Whilst evaluating the data from the APEX Pro, I spotted some inadequacies that I've not seen mentioned before now.

I've written them up for you all (link below) and within the page, I've also provided links to the detailed analysis, etc.

logiqx.github.io/gps-guides/devices/coros/apex-pro/data-issues.html

The gpstc has done some evaluation of the corros, and we have come to a similar conclusion as you

Select to expand quote

decrepit said..
The gpstc has done some evaluation of the corros, and we have come to a similar conclusion as you

I'd be interested to know if the Doppler inadequacies are also evident on the Fenix 6, Suunto 5/7/9 and any other devices using the Sony chip.

I don't have any data to check those watches but if anyone can provide applicable tracks to me, I'd gladly take a look!

I've now confirmed that this affects the VERTIX as well.

My website has been updated accordingly.

Good for you, now those blokes who are so keen to run with any old supermarket watch can find the reasons the GPSTC have standards for GPS receivers and Daffy and Co won't need to repeat themselves so regularly.
Nice.

I've found the Apex pro is about a knot slower in 2 sec speed and at times .5 slower in 10 sec average compared to GW60, if Vincet had posted with the GW60 his average would be 50knts and not 49.56, just saying.

Select to expand quote

choco said..
I've found the Apex pro is about a knot slower in 2 sec speed and at times .5 slower in 10 sec average compared to GW60, if Vincet had posted with the GW60 his average would be 50knts and not 49.56, just saying.

That is my general experience as well, although oddly I had some slightly better results on my COROS (faster than GT-31 and GW-60) on both occasions when I visited West Kirby a month or so ago.

I need to go back to those tracks and look more closely at why that was the case.

I've finally managed to get hold of some data from a Fenix 6X and a GW-60 during the same session. It provides some interesting insights into the behaviour of the Sony chip; features shared by the Fenix 6, APEX Pro and VERTIX.

TLDR - Repeated speeds are probably (imho) due to the Kalman filter in the Sony chip rather than COROS / Garmin firmware. 50 mm/s resolution (AKA quantisation) is only evident in COROS tracks as the Fenix 6X is 10 mm/s.

Here is my full write up / notes:

logiqx.github.io/gps-guides/sessions/contacts/gwys/

I can't find any technical documentation for the Sony CXD5603GF, so I have no idea what it has in the way of available navigation modes or track smoothing (on/off). COROS may be able to fix the quantisation issue (if so inclined) but maybe not the Kalman smoothing + repeating speeds.

Anyhow, I thought it interesting to confim the same artefacts are present in the Fenix 6, almost certainly a Sony "feature".

Select to expand quote

K888 said..
COROS may be able to fix the quantisation issue (if so inclined) but maybe not the Kalman smoothing + repeating speeds.

Those repeating-speeds are a result of dead-reckoning.

It may be able to be tested for a given device:
1. find a tin can big enough to completely contain the gps
2. walk in a straight line for say 10m
3. turn 90deg and at the same time insert the gps into the can - ideally do these both steps in under half a second
4. walk a bit further
5. synchronise this effort with a stopwatch

... at the point of turning 90deg, if the trackpoints continue in a straight line, then dead-reckoning is enabled.

Here's a closer look at the crash artifact:

The positional speed also stays high for the 4 points with identical doppler speeds. Then, both speeds drop down to 0 for 7 readings, a clear indication that the GPS was submerged after the crash. When reception is gained again, the first fix is about 2 1/2 points back relative to the travel direction. That's a clear indication that the last 3 points are not real, and due to something like dead reckoning or an "overoptimistic" Kalman filter.

A typical example that today's GPS are designed for cars, where reception often drops briefly under bridges etc, and keeping the speed constant when reception drops is the most reasonable thing to do. In GPS units that are based on u-blox chips, similar behavior can sometimes be seen after crashes.

It's worth noting that repeated speeds are seen both during crashes and during regular runs. Sometimes I see exactly the same speed for 6-12 seconds on my COROS. My COROS track from the same day at that location is full of repeated speeds and it's boosted my results, when compared to the 2 x GT-31s and GW-60 that I was also wearing.

My suspicion is that dead reckoning using SOG and COG is one of the main estimates feeding into the Sony Kalman filter, and it somehow ends up in a state with too much confidence in the dead reckoning. This could cause the Kalman to prioritise the dead reckoning over the latest observations; latest SOG + COG and positional estimates. When it looses the signal the Kalman filter just thinks "meh, I'll assume nothing has changed".

Another possibility is the pursuit of super-low power usage may mean that the Sony chip is cheating at times and relying too much upon dead reckoning, rather than generating new observational readings every second to feed into the Kalman filter. Seems impossible to know without insider info.

Like boardsurfr says, probably works well for a car or bike but not so good for windsurfing.