The
Mini 3 Pro is definitely superior in low-light and night time video recordings. The 1.7 aperture more than makes up for the smaller sensor. I've done numerous tests with both.
Air 2s can't come close, even using a noisy 6400 ISO setting.
Good points, but there is some confusion, on both of our parts.
First, the
Mini 3 Pro has a larger sensor than the
Air 2S, not smaller (my mistake). However it's pixel count, on the surface, seems higher, which would seem to mean mean the pixels are smaller, thus collecting fewer photons, resulting in higher noise and worse low light performance. That would normally be a problem, but for two other factors. First, the lens is a stop faster. Not a lot, just one stop, but in the right direction. Second, there's clearly some processing going on, and we're not looking at just the raw pixels. You can see this by noting the way DJI specifies the resolution: "Effective Pixels: 48MP". What's an "effective pixel"? The result of processing raw pixels. So is the actual sensor resolution higher? I doubt it, it's probably close to the same, or less than the "real" resolution of the
Air 2S, and more likely it's a processing trick to produce an "effective" higher resolution from fewer pixels. If that weren't true, the actual performance per pixel would be much worse than the
Air 2S.
Then there's the actual pixel size. The
Air 2S has 2.4um pixels, and while never stated in official DJI specs, it has been published that the
Mini 3 camera also has 2.4um pixels. So, while the data is of questionable parentage, no real advantage there either way in terms of photon capture. However, if the pixel size is the same but the sensor is slightly larger, that means there are more of them, right? Oh boy, does this get confusing.
Because of the processing going on in the
M3P, and the way DJI states the specs, it's really hard to tell what the actual pixel size and resolution of the sensor is. I would guess than the 12MP mode is telling us something, possibly that the raw unprocessed sensor is about 12MP, which would result in physically larger pixels, capturing more photons, and a true advantage in low light, but a slight disadvantage in unprocessed resolution. But hey, I'm all for less pixels and better low light, I wish we could do more of that in DSLRs, frankly. But again, image processing to the rescue.
So, larger sensor, fewer (but larger) pixels, faster lens, and image processing, and you've got your low light performance on the
M3P.
The slight negative compromises are, you give up 5.4K video (not really a huge deal), and you give up actual d-log, and get cinelike instead, also probably not that big a deal.
So yes, the camera is better in low light, but that doesn't make it better in every way. I haven't seen any tests, and haven't done my own, to see just how good that image processing is. Most processing that has to happen fast enough for video has to be compromised, but I guess we'll see.