Straight to it: instruments that behave perfectly on the pontoon and drop out at sea are almost never faulty instruments. Motion, vibration, heel and engine load all change what the wiring and the network are doing, and that is where the fault lives.
The pattern of the dropout is the most valuable diagnostic information you have, so it is worth recording carefully before anyone starts pulling connectors apart.
Note whether one display goes blank or every display loses one sensor. Note whether it happens on a particular tack, when the engine starts, when the autopilot loads up, or after an hour of running. Note whether the units reboot or simply show dashes.
Each of those is a different fault family. One instrument dying alone points at that instrument's own supply or its drop cable. Every instrument losing the same sensor points at the sensor or its branch. Everything rebooting together points at the supply feeding the whole network.
Electronics are sensitive to voltage in a way that lights and pumps are not, and many will reset below a threshold rather than dim gracefully. A bank that sits at a comfortable voltage alongside can sag noticeably once the autopilot, fridge and pilot pump are all working, and if the instrument feed already suffers voltage drop over a long thin run, the combination is enough to cause resets.
Alternator noise is the other common culprit. Instruments that reboot the moment the engine starts, or that lose data when engine revs change, are usually reacting to a poor supply rather than to anything wrong with the electronics themselves.
On an NMEA 2000 backbone, one bad connection can take out data from devices nowhere near it. Missing or duplicated terminating resistors, a backbone powered at the wrong point, drop cables that are too long, and cheap non-marine connectors are the usual offenders.
The classic intermittent is a connection that is mechanically sound at rest and separates under vibration. Crimps done on domestic ferrules, screw terminals that were never re-tightened after the first season, and cable runs left unsupported so they swing against a bulkhead all produce faults that appear only at sea.
Mast wiring deserves particular attention because it takes more movement than anything else on the boat, and the deck gland is a favourite place for water to enter and corrode a wind transducer feed from the inside.
If it is only depth that disappears, look at the transducer before the electronics. Fouling on the face, air under the hull in a seaway, and a tired cable at the through hull will all produce readings that vanish in exactly the conditions where you want them. Wind data behaves similarly when a masthead unit has water in it.
Substituting a known good sensor is conclusive, but it is not always practical afloat, which is why the wiring and network checks come first.
Intermittent faults reward systematic testing and punish parts swapping. Measuring voltage at the instrument under real load, checking backbone resistance, and inspecting the physical run in the places where it moves will normally find the cause in one session.
We do this as fault finding work across the South Coast, and where the network has grown piecemeal over several seasons we will usually recommend tidying the backbone properly rather than patching one more tee into it.
Alongside there is no motion, the engine is off, and load is low. Remove those three and the marginal connection or the marginal voltage that was always there finally shows itself.
Rarely. Most of these faults are supply or network faults, and the instruments themselves turn out to be serviceable once they are fed and connected correctly.
Yes. Incorrect termination is one of the most common causes of intermittent NMEA 2000 behaviour, and it is quick to check with a meter across the backbone.
Losing depth or wind in pilotage water is a navigational risk rather than a fire risk, but it should still be resolved before relying on the system offshore.
We diagnose instrument and network dropouts under real load rather than at the dock, and rebuild backbones that have grown untidy.