Jedi wind instrument experiment

[s/v Jedi]

Another project in the pipeline, that has been annoying me enormously, is to get a good wind instrument and not get my boat blown up when lightning hits it.

I’m not even gonna try B&G again, it seems all the new instruments fail, both wired and wireless.

I’m not even gonna try the new Maretron sensor… my old one magically still works and I believe the new one will be more reliable (and the humidity sensor replaceable) but it is too slow.

I almost ordered the LCJ Capteurs instrument but they were sold out and it’s still slow… even the high speed one. I believe this is a good quality unit but the design is becoming dated.

I have now decided to use the Spanish sensor from Calypso Instruments. It is a relatively new design that I think is doing well and it supports 10Hz frequency.

I’ll probably order it with the 1-meter high pole, carbon or aluminium, and with NMEA0183 output.

Instead of using a NMEA0183 to NMEA2000 converter at the base of the mast, I am going to use a RS-485 to IP wifi converter that makes a connection to my access point and IP encapsulates NMEA0183 data to/from UDP packets with my Yacht Devices NMEA2000 to Ethernet gateway on the other end.

This wireless link is just for creating complete galvanic isolation to prevent lightning from taking out all my electronics. The last hit we got took out my Maretron heading and USB gateway (that’s a boat-buck right there) as well as many other devices needed a factory reset to start working again.

Here to I use an isolated power supply. I contacted Yacht Devices about this and of course they recommend to use two of their wifi gateways, which will work great, but I rather use just one access point for everything to keep wireless better under control.

So here is the diagram I came up with for now. The wifi access point, router, NMEA to Ethernet gateway are all already on board so it’s just the new sensor and the IP converter, which also has Ethernet by the way, very interesting.

What worries me is that they claim t is bidirectional but there’s only two conductors for the rs-485 interface. I suspect only rs-232 will be bidirectional but I can simply put a mini dongle on that port to convert it to 4 conductor rs-485.

[s/v Jedi]

You could use the same setup with other sensors that use NMEA0183 to come down the mast and a converter to get to NMEA2000, like B&G and LCJ Capteurs. The Maretron, Furuno and Veratron sensors use NMEA2000 which means you must bring the NMEA2000 backbone all the way up to the mast, including a terminator.

Maretron sells a nice inline terminator for this which eliminates the T-connector but it’s still a hassle to bring that backbone that far up and I feel their solution is better suited for motorboats.

[dannc]

Thank you for this design and commentary.

Later,
Dan

[team karst]

hmmm; 232 is unidirectional for data, although it has control wires if not SW control. 422 is bi 485 is basically a multi-drop 422; that is, its addressable to accommodate many nodes.

The 2W layouts should really include the REF wire, making it 1Pair plus 1. Without a REF, one can exceed the common mode spec on the transceivers.

I think LoRa would be a good wireless start for a lightning hardened system. But, it will mean batteries here and there.

[Dockhead]

Originally Posted by s/v Jedi

You could use the same setup with other sensors that use NMEA0183 to come down the mast and a converter to get to NMEA2000, like B&G and LCJ Capteurs. The Maretron, Furuno and Veratron sensors use NMEA2000 which means you must bring the NMEA2000 backbone all the way up to the mast, including a terminator.

Maretron sells a nice inline terminator for this which eliminates the T-connector but it’s still a hassle to bring that backbone that far up and I feel their solution is better suited for motorboats.

Speaking from experience -- don't put the N2K cable up the mast. Bring the serial cable down.

Also from experience -- forget the Airmar and Maretron sensors -- they were not designed for this purpose and are far too inaccurate -- just check the specifications.

I'm on my second LCJ Capteurs sensor and these are good, but I have yet to get one working as well as the B&G 7-series tall wind sensor processed through a B&G H5000 computer on the boat I crossed the Atlantic on last year. That was the cat's meow. Maybe the LCJ Capteurs sensor will give similar results when I figure out all the calibration and compensation and finally get decent boat speed data. We'll see.

I'm using this one so far with a Windy Plug N2K converter which works well, although I will experiment also bringing the 0183 directly into my H5000 Hercules computer.

If you want any chance at all to get something resembling valid true wind data you need accurate boat speed, which is hellishly difficult to do, and you need to invest a considerable amount of time and effort into calibration.

And you will want to compensate the wind sensor data for heel, upwash and boat motion. That's what I bought the H5000 for.


Another pro tip -- get the sensor as high above the mast top as possible. It makes a really big difference. The upwash and turbulence from the sails reaches up pretty high and buggers the data on certain points of sail. Mine is 1.5m tall which I think is about the minimum for a rig like mine.

[s/v Jedi]

Originally Posted by team karst

hmmm; 232 is unidirectional for data, although it has control wires if not SW control. 422 is bi 485 is basically a multi-drop 422; that is, its addressable to accommodate many nodes.

The 2W layouts should really include the REF wire, making it 1Pair plus 1. Without a REF, one can exceed the common mode spec on the transceivers.

I think LoRa would be a good wireless start for a lightning hardened system. But, it will mean batteries here and there.

I need to dive in deeper to this. NMEA0183 is rs-422 and there are converters doing this with 4 conductors but the first ones I found had Ethernet only, no wifi.

I’m not even sure what comes out of the wind sensor yet, but I do see many of these rs-232 dongles for converting to rs-422 (https://www.amazon.com/Communication.../dp/B06ZYHPB23) so I think I can simply use the bidirectional single ended rs-232 interface of the IP converter with the dongle adapting it to rs-422.

Is LoRa fast enough? It’s on my list to build a couple of those text message gadgets with it…

[s/v Jedi]

It turns out the Calypso actually has a real rs-485 port with A and B conductors only so the converter I selected is good to go.

For configuration they use a USB converter and special software. The instructions show they catch a startup sequence from the sensor after powering it up to bring it into configuration mode that allows sensor data rate to be set from 0.1Hz to 10Hz and also update the firmware itself.

I really like the manual because it’s clearly written by someone with the same background in electronics as me

Here is a screenshot from the manual showing the use of the power and data conductors during configuration:

[s/v Jedi]

Originally Posted by Dockhead

Speaking from experience -- don't put the N2K cable up the mast. Bring the serial cable down.

Also from experience -- forget the Airmar and Maretron sensors -- they were not designed for this purpose and are far too inaccurate -- just check the specifications.

I'm on my second LCJ Capteurs sensor and these are good, but I have yet to get one working as well as the B&G 7-series tall wind sensor processed through a B&G H5000 computer on the boat I crossed the Atlantic on last year. That was the cat's meow. Maybe the LCJ Capteurs sensor will give similar results when I figure out all the calibration and compensation and finally get decent boat speed data. We'll see.

I'm using this one so far with a Windy Plug N2K converter which works well, although I will experiment also bringing the 0183 directly into my H5000 Hercules computer.

If you want any chance at all to get something resembling valid true wind data you need accurate boat speed, which is hellishly difficult to do, and you need to invest a considerable amount of time and effort into calibration.

And you will want to compensate the wind sensor data for heel, upwash and boat motion. That's what I bought the H5000 for.

Another pro tip -- get the sensor as high above the mast top as possible. It makes a really big difference. The upwash and turbulence from the sails reaches up pretty high and buggers the data on certain points of sail. Mine is 1.5m tall which I think is about the minimum for a rig like mine.

I think I have different requirements. I want a reliable sensor like you, but I don’t care so much for absolute accuracy, even though this Calypso claims 1 degree direction and 0.1 m/s speed accuracy. What I want is rapid data rate to get my autopilot to steer to a wind angle reliably.

I don’t need it to steer a 175 degree run without gybing but I want it to steer a 160 degree broad reach without gybing. Upwind I expect better performance because our boat is balanced so well I almost don’t need an autopilot to steer itself upwind.

(I used to sail a Dehler 42 that could sail effectively upwind for hours on end without touching the wheel! A Dehler! )

I tried autopilot steering to a wind angle using the Maretron sensor and it worked for one second only I think it is accurate enough but way too slow with data rate.

[s/v Jedi]

They now also have a NMEA2000 version that requires the backbone cable up the mast, so you can choose any configuration, even I2C or analog. Here is the manual: https://calypsoinstruments.com/web/c...&download=true

[team karst]

I suspect autopilots use PID algorithms, and these need to be tuned. There are competing goals. Response speed may be at odds with stability. Over damped setting may induce sluggishness and settling errors. My raymarine manual calls for “critically damped” response, which is not particularly fast. But then, the actuator is not constantly moving either. Tying to a wind direction sensor may demand a lot of smoothing and thereby slowing response. Sounds like a project in control theory. [emoji1362]

[s/v Jedi]

Originally Posted by team karst

I suspect autopilots use PID algorithms, and these need to be tuned. There are competing goals. Response speed may be at odds with stability. Over damped setting may induce sluggishness and settling errors. My raymarine manual calls for “critically damped” response, which is not particularly fast. But then, the actuator is not constantly moving either. Tying to a wind direction sensor may demand a lot of smoothing and thereby slowing response. Sounds like a project in control theory. [emoji1362]

I expect that as well and was very happy to see that this Calypso instrument allows to change that in settings so at least you have a chance to get it right.

But the B&G, when it works, works well with the AP. It does 10Hz updates so when I get that and can set smoothing parameters during testing, I think it’ll work. Hopefully for longer than a year which seems when the B&G sensors fail.

[catalystcat]

Critically damped would be great for long haul sailing. It's basically when you use the minimum amount of damping (usually negative feedback) that allows response without getting into oscillation. For control nerds (e.g., me), that's when the poles start to reach 0 on the X axis.

For some cases, that's the best answer, in that it gives the lowest power consumption, to a first order, and least amount of stress on the AP/steering. In other cases, a different damping may be better. E.g., damping ratio of 0.7 instead of 1 for fastest settling on target.

Not to induce too much thread walkoff, but once you move the physical response into the software, you open a lot of interesting possibilities.
And, this is kind of a good data accumulation and analysis problem.

For any given condition and wind angle, each boat will have a response, both steady state and in transition. Of course, "condition" could be very broad. It might encompass wave height and period, secondary wave height and period, sail parameters, current, wind variability in angle and speed, and a bunch of other things. For each case, there is some perceived optimum response in terms of angle, sail configuration, damping factor, etc.

Normally, sailors solve this by learning our boat over a range of conditions. So, for Jedi's upwind case of well balanced with a consistent wave and wind pattern, I would expect a very slow, highly damped steering response (not to be confused with sampling interval). It would be constrained by perceived gybe risk vs ABS (angle-180). So, a negative feedback could apply as one nears the gybe angle.

I could imagine building something like a polar diagram to inform the AP response parameters. Of course, the underlying data is WAY more complex and the optimization parameters (e.g., power consumption vs SOG vs ride/comfort) make it geometrically even more complicated.

For individual boats, it seems like learning the boat will dominate. But, for boats having a lot of commonality, one could feed this to a machine learning process and get a set of recommended parameters. (I know, we're supposed to use AI instead of ML these days).

I would think some trendy boat builder or modeler or bored/inquisitive boat nerd would try to see if they can either get enough data to make the ML useful, or generate enough synthetic data to make it work. It may be error filled. But, it would be fun and probably make for good marketing.

[team karst]

I also think AI could solve this puzzle.
Wasnt there an old AP that had a seastate control?
Very likely the optimum PID parameters change with heading and speed and wave direction. I wonder if products are beginning to accommodate that.
I just scraped by collegiate control theory. Poles and zeros and transforms [emoji33].

[s/v Jedi]

I used to have a WH-Smith analog AP which had a cluster of control knobs where you set things like gain and counter rudder etc. Then it had a removable panel behind which it had a row of miniature potentiometers where you could change every possible setting.

After years of tinkering with it I was completely familiar with every control and got it to work perfectly. Absolutely fantastic.

Then I bought a Robertson digital AP. During installation it did it’s thing, asking me to turn a wheel now and then, to go out and run at different speeds so it got to know the boat and then it said it was ready for the job. 30 minutes and it was ready. I activated it and it steered better than the WH Smith.

I sometimes think how I would love to write my own AP but then again my list of projects is long and my current B&G has Robertson blood running through it’s veins so I guess it is good enough. Shame these wind sensors are all so troublesome.

I hooe the new setup will work as planned and will be rock solid reliable. I got many more projects

[qwert]

Would this be any good to you?

https://wi-rb.com/

They started off with a very crude instrument delivered two years after I
ordered and paid for it. That was about ten yrs ago.

Now it seems that the "sailtimer" might be quite good. Its certainly awash with good ideas, some of which might suit you