Crazy GPS log/kmz (hundreds of 6000km jumps mixed with actual data)

Hello.

I’m new in this forum, also new in flying/modding quadcopters and, to make it even worse, English is not my mother tongue (so please, please, be patient :-s)

Bought a CX-20 (with APM) a few weeks ago, and after a few short trial flights (and of course some glue and spare parts) I downloaded the dataflash logs to analyze them. I was very surprised to discover that the CX-20 can travel through hyperspace, jumping from Spain to the middle on the Atlantic Ocean and back some hundreds of times in ten minutes.

In spite of the fact that I’m trying to fit seats on it for my wife and I and spend next weekend in Jamaica, I am curious about this amazing skill of a 240€ aircraft (I bet Lando Calrissian paid a lot more for the Millenium Falcon). Any help will be very appreciated.

In the logs, latitude jumps about 0,21° (23,33km) and/or longitude jumps about 53,69° (almost 6000km) each time, to immediately return to the actual position.

The flight experience was not affected at all, didn’t notice anything that looked strange (to a newbie as me). The quad was responsive and more or less stable in all modes, including loiter. After editing the log (erasing the lines with wrong lat. or long.) I succeeded to make a new kmz file that showed my actual drunk-hornet style flight path.

I tried a mod I found in Youtube, moving GPS to the roof of the upper hull and covering it with aluminum tape, but obtained the same jumps.

If I can find how to do it, I’ll upload some screenshots and the corresponding log.

Best regards.


Link to the log: https://drive.google.com/file/d/0B1DPrLm2DIC3MHJ3ZXI3UV9GUmc/view?usp=sharing

Screenshots of some flights (last one related to the log) and Lat/Long graph:

Changed GPS unit, same issue… But I noticed that telemetry logs are OK, just dataflash logs are bad. Maybe the problem is in the log recording? Or maybe tlogs are somehow filtered/fixed? Or just the lower sampling rate makes the (relatively few, compared with good data) “jump points” to be missed by tlogs?

Does anyone has the same problem, or any idea about it?