Thursday, October 07, 2010

Sony Introduces 1.12um BSI Pixels, New Sensors, Modules

Sony announced the commercialization of two new "Exmor R" BSI sensors: 1.12um-based 16.4MP IMX081PQ and 1.4um-based 8.1MP IMX105PQ, both specifically designed for mobile phones.

The most interesting one, IMX081PQ, is presented as the world's first type 1/2.8-inch BSI sensor which realizes 16.41MP effective resolution, and adopts unit pixel size of 1.12μm. Sony says to implement "a unique formation of photo diodes optimally designed for fine pixel structure to realize a CMOS image sensor with high resolution, high sensitivity and low noise". The picture below shows the resolution improvement:

The second sensor, IMX105PQ, is a type 1/3.2-inch BSI sensor which realizes 8.13MP effective resolution and adopts a unit pixel size of 1.4μm. The 20Lux illumination comparison picture is below (using the newly announced camera module):

Furthermore, Sony will commercialize IU081F and IU105F2 AF camera-modules which include the two new sensors. IU081F is said to be the industry's smallest and thinnest AF module (W10.5 X D10.5 X H7.9mm) and is equipped with the 16.41MP sensor. IU105F2 adopts the 8.13MP sensor, and is said to belong in the industry's smallest and thinnest size class (W8.5 X D8.5 X H5.67mm):

The table gives a few spec details foe the new sensors and modules:

Model NameIMX081PQIMX105PQ
Number of effective pixels4672(H)×3512(V)
16.41M pixels
8.13M pixels
Image sizeDiagonal 6.5mm (Type1/2.8)
square pixel
Diagonal 5.76mm (Type 1/3.2)
square pixel
Unit cell size1.12μm□1.4μm□
1/2 sub sampling30fps30fps
1/8 sub sampling120fps120fps
HD mode1080-30P / 720-60P1080-30P / 720-30P
Sensitivity205 digit (10Bit)310 digit (10Bit)
Saturation signal820 digit (10Bit)1023 digit (10Bit)
Power supplyAnalog2.7 +0.2/-0.1V2.7 +0.2/-0.1V
Digital1.2 ±0.1V1.2 ±0.1V
Interface1.8 ±0.1V1.8 ±0.1V
Major function3-wire serial communication, I2C, serial data output, supports flexible input clock
OutputMIPI 4,2,1 LaneMIPI 2,1 Lane

Camera modules data:

Model NameIU081F (16.4MP)IU105F2 (8.1MP)
Module size*210.5(W)×10.5(D)×7.9(H)mm8.5(W)×8.5(D)× 5.67(H)mm
AF actuatorVoice coil motor
Lens construction4 groups 4 elements (Plastic)
Focal length (35mm conversion)28mm
Camera outputMIPI (4 Lane)MIPI (2 Lane)

The samples availability and prices are given below:

Model NameShipment date (Plan)Sample price
Type 1/2.8 16.41 effective megapixels
back-illuminated CMOS image sensor
January, 20112,500 JPY
Type 1/3.2 8.13 effective megapixels
back-illuminated CMOS image sensor
April, 20111,500 JPY
Type 1/2.8 16.41 effective megapixels
Lens module "IU081F"
March, 201112,000 JPY
Type 1/3.2 8.13 effective megapixels
Lens module "IU105F2"
April, 20118,000 JPY

At the end of 2010, Sony plans to start the mass production of BSI sensors, including those for mobile phones announced today, at Sony Semiconductor Kyushu Corporation's Kumamoto Technology Center, on 300mm wafer lines. Sony already announced on Sept. 1, 2010 the investment of approximately 40 billion yen in Kumamoto Technology Center to increase production capacity for CMOS image sensors.

Sony Semiconductor Kyushu Corporation's Kumamoto Technology Center


  1. Excellent work and they defend well their last bastion !

  2. I note SONY uses F#2.4 for BSI and F#2.8 for FSI...not so good for sensitivity comparison.

  3. I believe they are presenting it as a part of the whole system improvement, including the sensor and the lens.

  4. does anyone know if this 8mp is the one that's supposed to be in the iphone 5? it would make sense because of the improved lens. TIA

  5. I agree with the first comparison comment. Using lenses with different f-numbers obscures how much improvement there is in moving from FSI to BSI, which is a key factor in determining the price premium for the BSI product.

  6. The BSI sensor allows for a lower F#, therefore BSI enables a better system. These issues are tied together. It doesn't obscure how much improvement there is, it demonstrates how much improvement there is. Arguably the most important aspect of BSI is the lower F#. Going forward, sensor performance is about the system that it enables. The days of simple comparison are gone.

  7. Made the comment about F#. "Days of simple comparison are gone" ? Days of simple comparison are gone from the beginning, sorry, this is not related to BSI. Of course we're talking about a camera system at the end. However, if you are an engineer in the CMOS sensors field, you know we cannot compare two sensors using two different F#, that's it, considering sensitivity especially.
    About BSI -as it is today-, we could also talk about white balance issues (part to part variations related to silicon thickness) and also color shading related to crosstalks. Go to google and look for iphone4 IQ issues then you'll find somes nice demonstrations made by users (using a camera here, not only a sensor...). Having said that, I still believe BSI architecture will improve in next generations (why not in iphone5) and is by construction an extremely goog approach.
    At the end this comparison is marketing oriented, not technology oriented, but why not. Since Steve Jobs has communicated around BSI on the iphone4, everybody in the field can understand it. Again, this is not a correct comparison from a pure technical point of view.

  8. Anybody can comment on the price? Only $18 for a 8M sensor. I thought Sony's BSI was expensive. And this is only sample price. If someone order 1 million units, the price could be much lower.

  9. Sony gives sensitivity and saturation with unit of "digit"! It is even hard to compare across its own products. Previously,some parts use "mv", and some parts use "e-", which were making more sense. Now they use "digit", gosh. Does anybody know how this version of BSI compare to its own 1.65um BSI shown in 2010 JSSCC in terms of sensitivity? Is it scaled by pixel size?

  10. Is this the first ever 1.12um pixel product ready for sampling? I've not heard anybody else sampling 1.1um pixel products. I did see something on 1.1um pixel from OVT(BSI) and Aptina but I think they were just some images for pixel research chips not products...We should celebrate the fact that cmos imaging is finally getting to the next pixel node...

  11. The rumor is that Sony has been demoing its 1.1um pixel to selected customers since February:

    Yes, I agree, it seems that Sony is the first to sample this pixel generation.

  12. @ The BSI sensor allows for a lower F#, therefore BSI enables a better system.

    With all due respect, your comment is incorrect. The key specification you probably mean is the z-height, not the f-number. If you go learn what z-height and f-number are and how BSI influences z-height you will understand my comment on the images.

    Even if you say, "Yeah, yeah, I meant z-height" I will still disagree with the rest of your comment that I didn't quote. It's not the case that z-height dictates performance and performance is all that matters. For the target audience, the main issue will be value - whether the improved performance relative to alternative chips is worth the higher price Sony will likely set.

  13. Comparing two images from from F2.4 and F2.8 as shown here is truly not fair. Larger aperture lens(with same focal length) increases sensitivity while decreasing depth of field.
    It is surely a merit to allow bigger aperture with same height in that it increases flexibility of camera system design. However, same or equivalent lenses should be used for sensitivity comparison.

    It is not comfortable to see such deceiving material from a leading company like Sony.

  14. No, I did not mean z-height. I meant F/#. And indeed it is F/# that matters. Regardless of the z-height, F/# tells you the photons per unit area. And it would be unfair to compare a BSI sensor to FSI at the same F/#. What if they were the same? Story not told. At these focal lengths, depth of field is already large and auto-focus is already used. Everyone wants a lower F/# at these dimensions if they can get it. BSI gets it. BSI is being used for the small pixel generation. It is all about F/#. CDM will understand next year.

  15. I agree with anon on F. Comparing an FSI sensor to BSI across F is the most useful. You don't want to compare at the same (high) F and then ask what is the value of the BSI sensor. FSI with microlenses can be equivalent or better than some BSI implementations at high F. The real reason for moving to BSI is to maintain or decrease F as pixels scale down. Sony is just demonstrating exactly why we use BSI by not using the same F. Asking to compare at the same F is interesting, but doesn't tell you the value of BSI. It is one data point, and not the most useful one. BSI allows more lens options.

  16. One can make a telecentric lens with low F/# and small CRA, so it fits both FSI and BSI. It's not like only BSI allows low F/#. The advantage of BSI is that it allows a larger CRA, so the bright lens can be made smaller and/or cheaper.

    So, if we'd like to include the lens in this comparison, one needs to include some size and price considerations. And lens distortions as well, as it's also a part of the optimization.

    So, my proposal is to keep the spec and make comparison at the same F/#, while keeping in mind that the larger CRA in BSI sensor enables either cheaper or brighter lens, depending on the system considerations.

  17. Think about this carefully. We are talking about F/#, not CRA. CRA is another advantage of BSI. Only consider on-axis pixels. Look at what happens to the FSI pixel when you try to use an F/# that is smaller than the ratio of pixel pitch to effective stack height. A significant advantage of BSI is that you can use a lower F/#. It is simple to see where the FSI pixel falls apart on this topic. An FSI pixel that uses waveguide and TIR, has a separate analysis. A telecentric lens is not helping on-axis pixels use a lower F/#.

  18. Seems like Sony just presented information for two different camera modules. Doesn't sound like they were trying to compare BSI v. FSI for that part of their announcement. If it was an image sensor announcement, then they should roasted. In this case they get a pass from me.

    12,000 JPY for a sample of the module! I wonder how they arrive at that price? I mean, it could be free, or it could be 120,000 JPY and it probably would not affect their business as long as they target top tier handset makers.

  19. One of these (apparently the 8MP) is supposed to land in the 5G iPhone. OVT wants to hold on. A faster lens makes all the difference.

  20. @ "Think about this carefully... A significant advantage of BSI is that you can use a lower F/#.

    Yes, I see what you mean and agree with you.

  21. Looks like Sony has surpase Omnivsion's technology leadership. Im wondering if when Omnivsion will have 1.1um BSI.

  22. The observation about stack height relative to pixel pitch is a fair point.

    If f/2.8 is the lower limit for the comparison FIS sensor, one extra image - FSI + f/2.4 - would illustrate the advantage of the new BSI product quite emphatically. To me that seems like an easy and useful way to get the marketing message across. Of course, it may not be easy to cut off a module top and put in your own tiny custom lens.


  23. That one extra image you suggest would be difficult to for people to grasp and would not be a good marketing message because of the explanation required. It is clear for everyone to see that they get a brighter image with BSI. If you take that one extra image at the same lower F/# where FSI falls apart, you have to start describing what it is about the image that fell apart. color mixing, sharpness, off-axis shifts. Then throw in the image processing which has to tune up to a new lens that is not going to work. Sony did not need to teach where FSI falls apart. They just show where BSI moves forward and they did a good job.

  24. Is there a place where we can purchase the Sony modules?


All comments are moderated to avoid spam and personal attacks.