Wednesday, April 28, 2010

Nokia N8 Features "The Biggest Sensor In A Mobile Device"

Nokia Conversations published a conversation with Damian Dinning, Nokia’s main man for imaging describing the camera portion of the newly announced 12MP N8 camera phone. The official HW spec for the phone does not tell the sensor size, while Damian says it's "the biggest sensor in a mobile device". One of the tweets seemingly belonging to Nokia insider reports that the sensor size is 1/2-inch. GSMArena claims the sensor size is 1/1.9-inch without disclosing its sources. These numbers translate into about 1.6-1.7um pixel size.

Nokia Conversations writes:

"The team behind the Nokia N8’s camera want to set new benchmarks. Not just for the quality of the still image camera, but the video recording as well. Across both uses, the team have been optimising each specific element of the device’s camera for as many use cases as possible. Cracking Improving the low light performance nut is one of the hardest challenges for mobile device cameras. Naturally, that’s where the camera team started. This time though, the effort was focussed not just on still images, but on video too.

In dramatically improving low light performance, the team were able to improve performance in bright conditions as well.
"

N8 records 720p/25fps HD video. The images it produces (both still and moving) are said to "sport a surprisingly high degree of detail, but without any much of the noise you see in images from some other devices". The sample pictures and video seem to confirm that.

The phone is expected to go on sale in Q3, 2010.

19 comments:

  1. Great image quality !
    But there is no low light photos :(

    ReplyDelete
  2. The pixel size is 1.75um and output resolution 4000x3000pixels --> optical format is 8.75mm and 1/1.83" in inches (there are naturally some extra border pixels but they are not calculated in this optical format).

    We are not saving in sensor silicon :-)

    Best regards,
    Eero S

    ReplyDelete
  3. Eero,

    Can you tell us which companies make these gorillas for you?

    YI

    ReplyDelete
  4. Good try:-)

    Naturally we only use world leading image sensor suppliers that deliver great results.....

    -E

    ReplyDelete
  5. Eero, what is the minimum illumination level for 720p at 25p ?? Thanks !

    ReplyDelete
  6. I will not give any more details yet since we have still time to optimize the performance but I guess everyone here can estimate what you get when you combine the best 1.75um pixel in the planet and F2.8 optics.

    Eero S

    ReplyDelete
  7. I would think a pixel "in the planet" cannot see much of anything.

    ReplyDelete
  8. Best camera ever built for mobiles...and we've made it!! ;)

    ReplyDelete
  9. Hi, if this sensor is from Sony, is it a CMOS Exmor R? = Back-light sensor?
    A best optimized 1.75um pixel could be a Back-light sensor...

    And this is very curios...
    Sony only have one (with two versions) Exmor R, The IMX50, a CMOS Exmor R 10MPx and 1/2.4" type.
    The pixel size of this sensor is 1.75um, the same Nokia N8 camera sensor pixel size.

    My conclusion is:
    This is a new Sony Exmor R sensor, the same physical pixel size (1.75um) than before Exmor R and probably with better design and technology... and performance... on a bigger chip sensor (1/1.8")

    Sony has a 1/1.8" Exmor CMOS sensor (non-backlight) with 6Mpx on Casio EX-F1 high speed camera, therefore Sony already has that type size sensor chip.

    Right...?

    ReplyDelete
  10. A new wave of mobile photo chips. Nice! I wonder whether it is the same type as the iPhone 4 (1,75 um pixel, backside-illuminated).

    ReplyDelete
  11. Hi, i am very anxious 2 buy this devise .. When it suppose to cm in India ? any body have idea?

    ReplyDelete
  12. good question to eero : has the n8 a back-light sensor ?

    ReplyDelete
  13. N8 is F/2.8, right? That means roughly 1.8um resolution limit. In the best of possible worlds, where the optical parts are always manufactured to the optimum design (for free, obviously), it still means at least four pixel blur for each point.

    In my professional opinion, I doubt that the cheap optics are good enough to actually use the 12M pixels to their fullest. 6M is enough when the optical designer did his/her job properly.

    ReplyDelete
  14. > "That means roughly 1.8um resolution limit."

    Do you consider MTF-50 to be the resolution limit? I calculate 45% MTF at 1.8um Nyquist for luminance (278 cycles/mm) for a diffraction-limited f/2.8 lens.

    Personally, I consider 10% MTF as the maximum "useful" resolution limit. A more conservative definition might use the chrominance Nyquist (the higher spatial frequencies of blue and red filtered pixels) instead of luminance (demosaiced) and/or the Dawes limit instead of 10% MTF, but I don't find any real life benefit in that level of sampling.

    You may not normally consider anything as low as 10% MTF to be usable, but with deconvolution, USM, and/or other sharpening methods, it can boosted significantly. It depends on usable SNR and sufficiently fast/complex ISP.

    > "I doubt that the cheap optics are good enough to actually use the 12M pixels to their fullest."

    I don't think using "pixels to their fullest" is the optimal circumstance when it comes to resolution. My position (speaking as a non-professional) is that the condition that should be striven for is using the lens to its fullest. Since it's impossible to improve system MTF by changing diffraction, we should instead improve system MTF through sensor MTF (and the knockon effects of OLPF MTF, if one were to be used).

    Another consideration is that lens aberrations may only change the slope of the MTF -- not the extinction -- so the spatial frequency of 10% MTF may remain the same. That is, contrast drops off more precipitously than a diffraction limited lens, but it can flatten out nicely and have a long tail before you hit the Rayleigh criterion. The additional resolution can be very useful in aberration correction (lateral CA, distortion, lens-PSF-based deconvolution) and other processing (leveling/rotation, changing rectilinear projection, etc.).

    I'm not saying that 1.8um is always preferred over 2.5um -- there can be plenty of reasons to choose 6 MP over 12 MP (read noise, cross talk, angular response, etc.), I'm just saying that I think there is definite resolution benefits to sub-2.5um pixels with D.L. f/2.8 lenses, and probably even with aberration-limited lenses as well.

    ReplyDelete
  15. This product is from Toshiba - not Sony.

    ReplyDelete
  16. Still almost 23 times smaller than a full frame 36x24mm sensor. Do you think it is useable for anything serious :-)

    ReplyDelete
  17. nokia sucks.i bought a n8 and felt i am the biggest fool coz with 2 months the 12 mp camera stoped workin. i took it to the nokia care but they could not fix the camera.i am still using the lame mobile phone.just a couple of days and i am going to throw it away.

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.