OpenCPN development preview

[verkerkbr]

Sorry for the spaces between the sentences. No longer able to correct this error.
Everything is working perfect. SignalK server with direct input to OpenCPN.

The only thing that is missing now, is OpenCPN/SignalK for my Raspberry Pi systems.

SignalK is default running in OpenPlotter version 2.0 beta.


Bram

[Hakan]

Bram..
What do you mean with: " missing now, is OpenCPN/SignalK for my Raspberry Pi systems."
I'm running RPi4 and Openplotter 2.0.19 beta.
Signal-K, Serials and more are there. K-Plex to come. Works fine. It's from that SK-server I tested my Win10 connection.
The OCPN, v 5.0, on Openplotter is TCP connected to SK and that's is good enough, or? All extra HTML Dashboards available via SK.

[verkerkbr]

Hi Hakan,

OpenPlotter is working very well here. And have also used SignalK server to connect to other systems.

Where I am interested in, is direct SignalK input in OpenCPN on my RPI.

Things are allready working on my other Linux system with Xenial and Bionic 64 bits versions.

It allready works for some time. But not yet with direct SignalK input.

Best wishes for 2020 there in the dark and cold Skandinavian winterdays.

Bram

[Hakan]

On my RPi4 with OpenPlotter I've not yet forced it to use 5.0.522 but I've built it on my RPi3. Works fine, found the SK server on the RPi4. see shot.

[Hakan]

And Bram BTW:
TransmitterDan made a package you may try?
https://github.com/OpenCPN/OpenCPN/issues/1618


And thanks for greeting. We're not that apart. And the winter climate is equal as well I think. Rainy and gray and no ice around the boat.

[verkerkbr]

Originally Posted by Hakan

And Bram BTW:
TransmitterDan made a package you may try?
https://github.com/OpenCPN/OpenCPN/issues/1618

And thanks for greeting. We're not that apart. And the winter climate is equal as well I think. Rainy and gray and no ice around the boat.

Hakan and TDan,

thanks for the efforts. Tried the beta. But I am allready changed to Bionic and Buster on my RPI's. This version does not work, to many missing files.

Waiting for an "official" beta setup from Dave.

Everything is working great on my other 64 bits systems, including the SignalK server.

Regards, Bram

[transmitterdan]

Yes, it’s only for Stretch.

[RhythmDoctor]

Originally Posted by RhythmDoctor

...The second <bug> was described in some prior messages. I can create a Flyspray entry for this if it helps for tracking - just let me know. This one is particularly important to me, because when in course up mode it required me to repeatedly turn off lookahead mode to re-center the "own ship" icon, then turn it on again. The frequent need for screen taps on my tablet made the feature pretty much unusable. Please see this description, and subsequent replies:

http://www.cruisersforum.com/forums/...ml#post2955674

Hi all, and Happy New Year!

I hope that Dave and team will give this bug fix a high priority in the new release version.

The new (in v5.0) feature to allow the user to place his Own Ship icon anywhere on the pane (which I refer to "green dot look ahead mode") does not work properly when OpenCPN is in Course Up Mode. It appears to work OK in North Up Mode, but there appears to be a problem with the calculations that causes the boat to drift when the boat changes direction, causing the chart to rotate. This is especially a problem for sailboats that need to tack upwind.

I think this feature is especially important for Course Up Mode, since one reason for using this mode is to keep the boat's progress vector aligned with the pane, especially if the pane or window has an elongated aspect ratio.

I went ahead and created a flyspray entry. Please go and vote for it if you would like to have a properly working "green dot look ahead" mode:

https://opencpn.org/flyspray/index.p...s&task_id=2638

[transmitterdan]

Originally Posted by RhythmDoctor

Hi all, and Happy New Year!

I hope that Dave and team will give this bug fix a high priority in the new release version.

The new (in v5.0) feature to allow the user to place his Own Ship icon anywhere on the pane (which I refer to "green dot look ahead mode") does not work properly when OpenCPN is in Course Up Mode. It appears to work OK in North Up Mode, but there appears to be a problem with the calculations that causes the boat to drift when the boat changes direction, causing the chart to rotate. This is especially a problem for sailboats that need to tack upwind.

I think this feature is especially important for Course Up Mode, since one reason for using this mode is to keep the boat's progress vector aligned with the pane, especially if the pane or window has an elongated aspect ratio.

I went ahead and created a flyspray entry. Please go and vote for it if you would like to have a properly working "green dot look ahead" mode:

https://opencpn.org/flyspray/index.p...s&task_id=2638

Are you sure you aren't a victim of chart rotation averaging? What happens if you turn off chart rotation averaging (Option->Display->Advanced) by setting it to 0. Also turn off course and speed filtering (Option->Connections). Does the "problem" still happen?

I think what you are seeing is an unavoidable conflict between "look ahead" and rotating the chart slowly because of filtering. If the chart rotation filter is slowing the chart rotate feature then the look-ahead function has to shift the boat around to maintain the look-ahead room ahead. Once the chart rotates the boat will drift back to the the original location. If you turn off chart rotation filtering then the boat should stay pretty close to the "green dot" position.

[RhythmDoctor]

Originally Posted by transmitterdan

Are you sure you aren't a victim of chart rotation averaging? What happens if you turn off chart rotation averaging (Option->Display->Advanced) by setting it to 0. Also turn off course and speed filtering (Option->Connections). Does the "problem" still happen?

I think what you are seeing is an unavoidable conflict between "look ahead" and rotating the chart slowly because of filtering. If the chart rotation filter is slowing the chart rotate feature then the look-ahead function has to shift the boat around to maintain the look-ahead room ahead. Once the chart rotates the boat will drift back to the the original location. If you turn off chart rotation filtering then the boat should stay pretty close to the "green dot" position.

Thanks for the suggestion. I wish someone had suggested this last August when I originally posted, since my boat was operational then. She is on the hard for the winter now, so I can only play back VDR logs.

I did adjust both settings to 0 and replayed the VDR log and, if anything, the problem may be worse. I suspect the problem would still exist for live data, since I'd expect the screen positioning calculations to work for VDR logs as well as live data.

I invite you and others to try my VDR log too. Go to the Flyspray entry to download it - it's a little too big to upload on this forum:

https://opencpn.org/flyspray/index.p...s&task_id=2638

It's easy to compare if you set one canvas pane to north up and the second canvas pane to course up. I usually put one corner of the Dashboard gauges on the OwnShip icon as a fixed marker on the screen, and watch the OwnShip icon drift away from it over time.

[transmitterdan]

It depends how fast you play the VDR file I think. I played your file at a speed close to real live speed and did not see anything “weird” with chart rotation filter set to 0.

With chart rotation set to 60 then sometimes the boat would instantly rotate 90 degrees (probably a tack) and then the chart would slowly rotate until the boat was once again pointing up on the screen. If the boat is not in the center and look ahead is on then it will try to honor the look ahead request by shifting the boat closer to the edge of the screen. You call that a bug but it’s just trigonometry at work.

Look ahead with the boat off center cannot 100% guarantee that a sudden 90 degree turn will keep the boat “frozen” to the spot on the screen. At least I don’t see how it’s possible. It takes some time for the computer to redraw the chart (which it does off screen in some systems) so it moves the ship to match the course. I don’t think this is any kind of bug.

What kind of computer are you using?

[RhythmDoctor]

Originally Posted by transmitterdan

It depends how fast you play the VDR file I think. I played your file at a speed close to real live speed and did not see anything “weird” with chart rotation filter set to 0.

With chart rotation set to 60 then sometimes the boat would instantly rotate 90 degrees (probably a tack) and then the chart would slowly rotate until the boat was once again pointing up on the screen. If the boat is not in the center and look ahead is on then it will try to honor the look ahead request by shifting the boat closer to the edge of the screen. You call that a bug but it’s just trigonometry at work.

Look ahead with the boat off center cannot 100% guarantee that a sudden 90 degree turn will keep the boat “frozen” to the spot on the screen. At least I don’t see how it’s possible. It takes some time for the computer to redraw the chart (which it does off screen in some systems) so it moves the ship to match the course. I don’t think this is any kind of bug.

What kind of computer are you using?

First, I've seen the reported behavior on two desktops (Windows 10), two laptops (Windows 10), and two Windows tablets (Windows 8.1). Wide variety of different levels of processing power, but same behavior on all of them. The latest tests are on my home desktop, a Dell XPS 8500, Intel Core i7-3770 @ 3.4 GHz, 16 GB RAM, NVidia GeForce 640 graphics card. Not a real new or gaming-grade computer, but PLENTY of power to keep up with the demands of OpenCPN.

I slowed down the playback rate for VDR, with the same result.

What I can't understand is this: The traditional Look Ahead mode works perfectly fine. In course up mode, the OwnShip icon will shift up and down if the boat speed changes (as expected), but the icon is always perfectly in the center, and whenever the boat returns to 6 knots speed, it's always in the same spot on the screen. Traditional Look Ahead mode works perfectly with any setting for rotational averaging or NMEA filtering. If rotation averaging is set to 5 seconds, an abrupt change in direction may cause the icon to move for about 5 seconds, but after that the icon always returns to the exact same pixel on the window.

So my question is: Since traditional Look Ahead works perfectly, what is it about the new enhanced Look Ahead that makes it impossible to implement with similar perfection? Is there something magic about one pixel location over another? I don't think there should be any difference - it should be equally possible with either. The key difference that I can tell is that with the new enhanced Look Ahead mode, the algorithm must ignore the length and direction of the ownship velocity vector, and focus solely on panning and rotating the chart around the selected pixel for the ownship icon. If this causes the far end of the ownship vector to go off the pane, so be it. To me this seems to be the key difference, but the latter should be easier to implement, because positioning does not have to worry about the changes to the ownship velocity vector.

[bdbcat]

RD...


Lets get this into Flyspray, with associated VDRs attached if needed. With this type of GUI bug, the eyes glaze over when reading a text-only description of the problem.


Thanks
Dave

[RhythmDoctor]

Originally Posted by bdbcat

RD...


Lets get this into Flyspray, with associated VDRs attached if needed. With this type of GUI bug, the eyes glaze over when reading a text-only description of the problem.


Thanks
Dave

Thanks, Dave, it's here with a VDR attached:

https://opencpn.org/flyspray/index.p...s&task_id=2638

Just run it in Course Up mode and see if you see the same problem I do.

[transmitterdan]

Originally Posted by RhythmDoctor

First, I've seen the reported behavior on two desktops (Windows 10), two laptops (Windows 10), and two Windows tablets (Windows 8.1). Wide variety of different levels of processing power, but same behavior on all of them. The latest tests are on my home desktop, a Dell XPS 8500, Intel Core i7-3770 @ 3.4 GHz, 16 GB RAM, NVidia GeForce 640 graphics card. Not a real new or gaming-grade computer, but PLENTY of power to keep up with the demands of OpenCPN.

I slowed down the playback rate for VDR, with the same result.

What I can't understand is this: The traditional Look Ahead mode works perfectly fine. In course up mode, the OwnShip icon will shift up and down if the boat speed changes (as expected), but the icon is always perfectly in the center, and whenever the boat returns to 6 knots speed, it's always in the same spot on the screen. Traditional Look Ahead mode works perfectly with any setting for rotational averaging or NMEA filtering. If rotation averaging is set to 5 seconds, an abrupt change in direction may cause the icon to move for about 5 seconds, but after that the icon always returns to the exact same pixel on the window.

So my question is: Since traditional Look Ahead works perfectly, what is it about the new enhanced Look Ahead that makes it impossible to implement with similar perfection? Is there something magic about one pixel location over another? I don't think there should be any difference - it should be equally possible with either. The key difference that I can tell is that with the new enhanced Look Ahead mode, the algorithm must ignore the length and direction of the ownship velocity vector, and focus solely on panning and rotating the chart around the selected pixel for the ownship icon. If this causes the far end of the ownship vector to go off the pane, so be it. To me this seems to be the key difference, but the latter should be easier to implement, because positioning does not have to worry about the changes to the ownship velocity vector.

What is happening is the ship rotates when the course suddenly changes as in a tack. Then lookahead mode says "hey we need to slide the chart around to maintain the lookahead distance". Finally, the chart will eventually rotate to put the ship back to looking up because you have selected "course up" mode. Then the ship will drift back to its original position on the screen because the chart is rotating and lookahead is always trying to keep up with that.

If the ship is in the center of the viewport then the ship doesn't move when it rotates or the chart rotates. But if the ship is off-center then something has to give.