Sony develops sensor capable of rendering color images at 0.005 lux

It can be dark indoors when the sun is shining outside.

Oct 25, 2014permalink

Tan68

Or it can be light inside and the power goes out.

"...uses extended exposure times rather than..."
I wonder what this really means... How long is extended exposure time. It is intended for use in cars.

Even if the number of frames per second is slow and would make choppy video, slipping this thing in a Google glass eyeglass would be more interesting than Google's product itself.
Night vision with color

Oct 29, 2014permalink

tripodfan

You'll see me in the morning. But I'll see you while you sleep. In colour, no less.

Nov 19, 2014permalink

Pete_CSCS

Whatever happened to that Panasonic sensor that could see in near darkness? DPR reported on it about 5 or so years ago.

They were going to put them in surveillance cameras but some of us had hoped they would trickle down to DSLRs and the like.

Obviously, it never happened.

Oct 23, 2014permalink

smakogon

As for SONY A 7s - I have an increadible high ISO test results: https://plus.google.com/u/0/113540123829280887772/posts/CKuXBfAh42Y?pid=6073189452363714706&oid=113540123829280887772

Oct 23, 2014permalink

Simon Stanmore

What I learned from this article ... There's not many photographers using DPR these days.

Oct 22, 2014permalink

TheEulerID:

You can use Wikipedia and work through the math. The point about infinities isn't real hard to see,

I'll look for the exact equation about an electron.

But division works this way, divide a constant by some number smaller than the constant and you'll get a bigger and bigger result. Drop that divisor to smaller and smaller fractions (or decimal numbers) and the result blows up to infinity. That's a localized (point) electron, with known momentum and position, ostensibly demonstrating the Heisenberg Uncertainty Principle.

I DID NOT raise the perturbation point.

The A7S isn't a particularly good high ISO camera.

My point remains that with a better understanding of "random" noise (something quantum mechanics invokes in place of understanding) there remains the real possibility of counting that other 35 percent of the photons.

And careful: the exclusive particle nature of the photon is more theory/belief (Feynmann's again) than established fact.

Oct 22, 2014permalink

Smeggypants

"The A7S isn't a particularly good high ISO camera."

HAHAHAHAHAHAHAHAHAHAAAAAAAA!! I love the comments section!

Oct 22, 2014permalink

Smeggypants:

Putting this more clearly: Compared to the D4s and the Df, the Sony A7S is not a particularly good high ISO still camera.
At ISO 25,600 it has typically Sony cyan and magenta banding in the shadows.

It’s colour is not particularly good, even with very good lenses and shooting at lower ISOs.

Your reaction is simplistic.

The A7S, like the Canon 6D, is a very decent high ISO camera, but it’s not a great high ISO camera. And put a good lens on the Canon and the colour be better.

Oct 24, 2014permalink

Another claim proven wrong by many controlled low light tests, where it shows less purple/cyan in shadows and less banding than both the DF and D4s. But I'm sure your magic RAW files that no one but you has ever seen, disagree. ;-)

Oct 27, 2014permalink

TrojMacReady:

No, you are wrong, and the problem is really easy to create with the A7S.

Sure the Df/D4s experience a similar problem, BUT at significantly higher ISOs. And my raw files that I've shot with all agree.

Oct 27, 2014*permalink

jkoch2

Color in "zero light" would surprise owls, dogs, or even bats. 1.7 megapixels might be fine for dashboard or rear-bumper cameras. It would be an improvement over what one typically sees captured by security cameras. The resolution might even be enough for small display of night party or selfie shots.

Is the super ISO contingent on modest resoluton? Would an 1/3" 8mp sensor be difficult to engineer with the same sensitivity without incurring lots of noise? Where photons are few, maybe ultra ISO can achieve only so much, after which point sensor size must increase.

Oct 22, 2014permalink

AV Janus

The end is in sight for high ISO battle.
One more nail in the FF coffin.
Next up; software DOF control with depth scene info from AF.

Oct 22, 2014permalink

That was done in the early 90's. At least on Minoltas.
The trick is to accurately guess the intention.

For instance, this fl and this focusing distance indicates head&shoulder portrait. Or landscape. Or macro. Or whatever.

Oct 22, 2014permalink

AV Janus

Today they could use many more hardware and software trick to give you best bokeh at the right circumstance!
I have a feeling Sony or Samsung will implement something under their "blur background" feature.

Oct 24, 2014permalink

Renzokuken

Got to love how aggressive Sony is with innovation and R&D.

Oct 22, 2014permalink

The sensor department, no so much the audio playback department, or raw file size from Sony cameras.

Oct 27, 2014permalink

wictred

can't wait to see the steve huff review

Oct 22, 2014*permalink

geo444

here is a comparison from Sony's Global site :

-left Sony imx224mqv : 1/60 sec - F1.4 at 72 dB max gain

-right Sony conventionnal CMos : 1/60 sec - F/1.4 at 48dB max gain

www.sony.net/SonyInfo/News/Press/201410/14-106E

72 dB gain = indiscernible Read Noise

;)

Oct 22, 2014permalink

geo444

Bravo Sony !!!

0.005 lux is really « near darkness »

= 135 times less light on subject than I-R 1/16 fc Low-Light samples

(1/16 fc = 0.0625 foot-candle = 0.6727444 lux)

strongly waiting for similar sensitivity in Astrophotography...

;)

Oct 22, 2014permalink

Boss of Sony

When will my minions announce the new cameras we've been working on? I suggest we announce the RX2 and A8.

Oct 22, 2014permalink

Boss of Sony

RX2 will have EVF, 36MP sensor, 4K video, and touchscreen. A8 will have 4K video, touchscreen, faster shot-to-shot-time, better AF, less noise, no compression.

Oct 22, 2014permalink

BOMBOVA

i would like to see this sensor used in a dash cam, i am in need of a good one, of this type.

Oct 22, 2014permalink

alan14

I can see this being used in CCTV applications, Axis have been selling cameras good to 0.05 lux for a few years, this will give Sony an edge in this market

Oct 21, 2014permalink

For some one keep bragging about canon can shoot firefly with some mysterious never-hit-market super camera.

Here is A7s shoot fireflies:
http://www.youtube.com/watch?v=eBdAXsPu3_o
http://www.youtube.com/watch?v=3gmUIoG7qQU

Not only a7s can see fireflies, it also record firefly reflection on water, and firefly-illuminated plants from distance.

Oct 21, 2014permalink

Rainer2022

I can remember a story about that super sensor from Canon, about 2 or 3 years ago.

Oct 21, 2014permalink

http://www.youtube.com/watch?v=eHnUtmenGQ0

that's just one year ago, which is a full frame, 2k sensor.
Now with every one move to 4K , I doubt canon will ever release this sensor

Oct 21, 2014permalink

bwana4swahili

That's kind of the performance out of Canon we've become familiar with... sadly! I've now moved to Sony and really like both the A7R and A7S. Yes, Sony has hype as well but also a product. :)

Oct 22, 2014permalink

bwana4swahili

I went back and rewatched this video. After having used the A7S I find the video rather old school, i.e.: a video of fireflies is no longer amazing! And the Canon color is terrible!

Oct 22, 2014permalink

grasscatcher

Well, geez! I shot some video of fireflies with my Panasonic GX7 back in July. Is it really that hard to do?

Oct 22, 2014permalink

Shiranai

Here's the canon video, which I find has not only a higher exposure but also less grain. I'd also say they used f4 rather than a fully opened lens.

Oct 24, 2014permalink

grasscatcher

So....would we see even lower lux if the sensor had, say, .3mpxl. You know, like the HTC concept of fewer, larger pixels to gather more light...

Let's hope the 'through the adoption of photo diodes that improve the conversion efficiency of light into electrons and through the installation of circuits that improve the conversion efficiency of those electrons into voltage.' technology makes it into consumer photography cameras.

Oct 21, 2014permalink

Canon EOS 60D

This sensor was developed for "Automotive use" and not photography. Camera geeks/nerds stop complaining. You're barking at the wrong tree. LOL

Oct 21, 2014permalink

It is a news story on digital PHOTOGRAPHY review, you seem to forget.

And the question if the sensitivity of sensors could improve significantly in the future is valid. And as I detail in my post below, this story provides no evidence it can.

Oct 21, 2014*permalink

Canon EOS 60D

Although it's a story on digital PHOTOGRAPHY review, it does not necessarily mean that it's about PHOTOGRAPHY. Dpreview posts a lot of stuffs other than photography.

Based on your conclusion below that this is just a marketing and the story does not show evidence of sensor sensitivity improvement, then there's no point for Sony to put this inside a photographer's camera and it will never be used or intended by Sony. Thus, non-photography related article.

Oct 21, 2014permalink

PVCdroid

It must be a reading comprehension problem then.

Oct 22, 2014permalink

ChapelThrill23

It might be now but I'd imagine what they learn from this will trickle down to cameras.

Oct 28, 2014permalink

DPR could have done a much better job when reporting this.

Let me fill in the missing facts:

1. 0.005 lux is -9 EV (almost exactly). The scene is pitch black dark indeed.

2. However, to judge the sensor, an EV figure w/o aperture and exposure time is meaningless.

3. According to Sony, it is F1.4, 1/60s exposure. -9 EV then requires ISO 6,000,000 to expose correctly. I.e., the image shown is ISO 6m.

4. That's 3.9 stop beyond e.g., the A7s highest iso level of 410k. But the image quality shown by Sony for a tiny 400x300 pixel image is terrible. While the A7s is still ok at 410k.

Therefore, no conclusion that this chip would be more sensitive than other sensors can be drawn from matrerial provided. Just marketing non info.

And btw, at quantum efficiencies already as high as 65% and read noise as low as 0.4 e- (A7s) rest assured that no miracles are left in this field if you cannot break the laws of nature.

Oct 21, 2014*permalink

To be fair, everyone can always do a better job.

Oct 21, 2014permalink

falconeyes:

“And btw, at quantum efficiencies already as high as 65% and read noise as low as 0.4 e- (A7s) rest assured that no miracles are left in this field if you cannot break the laws of nature.”

It should be possible to break the so called “laws of nature” with a better understanding of noise.

Laws of nature are based upon man made theories, which often don’t correspond to nature’s facts, for example the ostensive inability to know the position and momentum of an electron at the same time is well disproven by the existence of flash storage. Other facts also disprove this theory–and show the problem with relying on maths to entirey drive theory.

This means that the Heisenberg Principle is of limited usefulness in thinking about what electronics can do–there are other examples too.

Oct 21, 2014permalink

But the A7S sensor is also 50 times larger. Take 1/50th of a portion of an A7S image at its top sensitivity, pushed 4 stops. Then compare noise at the same output size. ;-)

Oct 21, 2014*permalink

bwana4swahili

"While the A7s is still ok at 410k." I own an A7S and image quality at 410k is ???...

Oct 21, 2014permalink

@HowaboutRAW. I am glad I never have to discuss random opinions like this when meeting with scientists. Scientists who are capable to understand the Heisenberg Principle from the ground up.

@TrojMacReady. That's correct. One would have to do a proper noise analysis and normalize wrt sensor surface. All I wanted to say was that the example provided doesn't mean much as it stands.

@bwana4swahili: Here is a A7s iso 410k image:
http://cdn.lowyat.net/wp-content/uploads/2014/04/30eb50f7-8ecb-4e2f-b8a8-46bfd22a9b83_zps0c4091e6.png
ok and even better after correcting the color profile for the magenta cast.

Oct 21, 2014*permalink

falconeyes:

The fact remains that there are problems with the Heisenberg principle, the math (theory) doesn't agree with all experimental and practical applications, pretty obvious.

Frankly, it is this dismissive attitude (taken by you) that impedes much progress in science and engineering.

Hint, any time math drives things to infinities or into the imaginary those are red flags for problems with the theory. And that's a very "ground up" apprehension of that theory.

Oct 21, 2014*permalink

It isn't my job to teach physics here, but since you ask ... If you can show the Heisenberg principle being broken by any experiment you'll be awarded a Nobel prize, I am almost sure. Recent work showing that measurement perturbations can be kept small are not in contradiction, as the Heisenberg principle does not state anything about measurement perturbations. Assuming it does was a misconception while popularizing the principle to non physicists.

Having said this, only one thing remains which is pretty obvious.

Oct 22, 2014permalink

Artak Hambarian

Howaboutraw: howaboutanobelprize?

Oct 22, 2014permalink

falconeyes:

Yawn, nothing I stated is particularly unknown or new.

That you're not familiar with these problems is a different question.

Slight perturbations are not the same as infinity, and the Heisenberg Principle predicts infinities with localized, stationary, electrons.

Oct 22, 2014permalink

HowaboutRAW (21 hours ago)

Wow. It's been a long time since I've read such ill-informed gibberish.

The fact is that quantum mechanics is the most well tested fundamental theory in physics. Every bit of electronics you use relies on that theory for its very design. The issue over the fundamental limitations of sensor technology come down to the quantised nature of light. Even if the "perfect" sensor could be designed that detected every single photon (along with its wavelength) and introduced no noise of its own, it would only perform perhaps 2 stops better than current bayer-pattern sensors. At that point, you are stuck with shot noise and no technology on earth can eliminate that as it's fundamental to light.

Of course, if you can some up with an alternative to the theory of quantum mechanics, be my guest (you might start looking at multiverses, as there are alternative interpretations of QE theory in that, and they probably don't help in this universe).

Oct 22, 2014permalink

TheEulerID:

And what you've done is confuse Heisenberg and quantum mechanics. Plank is not the same person as Heisenberg.

The fact remains that the Heisenberg Principle says that the position and momentum of the electron can't be known together (since the equations blow up to infinity), and the problem is that electrons do exist stably in known places.

So in fact you're not real familiar with quantum mechanics.

"'The fact is that quantum mechanics is the most well tested fundamental theory in physics.'" This is quoting you, but likely Feynman. And he's famous for getting rid of those infinities with rather suspect math (he admitted as such).

I suggest you look into quantum mechanics before simply quoting others.

Oct 22, 2014permalink

Heisenberg's uncertainty principle only says there's a trade-off between the accuracy you can know the two. That is, the more accurately you know position, the less accurately you can know momentum. This derives from the mathematical representation of the properties of quantum particles, and requires some knowledge of matrix mathematics. It's not changed in the 40 years since I first studied it. Also, there are other particle properties in such pairs.

It has nothing to do with measurement perturbation. That's the "observer effect". There's an unfortunate history of confusing the two, not helped by some popularisations.

In any event, it's unnecessary to invoke Heisenberg's principle. It simply follows from light being quantised and following a Possion distribution for random arrivals from incoherent light sources.

For reference, I graduate in physics at Imperial College, London in 1976. My name is Steven Jones, and if you care to check with them I'm sure they will confirm that.

Oct 22, 2014permalink

nb. I should add that if, as you state, flash memory disproved the uncertainty principle, then somebody else would have noticed.

Oct 22, 2014permalink

TheEulerID:

You've posted a way over simplified version of the Heisenberg Uncertainty Principle: I suggest you look at the math, the generation of infinities be easy to discern.

Right: "It has nothing to do with measurement perturbation. That's the "observer effect". There's an unfortunate history of confusing the two, not helped by some popularisations." But I was quoting.

"For reference, I graduate in physics at Imperial College, London in 1976. My name is Steven Jones, and if you care to check with them I'm sure they will confirm that." Yet again check the math, as the electron is more and more localized something blows upto infinity. (I more than think you missed something in your studies.)

In fact other people have noted the problems with localized electrons in various technology--including flash.

Escaping with "In any event, it's unnecessary to invoke Heisenberg's principle." doesn't do your argument anygood. Fact remains there are big problems.

Oct 22, 2014*permalink

OK then. Provide some links to credible sources that show there are major problems with the Uncertainty Principle (other than it's philosophical meaning, which is a rather different thing).

In any event, it was you that raised the uncertainty principle, not the OP. As I state, it's completely unnecessary to invoke the uncertainty principle to show that there is a limit on sensitivity due to the quantised nature of light. The A7S is already working at levels where the number of photons detected at photosites in darker areas can be counted on the fingers of one hand (and it detects about 65% of the photons that do arrive after passing through the bayer filter).

Oct 22, 2014permalink

TheEulerID:

You can use Wikipedia and work through the math. The point about infinities isn't real hard to see,

I'll look for the exact equation about an electron.

I DID NOT raise the perturbation point.

The A7S isn't a particularly good high ISO camera.

And careful: the exclusive particle nature of the photon is more theory/belief (Feynmann's again) than established fact.

Oct 22, 2014permalink

http://en.wikipedia.org/wiki/Uncertainty_principle

TheEulerID:

Here's Wikipedia on the Uncertainty Principle:

I suggest you solve the first equation for either term on the left; this means putting say a zero momentum in the divisor--result infinity on the left.

"Inside the museums infinity goes up on trial."

Oct 22, 2014permalink

@HowaboutRAW

There's no reason why you would want (or need) to put zero in either term. I'm not sure what relevance it has to photosensors either. Detecting a photon doesn't mean we know (or need to know) its precise location.

Oct 22, 2014permalink

Artak Hambarian

HowaboutRAW: it seems you are a curious individual. Keep being curious. However focusing on a statement that you do not clearly understand, is not so good for you. Existance of flash memory cannot in any way question quantum mechanics. Where from have you taken this false statement?

Oct 22, 2014permalink

@HowaboutRAW. We stretch this topic to the max, but it may be an interesting discussion for you anyway. From what you have posted in this reply section so far, I think, I am developing a feeling about where you are coming from.

Therefore, in all sincerity, do not believe everything which you can find written, only because you see fit. Understanding from first principles is much more important. It may take a couple years to work through all math but believe me, it is worth it.

As for infinities (singularities) and imaginary functions, don't believe for a second there is anything wrong with that. If you use sources claiming so, get rid of them. They obviously try to confuse you. E.g., the proper math treating of the kind of singularities you can encounter from the uncertainty relation is called "Functional analysis" in general and "Schwartz space" in particular (both have Wkipedia articles). Studying both was part of my third year education in Mathematical Physics. Learn, not read.

Oct 22, 2014*permalink

Artak Hambarian:

Electrons, though not single ones, and not charge, jump and are held in flash storage, therefore those electrons are localized and have known place and momentum.

My guess would be that there's some excuses about flash storage not yet using single electrons--wait 15 years.

Oct 23, 2014permalink

falconeyes:

Actually, most normal people learn this math between the ages of 12 and 14. It's real easy to understand.

But I'll explain it again: as momentum goes to zero in the denominator position blows up to infinity.

You've confused yourself by citing non-Heisenberg Uncertainty principle parts of QU. It's standard to rely on something you may indeed understand if the basic point, backed up, challenges your preconceptions.

I suggest Wikipedia for the Heisenberg Uncertainty Principle, and solve the first equation for one of the two terms on the left and let the other term go to zero. This isn't subtle of complex math.

Oct 23, 2014permalink

tripodfan

@HowaboutRAW. You claim that you can tell *exactly* where an electron is, because it's in a flash storage drive?

Oct 23, 2014permalink

tripodfan:

Um, I didn't say "'exactly'" but that's the idea; flash storage holds electrons in a very tiny space.

And it's not like flash storage is the only problem for the purported ghost like existence of electrons.

Oct 23, 2014permalink

@HowaboutRAW

Quite. Flash cells have a finite size, and that means there are no zeros involved as we don't know exactly where the electron is.

As it happens, the operation of flash memory relies completely on our understanding of the uncertainty principle. Clearing a flash cell relies on quantum tunnelling, and that is fundamentally based on the uncertainty principle. So, far from flash memory disproving the uncertainty, it's one of its most frequent demonstrations.

Incidentally, perceived problems with infinity (and dividing by zero) go back to at least Zeno. Modern mathematics has long dealt with those issues (although, it can't deal with the zeros at the bottom of a black hole, if you wish to go there).

Oct 23, 2014permalink

"Clearing a flash cell relies on quantum tunnelling, and that is fundamentally based on the uncertainty principle."

As does storing data in flash storage.

The problem is that the resulting place to which the electron jumps is very certain.

Now I think you mean that the dual wave/particle nature of an electron allows for quantum tunneling. Not a point I dispute, the problem is that under known, and repeatable, circumstances the electron becomes a local particle.

If flash proved the Uncertainty Principle, the electrons couldn't be used to store data.

It is reasonably certain, though there are some problems going back to Millikan, that electrons carry a charge of a discrete value.

Then:

"...perceived problems with infinity (and dividing by zero) go back to at least Zeno...Modern mathematics has long dealt with those issues", not really. Or yes if you mean sweeping the problem under the rug.

Oct 23, 2014*permalink

@HowaboutRAW:
I give up with you. Between you and people in the know, the gap is too large to communicate across if you don't try to learn.

Oct 23, 2014permalink

falconeyes:

The problem is you're just communicating distraction.

Indeed, it is safe to assert that with a better understanding of the interaction of light and electrons (meaning broadly matter and a better understanding of what an electron is) better low light sensors can, and will, be built.

However, particularly given your resistance to asking questions and finding answers, I don't see these much better sensors arriving in the next five years.

I remind you here, that Canon and Nikon still have not matched the optical quality of Zeiss and Leica lenses, though other manufactures have. And yes this means a different understanding of light than you'd cite.

Oct 23, 2014permalink

"However, particularly given your resistance to asking questions and finding answers, I don't see these much better sensors arriving in the next five years."

Wow. So @falcolneyes' particular resistance to asking questions is holding back a fundamental revision of the nature of matter and delaying the development of camera sensors for half a decade. I bet he never realised he was such a block to scientific and technological progress. All that by resisting asking questions. Such influence.

Oct 23, 2014permalink

I think the earth is flat. Anyone can see where the world ends, so it must be so.

Oct 23, 2014permalink

TheEulerID:

It's familiar block not just belonging to falconeyes. (He was just providing such a good example that I pointed out the connection.)

Oct 24, 2014permalink

Eleson:

And your "point" contributes nothing except a point of distraction.

Oct 24, 2014permalink

Jogger

Why is colour information so important in automotive applications? Wouldnt they just go monochrome and gain the efficiency of not having the colour filters.

Oct 21, 2014permalink

quezra

Maybe because it could have a lot more applications outside of automotive use if you can work color into it

Oct 21, 2014permalink

color is a great discriminator in pattern recognition. Remember that some pattern matching algorithm must identify something for this to be useful, like road, obstacles on it, trees, signs etc. Much easier with color information available. Moreover, it almost never is as black as -9EV in a real night. Not even in a starry moonless night.

Oct 21, 2014permalink

Jogger

ahhh.. that makes sense.

Oct 21, 2014permalink

I'd like it to see the difference between a red light and a green light. But then I drive a small car.

Oct 21, 2014permalink

junk1

I wonder if it has many clear (B&W) pixels to increase SNR similar to their LCDs.
A brighter reverse light would help too! Or how about a strobing reverse light synchronized to the sensor? Perhaps even strobing outside of visible light frequencies.

Oct 21, 2014permalink

RedFox88

Consumers, people just like color and it's as simlle as that.

Oct 21, 2014permalink

Roman Korcek

Sony
Helping Canon see the impossible.

Oct 21, 2014permalink

nunatak

key phrase: "uses extended exposure times rather than using multiple exposures."

to me this suggests it may be applied to fuzzy logic more than sharp imagery.

Oct 21, 2014permalink

Here is explain:

New WDR Function (IMX224LQR Only)

The IMX224LQR supports a DOL (digital overlap)-type WDR function. This function uses a method that outputs the data for three frames with different storage times line by line instead of frame by frame, enabling improvement of picture quality especially under low illumination compared to the existing multiple exposure WDR function.

Different lines can have different exposure, then combined to HDR pictures. It might not be ideal for still artistic photography, but might be good enough for shape recognizing and motion detection in dark environment from video feed.

HDR is pretty critical in driver-less vehicle. No one want their car see only washout zone under head-light and pitch black outside the coverage of head-light

Oct 21, 2014permalink

Timbukto

Just another sign that it will not be the profession or hobby of photography that pushes camera sensor technology forward. I imagine smartphones helped push it forward, and the next big push would be any future 'smart' device including smarter cars and vehicles. The industry for automated driving cars or other vehicles will be transforming. Unfortunately I'm guessing the big brother industry of smarter better security CCTV will also be looming as I imagine in the future they will no longer be inefficiently closed circuit, but openly distributed and data mined by big brother for effectively 'minority report' type technology.

Oct 21, 2014*permalink

(unknown member)

Coming to an Android in the very near future.

Oct 21, 2014permalink

papa natas

And my niece thought that by shooting the lights off her room when her bf comes to visit will continue to stop ME from keep recording....?
moowuaaaahahaha.
I don't see the greatness of this invention other than pushing the bar to a new dimension to help spying & peeping on others.
Oh no, someone would answer. Think about recording animal life.
Yeah, like if all those guys filming innocent animals copulating in the wild don't get a morbid kick out of it. Now even those species who mate in the dark won't be safe from their devious eye.
As Wye says...coming in the near future.

Oct 21, 2014permalink

daMatrix

This is great for navigation at night on a boat. Especially in and around water cities.
However boats must carry lights and a traffic radio is obliged for large / professional ships, smaller (read amateur) ships are literally guessing in the dark about size and distance of the (tiny) boat lights on the water (and some even are ignorrant of light rule) .

Oct 21, 2014permalink

ju_ju

Far from guessing there is a little thing called radar! Plus a wide spread use of AIS. Problem with using camera's for sighting other vessels is for one, a l;arge vessel needs time to change course and needs to know at a long distant where other vessels are. What distant will these cameras work? Especially in heaving rain? Fog? I do take your point on Amateurs but they would I suspect not know how to use any form of navigation, but having said that many amateurs are keen and clued up.

Oct 21, 2014permalink

alcaher

I dont know much about sensors but some people like thom hogan have written that mostly aps-c sensor (including sony) have reached its limits in terms of DR, Resolution and high ISO performance. If this is some kind of a new sensor technology, could we see more of this in next year's new aps-c cameras from Sony and nikon?

Oct 21, 2014permalink

dialstatic

People are always saying that something or other has reached its technical limits before a new paradigm comes to town and they're made to eat their words.

Oct 21, 2014permalink

@dialstatic,

not true for laws of nature. It is possoble to find limits for their validity, but once established, have never been broken. I.e., you can't escape the fundamental nature of photons.

Oct 21, 2014*permalink

schaki

Too bad about the pinhead size and while my 1.5mp Minolta Dimage EX1500 still is fully useable, 1.27mp or something around that is almost only good for front-camera on a cellphone.
The sensor in that old minolta was at least 1/2". We can hope that Sony is gonna introduce at least a 1/2.3", 1/1.7" and 1" of this new tech later on. 1/2.3" is still the standard in travelzoom compacts even though the 1/1.7" is getting to that point as well.

Oct 21, 2014permalink

mahonj

a: No-one drives around at night with their lights off - they drive around with them on.
Thus, you do not need this capability for automotive applications.

b: The image quality at +72db of gain is fairly awful - but what would you expect with +72 db gain in an image.

c: I think the news is the ability to do +72db if gain at all, or to get usefully clean images with (say) +36 db gain.

As far as I can see, this is about 2x as sensitive as normal sensors.

Which is something, but not driving cars around at night with the lights off.

Oct 21, 2014*permalink

T0nyK

Actually. Many high end cars have infrared night vision capabilities that are quite useful. My audi will warn me of people or animals near the road by placing a yellow box around them and alerting me before I see them with the naked eye, despite having my lights on. If this new sensor will enhance that technology, than that is very useful and interesting.

Oct 21, 2014permalink

(unknown member)

Think back-up camera at night, among other uses.

Oct 21, 2014permalink

Michael_13

Sure, Sony did this just for the fun of it. There must be no serious application at all. You must be right, Mr. Knowall!

Oct 21, 2014permalink

"As far as I can see" pretty much sums it up.

What if you instead had said:
"What kind of possibilities does this sensor open?"

Then you could have learnt about seeing outside the light beam, and much other stuff.
Tomorrows cars will be driven by computers, (well some are already.) And they need a lot of visual infomation to do that safely.

Oct 21, 2014permalink

grasscatcher

I live in deer country, drive many thousands of miles per year through deer-infested roads. My experience is that the drivers who hit deer are the ones who have their high beams on. I never drive with high beams - it blinds the deer and they freeze in their tracks. Low beams and the deer seem to have enough non-dazzled thought process left over to get out of the way long before the car reaches them. Thus, a system like this can alert you to deer without you paralyzing the deer and hitting them, IMHO.

As far as other technology, I see mostly uses for hunting, security cameras, and military uses (assuming they don't have something better already).

Oct 21, 2014permalink

stevens37y

for taking a photo of a black cat in a coal mine

Oct 21, 2014permalink

utomo99

We need better sensor on cameras, maybe 1 lux is good enough for now. and keep improving it. do not wait for 0.005 lux

Oct 21, 2014permalink

peterwr

"The ability to render color in near darkness is useful in automotive cameras that must make do with light on moonless nights."

Since most cars (and drivers) use headlights to deal with moonless nights, I'm guessing we're talking about *military* automotive cameras here. That and/or autonomous vehicles. Interesting, nevertheless.

Oct 21, 2014permalink

http://petapixel.com/2013/09/13/canon-debuts-exciting-prototype-sensor-exceptional-low-light-capability/

Headlights mostly point forward. Sooner or later, cameras such as these will become additions to or replace mirrors (in some cars they already are/do). And automatic obstacle detection (or even braking) systems can benefit from cameras covering a much larger angle than what the headlights cover too. "Eyes" all around the car.

Oct 21, 2014*permalink

LensBeginner

Can someone please enlighten me about what the previous generation sensors (even prototypes) were capable of doing?
I can get a feel of how much 0.005lux are, but I'm not familiar with the latest (now next-to-last) innovations...
0.01, 0.05, 0.1, 1..?

Oct 21, 2014*permalink

mrc4nl

Put this one in a smartphone please!

Oct 21, 2014permalink

sebastian huvenaars

Darker then a black steer's tookus on a moonless prairie night...

Oct 21, 2014permalink

Rob Bernhard

Drifting along like a tumbling tumbleweed.

Oct 21, 2014permalink

maurotandoi

Sony continues its improvments plan.It will be the leader soon.

Oct 21, 2014permalink

Just a Photographer

This sensor is developed by Sony components, which has 0 (zero) to do with Sony Digital Imaging. All they share is the brand name of Sony, but they are in fact two complete different companies.

Sony components is already market leader in selling sensors in the OEM market.

Don't be surprised if Sony will sell of their consumer division within a few years time and therewith their camera division as well. Sony is still much financially in troubled water as they are still rated 'junk' at Wallstreet.

Oct 21, 2014*permalink

naththo

Troll alert more likely. He doesn't own any Sony cameras. Nikon and Fujifilm only so far he owns. And another one owns only Fujifilm. Wrong place and the wrong time. I am here to smell fishy and sniff for any bull*

Oct 21, 2014permalink

Random Asian Guy

I think it's a little naive to say it has 0 to do with Digital Imaging. If it's useful and can sell and make a profit, divisions will cross.

It's almost as good as Canon and Nikon saying. There is no way a Electronic company like Sony and Panasonic could put digital sensors in SLRs.

Oct 21, 2014permalink

quezra

They aren't two different companies, they're two divisions in the same company. And if you read the Photokina interviews, there are obvious benefits, such as having some sensors made only for Sony cameras that they won't share with their competitors (e.g. the 1" sensor on RX100 took 2+ years before they let competitors have it, and then only after they had got a 1/3 stop brighter BSI version for themselves).

Oct 21, 2014permalink

Mike FL

A year ago, "Firefly Footage Captured in 0.01 Lux with Canon’s Amazing Low Light Sensor" as seen from the link below, but we have not see a single Canon being called "King of the darkness" as Nikon DF or Sony A7s does.

We do not know how long will take for a sensor from lab to production, but It is nice to see Sony also joins the game to speed up the process/lab_trial as result of competition.

When the time comes, I will very happy to buy a Pentax Q system equipped with (at least) 10MP sensor from Sony or Canon with low-light performance on par (may be better) than current Sony A7s and/or Nikon DF...

Can not wait...

Oct 21, 2014*permalink

Combine that Canon sensor with Canons statement about surveillance cameras and addressing that market.

Oct 21, 2014permalink

semorg

1/3 inch sensor at 1.27 megapixels = 63.5 megapixel Full Frame sensor!

Oct 21, 2014permalink

Mikael Risedal

a prototype, nothing real or new . same old read out.

Oct 21, 2014permalink

"every mfr is guilty of this."

That's why Toshiba, Sony and Nikon are not able to get a useful 14 stops (or more) of DR out of their sensors either after passing through the ADC's and the rest of the chain. Oh wait.

Even some APS-C cameras manage to output around 14 useful stop of DR.

Contrary to the DR numbers you quoted, which no one can ever extract from those cameras, due to their noisefloors, never mind banding further diminishing useful output.

Oct 21, 2014*permalink

Zeisschen

Just let Sdaniella live in here own patented world. She seems to own all all those Canon prototypes already. The fact that in real world others are already way over the horizon in terms of resolution and dynamic range with actual products we can buy is not part of her world.

Oct 21, 2014permalink

Hellraiser

Nope, wrong again. Canon doesnt have a single camera that has a DR over 12.1EV
http://www.dxomark.com/Cameras/Ratings/Landscape

Oct 21, 2014permalink

BarnET

Sdaniella sees impossible.
We see the truth and downfall of Canon

Oct 21, 2014permalink

Mikael Risedal

simple physics does not seem to be your best topic when it comes to signal noise, efficiency, color depth, resolution etc.

Oct 22, 2014permalink

"multi ISO"

I have yet to see examples of single shot Canon RAW files that can come close to 14 stops usable DR, without sacrificing half the vertical resolution (and introducing bad moiré and aliasing artifacts in certain areas along the way, or softening output in those areas by attempting to hide the flaws with crude blurring) or other firmware hacks and without the usual banding.

Prove me wrong, I'll be waiting.

As for DXO, they're actually being kind on Canon sensors, since they do not account for banding, which further limits the useful DR of many Canon cameras, beyond the numbers stated.

You're reaching, as usual.

Oct 22, 2014*permalink

Baloney. All manufacturers "cook" ftheir RAW files to some extend. If you've ever worked with dedicated astrophotography CCD systems, then you know what raw RAW looks like, with all the pixel flaws in all their glory. And do a whole lot of "cooking" yourself. Canon also bakes optical corrections in some of its RAW files.
Black blacks to hide shadow flaws is actually a known Canon jpeg trick to hide its shadow issues in many of its older cameras. For processed RAW files, it's a choice, since the output medium tends to have far less DR, so you compress your values into a gamma curve of your choice.

As for your RAW compression theories, that currently (visually) affects extreme transitions between certain values at worst. But uncompressed RAW doesn't magically bypass the Canon ADC's and therefore deals with the exact same read noise, thus DR limitations.

Still waiting for a valid single shot full resolution 14 EV Canon example to examine.

Oct 23, 2014permalink

Oh, a link about video and low res stills. Failed attempt.
Also, didn't you claim that 12MP do not suffice for a DSLR or even a compact? The stills in that link fall well short of even 12 MP. Never mind the other flaws mentioned and hack required.

But I admire your dreams. ;-)

Oct 23, 2014permalink

"don't ignore the fact what can be done with video (example) is same for hi-res stills."

No, it's not the same for stills, since the video is a low resolution sampling of the sensor, hiding most of the flaws that come with "multi ISO" reading of the sensor. Temporal differences also affect read noise. The ADC's will still give you the same (high read noise) DR results for full resolution stills as before and your only option is the one mentioned earlier,with all its flaws (half the vertical resolution, crawling moiré and aliasing or even extra blurring to hide those artifacts and a firmware hack).

Giving your RAW files a flat gamma curve, isn't magically going to improve your DR either, for ADC reasons mentioned above. It's the converter software makers that define the output curve,* regardless* of input (such a flat curve), so that reasoning ("catering to the consumer wish") doesn't hold.

Still waiting for that full resolution 14 EV Canon still example.

Oct 23, 2014*permalink

"I specifically said what CAN BE DONE with video (lo-res) CAN ALSO BE DONE with STILLS "

Read again, I said the principle in the video example is the same as "multi-ISO", just with flaws less visible due to the low res sampling and possible temporal NR from which video benefits. Hence why you can't have it work for full resolution stills, without cutting half the vertical resolution and the other downsides mentioned. Hence why you're not showing us (nor is ML without the points mentioned) any full resolution stills with 14 EV of DR. Despite "being in the know", as you claim it. It helps your credibility to stop making claims about Canon cameras that cannot be backed up by reality. Downsizing? You claimed that's basically cheating and that 12MP isn't even enough, let alone less.

We're not discussing other cameras here, you made claims about existing Canon cameras and have yet to back up your claims with a single full resolution still 14 EV shot. Quod non erat demonstrandum indeed.

Oct 24, 2014*permalink

Either you have trouble reading or just like to resort to straw man arguments. The video hides the flaws for being lower resolution and possibly temporal NR. Full resolution stills (without cutting vertical resolution, crawling moiré and aliasing or extra blurring) with 14 EV of DR is not possible with Canon cameras.

Oct 24, 2014*permalink

The flaws still apply to stills too, as does the major loss of vertical resolution (and color information) and the fact it requires a hack.

The "mono ISO" uncompressed RAW tricks only have to do with shifting the saturation curve a bit towards the right. Gaining between 0.3 and 0.5 EV isn't exactly the game changer you make it out to be.

And again, you said 12MP wasn't enough for a DSLR, yet here you're advocating downsizing to hide noise?

Your whole point about ARRI systems has nothing to do with your initial Canon claims to begin with. Others, such as RED, use different exposure times, which is quite hard (ahem) to do for photos from cameras that still have to rely on physical shutters due to their sensor designs.

Oct 24, 2014*permalink

star shooter

This sensor would be ideal for the Astronomical fraternity and hopefully inspire Canon, Nikon or even Sony, to bring out an Astro DSLR that has full corrected colour and sensitivity to equal or better the now defunct Canon 60Da.

Oct 21, 2014permalink

I'd love to see a full frame, H-alpha capable camera with astrophotography based features; an in-camera stacking feature would really cut it for me.

Oct 21, 2014permalink

Mikael Risedal

keep dreaming about Canon and their sensor tech.

Oct 21, 2014permalink

pfft, please. Canon sensors are what Sony sensors leave behind when they forget to flush.

Oct 21, 2014permalink

"if you know what you're doing"

If you're dreaming about analogue signals rather than useful data.

" or clean 1080p Low-Light video capture
what others have YET to do with stills or video"

If you close your eyes and ignore the A7S, which no current Canon on the market can equal in low light for both S/N ratio in video or stills.

Oct 21, 2014permalink

tkbslc

obviously not for photography.

Oct 21, 2014permalink

jaysonmc

The main significance here is the color portion. Our eyes cannot even detect color this low. It is akin to military saying shoot the guys wearing red vests and not blue ones in near total darkness. It is pretty amazing it can render color in this low of light.

Oct 21, 2014permalink

BodgeandDurn

Yes, but presumably you are talking about exposing for a few seconds to get that result? I suspect I could take an awful lot of ageing cameras and get the results you talk about. The significance of a sensor like this is that given the stated application (a car moving at speed) one imagines Sony are managing to get that color capture at tiny exposures.

Oct 21, 2014permalink

Boss of Sony

Congratulations to my minions.

Oct 21, 2014permalink

gargalamouth

In a full moon situation, you will certainly need an ND filter 3.0 and above.

_Sony rumor SRx_
The Sony A99.1* will be the first SLT camera with this new IMX224MQV sensor.
ISO range? 1 to 50 expandable up to ISO 100.(in case you left the lens cap on...)

* 99.1 is the percentage of light which goes to the AF array.

Oct 21, 2014permalink

I actually remember the QE on tiny sensors are over 70% years ago.
So I wonder how sony manage to increase sensitive by more than double, which makes QE well above than 100%

Oct 21, 2014permalink

Felix E Klee

Does the 70% refer to the QE of a full-color sensor, of a b/w sensor, or even of a single photo diode?

Oct 21, 2014permalink

Did Sony give any details on the shutter speed/aperture/equiv. ISO which was used to make this photo?

Oct 21, 2014permalink

yes, 72db gain (don't know what is the base ISO @0db though)
F1.4, 1/60s, 0.005 lux.

My guess is it is 409600 if base ISO is ISO100

Oct 21, 2014permalink

MASTERPPA

On my Sony 1/2 CAM, base ISO is 360 at 0DB and ISO720 at 6DB

A 1/3 chip may have a base ISO of 250 at 0DB and 500 at 6DB
My estimate is ISO 1,024,000.. Yes, ISO 1M (maybe 500,000)

Oct 21, 2014permalink

mpgxsvcd

According to Canon "They have a better sensor than this".

Oct 21, 2014permalink

ryansholl

Yeah, and according to my uncle he's got a better mousetrap. Turns out he's full of it but it makes him sound great at the time.

Oct 21, 2014permalink

mpgxsvcd

The astronomy community would be very interested in this sensor.

Oct 21, 2014permalink

rjjr

Not to mention law enforcement, the military and other government agencies.

Oct 21, 2014permalink

mpgxsvcd

I used a 1.3 Megapixel camera in 1999 and loved it. I could definitely still do ok with a less than 2 megapixel camera today if it could see in the dark.

Oct 21, 2014permalink

ryansholl

You don't need a 1.3mp sensor to see in the dark, the a7s is a credit card away :)

Oct 21, 2014permalink

misolo

It's a 1/3 sensor... scale up to 35mm-format size with the same pixel architecture and you have about 60mp.

Oct 21, 2014permalink

Giklab

Actually, more like 9 MP.

Oct 21, 2014permalink

misolo

1/3'' is about 7.2x crop. The ratio of areas is 7.2^2=51.8. The pixel count in a 35mm-format sensor would be 51.8*1.27=65.8.

Oct 22, 2014*permalink

Giklab

You are right, you do get ~66MP. Now I'm wondering what I multiplied by since I think I did square the crop factor...
Or maybe I didn't. Oh well.

Oct 22, 2014*permalink

autochrome

I hope someone updates the standards, because it's going to be problematic setting your ISO soon. ISO 819200, 3276800, 26214400, ...

Oct 21, 2014permalink

Camera manufacturers will probably just shorten it to 820K, 3.3M, 26M. They might also introduce 2-stop/3-stop intervals above a certain sensitivity. Though I don't think any camera will be hitting the millions any time soon.

Oct 21, 2014permalink

ryansholl

I really wish they'd just set ISO 200 to a standard of 1 and then go logarithmically from there. ISO 409600 becomes gain of 11, for instance, ISO 1600 a gain of 4, etc. Makes a hell of a lot more sense than dealing with the ridiculous numbers we're already dealing with.

Oct 21, 2014permalink