Patent: Expanded dynamic range using DPAF sensors

Canon Rumors Guy

Canon EOS 40D
CR Pro
Jul 20, 2010
10,779
3,158
Canada
www.canonrumors.com
Canon News has uncovered a patent showing how to increase dynamic range using DPAF equipped sensors.
Here’s a breakdown of Canon Japan Patent Application 2019-129491:
The basic idea is that during read mode, one half of the pixel is amplified at one level, and the other half is amplified at another level.  reading the two simultaneously and processing can increase the dynamic range of the sensor.  Basically dual ISO.
It should be noted that the patent does mention in-vehicle sensors, but can also be applied to any other type of sensor as well.

Continue reading...


 
  • Love
Reactions: 1 user

SecureGSM

2 x 5D IV
Feb 26, 2017
2,360
1,231
  • Like
Reactions: 3 users
Upvote 0
as it was discovered 3 years ago:


No, this is completely different. Current dual pixel RAW files contain the other pixel data in the secondary frame and the combined (by adding the two pixels' data together) in the main frame. But there is no difference in the amplification of the original dual pixels. With the dual gain system they could achieve a bigger than 1 stop boost.

I guess the maximum achievable benefit would be depending on the noise characteristics of the sensor. This would not help for photon shot noise, but certainly would help if read noise is a factor.
 
  • Like
Reactions: 4 users
Upvote 0
as it was discovered 3 years ago:

I remember that - it's an interesting feature that I'd considered using in some rare occasions, but it just doesn't seem practical in its current form since the DPRaw system wasn't designed for that (yet, I guess).

I shoot a lot of high DR scenes and have a 5D IV, but I've never gone through the trouble of using DPRaw in that way since it seems quicker and more reliable to just bracket. For me, I wouldn't enable DPRaw to extend DR unless I knew I'd be clipped otherwise to save card space just like I wouldn't bracket unless I knew I'd be clipped otherwise. The issue that keeps me from using DPRaw for increasing DR is I can't check if the file is still clipped on until I get the file off the camera and split it up into two exposures, I can only guess based on the assumption of an ~1-stop darkening in the highlights. With that said, I believe the histogram produced for captured images is really only related to the JPG preview so even that isn't a perfect measure of clipping either. Bracketing just seems more practical for conditions where the subjects aren't moving.
 
Upvote 0
Jan 29, 2011
10,675
6,121
No, this is completely different. Current dual pixel RAW files contain the other pixel data in the secondary frame and the combined (by adding the two pixels' data together) in the main frame. But there is no difference in the amplification of the original dual pixels. With the dual gain system they could achieve a bigger than 1 stop boost.

I guess the maximum achievable benefit would be depending on the noise characteristics of the sensor. This would not help for photon shot noise, but certainly would help if read noise is a factor.
Yes its amazing what can be learned when you actually read the article isn't it!

Current dual file RAW images give you plus one stop of DR in the highlights (which much to the Sony fanboy club puts the Canon sensor above the Sony sensor for DR at most ISO's), that is just a simple mathematical fact. Enabling dual amplification/sensitivity/ISO gives even greater mathematical potential.
 
  • Like
Reactions: 3 users
Upvote 0

Jack Douglas

CR for the Humour
Apr 10, 2013
6,980
2,602
Alberta, Canada
All this stuff would work unreliably in the areas where the information it is supposed to provide interferes with the information supposed to be provided by DPAF, i.e, in defocused areas with local contrast.
I guess you are right but isn't it the focused area where you're most interested in the detail via dynamic range?

Jack
 
  • Like
Reactions: 2 users
Upvote 0
[...]

I guess the maximum achievable benefit would be depending on the noise characteristics of the sensor. This would not help for photon shot noise, but certainly would help if read noise is a factor.
The simplest thing to increase DR with dual pixels is to make one pixel less sensitive electronically: Switch a series resistor (or Op Amp construction) between photo diode and capacitor structure to decrease the speed of charging. In dark areas the sensitive pixel gives DR and in bright areas the "blinded" one gives the DR. This would eventually result in less ISO, e.g. ISO 12 or ISO 6 for ultra high DR.

But I wonder if we will see some imperfections in lenses (flares / lower contrast by stray light) which aren't visible in sensors with less DR. But maybe on reason of the improved coatings Canon promotes in the currently released lenses?!
 
Upvote 0
Current dual file RAW images give you plus one stop of DR in the highlights (which much to the Sony fanboy club puts the Canon sensor above the Sony sensor for DR at most ISO's),

Unfortunately no, they don't. Have you actually used this DPRSplit app? On a couple of occasions I was able to recover some highlights.
But in general it doesn't really work as expected, merged files have no additional DR and it often creates an awful greenish colour cast in the highlights.
Good to have as a last resort, but I never rely on it. And I wouldn't recommend anybody to rely on it.
 
  • Like
Reactions: 1 user
Upvote 0
This is kinda similar to Magic Lantern's Dual ISO, where odd and even pixel rows are read in different ISOs. The file must go through a demosaic process but the result is very impressive, at the cost of some vertical resolution (proportional to stops difference). Maybe this is the dynamic range stuff the rumors were about.
 
  • Like
Reactions: 1 user
Upvote 0

sdz

CR Pro
Sep 13, 2016
262
209
Pittsburgh, PA
This would certainly be a differentiating feature that Canon would be able to pull over everyone else for a long time. However, for myself, I do not see the need. I never suffer the lack of ISO(I do animals) and more just sometimes miss focus by a few mm on occasion.

The needs:

1. More effective Dynamic Range is better than lesser Dynamic Range all else being equal.
2. To put large quantities of crow into the mouths of its critics.
 
Upvote 0
Unfortunately no, they don't. Have you actually used this DPRSplit app? On a couple of occasions I was able to recover some highlights.
But in general it doesn't really work as expected, merged files have no additional DR and it often creates an awful greenish colour cast in the highlights.
Good to have as a last resort, but I never rely on it. And I wouldn't recommend anybody to rely on it.

I have tried dprsplit and its outputs and it works very well.
You need to dual proccess the files to obtain the dynamic range benefits which is better than dual proccess one file. (Its like bracketing but with one frame).
 
  • Love
Reactions: 1 user
Upvote 0
Mar 25, 2011
16,848
1,835
All this stuff would work unreliably in the areas where the information it is supposed to provide interferes with the information supposed to be provided by DPAF, i.e, in defocused areas with local contrast.
Why do you say that? Did you understand the patent??

Looking at many of the comments, it is obvious that some have not read or did not understand the patent. Are they just commenting on what they imagine it might say!

Its not simple, its is very complex, it does not interfere with DPAF Autofocus, It takes sensor noise into account, its very well thought out. The patent has several different ways of accomplishing the process too.

Read it, it goes into step by step detail for each of the different methods. Timing charts are shown for the AF Mode and Readout mode for each case.
 
  • Like
Reactions: 2 users
Upvote 0
Mar 25, 2011
16,848
1,835
This would certainly be a differentiating feature that Canon would be able to pull over everyone else for a long time. However, for myself, I do not see the need. I never suffer the lack of ISO(I do animals) and more just sometimes miss focus by a few mm on occasion.
This is not boosting ISO, its increasing DR, even at ISO 100.

For myself, I definitely find cases where DR is totally inadequate. I'm not sure if this would fix those cases, but any help is good. I'd put this to the test right away. I've just been editing 2,000 shots from a theater event. Generally, the stage is not brightly lit, then, they hit a subject with white clothing with a spot while the rest of the stage is dim, it results in a exposure nightmare. Or, there is a light in a scene, perhaps a streetlight which blows out without dropping the exposure drastically on the fly. Then, in post, there is always more noise than desirable for those shots.

I toss far more images due to exposure issues than I do with poorly focused images. The DR of my 5D MK IV and my EOS R is very good, but is still not able to capture a significant percentage of my shots unless I underexpose. Sometimes, I set EC to -2 or even more.
 
  • Like
  • Love
Reactions: 3 users
Upvote 0
Mar 25, 2011
16,848
1,835
  • Like
Reactions: 1 user
Upvote 0