Averaging just one component of a 'velocity vector' doesn't work as you have noticed. This applies to [set;drift], [SOG;COG], [TWS,TWD], and more. You may reduce the apparent noise a little by averaging your two cartesian velocity components independently as you suggest but the answer will continue to be incorrect.

Another important question is whether we should be working with 'velocity vectors' at all. After all it's the 'position vectors' that contain the noise in the first place.

And have you ever noticed that our GPS delivers both position coordinates (which bounce around a little even when we are not moving) as well as our instantaneous [SOG;COG] vector based on position changes from one measurement to the next. Some higher end GPS units apparently incorporate accelerometers, magnetometers and gyros with sophistocated predictive filters to give much better estimates of position and velocity.

Ideally I would like to see some of this sort of functionality available within Expedition.

Tnx for the valuable insights!
John, you say that it is implemented.  But in the Exp 10.7.6 channel settings I still see separate damping for set and drift.   How does it work?

I thought I responded to this earlier but now I get "Bad request. The link you followed is incorrect or outdated." when I follow the link in the notification email. Hmmm. So, let's try again ...

Shouldn't all 2D vector quantities be damped/averaged/filtered as 2D vectors? Damping the magnitude and angular components independently simply makes no sense.

For instance, 'current' is a vector comprised of [set,drift]. Similarly, 'true wind', 'apparent wind' and 'ground wind' are vectors are based on [TWS,TWA], [AWS,AWA], and [GWS,GWD] respectively. [SOG,COG] could also be considered a vector but both are received independently from the GPS so the characteristics of the GPS (incl. internal damping/aveaging, sampling rates, etc) need to be considered here.

Ideally, all of the basic corrections/compensations/etc should be completed before the data is presented to Expedition in the first place but this is often not possible. But that said, true vector-based smoothing/averaging/filtering would be a great improvement over what we have available today.

In the case of measuring SET and DRIFT, the position (LAT,LON,SOG,COG) noise from the GPS can be of the same magnitude as the actual SET and DRIFT. Hence here cartesian filtering is of significant advantage.

When filtering other data this might not be the case and cartesian or polar filtering typically doesn't make much difference.

On the other hand, look at the case where you would like to know the average windspeed GWS (e.g. over a year) at a venue. Assuning a balanced windpattern cartesian averaging would possibly render an average windspeed of 0 whereas polar averaging would give you th typical windspeed of that venue which is probably what one would be looking for.

As for me, I see no reason to switch to cartesian averaging for the other measurements.