GPS about 30m off

You are to quick with your response JungleJim, I hadn't finished 

Yes, I took the data from the .FIT files. I wrote software to derive the data from the .FIT files with cm accuracy.
I converted WGS84 coordinates to the Dutch Map Datum (RDNew), which expresses the coordinate as Northing and Easting, in meter. I believe this is more accurate than calculating distances using Haversine, since the Dutch Map Datum is especially for Dutch territory. It also made calculation easier (pythagoras...). All in Matlab 
RINEX is on my bucketlist. I need some conversion software that derives coordinates from the raw satellite distances. RTKLib requires an academic degree it seems...

Great, you seem to know your stuff

RINEX data analysis would indeed be interesting. Early firmware versions of the GPSMAP 67 produced RINEX data that was simply incorrect (missing bands, duplicated data to name a few). Garmin fixed it eventually, so it would be interesting to see if the data is actually correct and useful.

Hopefully Garmin engineers can do something with the info in your blog post.

PS. Groetjes vanuit Nederland

I’m getting 110 meter or more drift too.  Do you know if Garmin is working on a fix?  It’s definitely not normal.  Did anyone get a replacement and have the problem fixed?  It would be helpful if the Garmin engineer(s) explained the problem and people could decide for themselves if it’s worth the risk to have that much error in an SOS device with both L1 and L5 enabled.  Oh the irony. 

I submitted the error to Garmin Support in the Netherlands (NL) and provided all kind of proof. They indicated it is 'by design'. Then I turned to Garmin Support in the US. They were aware of my submission to Garmin Support NL. They don't see it as an error. But in the end they offered me a replacement, which I did not take (because I don't believe this solves the problem and you get a repaired one).

I took the GPSMAP 67 and Fenix 7 to Madeira on a number of hikes mountainous area. On a number of to and fro hikes I counted the times it deviates more than 10 m. The GPSMAP 67 performs worse than the Fenix 7. See my blog, at the end of the page.

Apparently when the going gets tough, this tough doesn't get going...
I sold my device on the second hand market, highly disappointed.

Thanks for the reply and testing. One of the things it appears they’ve done is altered the gain on L5. L5 is supposed to have twice the power as L1, but they’ve tinkered with it (I assume to save battery).  I just wish they’d tell customers that.  You’re probably right about the sensor fusion algorithm being messed up.  I can’t help but wonder if the Garmin software engineers know that the Department of Defense has GPS L5 marked as unhealthy and they have to manually override that in the firmware.  I’m going to give a couple more days before I decide whether or not to return it. 

I know this is an old post, but I wanted to add that a year later, with the 67i, that I see similar errors. I'm using the unit in Yuma, Az. returning to the same sampling points in Nov 2024 and Jun 2025. We are planning to return to the same points in Jun 2026. So I will be curious, what results I get when compared to the mapped point (ArcGIS Pro) and a new Bad Elf Flex GPS unit. 

Hello I really enjoyed reading your blog posts frim your previous posts.

When I read the part about deadband filtering, it reminded me of something I noticed during another comparison that I made in my free time while doing roller skating. 

I own numerous Garmin watches and at that point I was using my Forerunner 955, Fenix 7X pro, and tactix Delta which is like a Fenix 6x pro.

Rollerblading was not included in the default list of available activities, so I customized the track me activity to record it. 

It was surprising that the forerunner used to create a very jagged track, while the fenix 7X used to make a very smooth track. They are both identical watches as far as the hardware and even software features go.

If I would record a walk activity the tracks came out the same. 

I think the Forunner was using unfiltered data during the track me activity.

I think it is because forerunner devices are designed for faster response to instant pace changes so the GPS pipeline may be faster with less averaging.

The tactix also had smooth track me trails but it uses a Sony chipset that was less precise and would eat up a lot of real deviations anyway.

For all activities other than track me, I do not recall the jagged saw tooth like tracks.

I have sold the Forerunner and replaced it with an instinct 2X as my black plastic watch. This seems to create slightly smoother tracks than the Fenix 7X. Probably due to energy and memory conservation.

However I still have the Track Me activity files from my 955. I will be happy to share them if you are interested. 

I have left a message on your website contact me link. Look forward to connect thank you.