Detect when engine is on from NMEA data?

[conachair]

Originally Posted by mglonnro

Sounds like a good idea

I wonder if my Raspberry is aware of its input voltage

If you run openplotter on the raspberry you can connect an ads1115 voltage sensor - this is old doc but V2 openplotter with full documenttion is promised soon>
https://sailoog.gitbooks.io/openplot...n/ads1115.html


It then will send very accurate voltage as signalk , which has an app to save to influxdb database, or file, or send to the web - extremely versatile. Same with temperatures using ds18b20 thermometers. I've 3, engine head, alternator and exhaust elbow.
Signalk on a Pi is just great

Also I've a few projects on the go , one is a pcb from JLCPCB which hopefully will provide 8 channels of voltage, 2 x ammeters, barometer, RPM, 2 x nmea muiltplexed, thermometers and variable voltage and I need to add a few controllable switches.
https://easyeda.com/editor#id=9b753b...5190814850862c

Boards won't cost much , they solder all the surface mount bits, just need to add a ESP32 and solder on some connectors. It then sends all the data as signalk (or straight to an influxdb database) over wifi to the Rasp Pi. Been running similar for a while, works really well.






https://easyeda.com/editor#id=|9b753...5190814850862c

[Breaking Waves]

Originally Posted by mglonnro

The point of all of this is to draw a polar diagram with wind angles, speeds, boat speed.

Sounds like you need to just do a few passes at the data to scrub it as well as can be. Error toward washing out (rather than keeping) suspect data.

1. wash out any data where you remember motoring.
2. wash out any data with boat speed > apparent wind speed, and any with apparent angle < 25 degrees (and the surrounding data points)
3. run the resulting data set against this polar , and the one linked bottom of post, and be extremeley skeptical about any data points which are > 100% of that polar (and the surrounding data points)
4. Create a polar based on the resulting data set and examine any performance outliers (look at heel angle for instance) to see if any look like motoring - wash out if suspect.

I think that would get you pretty close. Polars are never ever 'exact' in any case - depend on sea state and crew and such, so 'close enough' is the best objective.

also you can get a quite good polar from racing rule measurement data - like fyi - hanse 388 M L-keel polar by ORC data

[mglonnro]

Originally Posted by Breaking Waves

Sounds like you need to just do a few passes at the data to scrub it as well as can be. Error toward washing out (rather than keeping) suspect data.

1. wash out any data where you remember motoring.
2. wash out any data with boat speed > apparent wind speed, and any with apparent angle < 25 degrees (and the surrounding data points)
3. run the resulting data set against this polar , and the one linked bottom of post, and be extremeley skeptical about any data points which are > 100% of that polar (and the surrounding data points)
4. Create a polar based on the resulting data set and examine any performance outliers (look at heel angle for instance) to see if any look like motoring - wash out if suspect.

I think that would get you pretty close. Polars are never ever 'exact' in any case - depend on sea state and crew and such, so 'close enough' is the best objective.

also you can get a quite good polar from racing rule measurement data - like fyi - hanse 388 M L-keel polar by ORC data

Thank you for the encouragement and I will post back my results

[mglonnro]

Originally Posted by conachair

If you run openplotter on the raspberry you can connect an ads1115 voltage sensor - this is old doc but V2 openplotter with full documenttion is promised soon>
https://sailoog.gitbooks.io/openplot...n/ads1115.html


It then will send very accurate voltage as signalk , which has an app to save to influxdb database, or file, or send to the web - extremely versatile. Same with temperatures using ds18b20 thermometers. I've 3, engine head, alternator and exhaust elbow.
Signalk on a Pi is just great

Also I've a few projects on the go , one is a pcb from JLCPCB which hopefully will provide 8 channels of voltage, 2 x ammeters, barometer, RPM, 2 x nmea muiltplexed, thermometers and variable voltage and I need to add a few controllable switches.
https://easyeda.com/editor#id=9b753b...5190814850862c

Boards won't cost much , they solder all the surface mount bits, just need to add a ESP32 and solder on some connectors. It then sends all the data as signalk (or straight to an influxdb database) over wifi to the Rasp Pi. Been running similar for a while, works really well.






https://easyeda.com/editor#id=|9b753...5190814850862c

This is some cool stuff here! Thank you and I'll get back to it once I get the Raspberry from our boat.

[mglonnro]

Some progress.

The attached diagram is drawn from a few hours of real sailing (the first we did here in Finland). I think the AWA (which everything is based on to calculate TWA/TWS) needs some calibration, but the idea is here.

The coloured dots are TWA/TWS/SOG observations. green = TWS 0-6 knots, blue 6-15 knots, yellow 15-20 knots, red 20+

The outside line next to the dots is the max speed for that TWA.

The line even further outside is what the Korean Hanse 388 would have done in similar TWA/TWS. The difference lines between real and theoretical are: red = more than 1.0 knots, gray = 0.5-1.0 knots, green < 0.5 knots.

Detecting sailing time wasn't that hard. I realized I had a lot of timestamped pictures to help as well.

[Snore]

I guilty of not reading past post 15 or so. sorry.


Battery voltage would be an indicator. If voltage is > 12.x then it must be getting a charge.

If you have solar, this may require a diode. Another way is to sense off the tach on the alternator. Not sure how- but some of the smart guys around here might help.

[mglonnro]

Originally Posted by Snore

I guilty of not reading past post 15 or so. sorry.


Battery voltage would be an indicator. If voltage is > 12.x then it must be getting a charge.

If you have solar, this may require a diode. Another way is to sense off the tach on the alternator. Not sure how- but some of the smart guys around here might help.

Yes The voltage thing seems well worth trying. Unfortunately I didn't have any sensor or logging for that last season, but will see what I can do next season.

Wouldn't that be a useful NMEA message addition, btw? That you could send out a broadcast query for each connected device to report the voltage it sees? Or that they would broadcast it themselves periodically. Maybe this already exists? Didn't see voltages in my message database, though.

[mglonnro]

DONE!

I solved the original question as suggested, by looking at the data and reminding myself of when we did what.

In the end it was pretty easy to filter out the sailing periods!

I wrote a blog post about comparing what we did to Hanse's marketing material (about 90-94% of advertised speeds having less sail area!): https://nakedsailor.blog/2019/12/22/...ed-and-polars/

[conachair]

Originally Posted by mglonnro

DONE!

I solved the original question as suggested, by looking at the data and reminding myself of when we did what.

In the end it was pretty easy to filter out the sailing periods!

I wrote a blog post about comparing what we did to Hanse's marketing material (about 90-94% of advertised speeds having less sail area!): https://nakedsailor.blog/2019/12/22/...ed-and-polars/

Fantastic bit of work!!

How did you create the polar images from the data? Looks good, big well done.

[mglonnro]

Originally Posted by conachair

Fantastic bit of work!!

How did you create the polar images from the data? Looks good, big well done.

Thanks!

The polar image is drawn with my own nodejs code onto an HTML5 canvas.

First a piece of code that reads all the "millions" of points (once) from influx and generates the (averaged) polar points.

Then a piece of code that takes the average points and uses the basic canvas drawing stuff to draw the graph.

[teppokurki]

Fabulous work! Have you considered packaging and publishing your code as a (Signal K) webapp?

What wind sensor do you have - how accurate do you think it is?

[mglonnro]

Originally Posted by teppokurki

Fabulous work! Have you considered packaging and publishing your code as a (Signal K) webapp?

What wind sensor do you have - how accurate do you think it is?

Thank you!

And that's certainly a cool idea. I could package just the polar visualization stuff. And I'm sorry that I haven't found the energy boost to jump into Signal K yet!

The sensor -- it's a B&G 508. It actually broke all by itself later in the fall and I got a newer generation replacement for next summer. I'm pretty sure the data isn't spot on, but rather roughly in the ballpark.