Concerns

[karlblessing]

AI will even the score in less than 5 years.Your Apple/**** Tracy watch will equal full frame sensors.Well if the full frame sensor is not AI assisted.

Would have already been at that point if it were true. You can't really beat the physics of the lens that way.

 

[Droning on and on...]

What does the stacked sensor do for the video side especially in lower light or lower contrast conditions?

Do you know what he means by "stacked sensor"?

 

[old man mavic]

Apparently the stacked sensor is a CMOS ,with additional layers of circuitry stacked above the photodiode layer that give a better image quality and also a quicker response time that enables much more data to be collected ,so giving a clearer image
and it also improves low light resolution
just my take on the info i have found

 

[VenomXts]

I've seen a few reviews that I knew would be available at 9:05 am this morning. Apparently the smaller sensor on the Air 3 has new technology that more than makes up for the smaller size.
I've seen low-light comparisons between the new model and the Air 2s and it's absolutely no contest in favor of the Air 3. And some reviewers have gotten a legitimate 40 minutes of flight time already.
That's twice as long than I've ever seen on my now sold Air 2s. I bought the new model first thing this morning.

which video have you see on the low light side of things.. that's my biggest turn off in moving from my air2s to this and would love to watch the video.

 

[Macabre]

Apparently the stacked sensor is a CMOS ,with additional layers of circuitry stacked above the photodiode layer that give a better image quality and also a quicker response time that enables much more data to be collected ,so giving a clearer image
and it also improves low light resolution
just my take on the info i have found

None of that is given hower. Stacked sensor doesn't mean it will have better image quality over another type of sensor, nor it will have better low light or a clearer image.
It does however have a much faster readout speed so it's better for videos and shooting fast action with electronic shutter.

General rule of thumb when buying cameras is that if you buy a great camera and bad lens, you've wasted your money.
Often a lens is more important than the camera.

 

[wco81]

If the price is higher than the Air2S they must have some features and/or better video and photo quality which would make people buy the new product.

Or maybe they just think the Air 3 name will automatically make people think that it’s better than the Air2S and He Che worth more moneY.

 

[waynorth]

Apparently the stacked sensor is a CMOS ,with additional layers of circuitry stacked above the photodiode layer that give a better image quality and also a quicker response time that enables much more data to be collected ,so giving a clearer image
and it also improves low light resolution
just my take on the info i have found

Not quite correct. A stacked CMOS is similar to a BSI (Back Side Illuminated) sensor, but stacks the signal processor and its ultra-fast DRAM memory into the same silicon. This makes readout speeds very fast.
The circuitry is placed below the photodiodes. First are the microlenses, then the Bayer filter, then a light receiving surface, then the photodiodes, and finally two layers of circuitry.
A stacked sensor does not necessarily improve image quality by itself, and does not give a "clearer" image. But because of the much faster readout speed, the designers can improve signal processing and for example apply better noise reduction in low light/high ISO.

 

[karlblessing]

If the price is higher than the Air2S they must have some features and/or better video and photo quality which would make people buy the new product.

Often times that's not how it works. They may market differences as being better or just newer, but doesn't mean it must have something better.

For example, it now has omnidirectional obstacle avoidance, meaning sideways flight should now be protected to some degree. They didn't change the number or placement of sensors from the previous version far as I could tell but changed the lenses so that supposedly it can see more in a fisheye sense. How well it actually detects will be a different story we need to see in real world use. If might actually end up being inferior in the original directions, but still advertised as better simply because it's omnidirectional now.

 

[Torque]

which video have you see on the low light side of things.. that's my biggest turn off in moving from my air2s to this and would love to watch the video.

 

[KLRSKIR]

Finally, sensor size doesn't matter. This new camera is sweet!

 

[Droning on and on...]

Not quite correct. A stacked CMOS is similar to a BSI (Back Side Illuminated) sensor, but stacks the signal processor and its ultra-fast DRAM memory into the same silicon. This makes readout speeds very fast.
The circuitry is placed below the photodiodes. First are the microlenses, then the Bayer filter, then a light receiving surface, then the photodiodes, and finally two layers of circuitry.
A stacked sensor does not necessarily improve image quality by itself, and does not give a "clearer" image. But because of the much faster readout speed, the designers can improve signal processing and for example apply better noise reduction in low light/high ISO.

Thank you, @waynorth, I've become exhausted lately around here dispelling the sometimes laughable misinformation that gets disseminated – sometimes with such authority

And what you (very accurately) described can be found in seconds by googling "stacked sensor". Where people get the ideas they do about technology and engineering mystifies me. Magic? Astrology? Ghosts? I get it. Technology? What, do you people consult your psychic for technical explanations or something?

As for image sensor physics, size is not the top-line characteristic for producing performance. There is a naive, simplistic belief that exposure area is by far the most important factor in sensor performance. This is false. Very much so.

It's based on what was important with film. As a "sensor" film is very simple. You can't dynamically change it's sensitivity for example. ISO 100 film is ISO 100. Period.

The most important factors in sensor performance are dynamic range, and noise floor (the two are related, but different). A smaller sensor with the same resolution but wider dynamic range, higher saturation point, and lower noise floor will produce better results than a larger sensor. The same photon flux can be focused on a 0.77" sensor as a 1" sensor. Same photons. Assuming the same lens, there will be no more light "gathered" by the 1" sensor than the 1/1.3" sensor.

At that point things shift to the engineers. Technology keeps advancing. The new dual-preamp sensors get a lot of grief from the "professionals", but most of it is the usual cynicism of what they don't understand.

I'm old enough to recognize this as the same phenomenon that occurred with "audiophiles" when digital format CDs came out. There were some legitimate shortcomings when the technology was new. Of course! As always, those minor issues the audio "pixel peepers" were complaining about got fixed.

Does the relatively new dual-ISO technology in the Omnivision sensor produce a rec.2020 HDR compliant video? No. So what? It does produce a substantially wider dynamic range image than a larger sensor without it. That's the point.

 

[Droning on and on...]

For example, it now has omnidirectional obstacle avoidance, meaning sideways flight should now be protected to some degree. They didn't change the number or placement of sensors from the previous version far as I could tell but changed the lenses so that supposedly it can see more in a fisheye sense. How well it actually detects will be a different story we need to see in real world use. If might actually end up being inferior in the original directions, but still advertised as better simply because it's omnidirectional now.

This reminds me once again why non-engineers should avoid detailed discussions about engineering.

For example thinking obstacle avoidance has anything at all to do with sensor count, and placement. It doesn't.

Near perfect 360 OA in all directions could be accomplished with a single camera. Can anyone guess how? (engineers not allowed to answer).

What's the single most important component in the drone that influences the accuracy and function of OA? (again, you engineers keep your lips zipped )

 

[karlblessing]

So it seems DJI has confirmed that the sensors on the Air 3 is the same ones used on the Mini 3, just with updated software. So yea... that's a no for me.

 

[Felix le Chat]

Nope.

Comparisons between the Mini3P ans A2S have people lining up on both sides, the tech in the dual-amp Omnivision sensor is several generations more advanced than the 1" 20MP A2S sensor and demosaicing software can compensate for the larger interpolation errors with a quad-bayer filter.

There are valid counterpoints to many of these arguments, so there's plenty of debate. What isn't debatable is a direct comparison of sensor size as a means of comparing performance and capability is simplistic, and not very helpful.

Personally, I find I like the results from my Mini3P over my A2S more often than not. So, for me, I don't feel like I'm losing image capability with the Mini3P, and hence my A2S doesn't see a lot of action these days.

The 0.77" (1/1.3) mini 3/3pro sensor leaves my 1" 20mp Mavic 2 pro well behind as far as RAW still quality is concerned. Smaller doesn't necessarily mean poorer. The dual 0.77": dual focal length camera setup hanging beneath the A3 has guaranteed that as soon as all the usual first release bugs and foibles have been ironed out.... The Air3 + RC2 will be joining the rest of my girls.

 

[wco81]

So it seems DJI has confirmed that the sensors on the Air 3 is the same ones used on the Mini 3, just with updated software. So yea... that's a no for me.

The base SKU of the Mini 3 is $469. The base SKU of the Mini 3 Pro is $759. The base SKU of the Air 3 is $1099.

So the Air 3 uses the same sensor or the same-sized sensor as the Mini 3, which is less than half the price?

Or did you mean the Air 3 has the same sensor as the Mini 3 Pro, which is closer in price?

That's astounding either way, if they're using the same sensor, not just the same-sized sensor, across products with such a wide range of prices.

 

[waynorth]

So the Air 3 uses the same sensor or the same-sized sensor as the Mini 3,

I am not sure it's the same sensor. As far as I know the sensor in the Air 3 is a stacked BSI CMOS, while the sensor in the mini 3 is a non-stacked BSI sensor.
That might be a big difference, as a stacked sensor has much faster readout of data, and that can improve image processing dramatically, not least in noise reduction. If the engineers have done a good job, the image quality from the Air 3 could be surprisingly good for a 1/1.3 sensor.
And the examples I have seen so far have been very impressive.

 

[KLRSKIR]

Same "sized" sensor. Very different technology.

 

[RangerEd]

Doing a little research here's a statement from New Camera:
"So this is the main advantage of a stacked CMOS sensor: to improve the readout speed of the camera. The stacked technology doesn’t help the camera sensor improve its image quality in any way. It is just a tool to upskill the readout speed which in turn reduces the rolling shutter effect."
It's an easy few minutes read ... http://thenewcamera.com/what-is-stacked-cmos-sensor/

And here's an article from another technical company.

Why Stacked Sensor is the Future of the Camera Industry? - Tech Inspection

Why are stacked CMOS sensors considered to be the future of the camera industry? Let’s dive deep into the matter.

techinspection.net

My take is that it doesn't improve on the image quality. However, because of faster access times to the sensor, the camera is able to process all the pixels without resorting to pixel binning. In other words, a 48MP sensor will be processed as 48MP, and not reduced to some resolution that the camera processor can handle. Again, that's my take.

 

[Droning on and on...]

The sensor in the A3 also is back-lit, which DOES improve image quality.

This is not the same sensor as in the Mini3P, as I suspected (other threads). Not surprising, it's a newer generation of the Omnivision 1/1.3" sensor in the Mini3P, with some significant technology improvements.

Not surprisingly, initial reviews seem to be coming down on the side of better image quality, DR, and noise floor than the 3 year old 1" in the A2S.

Now, all things being equal, would a 1" version of this technology outperform the 0.77"? Of course. However that's not the debate over size we've been having.

Once again, it's not the size of your silicon that matters, it's how you use it

Also faster readout can't directly improve captured pixels, but it can enable higher quality video at higher frame rates where tricks like reducing color resolution and other means of decimation have been used in the past with slower sensors to read out at those rates.

As they say, it's complicated

 

[wco81]

One inch stacked and backlit sensors are available.

But they chose the smaller one, maybe so they can fit in the telephoto.