I took several test shots, trying to test the following variables:
- triggering with the on-camera shutter (2 sec. timer) vs. the app.
- the sharpness setting on or off,
- the auto angle correction on or off.
For each combination, I took several shots. I also took photos before and after shooting a video. Moreover I did the entire series twice just to see the repetition of any patterns. The gap between the two sets of tests was about 50 minutes.
For all of them, I stitched the shots using the Gear 360 Manager app, with the resizing setting turned off, and auto angle correction turned off (I reasoned that turning on auto angle correction could not possibly improve sharpness) though I don't know if that overrode the settings used when the shot was taken.
The ambient was cloudy weather, in late afternoon.
I found that the first photo was consistently sharper than the last photo, until I turned off the camera, left it alone, then turned it on again.
Let's look at 4 photos:
- the first photo from the first series (#410). ISO 100, f/2, 1/640.
- the last photo from the first series (#433). ISO 100, f/2, 1/400.
- the first photo from the second series (#434). ISO 250, f/2, 1/100.
- the last photo from the second series (#445). ISO 320, f/2, 1/100.
Comparing 410 and 433, we see that 410 is significantly sharper than 433, even though their exposures are almost identical (only 1/3 stop apart in shutter speed).
Comparing 434 and 445, we also see that 434 is significantly sharper than 445 (only 1/3 stop difference in ISO).
One possibility is that the blurriness has to do with shutter speed or ISO. To test this, I examined 433 and 434. Recall that 433 (the last in the first series) has a lower ISO and higher shutter speed than 434 (the first in the second series). If the blurriness had something to do with ISO or shutter speed, we would expect 433 to be sharper. But instead, 434 is sharper:
Another possibility is that this is some kind of stitching error. To examine this possibility, I compared the unstitched double-fisheye versions of 433 and 434 (433 on the left).
As you can see, even when comparing the unstitched versions, 434 is still sharper than 433. This means that the blurriness is with the original unstitched file. The blurriness is NOT due to a stitching error. It also appears that if there is a loss in sharpness in the image, the loss is not recoverable.
The only discernible pattern so far is that the first photo is definitely sharper than the last photo, and that the blurriness never reverses until after the camera is turned off. Based on this behavior, it appears that the most reasonable explanation is that the blurriness may be caused by the sensor and/or processor overheating. Going further, given that the Gear 360 does not stitch the images (stitching is done on a phone or desktop), I believe it is more likely to be the sensor overheating (rather than the processor).
In either case, IF it is true that the cause is overheating, one way to keep the Gear 360 images sharper is to avoid leaving it on unnecessarily. I will continue to investigate this.