Unlimited DR Camera invented by students at MIT...

[Antono Refa]

The sentence "by taking into account the number of resets for each bucket, the camera can figure out the relative brightness for each pixel" means the bucket spills over into a memory cell.

That is a neat idea, but means the DR is not infinite, but rather limited by the combination of pixel capacity and per pixel memory size, which isn't infinite.

 

[tron]

The sentence "by taking into account the number of resets for each bucket, the camera can figure out the relative brightness for each pixel" means the bucket spills over into a memory cell.

That is a neat idea, but means the DR is not infinite, but rather limited by the combination of pixel capacity and per pixel memory size, which isn't infinite.

OK it's not infinite but even an 8-bit counter per pixel would increase DR approximately by 8 stops (and similarly a 16 bit counter by 16 stops...)

 

[K-amps]

A limited version can be implemented without requiring shooting the sun and expect the bucket to be reset fast enough... the example shots had a lot of color fringing... I am sure it will get better, but the same issue happens now in HDR blending... this will be only as good as the tone mapping algorithm.

exciting times.

 

[tron]

A limited version can be implemented without requiring shooting the sun and expect the bucket to be reset fast enough... the example shots had a lot of color fringing... I am sure it will get better, but the same issue happens now in HDR blending... this will be only as good as the tone mapping algorithm.

exciting times.

Let's hope that the color finging was mostly due to the lens used...

Exciting times indeed

 

[Nelu]

Weird! The DR is unlimited but still, the last image still has blown-out highlights on the metal thingy...

Nelu

 

[rfdesigner]

This is impressive to see working in any way at all, very interesting but not the only way.

One issue with this method is you can lose linearity if resets are not perfectly accounted for (i.e. you assume 50,000e / well when some are 50,100e and others are 49,920e etc., you'd need good calibration and you'd need to know it still held true over time)

One method that could solve that is the electron count per pixel method.. you get rid of the ADCs altogether and just use one comparitor per pixel, whenever an electron turns up you record that as an electrical pulse and add it to your count for that pixel.

sCMOS sensors have thousands of ADCs (one per row), so millions of simple comparators isn't out of the question.. the bandwidths on all those comparitors would be exceptionally low compared to feeding all the data out of one gate, so noise falls substantially, once it's less than about 0.1e you can count electrons (noise power is proportional to bandwidth)

I should add this isn't my idea, it's just a proposed technique.

 

[jarrodeu]

Finally, a sensor that can match and surpass the range of film!

Jarrod

 

[privatebydesign]

Finally, a sensor that can match and surpass the range of film!

Jarrod

Seriously?

According to Eastman Kodak (1) they rated one of their best general purpose films at 13 stops, many slide films that were, and still are, the preferred emulsions for film users are between 6 and 8 stops (2). All digital cameras are comfortably above the slide film and many best the 13 stops of the best non specialised scientific films (3).

(1) http://motion.kodak.com/motion/uploadedFiles/Kodak/motion/Products/Camera_Films/Color_Negative/Product_Info/5213_SS_4pgs.pdf

(2) http://www.fujifilm.com/products/professional_films/pdf/velvia_50_datasheet.pdf

(3) http://www.dxomark.com/Cameras/Ratings/Landscape

DR is a complete red herring, as is resolution, digital kicked the butt of film in every metric long ago, just look at iso performance to see how badly film really does against digital on a technical comparison. Use film if you want to, but do it because you want to, not because you think it has some magic technical superiority to digital, it doesn't.

 

[stefang]

This is not much different from the idea to record the time after which a bucket is full, for which a patent already exists.
That idea will also lead to almost unlimited DR, because you can extrapolate the time to calculate the effective amount of light.

 

[leGreve]

To me it seems so simple an idea it's almost silly no one has tried it before. Then again, I don't know the feasability of mass manufacturing. This is the sort of thing a company like Canon should be writing a blank check to MIT to buy the rights to even if just to keep it out of competitor's hands until they can figure out a way to produce it for consumer uses.

Just imagine setting the perfect exposure for the shadows with total disregard for the highlights as they can be infinitely blown out and still fully recoverable. Holy cow.

Actually... one would more hope that someone like Elon Musk would buy the concept and make it open source so EVERYONE could use it, instead of a money grubbing un-innovating company like Canon.

 

[leGreve]

Weird! The DR is unlimited but still, the last image still has blown-out highlights on the metal thingy...

Nelu

Im sure they are really sorry that this test product not even remotely ready for mass production yet is not 100% perfect.

 

[Eldar]

I think this is intriguing and I wonder who picks it up and turns it into an industrialized product first. From a signal processing perspective, it sounds like a mouthful though ...

 

[Hector1970]

If Canon don't implement this in 5D Mark IV I'm refusing to buy it. ;D

 

[Dylan777]

Good reading. thanks for sharing

 

[jarrodeu]

Finally, a sensor that can match and surpass the range of film!
Jarrod

Seriously?

According to Eastman Kodak (1) they rated one of their best general purpose films at 13 stops, many slide films that were, and still are, the preferred emulsions for film users are between 6 and 8 stops (2). All digital cameras are comfortably above the slide film and many best the 13 stops of the best non specialised scientific films (3).

(1) http://motion.kodak.com/motion/uploadedFiles/Kodak/motion/Products/Camera_Films/Color_Negative/Product_Info/5213_SS_4pgs.pdf

(2) http://www.fujifilm.com/products/professional_films/pdf/velvia_50_datasheet.pdf

(3) http://www.dxomark.com/Cameras/Ratings/Landscape

DR is a complete red herring, as is resolution, digital kicked the butt of film in every metric long ago, just look at iso performance to see how badly film really does against digital on a technical comparison. Use film if you want to, but do it because you want to, not because you think it has some magic technical superiority to digital, it doesn't.

I am serious.
I'm assuming that you've been away from film for a while and are unaware of the advancements that have been made. I myself shoot almost all digital and but when I have time and I want the best, I shoot some film.

Dynamic Range
It is true that slide film has a limited dynamic range but when it comes to negative films like the new Kodak Portra 400, there is no comparison. As can be seen here, it has around 18 stops of DR. https://www.onlandscape.co.uk/2011/05/kodaks-new-portra-400-film/

Resolution
This also really depends on the type of film being used. Kodak Ektar in 35mm compares fairly well to the 21MP 5DMk II as can be seen here. http://www.twinlenslife.com/2011/01/digital-vs-film-canon-5d-mark-ii-vs.html
There some black and white films which provide a tremendous amount of resolution such as Adox CMS 20. Here you can see an example with the truck taken with 35mm film that was scanned in at 85 megapixels and tack sharp. http://www.adox.de/Photo/adox-films-2/cms-20-ii-adotech-ii/

High ISO
Here is a realm that digital clearly is superior to film which is why so many high ISO films have been discontinued. But not every situation calls for ISO 1600+

So when it comes to convenience and cost effectiveness digital is the clear winner but when it comes to dynamic range and resolution, that's when the answer becomes dependent on what type of film is being used vs the digital sensor. Film also comes in many sizes larger than any digital sensor.
Here are further tests that have been done on larger film formats. http://www.onlandscape.co.uk/2011/12/big-camera-comparison/
Jarrod

 

[kaihp]

This is impressive to see working in any way at all, very interesting but not the only way.

One issue with this method is you can lose linearity if resets are not perfectly accounted for (i.e. you assume 50,000e / well when some are 50,100e and others are 49,920e etc., you'd need good calibration and you'd need to know it still held true over time)

One method that could solve that is the electron count per pixel method.. you get rid of the ADCs altogether and just use one comparitor per pixel, whenever an electron turns up you record that as an electrical pulse and add it to your count for that pixel.

sCMOS sensors have thousands of ADCs (one per row), so millions of simple comparators isn't out of the question.. the bandwidths on all those comparitors would be exceptionally low compared to feeding all the data out of one gate, so noise falls substantially, once it's less than about 0.1e you can count electrons (noise power is proportional to bandwidth)

I should add this isn't my idea, it's just a proposed technique.

I'm right with you. There's going to be some uncertaincy about detecting the "pixel full" condition (no matter where you set it) and the time it takes to "flush" the pixel could result in lost electrons as well.
Finally, they will need to have a counter for each pixel as well, and those do not have infinite capacity either.

But definitely a novel and different idea. It will be interesting to see if they can make a product out of it.

 

[lion rock]

I'm right with you. There's going to be some uncertaincy about detecting the "pixel full" condition (no matter where you set it) and the time it takes to "flush" the pixel could result in lost electrons as well.
Finally, they will need to have a counter for each pixel as well, and those do not have infinite capacity either.

But definitely a novel and different idea. It will be interesting to see if they can make a product out of it.

Thus the extra bits just for keeping track of the number of "overflow" full buckets. So again, nothing special about the "invention", nor the "unlimited DR" claim. I would just say this is just an increase of bit number to give higher resolution of the photon count.
I'm NOT excited about the MIT claim as it stands now.
-r

 

[privatebydesign]

Finally, a sensor that can match and surpass the range of film!
Jarrod

Seriously?

According to Eastman Kodak (1) they rated one of their best general purpose films at 13 stops, many slide films that were, and still are, the preferred emulsions for film users are between 6 and 8 stops (2). All digital cameras are comfortably above the slide film and many best the 13 stops of the best non specialised scientific films (3).

(1) http://motion.kodak.com/motion/uploadedFiles/Kodak/motion/Products/Camera_Films/Color_Negative/Product_Info/5213_SS_4pgs.pdf

(2) http://www.fujifilm.com/products/professional_films/pdf/velvia_50_datasheet.pdf

(3) http://www.dxomark.com/Cameras/Ratings/Landscape

DR is a complete red herring, as is resolution, digital kicked the butt of film in every metric long ago, just look at iso performance to see how badly film really does against digital on a technical comparison. Use film if you want to, but do it because you want to, not because you think it has some magic technical superiority to digital, it doesn't.

I am serious.
I'm assuming that you've been away from film for a while and are unaware of the advancements that have been made. I myself shoot almost all digital and but when I have time and I want the best, I shoot some film.

Dynamic Range
It is true that slide film has a limited dynamic range but when it comes to negative films like the new Kodak Portra 400, there is no comparison. As can be seen here, it has around 18 stops of DR. https://www.onlandscape.co.uk/2011/05/kodaks-new-portra-400-film/

Resolution
This also really depends on the type of film being used. Kodak Ektar in 35mm compares fairly well to the 21MP 5DMk II as can be seen here. http://www.twinlenslife.com/2011/01/digital-vs-film-canon-5d-mark-ii-vs.html
There some black and white films which provide a tremendous amount of resolution such as Adox CMS 20. Here you can see an example with the truck taken with 35mm film that was scanned in at 85 megapixels and tack sharp. http://www.adox.de/Photo/adox-films-2/cms-20-ii-adotech-ii/

High ISO
Here is a realm that digital clearly is superior to film which is why so many high ISO films have been discontinued. But not every situation calls for ISO 1600+

So when it comes to convenience and cost effectiveness digital is the clear winner but when it comes to dynamic range and resolution, that's when the answer becomes dependent on what type of film is being used vs the digital sensor. Film also comes in many sizes larger than any digital sensor.
Here are further tests that have been done on larger film formats. http://www.onlandscape.co.uk/2011/12/big-camera-comparison/
Jarrod

You might be serious, but you are in denial.

Dynamic Range
Just taking a look at your first link that purports to prove Portra 400 has "around 18 stops of DR". have you actually looked at the strip? For instance why is the fourth image darker than the fifth image when it is exposed one stop more? Maybe the fact that they were shot when the sun was being covered more and less by moving clouds!

But lets not fret over that, lets rely on Kodak themselves who state, quite clearly in their Portra 400 data sheet (1), a Log Exposure range of 3.3-3.6, or 11-12 stops. Personally I trust Kodak's own measurements over uncontrolled random 'test' hyperbole.


Resolution
Your first issue with the examples you gave is scanning, it introduces all kinds of digital irregularities, besides, anybody that has drum scans done regularly, or did, knows full well you are not resolving more detail, you are just resolving the film grain more accurately.

Personally I have many 24"x36" wet prints from 135 format Fuji Velvia 50 and 100, some of the finest grained film which @ 1000:1 contrast ratio it is supposedly good for 160 lines/mm (2), they don't hold a candle to 21MP images at the same size.

The 5DSR has a 241 lines/mm.

As for your Adox film, it is clearly not what it is claimed to be. The truck image was scanned at 8000dpi, that is 314 lines/mm, nowhere near the 800 claimed, or the farcical 800 lp/mm they also claim, and it has no more detail it is just grain. Yes it is a bigger file than a 5DSR, but does it contain more detail?

Lets look at two crops from the same area from the Adox and a 5DSR, see below. Image 12 is from a 5DSR, image 13 is a same sized crop from the Adox, I know which I'd take, and the Adox needed four times the exposure.

High ISO
We agree on.


As a final note, yes film does come in larger sizes than digital sensors, well the ones currently available at prices we can afford, but stitching is free and easy and have you seen the resolution figures for large format lenses, they are nowhere near as good as those for 135 and medium format.

So when it comes to convenience, cost effectiveness, dynamic range and resolution digital is the clear winner. Like I said, there are many reasons to shoot film, but the main one is shoot it because you want to, because of the organic nature if the process, because you love the darkroom and it's smells feel and isolation, because you are in love with your 8x10 Wista and Berlebach..........




(1) http://www.kodak.com/global/en/professional/support/techPubs/e4050/e4050.pdf
(2) http://www.fujifilm.com/products/professional_films/pdf/velvia_50_datasheet.pdf

 

[Don Haines]

Finally, a sensor that can match and surpass the range of film!

Jarrod

Seriously?

According to Eastman Kodak (1) they rated one of their best general purpose films at 13 stops, many slide films that were, and still are, the preferred emulsions for film users are between 6 and 8 stops (2). All digital cameras are comfortably above the slide film and many best the 13 stops of the best non specialised scientific films (3).

(1) http://motion.kodak.com/motion/uploadedFiles/Kodak/motion/Products/Camera_Films/Color_Negative/Product_Info/5213_SS_4pgs.pdf

(2) http://www.fujifilm.com/products/professional_films/pdf/velvia_50_datasheet.pdf

(3) http://www.dxomark.com/Cameras/Ratings/Landscape

DR is a complete red herring, as is resolution, digital kicked the butt of film in every metric long ago, just look at iso performance to see how badly film really does against digital on a technical comparison. Use film if you want to, but do it because you want to, not because you think it has some magic technical superiority to digital, it doesn't.

This reminds me of the audiophile arguments about tube amps and transistor amps. So many people were so adamant that tube amplifiers were better and would attack those who disagreed with them. Put them on a spectrum analyzer and it was obvious that the transistor amps were far superior, but that had no effect on the tube enthusiasts who "knew" that the tubes were far better. No amount of scientific testing could ever change their minds because their minds were already made up. The reality was that they liked the way the distortion sounded.....

Same with film.... it is a different media and it looks different.... if that's what you want, then fine, use it. You don't have to try justifying it against digital or using ridiculously concocted scenarios.... if you like it, use it!

 

[privatebydesign]

Finally, a sensor that can match and surpass the range of film!

Jarrod

Seriously?

According to Eastman Kodak (1) they rated one of their best general purpose films at 13 stops, many slide films that were, and still are, the preferred emulsions for film users are between 6 and 8 stops (2). All digital cameras are comfortably above the slide film and many best the 13 stops of the best non specialised scientific films (3).

(1) http://motion.kodak.com/motion/uploadedFiles/Kodak/motion/Products/Camera_Films/Color_Negative/Product_Info/5213_SS_4pgs.pdf

(2) http://www.fujifilm.com/products/professional_films/pdf/velvia_50_datasheet.pdf

(3) http://www.dxomark.com/Cameras/Ratings/Landscape

DR is a complete red herring, as is resolution, digital kicked the butt of film in every metric long ago, just look at iso performance to see how badly film really does against digital on a technical comparison. Use film if you want to, but do it because you want to, not because you think it has some magic technical superiority to digital, it doesn't.

This reminds me of the audiophile arguments about tube amps and transistor amps. So many people were so adamant that tube amplifiers were better and would attack those who disagreed with them. Put them on a spectrum analyzer and it was obvious that the transistor amps were far superior, but that had no effect on the tube enthusiasts who "knew" that the tubes were far better. No amount of scientific testing could ever change their minds because their minds were already made up. The reality was that they liked the way the distortion sounded.....

Same with film.... it is a different media and it looks different.... if that's what you want, then fine, use it. You don't have to try justifying it against digital or using ridiculously concocted scenarios.... if you like it, use it!

Couldn't agree more Don. That is why I still own and occasionally use my 1VHS and my Mamiya 6x9.