Saturday, 29 June 2013

Holux M1000C inaccuracy

This table shows comparative features of GPS loggers and notes that the Holux is low resolution. Could that explain the disappointing log data?

I've collated this example of performance to illustrate my concern. The lack of alignment of the Holux is not due to the Google photo as I've overlayed the GPX on both OS and OSM mapping on different platforms with the same result.

Three GPS logging devices and their respective tracks along a suburban track near home. Mapping is OSM, Google Earth and OS 1:25000 respectively. Software was Garmin Connect, GPS Visualizer and Android Viewranger. 

Holux binary logs record latitude and longitude as 4 byte,  32 bit floats. This should allow 7 to 9 significant figures of a decimal number which is plenty for real world logging.  It is also worth seeing this illustration of truncation errors.
5 minutes of capturing NMEA sentences followed by logging. Plotted as divergence in degrees from the average position indicted from each recording.
I have compared the standard NMEA data against the float32 capture data; the quantization of the latter is very evident. However note that the errors of both fall within 0.0001 degree for nearly all of the time.

A 0.0001 degree ring around a point in the outdoor swimming pool at Hathersage. Most points would be well within this margin.

This Google Map shows that this variation is little more than the width of a road and so the 'theoretical resolution problem' does not account for the real-world logging errors seen.

No comments:

Post a Comment