- Joined
- Jul 5, 2014
- Messages
- 109
- Reaction score
- 20
Please excuse the long post...
I did an analysis of the DJI iOSD Mini and Flytrex data and it appears the Azimuth displayed by the iOSD is too high by about 4 degrees (based on an admittedly small sample set).
This could affect you if, say, you crashed your Phantom into a remote tree and don't have a tracker. Theoretically, the combination of Azimuth and Distance displayed by the iOSD (and captured by a DVR) *should* get you to within roughly 100 feet of the crashed Phantom at one mile distance. But with an azimuthal error of 4 degrees, this quadruples to about 400 feet.
Here is how I came to this conclusion. I hope I missed something because I want the iOSD reading to be correct.
I used RaceRender to overlay a flight video with both iOSD and Flytrex positional data:
[youtube]https://www.youtube.com/watch?v=UnA_0QwfT40[/youtube]
I took care to sync the main video to the iOSD video to the Flytrex data overlay. Since the Flytrex reads data right from the onboard GPS, I called this the raw "truth data". I reasoned that the iOSD is processing the same data so the two are internally consistent. The iOSD is computing the azimuth from the home point and current position of the "truth data".
I took four points from the video at various iOSD horizontal distances from home point:
1. 300 m (@1:04)
2. 601 m (@1:42)
3. 1200 m (@3:46)
4. 1470 m (@4:12)
Comparing these distance values to the Flytrex "truth" data shows only small differences in distance (4-7 meters) relative to the large horizontal distances (column D). This is good.
Now on to the azimuth data. At the four points in the video, I took the azimuth values from the iOSD and projected them by the displayed horizontal distance to determine the lat/lon (see http://www.movable-type.co.uk/scripts/latlong.html for how to do this). These values are in columns F and G (and also shown in the graphic as iOSD1-iOSD4). I compared them to the Flytrex-based lat/lon taken at the same time from the video (columns H and I). The Flytrex values are assumed to be "actual" position of the Phantom at that time (see Actual1-Actual4 in the graphic).
I used the home point and Flytrex positions to compute the "actual" azimuths for the four locations. These values are shown in column K. Subtracting the iOSD azimuths gives the discrepancy from truth (column L). It appears the actual azimuth is about 4 degrees less than the azimuth displayed on my iOSD with firmware v1.06, according to these data. Interestingly, the horizontal distances from home point reported by the iOSD and Flytrex are nearly the same. Only the azimuth is off. Btw, the magnetic declination at this location is +12.2 deg East. This *should* be accounted for by the algorithm that computes true azimuth. Even if it's not, it doesn't explain the 4 degree error.
Bottom line is that if you're using the iOSD as a "poor man's" GPS tracker, you might need to subtract a few degrees from the displayed azimuth to find your Phantom.
I did an analysis of the DJI iOSD Mini and Flytrex data and it appears the Azimuth displayed by the iOSD is too high by about 4 degrees (based on an admittedly small sample set).
This could affect you if, say, you crashed your Phantom into a remote tree and don't have a tracker. Theoretically, the combination of Azimuth and Distance displayed by the iOSD (and captured by a DVR) *should* get you to within roughly 100 feet of the crashed Phantom at one mile distance. But with an azimuthal error of 4 degrees, this quadruples to about 400 feet.
Here is how I came to this conclusion. I hope I missed something because I want the iOSD reading to be correct.
I used RaceRender to overlay a flight video with both iOSD and Flytrex positional data:
[youtube]https://www.youtube.com/watch?v=UnA_0QwfT40[/youtube]
I took care to sync the main video to the iOSD video to the Flytrex data overlay. Since the Flytrex reads data right from the onboard GPS, I called this the raw "truth data". I reasoned that the iOSD is processing the same data so the two are internally consistent. The iOSD is computing the azimuth from the home point and current position of the "truth data".
I took four points from the video at various iOSD horizontal distances from home point:
1. 300 m (@1:04)
2. 601 m (@1:42)
3. 1200 m (@3:46)
4. 1470 m (@4:12)
Comparing these distance values to the Flytrex "truth" data shows only small differences in distance (4-7 meters) relative to the large horizontal distances (column D). This is good.
Now on to the azimuth data. At the four points in the video, I took the azimuth values from the iOSD and projected them by the displayed horizontal distance to determine the lat/lon (see http://www.movable-type.co.uk/scripts/latlong.html for how to do this). These values are in columns F and G (and also shown in the graphic as iOSD1-iOSD4). I compared them to the Flytrex-based lat/lon taken at the same time from the video (columns H and I). The Flytrex values are assumed to be "actual" position of the Phantom at that time (see Actual1-Actual4 in the graphic).
I used the home point and Flytrex positions to compute the "actual" azimuths for the four locations. These values are shown in column K. Subtracting the iOSD azimuths gives the discrepancy from truth (column L). It appears the actual azimuth is about 4 degrees less than the azimuth displayed on my iOSD with firmware v1.06, according to these data. Interestingly, the horizontal distances from home point reported by the iOSD and Flytrex are nearly the same. Only the azimuth is off. Btw, the magnetic declination at this location is +12.2 deg East. This *should* be accounted for by the algorithm that computes true azimuth. Even if it's not, it doesn't explain the 4 degree error.
Bottom line is that if you're using the iOSD as a "poor man's" GPS tracker, you might need to subtract a few degrees from the displayed azimuth to find your Phantom.
