Previous news story    Next news story

Panasonic promises high sensitivity sensors using 'Micro Color Splitters'

By dpreview staff on Feb 4, 2013 at 17:50 GMT

Panasonic has devised a method to increase the sensitivity of image sensors, by replacing the near-universal colour filter array with prism-like 'Micro Color Splitters' to generate colour information. The key advantage is that all of the incoming light is directed to the sensor, instead of half or more being absorbed by the colour filter dyes. This promises to deliver images with less noise in low light. The development is published in the journal Nature Photonics, and outlined in a press release on Panasonic's website.

According to Panasonic, the dyes used in conventional colour filter arrays absorb at least half of the incoming light before it reaches the sensor's photosites. In contrast, the company's 'Micro Color Splitter' technology passes all of the incoming light to the sensor. In practical terms, this corresponds to a sensitivity increase of a whole stop, which should in turn result in a stop improvement in noise performance. In other words, a 'Micro Color Splitter'-based sensor should give similar noise at ISO 3200 as a conventional Bayer sensor does at ISO 1600, for example.

Perhaps the most interesting aspect of this development is that it may be the first serious alternative to colour filter array-based sensors since the Foveon X3 appeared over 10 years ago. Like any new technology there'll be plenty of development to do before it appears in a real product, but the prospect of a stop better high ISO performance is certainly a tantalising one.  

Press release:

Panasonic Develops Technology for Highly Sensitive Image Sensors Using Micro Color Splitters

Constitution and features compared with the conventional method

Osaka, Japan - Panasonic Corporation has developed unique "micro color splitters", which separate the light that falls on image sensors by exploiting light's wavelike properties. Applying them to actual image sensors allows bright color images to be achieved even under low-light conditions. This development makes color filters unnecessary by using the micro color splitters that control the diffraction of light at a microscopic level. Panasonic has achieved approximately double the color sensitivity in comparison with conventional sensors that use color filters.

Image sensors are used in devices like smartphones, digital still cameras and video cameras, as well in security, vehicle parking, office, and healthcare applications - anywhere, in fact, that digital imaging is needed. Conventional color image sensors use a Bayer array, in which a red, green, or blue light-transmitting filter is placed above each sensor. These filters block 50 - 70% of the incoming light before it even reaches the sensor. Progress is being made in increasing the resolution of image sensors used in mobile and other devices by reducing pixel size, but demand for higher-sensitivity cameras is also increasing. Panasonic has developed a new technology that can be applied to existing or future sensors to enable them to capture uniquely vivid color images.

The developed technology has the following features.

  1. Using color alignment, which can use light more efficiently, instead of color filters, vivid color photographs can be taken at half the light levels needed by conventional sensors.
  2. Micro color splitters can simply replace the color filters in conventional image sensors, and are not dependent on the type of image sensor (CCD or CMOS) underneath.
  3. Micro color splitters can be fabricated using inorganic materials and existing semiconductor fabrication processes.

This development is based on the following new technology.

  1. A unique method of analysis and design based on wave optics that permits fast and precise computation of wave-optics phenomena.
  2. Device optimization technologies for creating micro color splitters that control the phase of the light passing through a transparent and highly-refractive plate-like structure to separate colors at a microscopic scale using diffraction.
  3. Layout technologies and unique algorithms that allow highly sensitive and precise color reproduction by combining the light that falls on detectors separated by the micro color splitters and processing the detected signals.

Panasonic holds 21 Japanese patents and 16 overseas patents, including pending applications, for this development.

This development is described in general terms in the Advance Online Publication version of Nature Photonics issued on February 3, 2013.

More on the Technology

 Conventional method using a color filter  Developed method using a micro color splitter

1. Unique method of analysis and design based on wave optics permitting fast and precise computation of wave-optics phenomena

FDTD is widely used to analyze light in wave form, but its heavy computation workload has up to now made it impractical for designing micro color splitters. On the other hand, BPM is an effective method of fast computation, but it has lower precision than FDTD and cannot accurately simulate color splitting. This prompted Panasonic to develop a practical and original design method that permits fast and precise computation of wave-optics phenomena. This technology allows the precise modeling of optical phenomena such as reflection, refraction, and diffraction by modeling spaces in regions with different optical constants and applying BPM to the spaces. This method can be applied not only to the design of micro color splitters, but can be extended to the design of other nano-scale optical processing systems.

2. Device optimization technologies leading to the creation of micro color splitters that control the phase of the light passing through a transparent and highly-refractive plate-like structure and use diffraction to separate colors on a microscopic scale

Color separation of light in micro color splitters is caused by a difference in refractive index between a) the plate-like high refractive material that is thinner than the wavelength of the light and b) the surrounding material. Controlling the phase of traveling light by optimizing the shape parameters causes diffraction phenomena that are seen only on a microscopic scale and which cause color separation. Micro color splitters are fabricated using a conventional semiconductor manufacturing process. Fine-tuning their shapes causes the efficient separation of certain colors and their complementary colors, or the splitting of white light into blue, green, and red like a prism, with almost no loss of light.

3. Layout technologies and unique algorithms that enable highly sensitive and precise color reproduction by overlapping diffracted light on detectors separated by micro color splitters and processing the detected signals

Since light separated by micro color splitters falls on the detectors in an overlapping manner, a new pixel layout and design algorithm are needed. The layout scheme is combined and optimized using an arithmetic processing technique designed specifically for mixed color signals. The result is highly sensitive and precise color reproduction. For example, if the structure separates light into a certain color and its complementary color, color pixels of white + red, white - red, white + blue, and white - blue are obtained and, using the arithmetic processing technique, are translated into normal color images without any loss of resolution.

Notes:

1. Diffraction
 
Behavior of light as a wave on the wavelength (nanometer) scale. Various phenomena occur when a wave encounters an obstacle.
 
2. Bayer array
 
The arrangement of color filters used in most single-chip digital imaging sensors used in digital cameras, camcorders, and scanners to create a color image. The filter pattern is 50% green, 25% red and 25% blue.
 
3. Charge Coupled Device Image Sensor (CCD sensor)
 
A type of solid-state image sensing device for digital imaging, used in digital video cameras of all types. It has higher sensitivity and lower noise than other sensing devices.
 
4. Complementary Metal Oxide Semiconductor Image Sensor (CMOS sensor)
 
A solid-state image sensing device for digital imaging using CMOS.
 
5. Finite-Difference Time-Domain method (FDTD)
 
FDTD is a versatile modeling technique used to solve Maxwell's equations by spatial and temporal discretization.
 
6. Beam Propagation Method (BPM)
 
A numerical analysis technique in electromagnetics for solving the Helmholtz equation under conditions of a time-harmonic wave.

Comments

Total comments: 115
Aleo Veuliah
By Aleo Veuliah (Feb 10, 2013)

This is great. Well done Panasonic.

Hope that soon we can see cameras using this kind of sensors, preferably on the Micro 4/3 G System cameras.

3 upvotes
zorgon
By zorgon (Feb 6, 2013)

@falconeyes
This is not the same technology patented by nikon. Nikon uses dichroic mirrors which work by interference. The panasonic patent uses a splitter that works by diffraction. Comletely different technologies implemented in a completly different way.

3 upvotes
falconeyes
By falconeyes (Feb 5, 2013)

This approach has been patented by Nikon a few years ago.

Their patent uses dichroism to separate and divert colors at the sensel level which scales to very small photo sites.

0 upvotes
erikandmarcie
By erikandmarcie (Feb 5, 2013)

Summary of this article states:

"should give similar noise at ISO 3200 as a conventional Bayer sensor does at ISO 1600"

But actually, that's not true. ISO performance will be the same, all else equal. It's just higher ISO's won't be needed since more light reaches the photo site.

Comment edited 8 minutes after posting
1 upvote
yabokkie
By yabokkie (Feb 5, 2013)

forget the funny ISOs,
it's 3 dB better SNR at the same exposure.

Comment edited 8 minutes after posting
7 upvotes
zodiacfml
By zodiacfml (Feb 5, 2013)

Right, I didn't noticed that.

0 upvotes
the reason
By the reason (Feb 5, 2013)

Yabokie youre such a sad angry troll...does it make you cry that the sensor will match a full frame snr?

0 upvotes
Nerval
By Nerval (Feb 5, 2013)

yabokkie is not trolling here...
Noise is measured in DB, it's a logarithmic scale, 3DB is the full stop advantage DPR is talking about... So he's just stating a fact in mathematical value... Really.

1 upvote
Andy Westlake
By Andy Westlake (Feb 6, 2013)

@erikandmarcie: "But actually, that's not true. ISO performance will be the same, all else equal. It's just higher ISO's won't be needed since more light reaches the photo site."

This is a misunderstanding of the definition of ISO that's used for digital cameras. It simply links shutter speed, aperture, and JPEG image brightness. If two cameras give the same image brightness at a given shutter speed, aperture and light level, by definition they're shooting at the same ISO, regardless of the sensor's efficiency.

As a consequence, the fact that the 'Micro color splitter' sensor design is more sensitive doesn't mean the camera can use lower ISOs. It means more light is used to form the image under directly comparable conditions, which should give a stop better noise performance at any given ISO.

6 upvotes
erikandmarcie
By erikandmarcie (Feb 6, 2013)

@Andy. Interesting indeed, thank you. This must mean the new sensor will have a higher native/base ISO limit compared to conventional sensors?

0 upvotes
Fred Briggs
By Fred Briggs (Feb 10, 2013)

I would interpret this as a camera of this type, if set to the same shutter and aperture as a conventional camera, assuming equal sensor sizes, would have a stop greater exposure at the sensor. In conditions where there is not enough light this will mean less ISO amplification needed to create the correct image brightness. Accordingly a one stop lower ISO setting may be used, and less ISO amplification means a better signal to noise ratio.

In conditions where there is enough light to fully expose the sensor at the minimum acceptable shutter and aperture settings, then this would translate into the ability to use one stop faster shutter speed or one stop smaller aperture for the same exposure (at the sensor).

A sensor with this technology could be smaller than a conventional sensor but still give equal performance, and this may be the primary way the technology is exploited - i.e. current full frame performance at less than full frame size, and so on down the range of sensor sizes.

0 upvotes
yabokkie
By yabokkie (Feb 5, 2013)

it looks the design will have problems with large apertures and wide angles, nothing the current micro-4/3" cares.

1 upvote
itsastickup
By itsastickup (Feb 5, 2013)

What I want is massive dynamic range. I am still using film cameras for exactly this reason. There's only so far you can go with NDGrads, HDR etc.

1 upvote
Petka
By Petka (Feb 5, 2013)

How many stops do you get with film? Nikon D800 gives you 14 stops, D4 over 12 stops.

0 upvotes
dylanbarnhart
By dylanbarnhart (Feb 5, 2013)

I hope by "film cameras" you mean medium format, because 35mm film can't match dynamic range of today's DSLR.

0 upvotes
itsastickup
By itsastickup (Feb 5, 2013)

The dynamic range reported by DxO is entirely bogus. Yes, technically you can get 14 stops, but in practice it's more like 9.

With film there is no problem at all. Even the limited range Neopan 400 smashes the Nikon d7000 which has a supposedly high DR.

1 upvote
Nukunukoo
By Nukunukoo (Feb 5, 2013)

The immediate problem at the moment is resolution. At smaller photosite scales (read: Phonecams to DSLRs) it becomes incredibly difficult to manufacture to maintain the diffraction accuracy of the deflectors. Looks extremely promising. In the meantime, I'm expecting stacked sensors and pixel-binning (Like the Nokia 808) to be the next upcoming features on camera sensors.

Comment edited 2 times, last edit 1 minute after posting
0 upvotes
robenroute
By robenroute (Feb 5, 2013)

Stacked sensors: as in Foveon?
Pixel-binning sensors: as in Fuji's EXR sensors?

0 upvotes
Franka T.L.
By Franka T.L. (Feb 5, 2013)

well well, this is not the first time such use of optical prism. Nikon and I believe a couple others had similar setup in previous patent and design. what differs though is that the previous obes all center around doing a full spectrum recording on single collected site, where Panasonic is more about replacing the Bayer patterned filter with the prism. I wonder if and how this tech would be implemented in a real product ... would be very nice if Panasonic would give that bit more in depth

0 upvotes
Ken Phillips
By Ken Phillips (Feb 5, 2013)

Franka, the system does NOT use a prism, which is REfractive. It uses DIiffractive optics.

0 upvotes
Justtimthen
By Justtimthen (Feb 5, 2013)

Can't believe the filters have been losing that much light! Real glad that Someone's on to it though.
If the intensity of light onto the sensor is doubled I wouldn't be surprised if the end result looks even better than 1 stop particularly under the dimmest lighting conditions.
It will gather more light without decreasing DOF which in some curcumstances maybe very desirable so I think this is pretty big. Well done PanI

2 upvotes
Locknut
By Locknut (Feb 5, 2013)

By their very nature colour filters absorb light of the wavelengths they do not pass. So (simplistically) green absorbs blue & red, red absorbs blue & green etc. This means approx 50-70% of the light falling on the sensitive areas of the imager is absorbed and wasted.
This diffraction technique deflects the "unwanted" colour onto an adjacent pixel. Rather than being absorbed it is still used to create the image.
It is unusual for this sort of thing to be announced unless it has been implemented in some form. Hopefully we will see it in a product in a few years.

Comment edited 18 seconds after posting
1 upvote
Jun2
By Jun2 (Feb 5, 2013)

Hope I can get that in GH5

0 upvotes
Antony John
By Antony John (Feb 5, 2013)

Interesting concept.
Dyes as used in Bayer filters are not light stable and would degrade with time producing a colour shift. Wonder if anyone has ever experienced this?
Then again possibly the other mechanics of the camera would fail first before this becomes problematical with the number of actuations required to get to a noticable degradation threshold.
Now if only they could design photosites that were sensitive only to some specific light frequency bands (RGB) this would do away with the requirement for any 'pre-processing' of the light (Bayer array etc) to start with.

0 upvotes
Ulukai71
By Ulukai71 (Feb 5, 2013)

As far as I understand that would not make much difference:

White Light is parted into red, blue and green.
If you filter away red and blue, you have a loss of around 66% light.
If a certain (sub)pixel is just sensitive to green light, it ignores red and blue, which means it ignores around 66% of the available light.

0 upvotes
Nerval
By Nerval (Feb 5, 2013)

Not if you manage to have three micro prisms sending the light to three pixels... then each pixel receives this 33% of red, blue, or green, that it is destined to receive. That's what the bit about layout is, if you have the correct layout under your micro prism, you loose about zero light on the surface exposed.

But that's still very theoretical...
Waiting for them to make one and show it =P

While on a bayer array, most of the light is absorbed to let through only the expected wavelength but even the correct wavelength will lose a bit of intensity, so that's why you lose 50 to 70% of the incoming light at a pixel level on regular bayer array sensors.

Other detail, these micro-prism as good as they are are bound to lose light as well, especially if they work through diffraction... CF pattern of a wave after diffraction, or any "hair-split" diffraction with a ray of light.

Comment edited 3 minutes after posting
0 upvotes
Nerval
By Nerval (Feb 5, 2013)

Plus as some other noted, what about light incidence ? Some rays will have very specific incidence on extreme lenses, and MFT might be 'the most' "telecentric" system yet, it's still not actually telecentric...

But it's an interesting bit of technology. (More likely to find immediate use than Lytro's current tech.)

0 upvotes
the reason
By the reason (Feb 5, 2013)

Amazing the amount of sensor engineers we have on this site, that are so eager to disprove the sensor. Its a miracle someone here hasn't made a 60 mp noiseless sensor with 32 stops of dynamic range.

19 upvotes
Mssimo
By Mssimo (Feb 5, 2013)

How do they get blue or green?

0 upvotes
Mssimo
By Mssimo (Feb 5, 2013)

white + red
white - red
white + blue
white - blue are obtained and, using the arithmetic processing technique, are translated into normal color images without any loss of resolution....planned light room support in 2036. Foveon support still TBD

Comment edited 24 seconds after posting
1 upvote
Anadrol
By Anadrol (Feb 5, 2013)

green = white - red - blue...

2 upvotes
galecian
By galecian (Feb 6, 2013)

According to the figure, only W+R and W-R are provided !
So, the question is not answered here.
Is there a second raw with B or G ? (no mention of that in the article).

0 upvotes
wfektar
By wfektar (Feb 5, 2013)

First: thanks for letting us know. Could be interesting.

Second: I realize that this is a press release, which means it is obliged to spend a lot of high sounding words saying practically nothing. In which it succeeds (their note on diffraction is hilarious). So how about a little analysis: how it works (beyond the fluff), what might prevent it from working, that sort of thing. That is, a reality check of all the market-buzzword speak. Thanks!

2 upvotes
B1ackhat
By B1ackhat (Feb 5, 2013)

Unfortunately, Panasonic sensors usually show a level of noise at ISO 400 equal to that of ISO 800 on a Sony sensor. Also, I don't suppose these new sensors will do anything to alleviate Panasonic's JPEG color issues. Maybe Olympus will use them in the XZ-3. :)

0 upvotes
Adrian Harris
By Adrian Harris (Feb 5, 2013)

Strange that, - as my own Sony and Panasonic cameras produce results opposite to yours. The Panasonic produces very fine detail without the horrid colour noise that Sony always cleverly manage achieve with their own sensors!

1 upvote
Nerval
By Nerval (Feb 5, 2013)

In what world? And on which Calibrated screen?

Not trolling...
Three sony sensors, one panasonic sensor, on the bottom right. In which way does it have more detail?

CF Here :
http://www.dpreview.com/reviews/studio-compare#baseDir=%2Freviews_data&cameraDataSubdir=boxshot&indexFileName=boxshotindex.xml&presetsFileName=boxshotpresets.xml&showDescriptions=false&headerTitle=Studio%20scene&headerSubTitle=Standard%20studio%20scene%20comparison&masterCamera=nikon_d7000&masterSample=nikond7000_nrn_iso%206400&slotsCount=4&slot0Camera=nikon_d7000&slot0Sample=nikond7000_nrn_iso%206400&slot0DisableCameraSelection=true&slot0DisableSampleSelection=true&slot0LinkWithMaster=true&slot1Camera=sony_nex6&slot1Sample=dsc00032&slot2Camera=oly_em5&slot2Sample=p1010017&slot3Camera=panasonic_dmcgx1&slot3Sample=p1030067&x=0.09001346501952491&y=-0.9428337734423899&extraCameraCount=0

Comment edited 3 minutes after posting
0 upvotes
Nerval
By Nerval (Feb 5, 2013)

Not on the side of the bottle of Baileys, not on the green-violet fabric, not on the feather, not inside the box... Do not confuse with spurious resolution, like on the queen, you can tell from higher res cameras that there are many more lines than appears on this GX1 sample...

DPReview forum refuses the post of shortened url... Reason : Swear words... Seriously.

Right now Sony's line of production is about as good as it gets in terms of CMOS chips...

Comment edited 33 seconds after posting
0 upvotes
pocketuniverse
By pocketuniverse (Feb 5, 2013)

nice

0 upvotes
Lan
By Lan (Feb 4, 2013)

So, in summary, it's basically the 3CCD camcorder idea, but on a pixel level.

2 upvotes
erikandmarcie
By erikandmarcie (Feb 5, 2013)

Agree, except interpolation is still used in this case.

1 upvote
Nerval
By Nerval (Feb 5, 2013)

I thought exactly the same thing... As many others I guess.
But I mean their idea is kind of cool. Just not so sure it's that feasible.

0 upvotes
John Summers
By John Summers (Feb 4, 2013)

A Lumix teaser to go head on with the Sigma Merrill Fovern all color sensor.

Comment edited 42 seconds after posting
1 upvote
SigmaChrome
By SigmaChrome (Feb 5, 2013)

Well, they have got a lot of catching up to do. Foveon has a 10+ year start on them. And Foveon uses the properties of silicon itself to determine RGB values. Anyway, it will be interesting to see how long this development takes - AND if it ever gets to market.

1 upvote
migus
By migus (Feb 4, 2013)

Makes most sense iff integrated within the u-lens, to also compensate for the incident angle delta (center - edge). But it's a lab prototype, not mass product yet.

0 upvotes
schirmer
By schirmer (Feb 4, 2013)

Some mathematical details from nature article:
http://www.nature.com/nphoton/journal/vaop/ncurrent/extref/nphoton.2012.345-s1.pdf

Looks like a pixel pitch of about 1µ is required and the micro lenses have to ensure less than 10° angle of light at splitter. But this is only my clueless interpretation.

0 upvotes
ttnewton
By ttnewton (Feb 4, 2013)

>>> "...pixels of white + red, white - red, white + blue, and white - blue are obtained and, using the arithmetic processing technique, are translated into normal color images without any loss of resolution."

I wonder if no resolution is lost only providing the pixels are ~1/4 the area as for the conventional filter approach, because the final light entering a pixel actually derives from that pixel plus its adjacent pixels. For the same size pixels, it seems that sharing the light between pixels should result in a loss of resolution.

0 upvotes
Andy Crowe
By Andy Crowe (Feb 4, 2013)

I wonder if this approach will have the same colour problems as CYM filter arrays? It's a shame no-one managed to perfect that technology as it would have been an easy improvement over RGB arrays.

1 upvote
hugh crawford
By hugh crawford (Feb 5, 2013)

The fact that there is no such thing as magenta light is of course one problem with CYM filter arrays. CYMG has possibilities.

0 upvotes
micahmedia
By micahmedia (Feb 5, 2013)

http://www.dpreview.com/reviews/kodakdcs620x

(btw, if any DPR admins are paying attention, I found the product pages for a lot of older discontinued stuff seems to be pointing to the wrong places. I had to employ google to find that, since your internal links are broken.)

0 upvotes
Peter Heckert2
By Peter Heckert2 (Feb 4, 2013)

It is to consider, light does not come in from an recticular angle as shown in the advertisement images above.
It comes from all angles with bright lenses and if aperture or focal length (zoom factor) changes, the angles will change. So it must be much more complicated and sophisticated, if this should be an universal sensor for ILS cameras.

I suspect it is for webcams and cellphones only.

0 upvotes
stevens37y
By stevens37y (Feb 4, 2013)

Actually the microlenses (top lense on the diagram) collect the light from the different angles and direct it to the sensor area.

0 upvotes
Peter Heckert2
By Peter Heckert2 (Feb 4, 2013)

I would want to hope so, but no lens does change the angle of a light beam that goes straight through the center of the lens.

This is already a problem with Bayer sensors and wide angle lenses, false colors in corners can occur.

Sorry, I am a little bit pessimistic to see this soon in a Camera ;-)

0 upvotes
OttoVonChriek
By OttoVonChriek (Feb 4, 2013)

Four thirds standards require telecentric lenses

2 upvotes
6x9
By 6x9 (Feb 4, 2013)

OK, it is so nice... What is the trade-off?
I guess, this technology will not work equally well for all sensor sizes, as there must be limitations on the photocells (and/or deflector) dimensions.

0 upvotes
Equals Nothing
By Equals Nothing (Feb 12, 2013)

The way I understood it, they just have to keep the pixel pitch very small, so larger sensors are possible, but the rez would have to be massive. FF of 100mp?

0 upvotes
ThePhilips
By ThePhilips (Feb 4, 2013)

If it brings us even a step closer to the Foveon-like true-color sensor, I'm all for it.

Actually, this is a first time something what appears to be easy to manufacture is proposed.

But I'd love to hear how that "fine-tuning" of splitter supposed to work though. And the "white + red, white - red, white + blue, and white - blue" thing sounds to be quite prone to clipping in both directions.

3 upvotes
schirmer
By schirmer (Feb 4, 2013)

If one uses different capacity on 'sum' pixels than on 'sub' pixels the electrical signal wouldn't differ too much. But I think we'll see Panasonic patents quite soon. Otherwise they wouldn't have published this work now.

0 upvotes
Peter Heckert2
By Peter Heckert2 (Feb 4, 2013)

Possibly this must be calibrated to the lens in use and so could prevent usage of arbitrary lenses. So is this for fixed lenses only?

Comment edited 9 minutes after posting
0 upvotes
schirmer
By schirmer (Feb 4, 2013)

Not sure. It just replaces the color filter in the stack, so below the micro lenses so the micro splitters should behave similar.

0 upvotes
BJN
By BJN (Feb 4, 2013)

Why? The sensor should be agnostic to the light it receives.

3 upvotes
ThePhilips
By ThePhilips (Feb 5, 2013)

@BJN: "The sensor should be agnostic to the light it receives."

But not to the angle of the light hitting the sensor. Deeper the wells are, worse the light capture at edges. And now we got here also the splitter and anyway you look at it, it would be sensitive to the angle.

0 upvotes
OttoVonChriek
By OttoVonChriek (Feb 4, 2013)

What an excellent idea. And like all the best ideas, so incredibly simple!

I suspect the reason we are seeing this approach being applied first on four thirds is because it is easier to do with highly telecentric optics which characterise four thirds systems.

2 upvotes
Andy Crowe
By Andy Crowe (Feb 4, 2013)

I wouldn't call this a simple idea, it's probably very complex to manufacture tiny beam splitters like that.

Also they don't mention four thirds so it's still to be seen whether this technology will apply to large sensors or (like BSI) only compact sensors.

1 upvote
Cane
By Cane (Feb 4, 2013)

I am sure there must something to complain about with this, just not sure what.

21 upvotes
stevens37y
By stevens37y (Feb 4, 2013)

Interesting that nobody invented it earlier. The splitter can be a kind of microprism.

0 upvotes
Deleted pending purge
By Deleted pending purge (Feb 4, 2013)

Like Polaroid instant slidefilm... Possible.

0 upvotes
schirmer
By schirmer (Feb 4, 2013)

It's much smaller than a microprism: Size below the wavelength of light. So the dimension will be approx. in 100nm range. Not sure about available diffraction angle. If only small angle is possible small sensor pixel have to be used, otherwise the splitted wavelength will end on same pixel again.

0 upvotes
plasnu
By plasnu (Feb 4, 2013)

It is 2 TIMES better. The color filter is removed and the complementary color now reaches the sensor. Copmplementary color was blocked at the color filter on Bayer.

Comment edited 2 minutes after posting
2 upvotes
plasnu
By plasnu (Feb 4, 2013)

Breakthrough technology!!!!

1 upvote
JEROME NOLAS
By JEROME NOLAS (Feb 4, 2013)

When? How much...?

1 upvote
Roland Karlsson
By Roland Karlsson (Feb 4, 2013)

Interesting. Hmmmmm .... of course no one knows when or if this is going to materialise into real world sensors.

0 upvotes
Michael_13
By Michael_13 (Feb 4, 2013)

Excellent concept!
Would be great to have it in a camera, soon.
And please not only in µ43, but also in compacts like XZ-3 or LX8!

Hope they won't run into moiré problems...

1 upvote
Hugo808
By Hugo808 (Feb 4, 2013)

I bet my photo's will still be crap...

44 upvotes
Allan
By Allan (Feb 4, 2013)

Lol, mine too!

3 upvotes
BJN
By BJN (Feb 4, 2013)

Like your punctuation ;-)

4 upvotes
HDF2
By HDF2 (Feb 4, 2013)

Ya gotta love honesty.

3 upvotes
Michael_13
By Michael_13 (Feb 4, 2013)

... but crap with much less noise!

;-)

5 upvotes
ManuelVilardeMacedo
By ManuelVilardeMacedo (Feb 4, 2013)

Crap with incredibly accurate colours. I'm in.

2 upvotes
Tlipp
By Tlipp (Feb 5, 2013)

Right on Hugo808. Only the best would make that statement!

2 upvotes
CyberAngel
By CyberAngel (Feb 5, 2013)

I'll take a wild guess:
Huge is using (also) the Nokia 808 and
38391936 pixels

1 upvote
Emacs23
By Emacs23 (Feb 4, 2013)

Two lenses in the light path? That doesn't sound too good.

0 upvotes
markusm
By markusm (Feb 4, 2013)

These are micro-lenses. At this stage, you do not care too much about distortions, chromatic aberrations or any such thing. The only important thing is that the light travels in roughly the correct direction.

6 upvotes
Andy Crowe
By Andy Crowe (Feb 4, 2013)

Most (all?) large sensors already have microlenses.

0 upvotes
kb2zuz
By kb2zuz (Feb 4, 2013)

My first concern would be that it might lead to color reproduction problems. Foveon has poor color reproduction with certain colors as it makes a hard delineation between where red stops and where green starts and where green stops and where blue starts. The human eye over-laps these colors so there's some green in many reds and some blue in many greens, bayer filters mimic this overlap and have fewer problems with color reproductions. Without seeing specifics I can't say for certain that this technology will have problems, but I have a feeling that it is quite a possible concern.

5 upvotes
BJN
By BJN (Feb 4, 2013)

I agree. All I'm seeing here is white+red, white, and white-red. I don't pretend to understand how visible colors could be reconstructed using pink(?), cyan, and white but it does sound like it's the long way around to get green, the color we're most sensitive to. One stop of sensitivity is good, but only if color rendition doesn't suffer.

0 upvotes
MarkInSF
By MarkInSF (Feb 4, 2013)

They're only showing one row. In current Bayer sensors one row of photosites is r,g,r,g and the next is b,g,b,g. The row they're showing would replace one row, the next would instead split out the blue, so would be w-r,w+r,w-r,w+r and the next would be w+b,w-b,w+b,w-b. Each pixel consists of four photosites. In Bayer they are r, g on one row and g, b on the next.

In this the photosites would be w-r,w+r, then w+b, w-b. It takes slightly more computation to turn that into an rgb value (assuming you want one) , but what you need is there. Many video cameras have worked with analogous principles, measuring white, red, and blue. This is a bit more sophisticated. I've been hoping someone would come up with something similar. It's potentially a lot simpler to make than a stacked sensor, though it doesn't have all the advantages. It would still need an aa filter, though only in one dimension. Not a pro, so could be wrong about that.

3 upvotes
jl123
By jl123 (Feb 4, 2013)

Foveon disadvantages too. Also If fuji works out then 6x6 random could be used with this. no aa. j

0 upvotes
CFynn
By CFynn (Feb 5, 2013)

This has nothing to do with Foveon - completely different concept.

2 upvotes
Everdog
By Everdog (Feb 4, 2013)

I wonder how long it will take to get from lab to production.

2 upvotes
plasnu
By plasnu (Feb 4, 2013)

It wouldn't take a long time.

No no new manufacturing technology required.

1 upvote
vodanh1982
By vodanh1982 (Feb 4, 2013)

When's it supposed to come out? Next year?

0 upvotes
ageha
By ageha (Feb 4, 2013)

Most probably not.

0 upvotes
mandophoto
By mandophoto (Feb 4, 2013)

Well, this is very annoying. I promised myself after last year's camera acquisitions that those would be the last for at least 10 years. Now this?! Well this is just great, just great.

4 upvotes
b33g33
By b33g33 (Feb 4, 2013)

This seems to have all the benefits of Foveon with none of it's weaknesses.

Wouldn't this also improve color rendering and resolution over Bayer?

(From the press piece: "The result is highly sensitive and precise color reproduction. For example, if the structure separates light into a certain color and its complementary color, color pixels of white + red, white - red, white + blue, and white - blue are obtained and, using the arithmetic processing technique, are translated into normal color images without any loss of resolution.")

Hopefully they can get it to actually work outside the lab.

Comment edited 2 times, last edit 2 minutes after posting
4 upvotes
plasnu
By plasnu (Feb 4, 2013)

Wrong. You're misunderstanding.

It's just different color pixels than Bayer and will have the same (kind of) drawback as Bayer. This one also will require AA filter.

3 upvotes
ThePhilips
By ThePhilips (Feb 4, 2013)

@plasnu. I wonder if that Panny sensor can be somehow combined with the Fuji X-Trans...

1 upvote
Stefan Zeiger
By Stefan Zeiger (Feb 4, 2013)

I wouldn't jump to any conclusions before seeing results from a production model. Fuji's XTrans design does away with the AA filter but it still needs demosaicing, so you get higher resolution and different demosaicing artifacts than from a Bayer sensor. This one might be similar (with yet another tradeoff, and new challenges to makers of raw converter software).

0 upvotes
JadedGamer
By JadedGamer (Feb 5, 2013)

Not the same drawbacks as Bayer, since Bayer uses light absorbing dyes while this uses diffraction to guide "complementary" wavelengths away from the cell and onto neighboring cells, increasing the light reaching the (w+r/b) sensors.

Comment edited 2 times, last edit 3 minutes after posting
1 upvote
Zdman
By Zdman (Feb 5, 2013)

The split is happening at each photo site so there would be no need to demosaice. The problem could be with the math as a result of the overlap red on one side big whit blob in the middle and yellow on the other. The red and yellow sections are liable to be quite a small part of the deflected light so like foveon I suspect that at low light levels (where those small sections don't get enough light) colour accuracy and saturation will start to fall off.

0 upvotes
3systermuser
By 3systermuser (Feb 4, 2013)

so will new Oly OM-D successor use it?

1 upvote
ageha
By ageha (Feb 4, 2013)

No.

2 upvotes
Kim Letkeman
By Kim Letkeman (Feb 4, 2013)

I'm sure that Panny can't wait to outfit the next generation Olympus m4/3 camera with the latest breakthrough in sensor technology :-\

7 upvotes
Surefoot
By Surefoot (Feb 4, 2013)

Well their sensor division sells its production, to whoever. It may be Panasonic themselves, or an external company, in which case they do make a quick profit on it, instead of having to actually sell the cameras that use these sensors.

0 upvotes
agentul
By agentul (Feb 4, 2013)

also, no full-frame sensors either. Panasonic knows what "some people" have been saying about their cameras and lenses.

0 upvotes
brunobarolo
By brunobarolo (Feb 4, 2013)

With a one stop improvement, Panasonic may finally bring their sensors to the level where Sony sensors are today. Nice :-)

5 upvotes
3systermuser
By 3systermuser (Feb 4, 2013)

it is not just a stop better , according to Pana site , it will have 2 times better light sensitivity to any other kind of sensor tech available now including you mentioned Sony ones.
and Sony sensors are not really the best, I mean the new Toshiba sensors in the latest Nikons easily beat the Sonys.

I believe the D800 sensor is a Toshiba , this is why Sony has no access to it and using silly old 24.3mp junk used in the A99v and D600.
I have both A99v and D800E and know that the D800E sensor is much better , though as a whole camera I prefer the A99v.

Comment edited 3 minutes after posting
2 upvotes
Matthew Miller
By Matthew Miller (Feb 4, 2013)

"2 times better" = 1 stop.

17 upvotes
ryansholl
By ryansholl (Feb 4, 2013)

http://nikonrumors.com/2012/08/29/confirmed-the-sensor-inside-the-nikon-d800-is-made-by-sony.aspx/

Not that it matters much, but for the record and all...

Comment edited 12 seconds after posting
10 upvotes
OniMirage
By OniMirage (Feb 4, 2013)

I think the biggest detail improvement will occur with low light and high ISO performance. Being able to accurately represent color in low light is a huge plus especially if your lifting shadows. In addition high ISO performance receiving a boost may also receive the ability to resist color shifts and retain fine detail. If these two things are possible by this technology alone, it could be a big deal.

1 upvote
Emacs23
By Emacs23 (Feb 4, 2013)

3system: d800 has sony sensor. BTW, distribution noise figures of 24 "Toshiba" sensor are too close to Sony (other sensors differ much). And I was told the Nikon D3200 uses Sony sensor. This can only mean Toshiba licensed production of Sony.

4 upvotes
marike6
By marike6 (Feb 4, 2013)

@3systermuser According to Chipworks, the D800's 36 mp sensor is made by Sony. The Nikon D5200's APS-C sensor is made by Toshiba, and like the D800 sensor, it's absolutely fantastic, topping all other APS-C sensors tested by DxOMark.

3 upvotes
Kim Letkeman
By Kim Letkeman (Feb 4, 2013)

They are already close enough for practical purposes in RAW and pretty close in JPEG too. This has the potential to leap ahead of everyone by increasing sensitivity, resolution and color accuracy all at the same time. Can't wait to see it in a body ...

3 upvotes
plasnu
By plasnu (Feb 4, 2013)

1 to 1.5 stop better.

Comment edited 1 minute after posting
1 upvote
Revenant
By Revenant (Feb 5, 2013)

@Emacs23
D3200 and D5200 have different sensors, interestingly enough. According to Chipworks, D5200 has a Toshiba 5105 sensor, whereas D3200 has a Nikon NC81369R sensor (manufactured by Renesas.)
In other words, none of the Nikon 24 Mp cameras have the Sony IMX096AQL sensor from A77/NEX-7! And the three sensors have different architecture, meaning that it's not just a case of "rebranded" Sony sensors. For example, Toshiba uses copper metallization, whereas the other two use aluminium. Similar performance doesn't necessarily imply the same technology or manufacturer.

Here's the Chipworks teardown of D5200:
http://www.chipworks.com/blog/recentteardowns/2013/01/08/inside-the-nikon-d5200-dslr-toshiba-found/

0 upvotes
Victor Engel
By Victor Engel (Feb 8, 2013)

I'm skeptical this will have wide use. I see the potential for color cross talk that varies by lens and aperture.

0 upvotes
Total comments: 115