Patent: Dual Pixel Phase Detect AF While in AI Servo

[LetTheRightLensIn]

will dual pixel improve DR? or only AF? It would be nice if it would

I have no idea, but you can bet Canon is not sitting on their asses. They can see the danger of Sony, Panasonic, Fujifilm. It's been a game of leapfrog for a long time, and Canon will leap ahead soon. That's just my guess.

Perhaps, but it almost seem like they put their money on all sorts of dual pixel AF type stuff instead of improving low ISO image quality. It also probably costs less money to do with stuff than make the fancy new assemblies required to improve image quality.

 

[3kramd5]

I'd certainly welcome the demise of that flapping mirror.

So would I, except for the fact that all the alternatives are dramatically worse.

For the time being, yes. But display and battery technologies march forward. Eventually, EVF may come without a significant penalty, and when that happens, I'll sign up for sure.

 

[Mt Spokane Photography]

After reading thru the patent, which is one of those that tries to explain the operation of the camera starting with the creation of the galaxy and each step after that, here is my understanding. I can be wrong.

Simply put, when taking a photo using this invention:

1. the camera does a conventional phase detect and the lens focuses as directed

2. The mirror raises and the camera does a fast comparison of the live autofocus and the phase detect AF

3. The accuracy of the first measurement is compared and the focus may be slightly adjusted if necessary.


Then, there is the part explaining how phase detect accuracy can be affected by mirror bounce, sub mirror bending, humidity, and temperature. Apparently, a temperature / humidity detector is used to provide correction to the phase detect AF but not the live AF. That's something I did not know.


From this, I gather that the live AF can operate very quickly if its only looking for tiny corrections in the focus, and since focus has been carried out first by the conventional phase detect system, the invention will further tweak focus accuracy.

The outcome is that focus errors caused by the high speed of the flapping mirror, and submirror while trying to capture images with a high FPS will be corrected. There is not time enough to allow those moving parts to stabilize when shooting at 12 fps, and AF accuracy suffers.

Will it happen in a 7D MK II? My opinion is that we will see it because its consistent with earlier rumors.

 

[Steve]

Unfortunately, because lenses longer than 40mm are already non-retrofocal, they don't benefit. Thus, most of your lenses stay the same size, but now require an adapter that reduces IQ subtly by adding another flex point, and adds another piece of hardware that you have to manage when out in the field. That's barely even an advantage, and for many people, is a disadvantage.

Where exactly are you getting the idea that lenses longer than 40mm for mirrorless cameras require mandatory adapters?

 

[lol]

It may even be able to AF based on shape and colour of object, which would be very cool if it can be made to work reliably!

I believe such a thing already exists in the 1D X as well as many higher end Nikon bodies. They use the metering sensor to provide that info. Bit disappointed to see this got cut out when they made the 5D3. Really hope it makes the 7D2.

 

[Marauder]

It may even be able to AF based on shape and colour of object, which would be very cool if it can be made to work reliably!

I believe such a thing already exists in the 1D X as well as many higher end Nikon bodies. They use the metering sensor to provide that info. Bit disappointed to see this got cut out when they made the 5D3. Really hope it makes the 7D2.

Yes, something like this does exist for the 1DX. I'm just wondering if this patent possibly represents a new way of implementing this by using the image sensor. The 7D II, being designed as a wildlife/action camera would be a very logical product for such a technological innovation.

 

[LetTheRightLensIn]

After reading thru the patent, which is one of those that tries to explain the operation of the camera starting with the creation of the galaxy and each step after that, here is my understanding. I can be wrong.

Simply put, when taking a photo using this invention:

1. the camera does a conventional phase detect and the lens focuses as directed

2. The mirror raises and the camera does a fast comparison of the live autofocus and the phase detect AF

3. The accuracy of the first measurement is compared and the focus may be slightly adjusted if necessary.


Then, there is the part explaining how phase detect accuracy can be affected by mirror bounce, sub mirror bending, humidity, and temperature. Apparently, a temperature / humidity detector is used to provide correction to the phase detect AF but not the live AF. That's something I did not know.


From this, I gather that the live AF can operate very quickly if its only looking for tiny corrections in the focus, and since focus has been carried out first by the conventional phase detect system, the invention will further tweak focus accuracy.

The outcome is that focus errors caused by the high speed of the flapping mirror, and submirror while trying to capture images with a high FPS will be corrected. There is not time enough to allow those moving parts to stabilize when shooting at 12 fps, and AF accuracy suffers.

Will it happen in a 7D MK II? My opinion is that we will see it because its consistent with earlier rumors.

Could be super awesome for the action shooters. (maybe for video too )

I just hope they manage to be able to catch up for landscape stuff. :-\

I just hope they manage to be able to catch up for landscape stuff. :-\ I guess, if they don't mess up the video, one could sell off all Canon FF bodies, buy 7D2 for action and Sony A7R for landscapes??

 

[LetTheRightLensIn]

It may even be able to AF based on shape and colour of object, which would be very cool if it can be made to work reliably!

I believe such a thing already exists in the 1D X as well as many higher end Nikon bodies. They use the metering sensor to provide that info. Bit disappointed to see this got cut out when they made the 5D3. Really hope it makes the 7D2.

I wonder how well the Canon implementation of color tracking works though. When was the last time you heard someone with a 1DX even bother to mention it? It also requires that all AF points be used for tracking at once with the Canon at least.

 

[Dylan777]

It may even be able to AF based on shape and colour of object, which would be very cool if it can be made to work reliably!

I believe such a thing already exists in the 1D X as well as many higher end Nikon bodies. They use the metering sensor to provide that info. Bit disappointed to see this got cut out when they made the 5D3. Really hope it makes the 7D2.

Yes, something like this does exist for the 1DX. I'm just wondering if this patent possibly represents a new way of implementing this by using the image sensor. The 7D II, being designed as a wildlife/action camera would be a very logical product for such a technological innovation.

It calls EOS iTR AF. I have this feature turned off when shooting bird in flight. It works well on kids with colorful clothes.

 

[scottkinfw]

The prospect of improved AF is exciting no matter what the technology. I can't wait to see what comes and how it actually performs in the field. I also can't wait to see what bodies it comes it. Hopefully the new sensors won't be hampered by the new technology.

sek

 

[dgatwood]

Unfortunately, because lenses longer than 40mm are already non-retrofocal, they don't benefit. Thus, most of your lenses stay the same size, but now require an adapter that reduces IQ subtly by adding another flex point, and adds another piece of hardware that you have to manage when out in the field. That's barely even an advantage, and for many people, is a disadvantage.

Where exactly are you getting the idea that lenses longer than 40mm for mirrorless cameras require mandatory adapters?

You're subtly misinterpreting what I said. What I said was that there's no benefit to the shorter flange distance for lenses longer than 40mm, and as a result, most Canon lenses (particularly longer lenses) are built exclusively in EF or EF-S format, and require an adapter to use them on an EF-M camera.

I would note, however, that Canon just released a 55–200 in EF-M format. That at least makes EF-M a little bit less absurdly broken as a platform. Up until last month, EF-M was pretty much a sad joke....

 

[jrista]

I'd certainly welcome the demise of that flapping mirror.

So would I, except for the fact that all the alternatives are dramatically worse.

Agreed. A viable alternative hasn't presented itself yet.

 

[wickidwombat]

I would note, however, that Canon just released a 55–200 in EF-M format. That at least makes EF-M a little bit less absurdly broken as a platform. Up until last month, EF-M was pretty much a sad joke....

damn mine must not have got the memo that it was a sad joke as it still takes great photos...
so does my IR converted one

but then i use the 11-22 90% of the time and occassionally my rokinon 8mm on it

 

[vscd]

When you add all those things together, IMO most folks would be better off with a DSLR, using live view for those rare occasions when an EVF provides an advantage. And IMO most of the folks who are really better off with mirrorless (the folks who only use short lenses) would do just as well with a fixed-lens point-and-shoot.

Wise words, spoken calmly. That's exactly my point of view, but try to have a chit chat with a recent fuji-fanboy. Useless... ;D

 

[Sabaki]

It may even be able to AF based on shape and colour of object, which would be very cool if it can be made to work reliably!

I believe such a thing already exists in the 1D X as well as many higher end Nikon bodies. They use the metering sensor to provide that info. Bit disappointed to see this got cut out when they made the 5D3. Really hope it makes the 7D2.

Yes, something like this does exist for the 1DX. I'm just wondering if this patent possibly represents a new way of implementing this by using the image sensor. The 7D II, being designed as a wildlife/action camera would be a very logical product for such a technological innovation.

Certainly seems like something Canon should put into the 7Dii. It's in the D810 for crying out loud and that's not an action type body.

 

[lw]

Would be nice if this could be applied via a firmware update to the 70D.

If it is all in the algorithms, and not new hardware components, perhaps that is possible.

 

[2n10]

Would be nice if this could be applied via a firmware update to the 70D.

If it is all in the algorithms, and not new hardware components, perhaps that is possible.

If it is algorithms then yes a firmware update would work. If there are hardware differences it may take a little time to get it right.

 

[Steve]

You're subtly misinterpreting what I said. What I said was that there's no benefit to the shorter flange distance for lenses longer than 40mm, and as a result, most Canon lenses (particularly longer lenses) are built exclusively in EF or EF-S format, and require an adapter to use them on an EF-M camera.

So your post was only outlining the problems of the EOS M system and not mirrorless technology in general?

 

[Marauder]

It may even be able to AF based on shape and colour of object, which would be very cool if it can be made to work reliably!

I believe such a thing already exists in the 1D X as well as many higher end Nikon bodies. They use the metering sensor to provide that info. Bit disappointed to see this got cut out when they made the 5D3. Really hope it makes the 7D2.

Yes, something like this does exist for the 1DX. I'm just wondering if this patent possibly represents a new way of implementing this by using the image sensor. The 7D II, being designed as a wildlife/action camera would be a very logical product for such a technological innovation.

Certainly seems like something Canon should put into the 7Dii. It's in the D810 for crying out loud and that's not an action type body.

Yeah, if they can get it to work well with fast moving subjects, especially birds, it would be a big benefit! If!!

 

[dgatwood]

You're subtly misinterpreting what I said. What I said was that there's no benefit to the shorter flange distance for lenses longer than 40mm, and as a result, most Canon lenses (particularly longer lenses) are built exclusively in EF or EF-S format, and require an adapter to use them on an EF-M camera.

So your post was only outlining the problems of the EOS M system and not mirrorless technology in general?

That part was, anyway.