camera products and the future

1
camera products and the future
2
memory and processor
video, audio, temperature, chemical, tactile
ADC
communication
DAC actuators
electronic system of the future
3
why are products that can capture images
important?
this processor has a lot of logic allocated to
scene analysis
thats why people like to look at pictures
4
cell phone camera image 640 x 480
needs more pixels and more dynamic range
5
silicon sensitivity
human vision
6
pixel spacing limit 0.61 l ( f number )
about 3 um for f/8 in green light
7
volume / weight s3
CCD read energy s3
CMOS read energy s2
imager area s2
Imager cost s2
pixel count s2
f number L / D s0
pixel pitch s0
pixel area s0
pixel SNR s0
ISO s0
depth of field s-1
8
FOVEON X3 IMAGE SENSOR SUMMARY

• measures all colors in each pixel location, CCD
  sensors measure only one color
• uses a fundamental property of silicon for the
  color filtering, CCD sensors use plastic filters
  applied as a post process step

9
vertical color filter
lateral color filter
spot size
3 mm
Foveon X3
everybody else
Foveon X3 vertical color filter image sensor
captures 3X information from a
given spot
10
sun
pixel size
wavelength
lithography
transistor size
time
11
(No Transcript)
12
no noise reduction
ISO 1600
with noise reduction
13
James Janesick Scientific Charge Coupled Devices
SPIE press 2001
14
what camera consumers actually need

• 97 pictures 5x7 prints gt 2 M pixels
• consumers know they will get better pictures with
  more light
• good 99.95 digital camera

15
14 bit ADC big transistors
CCD spec imager specialized transistors
SOC camera chip
Ghz DSP tiny transistors
16
multi chip image module sensor
lens

imager
FRP
DSP
single element plastic lens in camera module
cannot correct for chromatic aberration optically
17
edge of field image before correction
after correction of lateral chromatic aberration
correction of lateral chromatic aberration is
easier if all color information from each pixel
site is available
18
the superficial plastic color filter is the
manufacturing Achilles heal for cheap CMOS image
sensors

• limits fab sources
• messy process with nasty learning curve
• color control difficult to manage
• color filters fragile

19
conclusions

• figure out what quality/cost consumers really
  need

20
conclusions

• figure out what quality/cost consumers really
  need
• determine what the real performance problem is

21
conclusions

• figure out what quality/cost consumers really
  need
• determine what the real performance problem is
• figure out how to make it cheaper

22
conclusions

• figure out what quality/cost consumers really
  need
• determine what the real performance problem is
• figure out how to make it cheaper
• dont listen to anybody, think instead