(VIRB) heading is static if speed = 0 in gps_metadata/160 even if camera is turning

Did a little testing, for VIRB gps_metadata/160 messages, comparing "heading" (field 5) to the "velocity" (field 7), which is a vector [x, y, z] as opposed to the scalar "enhanced_speed" (field 4).

Whenever enhanced_speed hits 0, heading/field 5 is static. But doing a rough angle from velocity x, y values seems closer to reality. A bit surprising that velocity still show a [low] displacement value when enhanced_speed is 0? Anyway, here are examples from the same FIT-file  (code after "->" can be pasted into the Julia repl, practical as an advanced terminal calculator!), with enhanced_speed = 0 then > 0.

ENHANCED SPEED = 0

// id:   4 enhanced_speed        : UINT32([0])
// id:   5 heading               : UINT16([20443]) STATIC
// id:   7 velocity              : SINT16([-5, -12, -2])
                           -> rad2deg(atan(-5/-12)) = 22.6198

// id:   4 enhanced_speed        : UINT32([0])
// id:   5 heading               : UINT16([20443]) STATIC
// id:   7 velocity              : SINT16([-4, -12, -2])
                           -> rad2deg(atan(-4/-12)) = 18.4349

// id:   4 enhanced_speed        : UINT32([0])
// id:   5 heading               : UINT16([20443]) STATIC
// id:   7 velocity              : SINT16([-4, -11, -2])
                           -> rad2deg(atan(-4/-11)) = 19.9831

ENHANCED SPEED > 0

// id:   4 enhanced_speed        : UINT32([523]) 
// id:   5 heading               : UINT16([20606]) 
// id:   7 velocity              : SINT16([-22, -46, -2])
                            -> rad2deg(atan(-22/-46)) = 25.5560

// id:   4 enhanced_speed        : UINT32([487]) 
// id:   5 heading               : UINT16([19301]) 
// id:   7 velocity              : SINT16([-10, -47, -2])
                           -> rad2deg(atan(-10/-47)) = 12.0115

// id:   4 enhanced_speed        : UINT32([413]) 
// id:   5 heading               : UINT16([28277]) 
// id:   7 velocity              : SINT16([-40, 9, -1])
                           -> rad2deg(atan(-40/9)) = -77.3196

I'm a bit confused. Velocity seems to be the better indicator of actual heading here. Is there something I'm missing as to how the speed/velocity/heading values are derived and which sensor may be involved?

Also trying the magnetometer readings at the moment, but I'm not quite there yet (have calibrated values, but need to test I've actually implemented the calibration correctly, right now it doesn't add up). 

The virb is a fascinating device with all the available sensor data.

Only thing I can contribute to this is that I recently learned that GPS devices sometimes use the gps signal doppler-effect to calculate speed instead of relying on time between two lat/lon points, this is because it doesn't need a full lock onto satellite signals, just listens to the drift like a fire-truck siren racing by and seemingly changing tones even though it's the same.

If a device isn't moving, and it is maybe using doppler-effect, maybe it can't also calculate heading/bearing?

But if it is accelerator based, well then yeah maybe it's a bug.

Yes, it really does make me wonder where heading in gps_metadata is derived from. If the gyro or magnetometer are involved there shouldn't be an issue. I have continuous readings from those, and the magnetometer readings are not static around the same timespans.

The doppler thing is pretty dope, by the way (had to to look it up).

It's disappointing that no other device saves the magnetic compass to the 1-second records though.

Maybe bouncing on the wrist is too much for a watch or footpod to know which way it's oriented vs magnetic north but you'd think it's technically possible with some serious R&D, someone will do it eventually.

It occurs to me that using Stryd's "air power" field and the lat/long between the last two points it is possible to calculate wind speed though. Not sure how wildly off the accuracy would be but maybe within a couple of mph