DALSA published white paper "
The Evolution of CMOS Imaging Technology" talking about its 5T global shutter pixel introduced in 1999. Other global shutter pixel architectures mentioned, dividing them onto "voltage domain" and "charge domain" approaches, whereas "voltage domain" is said to be worse in almost everything, except the "shutter leakage".
This is a funny post. Wanna see a charge domain global shutter running at kframes per second..
ReplyDeletesure... no problem!
Deletejust make the pixel array 80x80 and away you go!
Here is an example of 1024x1024, 1000fps "charge domain" global shutter sensor:
DeleteTwo-Stage Charge Transfer Pixel Using Pinned Diodes for Low-Noise Global Shutter Imaging
http://www.imagesensors.org/Past%20Workshops/2009%20Workshop/2009%20Papers/078_paper_yasutomi_shizuoka_univ_gs.pdf
Vladimir, please look at the linearity numbers of the paper you linked. At high speed you cannot transfer all the charges from the pinned storage to the FD. I think the main limitation of charge domain is indeed the "rolling shutter" type of readout i.e. the charge transfer from storage to FD is sequential...
DeleteTrue, sometimes the rolling charge transfer can be a bottleneck. To overcome it, one can pipeline the readout: while row N transfers the charge, row N-1 is read out, then row's N+1 reset level is sampled. This way the transfer time can be extended for the price of the column circuit complexity.
DeleteGetting back to the Shizuoka Univ. paper, the large signal non-linearity is not caused by the limited transfer time. Normally, incomplete charge transfer manifests itself in low-light small signal artifacts. The low light is quite linear in the paper. I can only guess about the large signal non-linearity reasons, but I do not believe it's because of short charge transfer time.
I found such "technology white paper" with strong commercial motive is quite silly.
ReplyDeleteAnd we love rolling shutter sensors that do not rely on any form of analog pixel memory like the global shutter ones do. Widespread of such sensors in cellphones stimulated software development to handle distortions - anyway such sensors provide the same amount of information from the scene, so it is up to processing to use it.
ReplyDeleteHere are the presentation slides (no text, just pictures) demonstrating the shutter artifacts on some global shutter sensors - presentation I made last fall at teh University of Utah:
http://wiki.elphel.com/index.php?title=File:UU_CompEngineerJrSeminar2011.odp (first link on the page).
Andrey Filippov
Elphel
Andrey,
Deletedo you happen to have some slides illustrating the effectiveness of software distortion corrections?
say... a "before" with the rolling shutter slants and "after" without them?
thanks
I did not start working on that part of software yet (just judging from the works by others available on th Net), so our focus was on static calibration of the camera system (aberrations + distortions), and the camera development itself. We do have the high bandwidth IMU logging (300Hz sensors, 12 bit@2500 samples/sec) and that processing will be the next step. Have to do that before the SIGGRAPH this August where we'll be showing the whole process.
DeleteWhat is the file format Andrey ?
ReplyDeletethansk !
There is a pdf version too - you may search https://www.google.com/search?q=Image_sensors_and_shutter_related_artifacts
DeleteIt is Open Office Presentation (http://www.openoffice.org/product/impress.html). This one does not use any animations, so PDF should be OK too - it is here - http://wiki.elphel.com/index.php?title=File:Image_sensors_and_shutter_related_artifacts.pdf
ReplyDeleteSorry to say that this presentation brings no useful information at all! Pitty
ReplyDeleteI'm sorry if you were not able to get anything of it. As I wrote - there is no text there, just images, all the rest was verbal.
Delete