kfh
 lvl.2
Denmark
Offline
|
I decided to perform an experiment based on a recording 2160p/50fps of a static environment. File recording format data sheet can be downloaded here: https://drive.google.com/file/d/1lDUZpTuRCHggkCuSdewy4W54gSCtd5_8/view?usp=sharing
The resulting file is a 4 second cut. The first frame is the original, and all the following frames are the result of combining two original files. File 1 is original frame 0 to 3:49. File 2 is original frame 0 to 3:48 offset 1 frame. File 2 is subracted from File 1.
This should expose ’next to nothing’ in a perfect world. Alas, the world is not perfect, so what you see in frames 0:01 to 3:49 is “residual noise” caused by the one frame offset. Noise is not a constant, but ‘wiggles’ into different places on the sensor, as well as displaying different levels, which will leave a kind of faint “base noise level” in each frame. If nothing exceptional happens, this will be fairly constant, but this is NOT the case. Each 30 frames, the residual noise rises abruptly, gradually lowering over the following few frames. I have only subtracted frame 0 to 3:58 from frame 1 to 3:59. No level adjustment. No contrast adjustment etc. Just plain digital subtraction. Since there is no movement involved - at all - this should deliver "next to nothing" - except a faint more or less random noise.
Try starting single frame movements (easy in Apple QuickTime) from frame 28, 29, 30, 31, 32 etc.. to get a feel for what to look for.
In itself, nothing dramatic, EXCEPT that this should not happen at fixed intervals in a static recording with no movements at all! The effect can ONLY be caused by a constant 0.6 second periodic “interference”. It helps to view in a low(ish) light environment. This is a constant effect starting at each 30 frames in the 50 fps recording. The transition from frame 29 to frame 30 (or 1:09 to 1:10 or…) leaves no doubt, that something is happening (especially, compared to the most recent frames before the transition). The “sudden rise in noise” also affects the border area in the blown highlights (horizontally, left half, slightly above mid image).
The “effect” will be more dramatic - akin to abrupt exposure shifts - in "normal" recordings in the darker areas with movements (I.e. leaves and branches in a forest at some distance), as exposed by several videos posted in this forum on Pocket 2.
The result is:
Frame 1: Original environment.
Frames 2 to end of file: Each 30 frames, the residual noise rises abruptly (0:30, 1:10, 1:40 etc at 50fps).
Original ‘untreated’ subtraction file: https://drive.google.com/file/d/1n5Aw3BuU-2jNsFB8uYA8SWO0qL5DC-LA/view?usp=sharing
Enhanced noise levels file: https://drive.google.com/file/d/1AJGcCKKb3EwqxC5f4b2i7Q0ZsYXf3__l/view?usp=sharing
Each file has a duration of four seconds, and especially in the “enhanced” file, the effects of “highlight” pumping (period 0.6 seconds) is clearly visible.
A thorough discussion of the “design flaw” can be found here: https://forum.dji.com/forum.php?mod=viewthread&tid=233809&extra=page%3D1%26filter%3Dtypeid%26typeid%3D765%26typeid%3D765
The flaw is admitted here: https://forum.dji.com/forum.php?mod=redirect&goto=findpost&ptid=233809&pid=2409396
Another view on the flaw can be seen here: https://forum.dji.com/forum.php?mod=redirect&goto=findpost&ptid=233809&pid=2424619
’nuff said ;-)
This is a clear design flaw. It may be remedied by firmware, but… if it will ever happen is anyones guess.
I think this can become a clear case of refund of full amount by (re)seller due to manufacturer design flaw established within two years of sale according to European consumer law, unless DJI gets their act together, and - quickly - releases a firmware update, that quickly remedies this flaw in performance.
Regards
P.S. The files will be removed from download on June 21, 2021.
|
|