This is the new RGBW pixel array of the patented Panasonic sensor.
Panasonic patented a new RGBW sensor design (Click here to see the document). It adds a White pixel on the usual Red-Green-Blue array. This should increase the dynamic range of the image. I heard from my sources that many companies are working on such an array and it is very likely that it is going to be used first in compact cameras.
Olympus also issued a new patent for 14mm, 16mm, 17mm and 22mm f/2.8 pancakes (Click here to see it). I believe the 14-16 and 22mm lenses are just modifications of of the already released 17mm f/2.8 pancake. If Olympus really plans to do such pancakes they should be at least as fast as the superb [shoplink 14796]Panasonic 20mm f/1.7 pancake[/shoplink]!
As usual with patents keep in mind that most of them are here just to protect intellectual propriety. A company may not use them at all in a real product!
P.S.: Links to the superb and very likely most sold m43 lens ever, the Panasonic 20mm f/1.7 at [shopcountry 14796].
76 Responses
f2.8 is not fast enough for m43, primes must be at least f2 (but f1.7 or 1.8 is even better !)
Why wouldn’t F2.8 be fast enough? I’d guess 90+ percent of all mFT-pictures are taken using a F3.5-F5.6 kit zoom. Speed is only one of several reasons for primes. IQ, small size and price are as important and you can’t get all at the same time.
all those gf1+20/1.7, gf2/gf3+14/2.5 kits never got used?
I’m sure gf1+20/1.7 and gf2/gf3+14/2.5 kits are sold and used, but the vast majority buy cameras with zoom today. Maybe the 90% is off, I dont have access to official statistics, but I almost never see people with fixed lenses while kit zooms are everywhere and some more advanced zooms like 18-200/14-140 etc are fairly common too.
Anyway, the point was that there are more reasons for buying a fixed focal length than just speed.
Because we need the possibility to have creamy blured background and to compensate the sensor performances by aperture of the lense ! f2 instead of f2.8 mean iso 100 instead of iso 200 ! And the current products show that it is possible to keep compactness with good aperture: pany 20 mm f1.7, oly 45 mm f1.8, oly 12 mm f2. The two latest are not pancakes but they are small and light.
JF.
When I look at something I can see 180 degrees around me. The peripheral image and the rest of the non focal area is clear and identifiable. It is not creamy, fuzzy, blotchy or full of strange unidentifiable shapes. See the hat picture in the next post on the 19mm sigma. A well balanced shot. Take the shot at wider than 2.8 and its a hat on a post, not a hat in a shop of straw products.
2.8 is way fast enough. You can do shallow depth of field shooting with 2.8 just fine.
At what FoV?
You need to be pretty close to the subject to get good isolation with 2.8 on MFT.
Not fast enough? Not any more with the new sensor.
5.6 eqiv on fullframe, this is almost stopping halfway down
If you always ‘think’ 35mm then you will never find anything new. Note, 35mm not FF.
To be FF it needs to be at least 10×12″ after all 35mm only exists because the ‘proper’ cameras were to heavy for the arthritis.
Now think that progression through and one has 43, M43 and Nikon V1.
Embrace the progression.
Aren’t all pixels on a sensor white anyway? But a color filter on top of each. Just means one will be without filter then.
Yea, but the filter usually only let 1 type of wavelength to go through. That means without filter, it would be 3 times as much light to go through.
This is why the rumored Leica M10 b/w without color filters will be a fantastic camera (if you’re into b/w). It would be a reason for me to buy one, until now Leica had no appeal to me.
How awesome that would be huh. If it has ‘only’ 1.5 times the dynamic range of todays sensor average that would be a game changer, imagine 2 or 3 times the range, it’d certainly have more flexibility than actual film has/had with dodging and burning, like the famous streetogs used to do in the early days. And no more ugly artifacts when simulating color filters, but then we need 14-bit raw as well.
Imagine this:
Instead of coloured filters, 3 picels with 3 levels of ND filter style for even greater DR…
Just saying 🙂
The is only W. pixler so not have colour filter on, but the maybe need not hightpass, lovpass and antialianse fiter on sensor.
I wouldn’t expect the white one to be completely without a filter, as the spectral sensitivity of a bare CMOS photodiode is skewed to the blue side. The “white” photodiodes will thus have to have an orange filter on top, reducing their overall spectral sensitivity to below three times the blue sensor in an RGB array. But please note that also the rest of Panasonic’s array isn’t RGB, but “W-B”, “W-G”, and “W-R”. This means it uses the substractive base colors YMC instead of the additive ones RGB. This will also increase sensitivity by roughly a factor of two. As, except for some early Leica digital compacts, YMC sensors haven’t found much use in the past, this layout presumably induces other problems, most likely with color rendition.
Yep, this is substractive colours sensor.
I just wonder how much this would affect to colour accuracy as Foveon basically has substractive colours with light of most colours happily spread to every channel which causes Foveon’s sensitivity problem. (gathers lot light but most of that is useless for calculating colour of pixel and has to be thrown away)
So maybe would be just easier to do that high end video cam standard method of beam splitting to different colours in one sensor package:
http://www.43rumors.com/foveon-like-sensor-patent-from-panasonic-flash-sync-patent-from-olympus/
I think this pattern is exciting. Each subtractive cell is surrounded by 4 white pixels from which the subtracted color can be estimated. And, the sensor has almost a 1 1/4 stop advantage.
Oops, it’s 1.44 stops I think, so almost 1 and 1/2
Panasonic really need to spit out a global shutter sensor. That will unleash new image stacking possibilities and limited DR will be a thing of the past.
They’re trying but global shutter on CMOS introduces a ton of issues N.B. I have no idea why this is. Massive spike in noise and reduced DR are not something that Panasonic can compromise with, especially when the only real benefit of global shutter is in video!
No doubt it’s what everyone’s working on but given that recent cinema-digital-backs are using radial mechanical shutters… I’m guessing a breakthrough is some way off.
Cheers Tom.
I think you aren’t seeing the photo possibilities. A Global Shutter can provide: much faster framerate, less mechanical wear (think timelapses at full resolution).
With much faster framerate comes Bracketing and Stacking which in their own turn can provide: less noise, more DR, more Resolution, more control on DOF.
Still think it’s all about video?
Yeah, because bracketing and stacking are already available with mechanical shutter while it will be possibly a thousand of years before electronic shutters could catch up with typical stills shooting exposures like 1/500 anyway (that’s a whopping 500 full XX-megapixel snapshots per second). This means an electronic shutter is not going to support a bracket of, e.g. [1/400 + 1/500 + 1/640]. Just around 1/30 or 1/60, which barely eclipses still shooting scenarios involving a tripod, I guess…
You’re right, improved mechanical wear and tear coupled with loosing yet another moving part are definite bonuses – plus cheaper manufacturing!
I’m not there when it comes to bracketing/stacking in the real world i.e. things move but I’m an absolute perfectionist. We do already have Sony’s “handheld twilight mode” and several “auto-HDR” functions though so it would be interesting to see how much a global shutter could improve these!
P.S. I’d love to see Olympus introduce a 4 EV stack mode
http://www.luminous-landscape.com/forum/index.php?topic=17775
Used it several times for indoor photography and the results are mind-boggling!
wondering whats special about the lenses in order to patent it, any optics specialists here? I wish Mr. W. from Dpreview would visit this forum too.
It’s there just to protect the specific lens design. No need for it to be extraordinary. Technology firms patent anything these days, even the most trivial aspect of a device.
really….
with this 4×4 pattern could panasonic remove the anti_aliasing filter, like fuji do with the x pro 1 ??
This is a normal non-random 2×2 pattern that only replace 1 green to white.
I’m all for RGBW CFAs but I think that it would be better incorporated as a 45 MP sensor with 2×2 pixel binning/over-sampling to produce a 10 MP image – equates to 12×8″ 300 DPI prints, ample for most photographers. Plus that sensor wouldn’t require an AA filter so enlargements would be similar to Foveon i.e. acceptable to 2x.
This set-up is perfect for 4/3s because the panchromatic pixel saturates far quicker than the respective RGB – use this on a full-frame and the W-channel is next to useless.
> This set-up is perfect for 4/3s because the panchromatic pixel saturates far quicker than the respective RGB
And why would that quicker saturating of unfiltered full spectrum sensitive pixel be less problem when pixel size is changed?
Problem stays same unless you make that white pixel smaller than others or decrease its sensitivity in way or another.
And more smaller pixels just means lower photosite signal more messed up by noise which is bad also for dynamic range.
I think perhaps this is what they are doing, but in a different way. The way I read it, the colour array is more like Rw Bw Gw Ww, using part colour part white elements. A clever dodge on Kodaks patent
I realise that there’s a paradox because all of the photo-sites are identical in size (correlation between RGBW pixels) but it’s more a case of using the W channel to counter the adverse effects of having smaller photo-sites, rather than boosting performance.
I’d like to see Olympus innovate on the sensor side and a 45 MP over-sampled design would do that (Nokia’s PureView and the GH2’s video performance are all the evidence that you need).
Basically I’ve tried to think of a way to replicate Foveon using a CFA. 2×2 sampling gives full colour pixels which is better than any de-mosaicing algorithm and makes an AA redundant – it’s all theory so until someone tries it I’ll never know but I think that it’s a valid concept.
2.8!? … too slow.
The proposed 17 mm is not the same as the old 17/2.8. This version has seven elements, whereas the old has six. The new 17 adds an extra element at the rear, wich has arrows underneath, that probably indicate that it moves for focusing.
Skip about lenses topic , new sensor for Pana ? Hmm , Looks promising , big thing on it’s way . Pana , please speed out the process !
ok panny.. I want that senzor in my bloody GH3! 😀
Yes, i think also this. 😉
Don’t forget the very highend camera too !
It should be over the top .
Please don’t get our hopes up. There is no indicator that Pana will release a camera beyond GH3. It was merely hinted a the possibility that the GH series itself gets more “pro”.
Those panckakes should be faster than 2.8.
It is the Olympus Walmart-Branded line of lenses….
Strange Focal length for primes. 14mm ok, 17mm ok – but 16mm & 22mm ?
What are the possible downsides of a RGBW arrangement? I can understand that the compounded pixel will become more sensitive, because the White one has no filter and catches more light.
But are there any downsides to it, such as reduced colour accuracy?
I think the downside is more complicated demosaicing algorithm. It amazes me that nobody has done this before. I thought Kodak had a patent on it.
Yes, reduced color resolution (and more chroma noise). There’s a reason why alternative Bayer arrays have been often proposed but generally not implemented. Certain camera makers (cough Canon cough) have been criticized for weakening their color filter selectivity to get “better” high ISO performance. See, for example, http://www.dxomark.com/index.php/Publications/DxOMark-Insights/Canon-500D-T1i-vs.-Nikon-D5000/Color-blindness-sensor-quality
The panchromatic pixel helps reduce luminance noise and the only hit regards chroma is in the green channel; if you already have a sensor that has good chroma noise performance, it could handle an RGBW CFA.
In a post above I suggested using pixel-binning/over-sampling rather than your standard de-mosaicing algorithms.
Panchromatic pixels are prone to blowing out thus they’ve been relegated to CCTV cameras and other small sensor applications – until someone can produce a sensor with independent pixel gain/ISO, we’re just treading water.
> Panchromatic pixels are prone to blowing
That effectively means all three colours channels will end up with heavy hit in noise.
Kodak’s experience with this array is quite the opposite
“Internal tests suggest this color filter array produces the same signal-noise ratio at sensitivity levels around 1.6-1.7 stops higher than those using Bayer pattern filters”
but since the panchromatics are being mixed with pixels designed to absorb 2/3 of the full spectrum they will be much less likely to blow than panchromatics mixed with RGB 1/3 spectrum pixels … would that be right?
Didn’t Kodak patent this a few years ago?
yes,
they called it “Truesense” which is actually a bag of technologies which RGBW was a part. That patent would have gone to Platinum Equity I guess.
However Panny’s patent differs, and is rather like Olympus pdAF patent for its bayer layer control we saw just a few weeks back, in that these pixels are part white. So more like Rw Gw Bw Ww
I would like a cam with all white pixels.
+1! It would be amazing to have a sensor that was designed for black & white photography (significantly increasing the effective resolution and dynamic range.)
Better yet, a camera with a removable/replaceable (or even one that could automatically swap) the color filter (and AA filter) would be simply astonishing. Of course, 99% of the general public knows not of what we speak. 🙂
You don’t know what you are asking for. Those filters require micrometric precision to be placed on the sensor, no way they could swappable by the user or automatically. It would be *much* easier to swap the whole sensor!
I wonder why they don’t address the issue of “dynamic range” by just eliminating the capacity (== charge storage) that is currently used per pixel. Instead of translating a charge level into a digital number in a second step, they could also put in enough transistors per pixel to just count the incoming photons as they arrive. A counter register does not really take that much transistors, but would be great to count all the “photon arrived at surface input gate” events, allowing for an unprecedented dynamic range.
German ColorFoto test, no 5 2012 — max score is 100. A top 10 list:
1. Nikon D4: 60.5
2. Nikon D700: 55
Pentax 645D: 55
Canon 5Dii: 55
5. Oly E-M5: 54
Nikon D3X: 54
7. Sony Alpha 65, NEX-7: 53.5
9. Sony Alpha 77: 53
10. Pana GX1: 52.5
Pentax K-5: 52.5
Not tested: Fuji X-Pro1 (the X100 gets 47.5), Canon 1DX + 5Diii, Nikon D800. The Oly Pen-E3 got 43.5, the Pana GH2 got 43.
The main criticism of the E-M5 is the texture reproduction above ISO 1600. DR is measured to 11.3 at base ISO.
Wow! Impressive.
Nice! Article is not available online yet, so I’ll try to get my hands on a magazine.
Dr of 11 is compact camera realm. I feel sorry for olympus
As far as I could see, they listed the DR of the D4 to somewhere between 9 and 10. The D4 is not exactly compact… Ok, maybe there was a misprint or maybe I mistook “Dynamik” for DR.
Ok, there were at least one obvious misprint in ColorFoto — one place, they referred to the “Dynamik” at ISO 100 — I assume they use 100 for all cameras that support this ISO number, and the lowest existing for others — e.g the E-M5.
Do people really want cheaper prime lenses? I doubt it….. how about faster please? I think people that aren’t willing to spend money on quality lenses will stick with their slow kit zoom, they will not spend money on cheap slow(ish) primes.
True that. F2 should be minimum imo. Shallow DoF is one of the main reasons to get a prime anyway.
Not too fussed if it is a pancake or not (personally)
Surely optical quality is the main reason for buying a prime? From there I’d guess size and weight.
Obviously a prime has a simpler optical design that allows for a faster lens and I’m all for that but DOF is all about lens focal length – hence 4/3s not being suitable for some projects.
+1
cheap pancakes for the tourists and big, high-quality wide aperture primes for the photographers
Those who have shot with the great FT HG zooms are aware that f2.8 is absolutely fine for everything if the lenses provide frame-to-frame sharpness wide open and are optically top class. They are good enough for isolatimg your subject and evereything else.
So, if these new pancakes – should they ever be produced – are optically in the same league as the HG Zuikos, f2.8 is abolutely fine for me. And for the cases when we need faster lenses, there already are the M.Zuiko 12 2.0, the Lumix 14 2.5 and 20 1.7, the Summilux 25 1.4, the M.Zuiko 45 1.8 – and soon the M.Zuiko 75 1.8.
Most defintely, not every lens in my bag has to start at 1.4 or 1.8.
The diagram makes it seem like a CMYW array.. W-R (cyan), W-G (magenta), W-B (yellow), and W. That’s actually pretty cool.
… and so thought the inventors of the other many alternatives (see: http://en.wikipedia.org/wiki/Color_filter_array and http://en.wikipedia.org/wiki/CYGM_filter ) – it’s not quite guaranteed this will result in better images, in the end.
Suddenly f2,8 is fine.. They play catch up with the new budget price sigmas
Could it simply be that 43 rumors is not one person, but many persons and that to some, 2.8 is fine while it is poor for others?
14mm to 22 mm pancakes: not very interesting… we have lenses like this yet: pana 14mm and 20mm, Oly 17mm…
I would stronly prefer pancake zoom (with better caracteristics than Pana 14-42X) or near-pancake wide lense (like 10 or 12mm) or small tele (40mm or 50mm)
new patent for 14mm, 16mm, 17mm and 22mm f/2.8 pancakes
—
Jeez, i am terribly excited.