compass interferences

[seandepagnier]

Earlier today I moved a stainless steel knife on the other side of a bulkhead from the pypilot compass sensors and caused the boat to jibe and knock me upside the head with plenty of blood and a headache.  I would admit my head should not have been in that location but I am mostly just glad my boom is made of wood instead of metal.

The compass bias was completely off by an earth field magnitude.    The earths magentic field is relatively weak and easily interfered with.

For example, a wire running to a solar panel that is several inches (15-20cm)  can cause the compass the change by significant amounts such as 5-10 degree course changes with each cloud.   Twisting the 2 wires greatly reduces the interference.

All sorts of objects (anything that a magnet sticks to) can also distort the field if they are moved relative to the sensors and this requires a recalibration each time.   Especially bad are magnets including electric motors and speakers and especially other floating compasses (moving magnets) which the software does not attempt to compensate in any way so they must be avoided.

Finally, the wires to the motor controller for power and the motor should all be routed so they do not run too close to the compass sensors.

[stelian]

First, I hope you are ok !

I wonder if something could be done to prevent this (in hardware, like shielding somehow, or in software). I never encountered this with a commercial pilot (but I haven't tried either), and I think pypilot is somehow too sensitive, it gets too easily confused...

[oilybilge]

Wow, bad luck Sean. Hope your head's recovered!

It's a timely post as I spent yesterday trying to find a place to mount my Pi/IMU. On a small boat it's amazingly difficult. Even the exhaust hose gives off a slight magnetic field.

Eventually I put it up forward under a berth. Now I just have to make sure nobody sits there with a phone in their pocket.

[mgrouch]

Feel better!

Few ideas. In future designs might be it makes sense to add
buzzer and trigger it before jibes and tacks.
(Assuming pypilot has true wind data).
Also give some time before jibing.

I’m using Capt Don boom brakes sailing downwind.
Prevents boom from slamming.

Fair Winds!

[Jean-Marc Douroux]

Hello Sean, I hope you are recovering well.

To avoid interference, I have chosen to put your Pypilot computer on a panel against the hull, 50 cm above the level of the chart table and 15cm below the deck and 30cm from the electrical panel where there is no no high current.

To disturb the magnetometers with this installation, it would be necessary to put metal outside the hull, 15 cm below deck level at the level of the computer.

[Techstyle]

Few ideas. In future designs might be it makes sense to add
buzzer and trigger it before jibes and tacks.
(Assuming pypilot has true wind data).
Also give some time before jibing.

I guess you could make it so the buzzer sounds if the boat passes through the wind (tacks or jibes), but there could be equal chaos when there is any significant deviation.  so maybe you add buzzer when deviation is more than X degrees (10, 15, 20?)

I have had issues with my heading sensor as well and wonder if it is possible to use the GPS heading either to error check the magnetic sensor or use dual GPS sensors to define a line as a heading sensor?

[seandepagnier]

The idea of the autopilot not jibing or tacking even on compass or gps course is a good one and I have thought of it before but not implemented.

It does require a wind sensor though and I didnt have one connected at the time. I was in compass mode. In gps mode it possibly would not have jibed either.

The newer version of pypilot in git does give warnings about magnetic distortions, but I also found a few more bugs in the calibration.

For example, when I was beating against the wind in the gulf stream, the waves were very bad, and the boat was thrown around enough holding a course that it confused the calibration into thinking it had made a 360 in a location on the earth much closer to the magnetic pole. Locking the calibration prevented this, but it shows more work needs to be done.

[stelian]

I have also seen some reports of pypilot re-calibrating itself even if the calibration was locked (https://www.hisse-et-oh.com/sailing/pypi...420f9f5ab5, in French)

[Jean-Marc Douroux]

In the case of french sailor, it is with an Openplotter configuration and an RPI4 whose 5V converter is 10cm from the IMU. Is there sometimes interference between the IMU and some 5V converters?