GPS about 30m off

As I already wrote above, I have the same problem. The Garmin support told me, that this is a normal behaviour. As I cannot belive that, I talked to my retailer to replace the GPSmap with a new one and he did.

Then I did the same test with a new GPSmap (without updating - Firmware v6.91) and it worked without problems over some days, no drift!. Then I updated the device to v7.60 and it still worked over a longer period. BUT: after updating it to the current v8.90 I instant had the problem again :-(

Starting with v8.xx, Garmin also updated GPS firmware version from v2.46 to v11.02. So I think that maybe it is a bug in the GPS firmware?

Sadly it is not possible (as far as I know) to downgrade the GPS to an older firmware version :-(

I downgraded the firmware to FW 7.60 (GPS FW 2.46). However this did not resolve the problem.

I wrote a detailed report on my blog.

Note: what you see in the images is that when you start moving, the coordinate seems to converge to the real location.

I got a message from Garmin Support: the handed the issue to a tech team and asked for more information.
I hope they resolve it.

Update

• When you leave the device stationary under less than ideal condition, the indicated location as is logged starts to drift right away, with potentially large deviations  (>100 m)
  Sometimes the device recovers when you start moving, sometimes a substantial residual error persists
  Only way to let the device recover is via a power cycle
  
  
• I mailed and called Garmin Support, provided information, report, photos, etc, etc. to their tech support
  Twice they pointed to the FAQ 
  Their reaction indicates that they don't understand the problem.
  I told them that this is highly disappointing, that I give up and buy another device.
  Hope they see the light one day and solve the problem...
  
  

Update:
I also contacted Garmin Support US (previous contact was with Garmin Support Netherlands). They indicate the behavior is 'by design'.
I am not going to put more energy in it and won't use the device.

scubajorgen said:

When you leave the device stationary under less than ideal condition, the indicated location as is logged starts to drift right away, with potentially large deviations  (>100 m)

I do not understand what is this all about. I do not know anything else then this for years, not only Garmin products, but other simple gps devices. When I put such device on top of a high house, meaning full sky view, such irregularities I am experiencing all the time jumps up to 400m around. With the current garmin devices which are operated together with a phone, there is also some interaction between the gps rx in the device and position calculated by the phone. I had some jumps of few km when on the sea on a cargo ship and suddenly, near the coast, the phone calculated its own position and this landed then on the explore website.

skyeye said:

I do not understand what is this all about. I do not know anything else then this for years, not only Garmin products, but other simple gps devices. When I put such device on top of a high house, meaning full sky view, such irregularities I am experiencing all the time jumps up to 400m around. With the current garmin devices which are operated together with a phone, there is also some interaction between the gps rx in the device and position calculated by the phone. I had some jumps of few km when on the sea on a cargo ship and suddenly, near the coast, the phone calculated its own position and this landed then on the explore website.

It seems that you are saying that you expect GPSses to be hundreds of meters up to kms off, but that is not my experience. 

I bought a 2nd hand GPSMAP 66sr, which behaves like expected (like my Fenix 7, Edge 1040, TomTom, etc): when left stationary it shows the coordinate with some error: 95% of the logged coordinates fall within 2.9 m when leaving it outside for 24 hours, unobstructed sky view.
For the GPSMAP 67 it is even better, 95% of the logged coordinates fall within 2.6 m.

This is in accordance with what garmin specifies in the FAQ of Garmin:

The GPS location accuracy of Garmin outdoor watches is around 3 meters (10 feet), 95% of the time. 

Next charts gives the average error of the measurements (top) and the radius of the circle enclosing 95% of the measurements (bottom) as function of the duration of measuring



However.... If exposed to less than ideal conditions (i.e. unobstructed sky view, vertical orientation; e.g. in the forest), the coordinate  as logged by the GPSMAP 67 starts wandering away when standing still, like Edwin started this thread with. Giving it a power cycle immediately resolves the error. If moving, the device is accurate, however I stood still in the forest yesterday for 20 minutes: this resulted in 12 m error! 

Even sometimes the GPSMAP 67 gets berzerk under ideal conditions as next picture shows giving a huge distances (24 hours of measuring).

In these pictures the black dashed circle is the 50% distance and the red dashes circle the 95% distance

In this case the device does not behave like stated in the Garmin FAQ.

I have no dual frequency device currently, but with normal single frequency device placed on the roof of a 4 storage building, the 'spider' contains certainly jumps of more then 100m during the day. This in central europe.

During one day, there will be some 5-6 over 100m spikes, the rest might be within some 30m.

If a device is connected to a phone, then things can become more problematic when the phone tries to interfere with the data provided by the gps device.. I was also surprised by this, but messenger connected to a phone did really interfere very much.

Measurement on GPSMAP 66sr and GPSMAP 67
I did an extensive comparison between the GPSMAP 66sr and GPSMAP 67. As well in unobstructed sky conditions as in a dark forest. See my blog page.

Bottom line: when moving both devices are very accurate especially for hiking and under less ideal conditions. For cycling use another GPS.

Reason of the issue and fix
As stated above the issue starts when the device is stationary.

I believe this has the following reason: Apart from the satellite signal the device also uses the acceleration sensor signal and/or digital compass for calculating your position, based on previous position. When you shake the device quickly from side to side, say 80 cm, you immediately introduce an error of a few meters. This cannot be explained based on satellite reception.

Your car navigation does something similar: if you drive into a tunnel (GPS signal lost) it uses velocity and direction to calculate your position in the tunnel.

This makes the device extremely accurate when moving, it records the slightest movements. For example if  you walk a square with sides of 2 m (2 paces) it is recorded. Next image shows a larger rectangle in the dark forest, which I walked 10x CCW (bright color) and 5x CW (dark color). The GPSMAP 67 is very precise.

Apparently if satellite reception worsens, it relies more heavily on the acceleration sensor. The abnormal drift when stationary is an error associated with dead-reckoning and probably originates from noise on the acceleration sensor signal, explaining the randomness.

Obviously the solution would be to reset the location based on satellite reception when the drift becomes to much. Or never leave out the satellites out of the equation...

Garmin

We all experience that Garmin does not acknowledge there is a problem. The GPSMAP 67 showing a completely wrong coordinate is 'by design'. 

They are willing to replace my device by a NOH (Newly Overhaul) though, but I do not think this will solve the problem.

Btw, does anybody have experience with NOH replacement? Are they truly as good as new?

Further research

The device is able to log the 'raw' satellite signals in RINEX format in RINEX files. If I can find a software that deduces the coordinates from this, I can compare this 'second opinion' to the coordinates logged in the GPX and FIT files.

Maybe this will convince Garmin, but I doubt it...

Interesting blog post! One question though, did you also analyze data from the FIT file the units record? Because that is the native format, and the GPX file is a derivative of the FIT file. Not saying there isn’t an issue but in order to exclude any other factors it would be best to look at source data.

Also analysis of the RINEX data of both GPSMAP 67 and 66sr could give additional insights.