Related

Compass True North vs Mag North + Weird Nav Problem

IamTC
over 2 years ago

Hello,

Here is a video detailing my problem: drive.google.com/.../view

My compass was calibrated today at a place where I intentionally went to get away from any possible magnetic interference. Also, the video compares with an iPhone X's compass which was also calibrated.

I have also recently did a hard reset in an attempt to resolve the issue.

Strange thing is that the mag north reads fine. I could have lived with this. BUT ... because of this discrepancy, the arrow on my track while doing the "Navigate" activity does not point correctly.

Other important stats of my watch: 13.3 firmware. it is an APAC model and under warranty (Serial number 6LL001460).

I wonder if there's a way I can install 14.3 firmware on my own? Seems like that might have solved it for some folks?

Thanks!

TC

/************************/

So here's the video showing the weird nav problem. I currently have my North Ref set to magnetic north, and maps set to "Track Up".

Almost tells me to do a u-turn. every. time. I. stop.

https://drive.google.com/file/d/163qJoQsC7aHhYRRjfZUyHONDL4-WamrF/view?usp=sharing

It's a simple route generated from Connect. Just a couple hundred meters of straight line.

But its quite obvious what's going on. The hardware is not powerful enough (understandable, since its a watch) to support things like Kalman Filtering. Otherwise, something like a Kalman filter would have kept this at bay. 

So when moving, the system gives more weight to GPS derived heading, all is good.

But when stationary, the system gives more weight to compass derived heading. Which somehow is quite messed up now.

  • 0 over 2 years ago in reply to PMO

    Hey PMO,

    Love a good discuss.

    If you see the video, a "jump" occurs within a second or two. Most of the Kalman filters I've seen would give a continuous behavior, slowly going hay-wired, i.e. no matter how good a filter is, give it rubbish and it will eventually give you rubbish, but gradually.

    Finally found it: The APAC release of v13.3 was last released on 27 Apr 2021 (www.garmin.com.sg/.../ .

    It would be nice to understand the firmware release cycle for APAC's Instinct Solar.

  • 0 over 2 years ago in reply to IamTC

    You are right. A pure Kalman will do that. However, to better cope with events not aligned with the predicted dynamics, often event-driven forgetting factors are used, which effectively "blow-up" the error covariance matrix, and thus force the filter to lose all confidence on the predicted position and go fully for the observation. In those cases, much more dynamic (indeed, very "jumpy") behaviors can be observed at the output of a Kalman, when those forgetting factors come into play and make themselves being felt.

    This is, however, purely academic. I do not think that this is what is going on. My guess is that, whatever type of filtering is being used (and it may be simply a low-pass, not a statistical filter), whenever the watch detects that the user has stopped, it switches to a compass reference, thus introducing a full discontinuity and discarding the previous filter history. It is probably a thing like "Did he stop? One, two... he did. Lets forget the previous direction based on the motion heading, and start anew, this time using the compass." (Did you like this pseudo pseudo-code?) :). Therefore, the jump would not be the output of a filter, but instead the discarding of the filter's output and switch to a totally different situation.

    But I digress. The important thing is to get your hands on the new firmware. In the non-APAC version we did not have a 13.30. Therefore, I have no idea on wether those two bugs (mag/true north and navigation whenever you go off-course) are corrected on it.

    Best of luck. Enjoy your Instinct. I finally feel (V14.10) that the bugs have mostly been ironed out, and am finally fully enjoying this great watch.