Tomasz Hemeprek from Physics Institute of the University of Bonn, Germany posted a long and interesting interview with Bart Dierickx taken at Workshop on CMOS Active Pixel Sensors for Particle Tracking held in Sept. 2014 in Bonn. Bart tells about how he founded Fillfactory and then Caeleste, different projects, few bits of image sensor history and also his views on image sensor future, and more.
Few quotes:
Q: How many different pixel design have you done in life?
A: I have designed myself at least 100 different imagers and as a team leader or CTO certainly more than 100. Many of those chips contained multiple variants of a baseline pixel. A large part must have been classic 3T and 4T pixel and an even larger amount of pixels which are “special”, with logic inside, or analog processing, or hybrid, or having some special feature like time gating or having certain functionality or redundancy. Every project is different. So if you ask, it will be around 1000.
Q: How the image sensors will look like in 10 years. Do you look at the mobile market what they are doing?
A: The pixel size race will stop, but what is not yet really there is a global shutter in a small pixel. That race is now going on. Today a 1µm pixel is a rolling shutter pixel: if you move your camera you get the jello effect. Global shutter pixels are the next big thing. They are racing again, maybe down to 1.5 µm? But also that race might stop or reverse, as happened with telephones: about 5 years ago mobile phones became smaller and smaller and everybody said that they would become so small that you would lose them. Suddenly they became bigger again due to the smartphones. All those smartphones have the size to fit in your hand and the hand does not scale.
Smaller and smaller phones come back: they are called wearable devices. So there is still a strong motion to make things smaller and smaller.
ReplyDeleteIt's not the hand that doesn't scale. Visible photons don't scale.
ReplyDeleteI fully agree. The reason because the pixel size has reach an almost limit around 1um pixel pitch is because smaller pixels will not sense the visible light. May be that this srinking may continue until half one micron, but not smaller.
Deleteon the other hand, the sensitivity is proportional to the pixel area, then using smaller and smaller pixels imply the need of more and more light to obtain images with the same quality per pixel. This is probably the reason because the srinking tendency has been inverted in some new products, like for example the iPhone 6 camera with pixels larger than the iPhone 5 or 5s.
Isn't it due to the diffraction limit that you cannot use smaller pixels? You could still use, but probably need to do some computation to retreive high resolution images
DeleteHow to solve sensitivity and SNR issues ?
DeleteThe wavelength of visible light extends up to 700nm, so a pixel size smaller than 2x / 3x this wavelength is already amazing. Few years ago 2um was supposed to be a physical limit.
ReplyDeleteWhy does/would wavelength cause any issues for a pixel pitch?
ReplyDeleteArnaud, the real pitch is 2X since we use CFA on the sensor. So 700nm X2 = 1.4um. That is very much people stop on.
ReplyDelete-yang ni
This isn't really true. Each pixel, independently of the CFA, will be affected by the diffraction phenomenon and if the hole size, which is smaller than the pixel pitch, is in the order of the photon wavelength or even smaller, a significant amount of incident light power will be reflected. And this is not a desirable effect. The sensitivity problem cannot be solved even with processing. As an example, note the size of the photoreceptors in the human retine. About 1um pitch the smallest.
Delete