Luminous Landscape Forum
Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: design_freak on April 10, 2013, 02:53:32 pm
-
Belgium-based CMOSIS have developed an image sensor that blows away the competition
http://www.cmosis.com/products/standard_products/chr70m
-
1: it is not a medium format sensor, so it does not really belong in this forum
2: it does not even seem to be made for a "photographic camera" in the usual sense of the word. It seems to be designed for machine vision and aerial photography.
-
blows away the competition
...You've seen pictures generated with said sensor?
Or from where are you deriving your very strong analysis here?
-
1: it is not a medium format sensor, so it does not really belong in this forum
2: it does not even seem to be made for a "photographic camera" in the usual sense of the word. It seems to be designed for machine vision and aerial photography.
It is most likely made for photographic application in that it's designed for a 35mm optical format. There is also a monochrome version... that is not a typical machine vision application.
I also think that a 70MP sensor has strong implications for the MF market as well as CMOSIS being a possible MFD sensor supplier.
CMOSIS is the company that makes the new Leica sensor. The Leica is getting excellent reviews.
-
Hi Doug,
Did you ever read the announcement of Phase One equipment?
Chill out :)
-
It has rather low dynamic range. Its intended use is for document scanning for example, it is not intended for photographic applications the way we think about it.
-
63db snr is not going to blow away any current tech...
-
63db snr is not going to blow away any current tech...
True. This sensor has analogue output rather than on chip ADCs. However CMOSIS filed two US patents on new on chip ADCs
-
with 3.1um pixel pitch you can't expect much even with on board ADC's.
-
with 3.1um pixel pitch you can't expect much even with on board ADC's.
D7100 sensor is 3.9 µm and has a dynamic range of 13.7 vs 14.4 D800 and 13.6 IQ180
It looks like new technology is pushing the quality obtainable with small pixel pitch..... the limitation will be diffraction.
-
D7100 sensor is 3.9 µm and has a dynamic range of 13.7 vs 14.4 D800 and 13.6 IQ180
It looks like new technology is pushing the quality obtainable with small pixel pitch..... the limitation will be diffraction.
It will be interesting to see what the limits will be in terms of resolution. Diffraction is relative to the format size, so it is no worse on a small sensor than on a large, just open up the aperture. Concerning depth of field vs diffraction it's a zero-sum game.
My guess would be that the limit will be in the optics and the required precision for those to get such high resolving power. Lens mount could have precision issues too, causing widely varying performance. It seems to me that the medium format has a better tradeoff in terms of format size to be able to squeeze out resolving power from optics. That is at least what I see currently, tech camera lenses resolve more pixels inside their image circle than the best tilt-shift lenses for DSLRs.
I've thought about the reason why that can be, I think it's not so much about that Schneider/Rodenstock would be better at doing optics than for example Canon, maybe it's even the opposite, but Canon need to make different design tradeoffs including stronger retrofocus (there's a mirror there, but even if mirrorless color cast which we accept in tech cams would not be acceptable in a 135 camera so retrofocus it is) and support larger aperture. A tech cam lens designer can give-a-damn about large aperture and care less about color cast and design just for high resolving power for a specific working aperture.
We'll see those high resolution sensors in cameras quite soon I think, and in the longer term it is best to outresolve both lens and diffraction with the sensor so we maximize what the system can get, and we don't get issues with aliasing and moiré.
From a tech camera perspective though I'm a bit worried that those sensor designs actually worsens color cast which makes those successful highly symmetric wide angle lens designs impossible. The trend in medium format has also been towards smaller pixels and worse color cast, but I'm glad than the new sensor in IQ260 actually take a step backwards from the IQ180 and works better with tech cam applications.
However, we see that when resolving power becomes really high the complexity of lens design increases, as we can see in Rodenstock wide angles. Part of their complexity is due to that they are retrofocus to reduce color cast with an IQ180 I guess, but a simple symmetric design only goes that far on the wide end. I don't particularly like that trend either, at some point resolving power is enough for practical image making and I would prefer focus on other aspects of the camera system rather than pushing towards more and more complex optics which increase cost (and weight and reduce robustness) in a more permanent way than complex electronics do. The complex lens designs also often have worse bokeh due to their high degree of correction. I'd like to see the traditional large format lens designs as well represented in the Schneider Apo-Digitar series live on.
-
I think that MF and the traditional makers are actually loosing grip on this technology lead. See here
http://cinescopophilia.com/astro-design-debuting-an-8k-camera-head-and-1080p-lcos-evf-at-nab/
...........AH-4800 Image sensor
: 2.5inch 33million pixels single plate CMOS............
Welcome CMOS Medium format - with 60 images a second !
Greetings from Germany
Stefan
-
Dynamic range seems to be the main problem of those global shutter CMOSes, if we want high end image quality? I wonder how long it takes before it will be at the level we have come to expect in photography.
I guess we'll have MF CMOS without global shutter before we have MF CMOS with it though.
I would not call the video advances as "technological lead" as their sensors won't work well for photography. And development pace in photographic applications in the smaller formats has been pretty rapid too. Hopefully it will spill over to MF rather soon, it shall be most interesting to see what changes a CMOS sensor will bring to how an MFDB works.
-
Hi Torger
well, as there is superexposure resolution enhancement, multishot noise reduction and HDR dynamic range extension, it does not have to happen in further
photon collection, it will happen in shere computing power in raw post. And thats actually available: NOW.
And I think this will be even higher quality than any silicon made straight approach.
And this also prevents any further problems with optical feed by diffraction and other little nuisances.
Greetings from Germany
Stefan
-
And doing some math - this CMOS does have 8.2 micron pixel size. 63,5mm:7680 pixel
for comparison an IQ180 has 5,2 microns and 27,4 square-micron size
8,2 microns will result in 67,24 square -microns, roughly 245% more caught photons. (2,5"= 63,5mm:7680 pixel)
I doubt somebody will be dissappointed by the dynamic range of that sensor (out of the box).
Regards
Stefan
-
Hi Stefan
Where did you find the sensor dimensions? It says 2.5" but I dont know what it means in mm, after all, 4/3" seems to 18x12 mm. have you found any other number?
Best regards
Erik
And doing some math - this CMOS does have 8.2 micron pixel size. 63,5mm:7680 pixel
for comparison an IQ180 has 5,2 microns and 27,4 square-micron size
8,2 microns will result in 67,24 square -microns, roughly 245% more caught photons. (2,5"= 63,5mm:7680 pixel)
I doubt somebody will be dissappointed by the dynamic range of that sensor (out of the box).
Regards
Stefan
-
This sensor would make a good match for the only 35mm lens capable of providing the required resolving power - the newly announced Zeiss 55mm f/1.4...
-
Hi Erik
1"= 25,4mm, the data is given in Pixel(the longer side counts for the size), there may be a slight variation because the 2,5" may apply to the full sensor size
(as the Dicomed Bigshot which was 60x60mm but the active pixels were only using 55x55mm)
AH-4800 Image sensor
: 2.5inch 33million pixels single plate CMOS
(SENSOR Developed by NHK Engineering System,Inc.)
[Active resolution] : 7680×4320
[Lens mount] : PL mount
[Output] : 12-channel parallel optical-fiber
[Dimensions] : 125(W) x 125(H) x 150(D)mm
[Weight] : 2kg
http://www.astrodesign.co.jp/english/news/news-20130405-1790.html
-
Hi Color Pikker
sorry that wouldn´t do as the Zeiss is for 35mm and this chip has 63,5x35,4mm
This needs MF lenses. It will also be totally uncritical in sharpness as any decent MF lens will feed the 8,2 Micron easily (that´s about the Pixelsize of the old 20 Mpix backs)
Regards
Stefan
-
http://www.reduser.net/forum/showthread.php?97745-8K-Camera-weights-just-2kg
-
Hi Color Pikker
sorry that wouldn´t do as the Zeiss is for 35mm and this chip has 63,5x35,4mm
This needs MF lenses. It will also be totally uncritical in sharpness as any decent MF lens will feed the 8,2 Micron easily (that´s about the Pixelsize of the old 20 Mpix backs)
Regards
Stefan
I was talking about the 70mp cmosis sensor.
-
I was talking about the 70mp cmosis sensor.
sorry - yep - that will be a tough one.
Regards
Stefan
-
2.5" sensor usually means something like 1/2.5" sensor, that is rather small... sensor makers have a funny way of relating inches to sensor size which I have never truly figured out.
For example, Nikon 1 has a "1 inch sensor", and the actual size is 13.2x8.8mm, I wonder where 25.4mm fits in there?
When 2.5" sensor is talked about in compact cameras the size is 5.76x4.29mm, but I doubt it would be that tiny. But I also find it highly unlikely that it would be 63.5 in diagonal. Is there an official statement of the actual exact measurements of the sensor?
-
Hi Torger
I´m pretty sure this is what it says, the camera is much too big for a small 1/2,5" and it sports full PL mount which also doesn´t make sense for such small stuff.
The datasheet is definite - it says 2,5".
http://www.astrodesign.co.jp/english/astro/News/NAB_Debut/ASTRONEWSE_AH4800_20130405.pdf
-
It sure looks like video is were all the game is, and probably because of the higher budgets.
-
Hi Torger
I´m pretty sure this is what it says, the camera is much too big for a small 1/2,5" and it sports full PL mount which also doesn´t make sense for such small stuff.
The datasheet is definite - it says 2,5".
http://www.astrodesign.co.jp/english/astro/News/NAB_Debut/ASTRONEWSE_AH4800_20130405.pdf
On the other hand, why would the camera have a CP.2 lens on it? Does it actually cover the supposed 63x35mm format, twice as wide and 50% taller than the intended image circle?
Edit:
2.5" sensor usually means something like 1/2.5" sensor, that is rather small... sensor makers have a funny way of relating inches to sensor size which I have never truly figured out.
For example, Nikon 1 has a "1 inch sensor", and the actual size is 13.2x8.8mm, I wonder where 25.4mm fits in there?
Ahh! Bingo! It is related to the TV tube inch measurement, not in fractional inches, but indeed in whole inches.
1" is technically defined as 12.8 x 9.6mm, which means that multiplying the size by 2.5x results in a size of 32x24mm, making it smaller than 35mm format in the horizontal dimension, and the final image is likely going to be smaller still, because it will use a recording ratio typical for video rather than 4:3.
All in all, the sensor size physical recorded image will be somewhat bigger than the standard S35 of cinema cameras, and even the Epic sensor which is 30mm wide. This means that not all PL lenses will work, and why the CP.2 is demoed, full-frame lenses are likely to have better coverage for the whole sensor.
-
P. S. Sorry, just noticed the post above mine. And 32mm wide would be natural, just fitting into the 33x26mm maximum field size limit of modern sensor fab. equipment to avoid on-wafer stitching.
AH-4800 Image sensor
: 2.5inch 33million pixels single plate CMOS
...
http://www.astrodesign.co.jp/english/news/news-20130405-1790.html
My guess is that this "2.5inch" is in the usual bizarre convention of video sensors, where the actual diagonal is about 2/3 of the stated number, like 4/3" meaning 22.5mm diagonal. So scaling up: 22.5mm * (2.5" / 4/3") = 42.2mm, fairly close to the 43.2mm diagonal of 36x24mm "35mm format". That seems a good size choice for a camera of such high resolution, which is likely to want access to higher resolution still-camera lenses.
-
In case anyone is wondering, for curiosity's sake, the inch size was developed way back before the 90's when video cameras used CRT tubes to capture the image instead of CCDs. Tubes were typically measured in terms of diameter in fractional inches, but the useable area was only a central portion of the tube, so for any given size, a smaller center crop was used.
Since by the late 80's a whole industry was already built around the size convention of tubes, it made sense to keep it when talking about digital sensors, so a 2/3" lens could stay 2/3" and not some new made up figure.
I do wonder why NHK used such an outdated sizing method for a large sensor camera.
-
probably you are right, these measurements are really odd and misleading.
But then: a 33 Mpix chip with that data is nowhere mentioned. Out of the blue ?
Hmmmm. ???
-
Here is something about those sensor sizes:
http://www.dpreview.com/glossary/camera-system/sensor-sizes
-
Thanks for the link Hening !
1/2.5" 4:3 10.160 7.182 5.760 4.290
But this would mean this camera has only a 5,7 x 4,2 mm large chip ?
I don´t think so !
-
1/2.5" 4:3 10.160 7.182 5.760 4.290
But this would mean this camera has only a 5,7 x 4,2 mm large chip ?
This sensor is described at 2.5", not 1/2.5" --- a size not on that list. By the way, some Sony data sheets for its "APS-C" sized sensors describe them with wording like "APS Size Diagonal 28.4 mm (Type 1.8)",
where "Type 1.8" s a synonym for the weird '1.8 inch' specification:
http://www.sony.net/Products/SC-HP/cx_news/vol28/pdf/icx413np.pdf
The size in fake inches or "Type" only refers to the diagonal measurement. For example, there are 2/3" sensors in different shapes, some 4:3 for stills, others 16:9 for HD video, with the common dimension being the diagonal of about 11mm.
-
Hi,
I think it was shown with a Compact Prime? That would be something 43 mm diagonal at maximum. 4/3 has a diagonal of 21.6 mm, 2.5 / 1.33 * 21.6 -> 40.5, slightly sub 135 full frame.
My guess is that the sensor size is around 35.4 x 19 mm, assuming 1080P side ratios.
Best regards
Erik
Thanks for the link Hening !
1/2.5" 4:3 10.160 7.182 5.760 4.290
But this would mean this camera has only a 5,7 x 4,2 mm large chip ?
I don´t think so !
-
Didn't we already figure the size thing out? It's a roughly 32x24mm sensor from a 33x26mm wafer process.
2.5" is 63.5mm, or 42.3mm diagonal according to tube sizing, which would result either in a roughly 36x20mm 16:9 sensor or 33x24.75mm 4:3 sensor, the latter of which sounds more practical from a manufacture point of view.
4:3 is not impractical for video, because as we can see in the Alexa, the extra sensor space can be used to emulate the full area of 35mm film stock, which allows the use of anamorphic lenses without any crop. With Cooke having recently announced a line of anamorphic lenses, we may just end up seeing more video cameras that have 4:3 sensors.