Home · Register · Search · View Winners · Software · Hosting · Software · Join Upload & Sell

Moderated by: Fred Miranda
Username   Password

  New fredmiranda.com Mobile Site
  

FM Forums | Canon Forum | Join Upload & Sell

1       2      
3
       end
  

Canon Files for Dual-Scale Column-Parallel ADC Patent
  
 
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.3 #1 · Canon Files for Dual-Scale Column-Parallel ADC Patent


snapsy wrote:
Isn't this what the Magic Lantern firmware uses in their recent Dual ISO feature? If so it's been implemented on Canon chips for some time (but not used).


Not really since only a few chips allow it and even then it's an entire line so you get reduced resolution and other problems. This is per pixel. It's also remarkably like what Emil Martinec posted to here and DPR a few years ago as a suggestion, noting that if you combine shadows in a scene shot at ISO1600 with the brighter tones shot at ISO100 you get quite a fine image with good DR on current Canons.



Aug 29, 2013 at 02:05 AM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.3 #2 · Canon Files for Dual-Scale Column-Parallel ADC Patent


Paul Mo wrote:
Autofocus.


Fair reason. Many don't use AF for video, but it can be helpful for at times, no doubt and I wouldn't mind if the 5D3 had usable video AF. I still prefer the compact package though and not needing to lug around a second giant system and spend doubly and I doubt it performs as well as a 5D3+ML for image quality though either or even a 1DX. Once they move PDAF to the higher-end bodies they should AF as well as the video cam too, of course that has yet to happen.



Aug 29, 2013 at 02:10 AM
johnctharp
Offline
• • •
Upload & Sell: Off
p.3 #3 · Canon Files for Dual-Scale Column-Parallel ADC Patent


It took a minute to dawn on me- but wouldn't this be a perfect application for DPAF? With one half shooting at a low ISO and the other at a higher ISO, intelligently combining the result for a standard resolution RAW or providing a 'double-resolution, dual-ISO' RAW file instead, would be very interesting.

Hell, I have to wonder why Sony isn't trying this with Exmor (or aren't they?). Splitting the pixels on a 24MP APS-C or 36MP FF sensor would result in a still very usable 12MP crop or 18MP FF image, with significantly higher highlight and shadow detail, wouldn't it?



Aug 29, 2013 at 02:12 AM
Matt Grum
Offline
• •
Upload & Sell: Off
p.3 #4 · Canon Files for Dual-Scale Column-Parallel ADC Patent


Fred Miranda wrote:
Thanks for posting this skibum5.
It's just a patent and may never see production...but it gives us hope that Canon will start paying attention to DR.


Things are going to get very interesting when the full frame Sony NEX comes out with a Metabones autofocus enabled adaptor for EF glass, particularly for landscape photographers if it has 36MP or more, 14 stops dynamic range and no AA filter...



Aug 29, 2013 at 12:40 PM
alundeb
Online
• • • •
Upload & Sell: Off
p.3 #5 · Canon Files for Dual-Scale Column-Parallel ADC Patent


johnctharp wrote:
It took a minute to dawn on me- but wouldn't this be a perfect application for DPAF? With one half shooting at a low ISO and the other at a higher ISO, intelligently combining the result for a standard resolution RAW or providing a 'double-resolution, dual-ISO' RAW file instead, would be very interesting.

Hell, I have to wonder why Sony isn't trying this with Exmor (or aren't they?). Splitting the pixels on a 24MP APS-C or 36MP FF sensor would result in a still very usable 12MP crop or 18MP FF image, with significantly higher highlight and shadow detail, wouldn't it?


There are severe disadvantages with split ISO compared to Exmor even if you get the resolution. With split ISO you can only use all the light in the overlapping tone range. In the highlights and shadows, half the light is thrown away, meaning one stop worse photon shot noise. Not so clean shadows and not so clean well-exposed skies or high key tones.



Aug 29, 2013 at 12:52 PM
johnctharp
Offline
• • •
Upload & Sell: Off
p.3 #6 · Canon Files for Dual-Scale Column-Parallel ADC Patent


alundeb wrote:
There are severe disadvantages with split ISO compared to Exmor even if you get the resolution. With split ISO you can only use all the light in the overlapping tone range. In the highlights and shadows, half the light is thrown away, meaning one stop worse photon shot noise. Not so clean shadows and not so clean well-exposed skies or high key tones.


I think... you missed the point entirely.



Aug 29, 2013 at 02:34 PM
alundeb
Online
• • • •
Upload & Sell: Off
p.3 #7 · Canon Files for Dual-Scale Column-Parallel ADC Patent


johnctharp wrote:
I think... you missed the point entirely.


Statements like that are quite arrogant and not very helpful.

So you didn't mean one half of the sensor shooting at high ISO and the other half at low ISO?
Or what did I miss?




Aug 29, 2013 at 03:39 PM
 

Search in Used Dept. 



johnctharp
Offline
• • •
Upload & Sell: Off
p.3 #8 · Canon Files for Dual-Scale Column-Parallel ADC Patent


alundeb wrote:
Statements like that are quite arrogant and not very helpful.

So you didn't mean one half of the sensor shooting at high ISO and the other half at low ISO?
Or what did I miss?


And I'm not trying to be arrogant or not helpful- but I can't really figure out where you're coming from, though it sounds like you're talking about ML's implementation of dual-ISO, where I'm talking about using it on the '40MP' Canon APS-C sensor in the 70D and the higher-resolution Exmore 24MP and 36MP sensors.

On the Canon DPAF sensor, it makes sense, as you'd still get the full light intensity- you'd just be measuring it in two different ways, for each photo-site. While this is similar to what ML is doing with Dual-ISO, it's different in that there are twice as many photo sites, and it'd be a first-party implementation and not a third-party hack (supposing ML doesn't enable it in the 70D before Canon).

I think that there's real hope there for Canon to augment their lower native per-pixel DR with this method on DPAF sensors by intelligently reading each photo-site half at various ISO speeds, providing greater sensitivity in darker areas, while lowering sensitivity in the highlights, effectively widening the DR, and creating a 'single-shot HDR' style RAW that records far more detail than we've yet seen.

For the Exmor, my only observation is that the resolution is already very high- and it's mostly useless, as neither Sony, nor Nikon, nor Pentax, nor anyone else that ships Exmor-equipped cameras has a robust line of lenses that can truly take advantage of that extra definition. Thus, it would make sense to set up a dual-ISO mode that halves the resolution and uses adjacent photo-sites to build a higher-DR, half-resolution RAW, at 12MP for the cropper and 18MP for the FF. You could guarantee that while you're getting lower output resolution, you'd still be close to the effective resolution of the most popular lenses (aside from the 14-24, and maybe a few primes, especially the Sigmas), but you'd still have a plenty usable image that holds significantly more contrast detail.

I'm not saying that it's feasible, and it does seem that Nikon cameras at least are fairly gimped when it comes to on-body processing compared to Sony and Canon; and Sony doesn't make an FF MILC or DSLR with their top-end sensor, just the SLT A99, so it's not something I'd expect to see soon- though it'd certainly be cool to see it in a next-gen Pentax K-5 III(s) using the 24MP APS-C Exmor.



Aug 29, 2013 at 09:42 PM
buggz2k
Offline
• • •
Upload & Sell: Off
p.3 #9 · Canon Files for Dual-Scale Column-Parallel ADC Patent


I really wish they'd hurry up and use all this good stuff they are holding back/out on.
Sony is kicking ass.
And I just LOVE my FujiFilm X-E1 system (body and lenses) with XTrans sensor!
While I still hang on to my beloved 5DMkII, I don't see me buying another Canon body until something TRULY kick ass and state of the art comes out.
Tired of the meager old "good enough" from Canon.



Aug 30, 2013 at 01:35 AM
johnctharp
Offline
• • •
Upload & Sell: Off
p.3 #10 · Canon Files for Dual-Scale Column-Parallel ADC Patent


A new sensor pattern, either non-Bayer, or full-on Foveon, would be nice, if it would mean that they could remove the AA filters completely. A Foveon design with similar-sized photo-sites to their current sensors would be incredible, really, as you know the RAW converter houses would be on top of that; they'd definitely get better support than Fuji is getting for X-Trans and Sigma for the FoveonX3.

But that 16MP X-Trans is definitely something amazing. Along with actually good lenses (I had a brief moment of Pentax envy the other day, the K-5 II is a nicely setup camera, until I looked at the performance of all of those neat pancakes...), Fuji is really only behind on autofocus performance- just like everyone else that is doing CDAF or HAF, including Canon on every Live-View implementation except on the 70D.



Aug 30, 2013 at 01:55 AM
Pixel Perfect
Online
• • • • • •
Upload & Sell: On
p.3 #11 · Canon Files for Dual-Scale Column-Parallel ADC Patent


johnctharp wrote:
Fuji is really only behind on autofocus performance- just like everyone else that is doing CDAF or HAF, including Canon on every Live-View implementation except on the 70D.


Well Olympus and Panasonic aren't behind on CDAF, they are way ahead of the others with Fuji and Samsung trailing even Sony. Of course the bug bear of CDAF is still moving objects and a scheme like Canon's DPAF would be preferable for mirrorless cameras or at least a hybrid scheme.



Aug 30, 2013 at 02:01 AM
alundeb
Online
• • • •
Upload & Sell: Off
p.3 #12 · Canon Files for Dual-Scale Column-Parallel ADC Patent


johnctharp wrote:
On the Canon DPAF sensor, it makes sense, as you'd still get the full light intensity- you'd just be measuring it in two different ways, for each photo-site. While this is similar to what ML is doing with Dual-ISO, it's different in that there are twice as many photo sites, and it'd be a first-party implementation and not a third-party hack (supposing ML doesn't enable it in the 70D before Canon).

I think that there's real hope there for Canon to augment their lower native per-pixel DR with this method on DPAF sensors by intelligently reading each photo-site half at various
...Show more

There is a principal difference between "Exmor" and "the patent" on one side, and "ML" or "Dual pixel Dual ISO" on the other side with respect to light gathering. While "the patent" records the full light intensity before ISO split, "Dual Pixel Dual ISO" does split the light before photon conversion. There is no way you can get Dual Pixel to record the full light intensity in two different ways. A photon does not convert twice in a dual pixel, it either goes into one half-pixel or the other, meaning each half-pixel only sees half the light.

If the Dual-Pixel implementation was in such a way that the two pixel halves could be electrically merged independent of the read-out transistor connected to each photodiode, this infrastructure could be an implementation of "the patent". But since both readout channels need to be activated just to get all e- converted photons out of the photosites, it can not be used to implement "the patent".

Yes there are some benefits to a "Native ML" implementation, like no loss in resolution (full 20 MP resolution), but that will still be at the cost of halved light sensitivity in the highlights and shadows.



Aug 30, 2013 at 04:40 AM
johnctharp
Offline
• • •
Upload & Sell: Off
p.3 #13 · Canon Files for Dual-Scale Column-Parallel ADC Patent


Good response, thanks alundeb!

I wasn't aware that DPAF currently only reads one half of the pixel, instead of both halves- but that does give us hope that Canon could revise the technology, especially with larger FF pixels, to give both DPAF and full dual-ISO support at the same time.



Sep 01, 2013 at 08:52 PM
1       2      
3
       end




FM Forums | Canon Forum | Join Upload & Sell

1       2      
3
       end
    
 

You are not logged in. Login or Register

Username   Password    Retrive password