Dropout of Pitch/Roll, Lat/Lon G's

Could you get us the FIT file for your data as well, in addition to the original GPX file?

You can get the FIT file by right-clicking on the clip in the Clip Library, and selecting 'Properties', then go to the GPS Info tab.

You should have two export links, one for GPX and one for FIT. The GPX you sent looks fine, so we suspect it's the FIT that is giving us trouble.

Fit file for the shorter video uploaded to the same dropbox folder. Export was done with 2.7.0.

Could you get us the FIT file for your data as well, in addition to the original GPX file?

You can get the FIT file by right-clicking on the clip in the Clip Library, and selecting 'Properties', then go to the GPS Info tab.

You should have two export links, one for GPX and one for FIT. The GPX you sent looks fine, so we suspect it's the FIT that is giving us trouble.

For troubleshooting purposes, if I wanted to try to track lean angle (pitch/roll) in another app, what would I map to the VIRB GPX file?

For troubleshooting purposes, if I wanted to try to track lean angle (pitch/roll) in another app, what would I map to the VIRB GPX file?

We determine pitch and roll by applying equations to the accelerometer data written in the GPX file.

So far, we have been unable to reproduce the g-force cutting out.
What we can reproduce is the nonsensical information for pitch and roll during your motorcycle ride. Allow me to try to explain why this is.
When we first implemented pitch and roll using the accelerometer data from the VIRB, we began to notice that an accelerometer wouldn't accurately portray pitch and roll during certain activities. Because we only have an accelerometer available, our pitch and roll equations have to rely on the only force being applied to the camera is gravity. During slower activities, this tends to be an OK assumption therefore equations tend to result in reasonable data. HOWEVER, certain activities (like motorcycle riding) depend on other external forces to function. This is why you can lean your motorcycle in a turn and not fall over. During these activities the camera's accelerometer would be reading the combination of all these external forces, instead of just gravity. As stated earlier, our equations rely on the only force involved to be gravity, so these activities would result in nonsensical pitch and roll data displayed on your video.

The bottom line: The only way we could produce reasonable pitch and roll data during every activity would be by adding an additional sensor, a gyroscope.

As far as why accel/braking isn't the most reliable, this is most likely due to vibration on the device due to the road. This is probably something we could filter out on VIRB Edit's end.

We determine pitch and roll by applying equations to the accelerometer data written in the GPX file.

So far, we have been unable to reproduce the g-force cutting out.
What we can reproduce is the nonsensical information for pitch and roll during your motorcycle ride. Allow me to try to explain why this is.
When we first implemented pitch and roll using the accelerometer data from the VIRB, we began to notice that an accelerometer wouldn't accurately portray pitch and roll during certain activities. Because we only have an accelerometer available, our pitch and roll equations have to rely on the only force being applied to the camera is gravity. During slower activities, this tends to be an OK assumption therefore equations tend to result in reasonable data. HOWEVER, certain activities (like motorcycle riding) depend on other external forces to function. This is why you can lean your motorcycle in a turn and not fall over. During these activities the camera's accelerometer would be reading the combination of all these external forces, instead of just gravity. As stated earlier, our equations rely on the only force involved to be gravity, so these activities would result in nonsensical pitch and roll data displayed on your video.

The bottom line: The only way we could produce reasonable pitch and roll data during every activity would be by adding an additional sensor, a gyroscope.

As far as why accel/braking isn't the most reliable, this is most likely due to vibration on the device due to the road. This is probably something we could filter out on VIRB Edit's end.

Thanks for the detailed info. I appreciate the VIRB Edit team looking in to it. It sounds like one of my main goals for purchasing the camera cannot be realized. That's disappointing considering the camera cannot be returned.

I look forward to whatever you guys can do with VIRB Edit in the way of filtering to address other data.

Thanks for explaining the calculated roll and pitch.
It was something I was expecting not to work on a motorcycle.
It probably also doesn't work in an airplane taking a coordinated turn.

Thanks for explaining the calculated roll and pitch.
It was something I was expecting not to work on a motorcycle.
It probably also doesn't work in an airplane taking a coordinated turn.

Yes that is right. A gyro is necessary for these kinds of activities.