TWA is misbehaving on starboard tack

[sergei]

It was noted that TWA is kind of bimodal when boat sails on starboard tack, e.g. https://youtu.be/DWoVmSsBorY?si=e6ahgVSJjIcPhygu&t=330

It oscillates between expected value and about 90 degrees

[sergei]

Collected the log file with the MHU output awa-moving.log

[sergei]

Here we see four plots:

• Red, Green, Blue - The AWA angle computed using the value of the corresponding MHU wire (by taking it's asin() )
• Yellow - Value selected by FW for the final output

As one can see on starboard tack the green wire is most behaving, but we oscillate between choosing red and blue

[sergei]

Current algorithm divides the entire range into three 120 degrees zones and picks one wire to perform arcsin() operation on it. There are two problems with it

• The input tends to saturate when close to the rail (see scope shot below)
• The arcsin() is not too sensitive at the extremes, it's preferable to use the mid range of it
• We don't benefit from any redundancy

[sergei]

New algorithm is proposed.

• Compute estimate of AWA using all three inputs
• Resulting AWA is a weighted sum of these three, where the weights are proportional to the proximity to zero. This way most linear part of arcsin() will get most weight

[sergei]

Now the plot of AWA with the new algorithm (black)

[sergei]

Test of C++ code on the log file is promising: Next step is test on the water on the next Wednesday night.