GPS Accuracy

Why not post the question in that Facebook thread? While you're at it, ask him whether he's still using Smart recording, something I reported seeing in his methodology last year. At that time, he wrote that he saw no difference between Smart and Every Sec.

I just posed the question about 1sec vs smart on the facebook thread. Let's see what he says.

Personally, GPS performance has been far from satisfactory for me.

https://connect.garmin.com/modern/activity/1643505277 (quite OK but the track is shifted towards the middle by the turn)
https://connect.garmin.com/modern/activity/1643503973 (this one is quite bad- the trees at the end cause a big shift+the turn just before lap 2 has me making the turn in a completely wrong spot.

Fellrnr's reply to my question about smart vs every second recording: "I tested and found no difference"- what a pile of rubbish!!

Here's my reply to his:

"Interesting..I find smart recording to be greatly inferior to every second recording due to the infrequent polling of GPS position data. Garmin used to have smart recording only devices on the market, which is no longer the case for a reason! Would you mind posting an activity that you recorded with both?"

Fellrnr's reply to my question about smart vs every second recording: "I tested and found no difference"- what a pile of rubbish!!

Here's my reply to his:

"Interesting..I find smart recording to be greatly inferior to every second recording due to the infrequent polling of GPS position data. Garmin used to have smart recording only devices on the market, which is no longer the case for a reason! Would you mind posting an activity that you recorded with both?"

This is incorrect, "smart recording" only reduces the number of points saved in the track (maybe useful back in the day for storage space considerations), it has no impact on the frequency of the GPS polling that remains at 1Hz, so no impact on distance/pace, etc...Now of course one should keep that mode in mind when looking at the actual tracks, especially when there are turns.

@FlipStone - you mention "weird exceptions", did you also notice irregular laps, different distances ?

NY, you know that it computes the GPS position once a second even in smart recording mode, yes? We know for sure that models of the 610 and 310xt era just recorded a subset of the points in smart recording, but those points were in exactly the same place as they would have been with every second recording. I don't think Garmin have ever said whether there's any extra filtering on the points recorded to the FIT file in the current watches. Either way, though, the total distance as shown on the watch will be exactly the same.

Trying to reconcile the posts in this thread is giving me a headache. Apparently the F5 doesn't agree with the 235 and M400, and they are much better. And also fellrnr says the F5 is rubbish, and he is definitive. Except, he says the GPS on the 235 and the M400 are "distinctly grim" and "stunningly bad" respectively, giving them scores about two-thirds of what he gave to the F3.

Don't get me wrong, I'm happy users are looking closely at it, but I'm not sure if it's getting me any nearer a decision yet.

ETA: I started typing this before Webvan posted...

This is incorrect, "smart recording" only reduces the number of points saved in the track (maybe useful back in the day for storage space considerations), it has no impact on the frequency of the GPS polling that remains at 1Hz, so no impact on distance/pace, etc...Now of course one should keep that mode in mind when looking at the actual tracks, especially when there are turns.

@FlipStone - you mention "weird exceptions", did you also notice irregular laps, different distances ?

As english isn't my native language it can sometimes be hard to find the proper words ;) I guess I just meant irregularities, weird 'scribbles', sudden 'jumps'. (See how I keep using the ' to emphasize I'm not happy with any of those words ;) ).

But the activities are really too small to draw any real conclusions, lttle over 1k for the walk and less then 3k for the bike ride. I'll add them here anyway to show what I mean:

735XT Walk:
https://connect.garmin.com/modern/activity/1643223559

Fenix5 Walk:
https://connect.garmin.com/modern/activity/1643225207

The Fenix5 one shows a very messy track just over midway (when going underneath the waterlinieweg, I was almost standing still there though but the 735xt handles that way better).

735XT Bike:
https://connect.garmin.com/modern/activity/1643223527

Fenix5 Bike:
https://connect.garmin.com/modern/activity/1643225182

The fenix5 one shows a small jump upwards in the beginning (1:40-1:45 mark). Nothing extreme again but it was just a straight road, again the 735xt seems smoother and more consistent.

Planning on taking it for a ~ 7.5k run tomorrow maybe or the day after, we'll see how that goes.

Fellrnr's reply to my question about smart vs every second recording: "I tested and found no difference"- what a pile of rubbish!!

Here's my reply to his:

"Interesting..I find smart recording to be greatly inferior to every second recording due to the infrequent polling of GPS position data. Garmin used to have smart recording only devices on the market, which is no longer the case for a reason! Would you mind posting an activity that you recorded with both?"

That really casts a cloud over all of his Garmin tests. Smart Recording is the default, so I guess that's why he uses it.

I really wish I could think of a way to describe how bad using smart recording is for any sort of data analysis. I can't think of a comparable thing. But this was brought up before, and somehow keeps getting ignored every-time folks reference the data. It's one thing if somehow after a device reset/etc you forgot for a couple runs/rides to set it back. Crap happens. But on purpose?

In general, smart recording records every 4-7 seconds. Sometimes more, sometimes less. But on average, 4-7 seconds is a safe bet. So, you can either have 1/7th the data for 'statistical analysis', or 7 times the data.

If someone only gave me 1/7th the data to play with, I nor anyone else would be able to make very good accuracy determinations.

(Side note: Yes, I still think it's incredibly stupid that Garmin defaults to Smart Recording - mostly because it's a self-inflicted wound at no tangible cost in the grand scheme of data storage costs that they have. They'll lose more in revenue from one day of people loading his data set tests using smart recording, then their entire year of data storage costs for smart recording files.)

It would be easy enough to settle this. Someone go ask him to halt his test of the Fenix 5, and restart it with Every Second enabled. He's retested watches before.

NYanakiev7??


If things turned out differently, he'd certainly have some egg on his face. Might be a disincentive for him to retest.

Simple to understand that errors that may be minor statistically in data can easily be exacerbated when using a reduced data set for analysis regardless of whether the reduced data set contains some of the same exact data as the full set.

Ridiculous to say that when performing any analysis that having anything less than the full data set is adequate.