Related
Options

Instantaneous pace steady but slow (possibly since 19.20 software update)

stephen.e.cox
over 3 years ago

I've noticed a persistent issue in the last couple of weeks. I didn't, unfortunately, notice when my software updated to version 19.20, so I can't say for sure that this is just related to the recent software update. My instantaneous pace during runs looks pretty steady, but it will be persistently low for long portions of runs. For example, today during an easy trail run (flat terrain) I watched my instantaneous pace during a full kilometer of running at about 5 min/km pace. The whole time, the watch showed my pace between 5:40 and 6:30, which is implausibly slow. When the auto lap completed for that kilometer, however, it showed a lap pace of 4:55. 

In the past, the instantaneous pace would only be that far off when I had a GPS issue, and it would usually overcorrect in the other direction when trying to correct my location. Now, there is nothing wrong with my recorded GPS track, and the lap paces all come out fine, so I am guessing this is a result of a heavyhanded attempt at quieting the occasional spikes in instantaneous pace that used to occur. Has anyone else noticed this recently, or is it possible there's just something wrong with my watch?

  • 0 over 3 years ago in reply to tsprks

    Just by raising and stopping the movement of your arm you mess with the accelerometers and the instant pace...
    Sorry...
    Keep a steady/consistent arm movement and speed, preferably in a straight line (always huge problems in crossroads or turns, of course), and you will see a consistent pace...sooner or later... ;)

  • 0 over 3 years ago in reply to Paulo_Goulart
    Paulo_Goulart said:
    Just by raising and stopping the movement of your arm you mess with the accelerometers and the instant pace...

    The fact that instant pace relies that much on accelerometer is what wrong with it in the first place. Garmin makes all sorts of assumptions that I think are wrong, especially when applied to trail running.

    For example, I noticed that when I transitioned from fast level running on an open terrain to climbing a steeper terrain on a wooded trail, initially my pace was horribly wrong - see the example above. Certainly the pattern of acceleration is quite different when climbing a steep trail vs. level running. It takes way too long for Fenix to adjust - up to 10 minutes. I see the same kind of issue in a beginning of each run, where my Fenix pace is reliably too slow. I see the same kind of issue when transitioning from uphill to downhill. 

    I think it is still the right thing to fuse GPS and accelerometer data, but calibration of accelerometer should be done on a much shorter rolling interval - perhaps 30-60 seconds rather than ~5-10 minutes that it seems to use based on my observations.

    It is wrong to assume that GPS isn't accurate. GPS data is certainly way more accurate than accelerometer data but it is not as precise. See https://en.wikipedia.org/wiki/Accuracy_and_precision for the difference between accuracy vs. precision. Pace should be primarily based on GPS data and only use accelerometer to filter / smooth data in a very short term. But it seems Garmin has it the other way around based on how long it takes for the pace to adjust. So basically the way Garmin has pace now is precise but not accurate.

  • 0 over 3 years ago in reply to AndersB

    So still no fix on current pace also on 19.20. I am very disapponted with that, and I'm considering to switch to a FR945 instead of a F6.

    I hope I am wrong, but my fear is that GPS accuracy on F6 isn't good enough to get solid and decent pace/distance in running.

    Otherwise I can't find a reason why they can't replicate the forerunner's series performance for pacing/distance, which seems much better.

    To be honest I didn't try your CIQ field based only on GPS data, but in general GPS during running seems always very poor precision compared to biking.

  • 0 over 3 years ago in reply to Mattbest

    I'm not sure if there have been improvements to the distance and/or pace. I have done 12 runs with 19.20 and 19.21 and it have "felt" better while running but I haven't put so much analysis into the data. But yesterdays run (when I switched the device to outer wrist instead of inner wrist) was the worst in a very long time. Maybe the accelerometer was confused by the "new" placement and gave faulty data. I'm not sure. Have been working on my data field and will try to make it even better soon.

  • 0 over 3 years ago in reply to Mattbest

    I use the “Race screen” data field. It’s ok. Better than Garmin (much quicker to update).

    But I really do not care, I run since my 13s (I’m 53) and I know exactly my pace, much better than any watch…I never failed myself (in that respect)…

  • 0 over 3 years ago in reply to Mattbest

    Just released a new public beta (version 0.9.26) for my data field. Have improved the Distance summary so it also shows the distance calculated by the device Pace. So it will be easier to see that it in average reports to slow Pace. This time I ran a little faster and the built-in distance (3D was on) was really close to my own algorithm. But both is to short compared to the controlled distance. Will work more on the algorithm another day.

  • 0 over 3 years ago in reply to AndersB

    My feeling is that between 5'15" and 6' min/km Garmin algorithm gets totally confused, it gradually improves with increasing speed (becomes good starting from 4'30" min/km).

    Do you have the possibility to use a Forerunner in the other wrist and compare your data field also with a more reliable watch?

  • 0 over 3 years ago in reply to Mattbest

    Yes, that's my impression also. That it works less good in lower speeds and when the cadence and running technique varies in the activity.

    No, I don't have any Forerunner except for my old Forerunner 310XT but I have a friend who runs with a Forerunner 945. Will see how his next runs looks like. 

  • 0 over 3 years ago in reply to Mattbest
    Mattbest said:
    My feeling is that between 5'15" and 6' min/km Garmin algorithm gets totally confused, it gradually improves with increasing speed (becomes good starting from 4'30" min/km).

    I think Garmin's algorithm assumes that the speed is proportional to cadence (number of steps per minute). It assumes a constant stride length and doesn't take variable stride length into account. That is pretty evident when looking at treadmill runs when there is no GPS and only accelerometer based speed is used. On treadmill I often do a very gradual speed ramp up where I start with a very slow speed and increase it in equal increments every 30 seconds. It is obvious, looking at the graphs afterwards, that the speed is not proportional to cadence. According to Garmin resulting data the speed either increases very rapidly or plateaus. Sometimes it even drops down. In reality the stride often lengthens when increasing speed but Garmin doesn't seem to sense that.

    I think the same issue applies to outside running. I think the algorithm does calibrate stride length, but that takes rather long time. So the same issues with the pace inaccuracy happen when transitioning from one type of terrain to another, making sharp turns, etc. 

  • 0 over 3 years ago in reply to Mattbest
    Mattbest said:
    I'm considering to switch to a FR945 instead of a F6

    I suspect that Garmin has now pushed the same algorithm to Forerunner series. Here is the same issue reported by FR 745 users. They say it started recently.

    https://forums.garmin.com/sports-fitness/running-multisport/f/forerunner-745/273667/forerunner-745-pace-issue/

    I was considering FR 745 because it is supposed to be the most accurate Garmin watch, but after seeing that thread I won't buy it.