Last bets regarding image quality vs 2S and Mavic 3

[karlblessing]

Thank you for the comment. Keep in mind that for the photo portion of our comparison all the footage is recorded at 16 bit (on both cameras). The 8bit and 10bit sampling only apply to videos. I just checked our RAW data and all the footage you see in our videos was recorded at 8 bit with the Air2S. I will double check but I'm pretty sure the 10-bit sampling is only available in LOG footage, which we did not use for this comparison.

More likely 10 or 12-bit native , but your only choices after getting it into a raw processing software is to either put it out as 8 bit or 16 bit supporting format. Not a lot of cameras provide 14-bit raws (and only very very few are going to give you a full 16bit gamut out of the sensor).

Since the Mavic 3 is doing 12-bit for the dng raw files I doubt either the mini 3 or air 2s exceed that and are likely 10bit or 12bit for the raw photo capture too.

Also since both wouldn't do more than. 4:2:0 on the chroma subsampling for video, even if either were higher than 10bit on photo capture, a lot of the information is lumina and not chroma, sharpeness/focus would still be able to be compared if the contrast was matched at the same shooting mode.

 

[test2000Anafi]

Also; the footage that real pros (with mini3 pro) put out exponentially exceeds what even 100 amateurs put together will ever produce, while it’s those amateurs that argue 24/7 about which log settings trumps which and why you never should use 14 bit over 12 bit and whatever lol. I’m guilty of this as well but 95% of all forum talk can be distilled into ‘pride of ownership’.

You saw it on one of these boards. When mini 3 appeared sharper they were like “well, sharpness is not always good because sometimes it looks too good!”
Whereas if their own drone was sharper they would be saying
“Omg look at that blurry footage of the other drone!! You need real sharpness”

What they say has no real baring so it would be cool if some comparisons did not have the labels on them. That way the confirmation bias and pride of ownership would not muddy the discussion as to what is the better image

 

[pilotinstitute]

More likely 10 or 12-bit native , but your only choices after getting it into a raw processing software is to either put it out as 8 bit or 16 bit supporting format. Not a lot of cameras provide 14-bit raws (and only very very few are going to give you a full 16bit gamut out of the sensor).

Since the Mavic 3 is doing 12-bit for the dng raw files I doubt either the mini 3 or air 2s exceed that and are likely 10bit or 12bit for the raw photo capture too.

Also since both wouldn't do more than. 4:2:0 on the chroma subsampling for video, even if either were higher than 10bit on photo capture, a lot of the information is lumina and not chroma, sharpeness/focus would still be able to be compared if the contrast was matched at the same shooting mode.

The RAW photo EXIF indicates 16-bit depth for both cameras. Definitely 4:2:0 subsampling for the videos.

 

[karlblessing]

The RAW photo EXIF indicates 16-bit depth for both cameras. Definitely 4:2:0 subsampling for the videos.

I highly doubt it's actually 16bit, regardless of what the exif says, some of the high end Canons will say 16bit in the exif but is actually 14bit.

You can't rely on the exif because it's either going to say 8bit it 16bit.

Having either of those drones actually be 16bit would be unprecedented and very expensive to provide, not just in a drone but in a camera in general.

DJI Mavic 3

www.hasselblad.com

If you have access to a mavic 3 you may notice the exif says the same but is actually 12-bit as mentioned above.

 

[pilotinstitute]

I highly doubt it's actually 16bit, regardless of what the exif says, some of the high end Canons will say 16bit in the exif but is actually 14bit.

You can't rely on the exif because it's either going to say 8bit it 16bit.

Having either of those drones actually be 16bit would be unprecedented and very expensive to provide, not just in a drone but in a camera in general.

Yeah that's what I thought too. How can you ever be sure what you get though? This is important information we want to include when we present our data so viewers have the right information to make a decision.

 

[karlblessing]

Yeah that's what I thought too. How can you ever be sure what you get though? This is important information we want to include when we present our data so viewers have the right information to make a decision.

Well one way is logical to assume that they're not going to give you more than their very expensive Mavic 3, which is outlined as 12-bit by Hassselblad.

DJI Mavic 3

www.hasselblad.com

You can know for sure by knowing exactly what sensor is used. The sensor whitesheet outlines those numbers.

 

[boldblue737]

Thank you for the comment. Keep in mind that for the photo portion of our comparison all the footage is recorded at 16 bit (on both cameras). The 8bit and 10bit sampling only apply to videos. I just checked our RAW data and all the footage you see in our videos was recorded at 8 bit with the Air2S. I will double check but I'm pretty sure the 10-bit sampling is only available in LOG footage, which we did not use for this comparison.

Thank you, I assumed as much. In time I’m sure you and other reviewers will test head to head LOG files with LUTs and grading. You can see in your 4x zoom how much smoother the Air 2S footage is (less noise) and any slight variation in sharpness I think can be remedied in post. Add in the LOG/10 bit and color grading should have better results on the Air 2S, but honestly how many do that regularly on any drone let alone a mini sub $1000? The Mini 3 is great.

 

[EpicFlight]

Thank you for the comment. Keep in mind that for the photo portion of our comparison all the footage is recorded at 16 bit (on both cameras). The 8bit and 10bit sampling only apply to videos. I just checked our RAW data and all the footage you see in our videos was recorded at 8 bit with the Air2S. I will double check but I'm pretty sure the 10-bit sampling is only available in LOG footage, which we did not use for this comparison.

It may be more apples/apples to cripple the 2S for your comparison, but it's pretty useless data for what is being debated: Which drone produces higher quality footage?

That requires comparing each at their best, which means graded, 10-bit, 5.7k video on the 2S.

Which beats the snot out of anything that's been posted from the Mini 3, so far.

 

[MaximusCZ]

Given exif can report higher than actual bit depth, I presume its, sadly, not possible to exif report smaller than actual bit depth? I ask because I saw the sensor number (of mini 3 cam) somewhere, and there listed Omnivision OV48C have this in its features on manufacturer's site:

Supports output formats:
– 10-bit RGB 4C non HDR
– 10-bit RGB Bayer non HDR
– Up to 14-bit Bayer HDR

 

[globetrotterdrone]

I think you can write any flapdoodle into the EXIF you like, but DJI will still certainly sync their entries to the hardware that produced it logically.

Higher bit depth might just derive from the fact, that the data is stored in a "bigger container", like you could number 1 specify in short, float, double, long double in data types and might not specify the original depth it was recorded. But I may be mistaken here.

 

[karlblessing]

Given exif can report higher than actual bit depth, I presume its, sadly, not possible to exif report smaller than actual bit depth? I ask because I saw the sensor number (of mini 3 cam) somewhere, and there listed Omnivision OV48C have this in its features on manufacturer's site:

The 14 bit Bayer is probably how they're achieving 48mp like with the original Air 2, basically a sort of trick to capture neighboring days to up sample more color data, which is probably only used in specific applications, rest of the time saves raw 10bit sensor data in a large container. (Sort of how some cameras do sensor shift merges to create a higher resolution file with higher bit depth greater than the sensor is capable of in a single capture ).

The Sony sensor they used in the original Air 2 had a feature where it would do a quad pixel capture and use that data to create lower resolution raw that had better color fidelity and lower noise due to using the data of 4 pixels as a single capture point.

Far as cleaner noise on the air 2s , they do a Temporal denoiseing to the raw data in drone before saving the raw file. If that is always on it could soften results.

For the consumer aimed benefits they should do both cameras in auto capture to jpeg, since even if you could produce a better image in raw out of the air 2s it's not going to be something the point and shoot equivalent folks are going to care to utilize, they'll shoot raw and then ask why is it so flat and soft in the raw software defaults.

I do wonder how it would compare if the same software processing is done in the air 2s or if the mini 3 had the ability to turn off the increased sharpening beyond the sensor resolution.

 

[EpicFlight]

You saw it on one of these boards. When mini 3 appeared sharper they were like “well, sharpness is not always good because sometimes it looks too good!”
Whereas if their own drone was sharper they would be saying
“Omg look at that blurry footage of the other drone!! You need real sharpness”

Stop misrepresenting what others have said. No one said that, not even close.