Friday, October 14, 2011

Chipworks Identifies Sony Sensor Inside iPhone 4S

Chipworks' teardown of iPhone 4S reveals that its primary 8MP sensor is made by Sony. Still, there is a chance that the sensor is dual sourced and Omnivision might appear in other iPhone samples or at later time. Omnivision shares dropped by 9.4% after Chipworks made its report available.

iPhone 4S primary camera module

Update: iSuppli reports that iPhone 4S has a five-lens camera module, a first in any smartphone analyzed by the IHS iSuppli Teardown Analysis service. Previously, the state-of-art has been a four-lens approach.

Update #2: Oppenheimer published report on likelihood to see Omnivision sensors in future iPhones.

Update #3: Chipworks published reverse engineering report of Sony sensor. The chip is identified as IMX-145, 8 MP, 1.4um pixel pitch, BSI CMOS image sensor.

23 comments:

  1. Too bad it had such impact on the OVT share price, but nevertheless, good work Chipworks! You put that baby to rest.

    ReplyDelete
  2. Eric.. this probably made your day.. congratulations

    ReplyDelete
  3. If it had been a Samsung sensor inside, or perhaps even Aptina, that would have made my day.

    I don't like seeing shareholders lose value, but it is good when stock price represent actual value rather than rumor-mill value.

    ReplyDelete
  4. I wish it was Aptina's sensor inside!

    ReplyDelete
  5. Does any one know who makes 8MP BSI (w/IR filter) sensors.

    thanks

    ReplyDelete
  6. I don't think any one in the CIS industry was shocked to see this news. It merely validates the rumors for the last 18 months.

    http://image-sensors-world.blogspot.com/2010/04/apple-to-use-sony-sensors-in-2011.html

    http://uk.ibtimes.com/articles/130324/20110404/omnivision-apple-sony-iphone-5-camera.htm

    ReplyDelete
  7. According to OV F1Q 2011 Earnings Call Transcript (from seekingalpha.com), it was very clear that the initial volume are not from OV

    __________________

    Ray Cisneros
    ......
    And so again, we are focused, our intent is to execute on our plan and that’s the new adjustment right now at the tail end of Q2.



    Betsy Van Hees – Wedbush Securities

    Thanks, Ray. But when did you see this occur and when did this problem start with BSI-2 that caused this delay in the ramp cycle?

    Ray Cisneros

    Right. I think the recent events

    ReplyDelete
  8. is this an AF module??

    ReplyDelete
  9. Who is in the secondary camera?

    ReplyDelete
  10. now is ov, future will be Samsung

    ReplyDelete
  11. Is it really true that iPhone4S 8MP camera is the first to adopt a 5 elements lens design? I thought current 8MP camera modules are limited to use 5 elements, because with 4 elements you would not be able to reach the level of image quality needed. Does the 8MP Samsung or HTC phones use only 4 elements?

    ReplyDelete
  12. iSuppli has teared down 8MP Samsung and 8MP HTC:

    http://www.isuppli.com/Teardowns/Pages/Samsung-Galaxy-S-II-GT-i9100-Mobile-Handset-Hardware-Analysis.aspx

    http://www.isuppli.com/Teardowns/Pages/HTC-Sensation-4G-Mobile-Handset-Hardware-Analysis.aspx

    So, when iSuppli tells that iPhone lens is the only one with 5 elements, it probably knows what it says. iPhone 4S features quite bright F2.4 lens - may be this is why more elements is needed.

    ReplyDelete
  13. vlad or anyone that can help on this, i'm trying to determine based on published information whether the sony or ovt sensor has "better" sensitivity. the problem i'm having is that ovt uses 848 mv/lux-sec for its sensitivity spec and sne uses 310 digit (10 bit) for its metric. is there a way to convert one to the other so that both can be compared using the same metric? any comments you could make on the real meaning and value of these metrics to any comparison of sensitivity would be greatly appreciated.

    thanks in advance

    ek

    ps: hopefully you guys might get a charge out of the kind of things the financial analyst community considers in analyzing potential revenues and profits from a sensor and how easily they can get thrown for a loop on something like this. kind of like the mistake having to do with comparing active array size of ovt part to effective megapixel resolution for sony part and saying that since apple advertised 3264 X 2448, only ovt part could be in new iphone. lol

    ReplyDelete
  14. I afraid there is no simple way to compare these numbers. As I wrote before, even if they are in the same units, the direct comparison is needs more data and quite long optimization of color conversion matrix for each sensor. More than that, even after these calculations the answer might be ambiguous, like one sensor is better in incandescent light, while the other is better in fluorescent.

    And all this is just numbers, without subjective evaluation. These numbers assume that all kinds of noise are equal, while in fact they are not. It's known that structured noise, like column- or row-wise one can be much more irritating that random Gaussian. Or salt-and-pepper noise is much more intimidating, but easily maskable by modern ISPs. What do you do with this?

    In short, there is no simple answer. Even if all data is readyly available, the correct comparison takes a lot of time and effort.

    ReplyDelete
  15. thank you very much vlad for preventing this dog from spending any more time barking up the wrong tree. all you had to do is mention the metrics assume that other variables (noises) are held constant (equal) and i understand how that devalues the metric, for example, by comparison to the "et ceteris paribus" assumption from the discipline of economics basically renders the information unrealistic and demotes the metric to the status of a tool at best. hopefully for consumer benefit, some day the industry will agree on metrics that will allow easy comparison by consumers of metrics having to do with sensitivity and other attributes that consumers feel are most important.

    thanks again

    ek

    ReplyDelete
  16. The best metric so far is Nokia SNR10 figure. It's still far form ideal, as:

    - it talks about camera modules together with optics. From one hand it's important as lens parameters change a lot in sensor behavior. From the other hand, it's not a pure sensor comparison.

    - it treats all the noises in the same manner. Some noises can be effectively reduced by the subsequent ISP, like salt and pepper, while others penetrate through ISP pipeline unsuppressed, like some kinds of structured noises. In fact, ISP should be a part of comparison, like lens - but then we are not comparing just sensors.

    - the illumination spectrum is considered to be daylight, while most low light scenes are either incandescent or fluorescent. This might affect the last 5% of the figure, but in case of close run one one might choose a wrong winner.

    There are other minor inconveniences with SNR10 metric, but it's still the best I know of.

    ReplyDelete
  17. thanks vlad. excellent info.

    ek

    ReplyDelete
  18. I've just re-read this and one thing was wrong:

    "This might affect the last 5% of the figure, but in case of close run one one might choose a wrong winner."

    In fact, the absolute SNR10 figure changes a lot with illumination type, much more than 5%. The sensors might have the same SNR10 at daylight, while in incandescent light one of them would be better. The 5% relates to the possible difference between these two daylight-equal sensors.

    ReplyDelete
  19. Anonymous said ...

    iphone4s is OV (Howe) camera, the people of Taiwan know that OV (Howe) took over
    iphone4s front and rear camera (VGA + 800 megapixel BSI - 2) in order !!!!!
    I am Chinese.
    http://news.chinatimes.com/tech/11050901/122011101100305.html
    ---------------------
    http://www.eepw.com.cn/article/124574_2.htm

    ReplyDelete
  20. @ "iphone4s is OV (Howe) camera, the people of Taiwan know that OV (Howe) took over
    iphone4s front and rear camera (VGA + 800 megapixel BSI - 2) in order !!!!!
    I am Chinese.
    http://news.chinatimes.com/tech/11050901/122011101100305.html
    "

    When will these "800 megapixel BSI - 2" products be coming to a store near me? :D

    ReplyDelete
  21. thanks for the clarification vlad.

    have a nice weekend.

    ek

    ReplyDelete
  22. @Anon 1:41

    I'm sure there are other modules with the OVT sensor inside of them, but it doesn't mean they are the only one.

    ReplyDelete

All comments are moderated to avoid spam.