There is a lot of buzz about the improved camera that will be available in the iPhone 5S released this month. One of the surprising announcements from Apple was the increase in pixel size from 1.4um to 1.5um and the statement “bigger pixels = better picture.” This got us thinking about a topic we previously discussed in detail about why cameras for consumer electronic products are not used for machine vision or global security applications.
Of course consumer cameras are are not designed to meet the 24/7 demands of industrial or surveillance systems, but what about the high resolution image sensors with very small pixels (less than 3 um) used inside? Is the picture from sensors with 1.5um pixels good enough outside of the consumer market?
The larger pixel image sensors (greater than 5.5 um) can allow for the best accuracy (i.e. Full Well Capacity and Read Noise), but they also result in the highest costs due to large sensor sizes (silicon real estate consumed) and additionally expensive optics. Larger pixels are still used in the scientific market, but the trend in other markets has been towards much smaller pixel sizes. Smaller pixel image sensors reduce the cost of the camera because of the camera size, or provide more pixels inside the same camera and optics leading to higher resolution, but is the resulting image really good enough?
For a detailed discussion on the pros and cons of smaller pixels in machine vision, click here.
As for global security applications, there is a desire to go to smaller image sensor sizes to have smaller optics to meet reduced size, weight, power and cost (SWaP-C) requirements. So why aren’t image sensors with pixels below 4 um being used?
There has been huge funding in smaller pixel development driven by the cell phone camera market to shrink the pixel size and to significantly reduce the noise level. For instance high quality CCD sensors from 10 years ago had a read noise of 20 electrons. Today CMOS sensors with rolling shutter can have a read noise of 2 electrons. Because the noise level dropped 10 times, only 1/10th the photons are necessary. The pixels can be shrunk dramatically and still have acceptable low light performance.
Smaller pixels (less than 3 um) now mean there can be full-HD on 1/3” and 1/4” CMOS image sensors. These have good low light performance, but the available optics for global security and industrial applications are not matched in quality yet. Also, 1/3” CMOS sensors and smaller utilize a rolling shutter as compared to global shutter available in CCD and ½” and greater CMOS image sensors. Currently rolling shutter is not well received in the industrial and military applications as a blurring effect occurs with movement. Global shutter CMOS is more complex and will take further development to shrink for smaller pixels. Further testing may show that rolling shutter is acceptable in outdoor, moving applications with higher frame rates, such as 100 fps, so that the blurring effect is not evident.
One of the reasons the images from cell phone cameras with only 1.4um or 1.5um pixels are so good are the optics that are used. These cameras use plastic optics rather than glass. The plastic is flexible and can be made into any type of surface desired where glass only comes in a spherical shape and those available for image sensor sizes smaller than ½” are not yet good enough for demanding imaging requirements. Plastic lenses are currently not practical in lower volumes (outside of the consumer market). And, while plastic lens help meet the lower weight requirements; there could be problems with temperature extremes and anti-reflective coatings to limit their use in the military market.
In our first discussion of small pixels we detailed how sensors typically use micro-lenses to focus the light onto the active part of the pixel (the part without transistors). With a 2 to 3 um pixel, only about 1 um is sensitive. This is also referred to as the “pixel straw”. The light coming in can go into the wrong pixel or into the wrong photo diode known as optical crosstalk and electrical cross talk. This will result in worse modulation transfer function, or MTF, and therefore a less sharp image. Crosstalk also ruins color reproduction as say some red light goes into a green pixel, etc.
Back-side illumination (BSI) technology offers significant improvements over micro-lenses to eliminate the electrical and optical crosstalk. These were previously only available in the very large consumer market or the very expensive scientific market, but are now becoming available for the industrial and security markets.
So to summarize, small pixels (2-3 um) on CMOS image sensors offer the following pros and cons:
- Lower costs and weight through smaller sensors, optics, and cameras
- Increased resolution with same sensor size and optics
- Sensitivity in dim light is acceptable
- Color reproduction can be good enough
- Give up global shutter (get rolling shutter)
- Slower frame rates
- More noise
- Lower full well capacity
- Lower MTF
- Optical/Electrical crosstalk
- Lower quality available optics
There is a trend towards using smaller pixels for machine vision and global security applications leveraging technology developed for the high volume consumer camera applications. Our conclusion based on a thorough analysis is that with pixels less than 3 um, too much functionality and performance is sacrificed for use outside of the consumer market (FOR NOW).
It will take some time before the extremely small pixels have high enough performance to make it into applications such as FPD inspection, electronics metrology, border security, homeland security, situational awareness and many others.