I guess most of you know how car maker (particularly German car makers) are focusing on image sensor technology to create save cars. And according to Nikkei Sony is ready to conquer the automotive image sensor market too. Aptina and OmniVision are the current leaders but Sony aims to eventually pick up 50% of the market.
The interesting news here is that Sony developed a new special sensor for automotive applications which is “10 times as sensitive to light as existing chips, allowing for easier detection of objects in dark environments. The sensor can provide a color display of the surrounding area even in moonlight, enabling it to detect obstacles more readily than a driver can.”
The sensor will be ready for late 2015.
via Image Sensor World.
0 Responses
I´m working within the vehicle safey industry and I would say that Volvo is leading the pack, followed by German and Japanese premium brands. EuroNcap is shifting is focus more and more towards active safety features (“accident prevention”) rather than passive safety as they would like to improve safety of new cars without increasing weight (CO2 emissions).
From 2015 and on, EuroNcap will only award 5 stars to cars which are equipeed with “advanced safety features”, ie autobraking, road sign detection or similar technologies.
I think Sony is doing the right thing, within a few years all premium car will have either cameras, radars or both. IR sensors for night vision are extremely expensive and if one could use a camera sensor instead, this would be a hit.
Uuuum, are you aware of the tech packed into the Infiniti Q50? Sorry, who is leading the pack?
FF form factor: 4MPix res?
Speaking of sensors the specs have surfaced for the new Canon EOS 7D senor coming oyt in September.
It will be a new technology multi layer sensor.
Unfortunately it will only be 24mp.
Canon has been hyping rumors of higher resolution and lots of megapixels but apparrently thier idea of lots of megapixels is 24mp.
This is dissapointing because I was hoping that Canons new chip would be revelutionary enough to force Sony to release something better .
One ray of hope…although unlikely…Canon did anounce a 120mp cmos H size sensor in 2010..so they could release a differrent camera with 120mp….this I would consider as an extreme long shot at best.
Sony is expected to release a 36mp a99II.
It is widely believed that Sony has a 54mp multy layer sensor in develpment….the Barry Sensor… I am hoping Sony will release a new A mount camera with this sensor soon as a competitor to Canons new multi layer sensor.
It would be in thier best interest to release it soon to reduce the impact of the new Canon multilayer sensor but if rumors of a 36mp a99II are true Sony is not likely to release a new better A mount camera in Sept to compete against thier own new camera.
Old rumors from several years ago hinted that this sensor might appear on the first hybrid A E mount camera?
50MP would seriously be pushing 35mm lens designs. I get why you’d want it for sports or wildlife, but for landscape we already have options- multistitch. Large format and the use of current DSLR’s can yield 130-150MP results with razor sharp detail. It’s also at a fraction of the cost of what Sony would probably charge for a 54MP camera.
the 50/1.4 Sigma ART is already there.
What you are forgetting is that in a multilayer sensor a single photosite is used for all colors wheras in bayer array 4 photosites are used to produce a color pixel.
A multilayer sensor will allow 3x the pixels with the same pixel size.
In the specific example of the 54mp sensor we are talking about it is 3 layers of 18mp and therefore the pixels are the same size as those of a 18mp bayer sensor.
In simplest terms tha actual resolution possible with a multiplayer sensor using the same lenses is 4x greater.
That’s why I sometimes drool profusely when I think about a 24MP Canon Foveon sensor. The dream is they’re talking 24×3. Not as crazy as it sounds because Sigma is very ambiguous on describing their own Foveon specs.
However, my expectation is much lower. I don’t really expect Canon’s new “layered” tech to be Foveon based.
Thats not how they count pixele.
24mp would be 8mp x 3 layers.
What you describe would be called a 75mp sensor.
It depends. Sigma (aka Foveon) does the math that way but if you count pixel sites, which is the way the rest of the industry does it, then the rumored Sony sensor would be called a 18mp sensor. Yes an 18mp sensor built this way would be of a higher resolution than even a 24mp Bayer sensor but it also wouldn’t achieve the resolution of a 54mp Bayer sensor. That’s been repeatedly proven with the Foveon sensors. Real resolving power of the Sigma Foveon sensor is about 1.5x – 1.8x (depending on who is doing the testing) of a Bayer sensor with the same number of pixel sites. I personally think we should count pixel sites and then let the resolution differentiate but marketeers are always going to do what gives their product the biggest numbers. So I’m sure Sony would call this so called 54mp sensor a 54mp sensor but the rest of the industry would call it an 18mp sensor. There’s going to be a lot of confusion among consumers which is exactly what the marketeers want. Perhaps we should start using a CIPA resolution standard rather than a megapixel count if we move into a world with multiple sensor technologies….
What he said!
That is not correct.
Bayer counts actual monochrome pixels.
A 4 pixel array is interpolated by DEMOSAICING to a single color pixel…the process is repeted 4x from differrent imaginary pixel locations.
A 24mp bayer sensor is actually 12mp green, 6mp red and 6mp blue.
A foveon type sensor of the same pixel count will actually have 1/3 more resolution…it will produce a true actual 8mp native resolution..not interpolate…when interpolated 4x..200% ….the image will be 32mp and be of an equivelent interpolation level as the 24mp file produced bu the bayer sensor.
The 54mp layered sensor is actually equivelent to a 74mp bayer sensor.
Look up Demosaicing to understand the process I just described.
I used the words pixel sites rather than pixels specifically to try to avoid that confusion. As you know color Bayer sensors use an:
RG
GB
orientation but each of those is a pixel site and the color is then interpreted for each in demosaicing. Yes there is a lot of color error in that but it is 4 pixel sites that result in 4 discrete pixels after conversion.
The sensor you describe as a 54mp sensor has 18 million pixel sites but each one is sensitive to all three colors rather than just one. That should provide more color accuracy but it does not triple the resolving power. The final image from that has 18 million pixels but the color for each is more accurate thereby giving a higher acuity image but it does not give you the same thing as a sensor with 54 million pixels and equating them is just not accurate. That, among other reasons, is why i like the idea of a resolution standard rather than a megapixel standard.
By the way, I didn’t mean to insinuate that you were equating 18mp to 54mp and I wasn’t clear in that above. but many, including those that read headlines like this do and, as you know it’s not a really accurate way to compare Bayer to multi-layer sensors. I wasn’t arguing with you at all just saying that when using the term megapixels to describe these two different sensor technologies we have to be very careful since we are really doing an apples to oranges comparison.
Oh and just FYI, I’ve known what demosaicing is for about 20 years and actually written demosaicing algorithms in the past 😉
Barry I’m afraid it is you who doesn’t understand demosaicing, look at the actual algorithms used. It is definitely not naive bicubic interpolation of each colour channel (which is what you have to do when you upsample a Foveon image).
In a Bayer sensor you can use the green pixel values to estimate the missing red and blue values and vice versa. This is possible due to the strong correlations between colour channels you find in a real scene.
Go and read the source code to DCRAW to understand what actually happens during demosaicing and report back as an actual expert:
http://www.cybercom.net/~dcoffin/dcraw/
I don’t have the link handy but I read elsewhere a description of how confusing Sigma was describing their sensor. If I recall, they had no less than 3 different meanings.
The confusion lies in the fact that ther is no common denominator to simplify the comparison.
Stacked sensors with 3 layers are being compared to a 4 pixel bayer array with twice as many green pixels as the other colors.
To top that off they are then interpolated 4x from 4 differrent locations.
Most people do not even realize that a 24 mp bayer sensor is really only 12mm + 6mp + 6mp…but they do realize 24mp foveeon is 8mp x 3.
They also usually do not understand that the 8mp file produced by a foveon is mathematically equivelent to a 32mp bayer sensor.
“the 8mp file produced by a foveon is mathematically equivelent to a 32mp bayer sensor”
That’s only true in the absolute worst case, that is if your scene contains pure red or blue light (and you can therefor only use one out of every four Bayer pixels).
If your scene contains yellow objects then you will record values from 8 million red pixels and 16 million green pixels. The red and green will be correlated, which means you can use the green values to accurately estimate the missing red values. If you have any grey objects then you have 32 million values to work with.
This all means that the effectiveness of Bayer vs. Foveon depends on image content. Testing on real images gives the Bayer sensors a 40% advantage in resolution given the same total number of detectors (i.e. an 18×3 stacked sensor will compete with a 40mp Bayer) in the average case.
I believe in canon hype when I see it.. So far the 7d 2 was supposed to come out 2x every year with revolutionary new AF, sensor, whatnot… What came? Nothing.. The closest to revolutionary Canon has done in years was the 70d dual pixel af..on a sensor that otherwise has the same lame sensitivity and dr of all the previous 5-6 year old Canon sensors. If they really come out with something new, it’ll take Sony and the rest maybe 15min to catch up..that’s a rough guess though.
I almost posted on this yesterday thinking about you but figured you’re gone on your outing shooting birds by now. Thus far, all the interesting pre-Photokina announcements I’m following are coming from Canikon and Sigma. In fact, not only do I find the soon to be featured at your local Wal-Mart a5100 uninteresting, I find the rest of SAR’s Sony dribbles actually depressing because now I’m expecting the worst regarding a-mount announcements.
It’s almost certain the Canon EOS 7D replacement – let’s call it the Mark II though it’s doubtful that will be the name – will have a “layered” sensor being hyped as revolutionary and not simply evolutionary like Sony sensors of late (yeah, a77II I’m thinking of your disappointing one). There are competing theories on what “layered” will mean. It appears to include Canon’s dual pixel technology however it pans out.
The most obvious assumption is it’ll be Foveon tech (I think Canon Rumors leans that way) since it’s the most common interpretation of “layered”. However, Sony’s done “layered” sensors in the past which were a whole different animal. I will say if Canon comes out with a 24MP Foveon sensor for the 7D Mark II it’ll still be exceptionally cool if done right and I doubt Canon will put out what’s expected to be a pro-level APS-C that’s half baked like Sigma’s Foveon bodies.
There are other interpretations on what Canon means by “layered” which get increasingly less interesting and are clearly merely evolutionary (like Sony’s “gapless” tech was). Since Canon has put so much time in developing this new senor I expect whatever they do to significantly boost IQ.
If it does, it’ll be interesting to see how Sony responds because this goes deep to the core of their strategy and far beyond them just competing SLT APS-C vs DSLR APS-C. Sony’s bread and butter is sensor tech they use across lines and sell to others. They can’t afford to be significantly leap-frogged by Canon. Will it force Sony to sh*t out a Barry Sensor in 2015??? ;^)
Sony is that crickets I still hear? [sigh]
MP count isn’t going to increase massively this generation, however we’re more likely to see the prices of cameras with high mega pixel sensors drop.
Apparently the A99II is a seriously well equipped camera (36mp, 12fps), that will sell somewhere’s between the D610, and D810 price points.
Plus Sony’s low light capabilities are rapidly evolving, and I wouldn’t be surprised if we see a 36mp SLT capable of A7 ISO performance within the next five years.
,,It is widely believed that Sony has a 54mp multi layer sensor in develpment….the Barry Sensor”
Sony is also working on a organic sensor….maybe Sony means this with the Barry Sensor?
Foveon and most multilayer sensors in development utilize organic…carbo based…materiels because they are very good atcabsorbing light.
Silicon based materiels have so far proven to be better at transmitting electrical energy.
Unlike thier competitors sonys patent utilizes both organic and silicon therefore possibly eliminating the shortcommings of purely organi sensors.
So you are correct..but partly correct.
thx!
if I heard it right, fuji is working on a single layer organic sensor
I see that they are…the same advantages found in organic materiels for a layered sensor woyld be relivent for a single layer…as well as the disadvantages.
I would be quite simply amazed if the 7D mkII was announced with a multi-layer sensor in September.
The only commercial multi-layer implementations we’ve seen so far have been a) very noisy beyond ISO100 and b) very slow in operation. The 7D is built for speed and users rely on high ISO settings for sports and fast shutter speeds with long lenses.
Canon is known for small incremental improvements between generations. They would have had to have made a quantum leap in this area to ensure speed and high ISO performance were not significant downgrades compared to the original 7D.
The rumored “new sensor technology” in the 7D mkII will likely be something to do with autofocus, just like the last piece of “new sensor technology” from Canon.
Many companies research things they have no interest in bringing to market just so that they can patent them to prevent competitors from profiting in these areas. I wouldn’t hold my breath for multi-layer sensors from anyone any time soon (except for Sigma, though they have already abandoned the 3 unique colour values per pixel design). But if it does happen, I highly doubt it will be Canon.
The 7D is not targeted at high resolution work. It’s a sports/action/wildlife camera. The rumored high resolution sensor from Canon is likely going to be used in a camera made for landscape and studio work.
Cars before cameras huh Sony. Gee z.
That’s why Sony’s imaging division is making massive profits each quarter (170 Million USD) while their competition posting losses. Plus they can apply everything they learn and apply it to the SLT, and CSC products lines.
Sony Semiconductor and Sony Consumer Electronics (of which the imaging division is a part) are two different companies within the Sony Corporation. The profits here would go to the former, not the latter. The imaging division doesn’t make sensors, they buy them from Sony Semiconductor, just like any other camera manufacturer.
The new small curved sensor with 20mm equivelent lens for cellphones is highly sensitve and a similar sensor would indeed be very usefull for autos.
Curved sensors are particularly well suited for low light and I would be surprised if this new sensor is not a curved sensor.
Sony writes not how much pixel has this sensor 😉
A Fullframe CMOS Sensor with 2MP is better than a curved sensor with 12MP 😉
But with the same MP is the curved better.
Smaller sensors have greater depth of field.
These sensors will be small for several resons.
Greater dof.
Lower cost
Space and weight savings.
10x as sensitive is marketing bull%%%t
* tested with equipment so sensitive it needs to be turned off 99% of the time and in conditions as specific as a light breeze on Mars under a full Earth.
10x is 3 stops, so marketing talk at all….If current car sensors do iso 1600. Sony says they made iso 12800 sensor
Who knows. This is refers to sensors used in cars, and who knows what the state of technology is in that arena.
I think not so. When Sony means 10x sensitive to the exist photo sensors is that ISO 819200 this value is possible.
So stupid Sony again. Why waste new sensor tech on a car. Idiots. Want to take down Canikon? Put this tech in cameras. Idiots.
The camera market is small compared to car market.
New regulations require every new car to have a backup camera.
Thats a new guaranteed market for cameras that far exceeds the standalone camera market.
They would be stupid NOT to develope new technolgy for this market.
Your comment is very…shortsighted.
Yes, it would be idiotic for Sony to take advantage of another business and prophet opportunity selling a device that parallels their current technology product line. Actually it would be idiotic not to.
There are more cars sold in the world every month than there are cameras sold in the world every year. Not stupid at all!
I see you are trying to set new standards for idiotic comments.
It is a smart move.
Sony has the technology to develop good sensor.
Sell more -> earn more -> more R&D fund -> make better sensor
It is a good news for Sony DSLR users.
Please engage your brain before start speaking. Imagine owning 50% of the car market. Heck, I’d jump or of the dslr business in a heartbeat!
The automotive camera market is huge! And car security is hardly a waste of new tech. Don’t you want to drive safely?
Why only for cars????????
Please make a camera with Fullframe (8MP-12MP) and this sensitive sensor and bring it 2015 to the market!
I waiting longingly for a camera with extreme sensitive to light!
Probably because it is just for automated detection and/or the actual IQ is crap compared to an actual camera.
Hm, possible.
But I found the extreme lowlight video sensor from canon works good….. , however has that only approx 2MP :-/
For me was that immaterial, I buy also a video camera with this 2MP, when it gives no other with more MP and this lowlight sensitivity…
I was years ago satisfied (only with the quality in good light) with my Nikon coolpix 995 and Canon powershot g3 . This cam’s has approx 4MP. Its suffices for me, when Sony release a ultra lowlight camera with this 4 MP
When I need for shots with more ligh more MP , it gives some good cameras for it 🙂
Get the A7s
I need a camera with approx the sensitivity from the canon video sensor. This 10x more ore approx ISO 812000 comes near to that what I want. The a7s is the best lowlight camera today, but for me a little to weak in very lowlight situations. Above all the noise in 204000 and above is sad. The violet light from the amplifier ,too.
I would bet a FS900 or alike full frame with the A7S sensor is in the works.
One of the worst parts of a car is the headlight. It blinds people coming from another direction, limits your view outside of the headlight’s fov, and provides a very small window of time to avoid something that does finally make it into the headlight’s fov. The fact that they’re still being used in every car is unbelievable to me. Hopefully this is the first step towards cars that don’t have headlights and instead simply have small lights to indicate their presence.
I’m sorry, physics says “no”.
Sensors are already achieving 50% quantum efficiency, that means half of all incoming photons are converted. Since you can’t detect more than 100% of the incoming photons a 10x increase in sensitivity is completely impossible.
I think Sony’s marketing don’t mean 10×50% but rather 3 Stops more (as Kiril Karaatanasov below write). The best camera a7s has real ISO 102400 and 3 Stops more are real ISO 819200 that is possible!, because.- the highest possible ISO is approx ISO 1000000.
A camera with real ISO 819200 is a dream 🙂
That’s not really sensitivity when you’re talking about amplification. The highest ISO settings are always implemented digitally but multiplying the numbers that come out of the ADC, so you can multiply them by any factor you like, it wont increase image quality.
In short ISO 819200 is perfectly possible, but since you’re not capturing any more light you wont get a meaningful image, just a whole lot of noise, just like if you took an ISO 102400 and pushed it three stops.
I’m read about the incoming photons. What is when the sensor has less pixel with a bigger pixel area. With a bigger pixel area comes with the same high ISO more photons to the each pixel. So is the picture in the ISO 819200 brighter but has not much more noise.
Am I right?
I think they mean ‘sensitivity’ in the marketing sense, i.e. ISO capability. So by offering ISO 819200 with some advanced NR algorithm, they would claim 10x better sensitivity. NR reduces details, but you don’t need to see details in this type of camera. As long as the camera can detect reasonably large objects, it will work fine.
In order for your claim to be true 50% of the actual energy frim the photons would have to be converted to electrical energy..not even close as the bayer array automatically filters out 75% just to get the correct colors to the correct pixel. We are at nowhere near 50% efficiency.
Then 100% of the elctrical energy would have to be converted to image detail…again not even close.
Claims made about how efficient our current sensors are and how close they are to perfection are greatly exagerated and only reflect our lack of ability to do it better.
The frequency response of each filter in the Bayer CFA overlaps significantly meaning the light loss is closer to 50% (no idea where you’ve got the 75% figure from).
Lets say you’re able to count and measure the frequency of every single photon (which we are nowhere close to being able to do yet) with 100% accuracy. That’s still only a 4x improvement in sensitivity.
Sensitivity and image detail are two completely different things so I’ve no idea what point you’re trying to make here.
Fundamentally if you only capture so many photons there are very hard limits to the signal to noise ratio you can expect. There will be no huge leaps in sensor technology that will allow noise free images in low light. I’m sorry.
Excellent news. Object detection algorithms benefit tremendously from color information. See Subaru’s Gen1 vs. Gen2 stereoscopic camera capabilities for example.
reading it…..comes into my mind the Canon sensor they did for very low light. A full frame sensor with 1920×1080 resolution.
It’s a pity at this time SAR have no more info’s about this sensor. I mean also, its a low MP Sensor. It is arguable whether, has the cam a smal (1Zoll) or a large (Fullframe) sensor.