortest maximum distance to power for farthest device that you can reasonably achieve. High draw devices should have their own power. I put most of my backbone next to the battery for very short power runs for most devices.
User avatar
RobertB
Admiral
Posts: 1863
Joined: Sun Jun 05, 2011 7:42 pm
Sailboat: MacGregor 26M
Location: Clarksville, MD

Re: How to solve the problem that is SeaTalkNG

Post by RobertB »

Well, almost there. I have installed all the components and just finishing up the power wiring.
The refrigerator and now the autopilot will now run on relays wired directly to the battery selector switch. The relays are switched by the breakers on the panels up high (autopilot will switch on with the radio and chartplotter, refrigerator has its own switch). This eliminates at least 15 feet of wire runs and the associated voltage drops. It has also fixed by refrigerator problems.

The NMEA 2000/SeaTalkng is as follows: A 5m backbone with a 5-way terminator each end. Above the aft dinette seat (in my shelves), I connect to one 5-way: the computer (1m spur), the sensor head (1m backbone extension, tee, 3m spur), and the network power (1m). In the pedestal, connected to the other 5-way are: the NMEA 2000 engine adapter (5m, end cut off and SeaTalkng end spliced on), the Garmin (1m adapter cable), and the autopilot control head (1m spur). So, network distance from power relay: Sensor 22 feet, control head and chartplotter 24 feet, engine 34 feet, and computer 7 feet (and 3 feet directly to the computer itself) . I am assuming the chartplotter, computer, and engine do not need power from the network backbone since these all have independent power sources. If a problem, I can always cut down the length of the engine adapter to remove the excess.

Will follow up soon with pictures of the install. The linear actuator is mounted to a block glassed to the hull and a custom stainless steel and PVC clamp assembly bolted to the steering cross member. A ball-lock pin keeps the actuator from jumping off the cross member bracket. The control head is mounted in a custom made cube that sits beside the chartplotter on a swivel base on top of the pedestal.

I am not sure the relays are a long term solution - I am rethinking locating the breaker panels down low near the batteries/ladder area to cut down on the long power runs - and to clean up the bird nest of wires since now most everything is installed. Will keep the original panel for the lights. Am also installing the AC main/2 branch panel down low this weekend - think it will work with the flush switches. I was lucky and purchased this panel http://www.amazon.com/gp/product/B00GZO ... UTF8&psc=1 for $78 (less than half price, lists for $212 and sells for $145) on Amazon.
Post Reply