Educational news: Teledyne DALSA blog publishes a nice TDI mode post with animated explanation. Adimec blog explains blooming and smear effects in CCDs.
Why previous and current frames? Why not simply from the current frame? The leakage photo carriers are confined within the channel stops and travel vertically above and below directions and appear as white columns. Am I missing something?
To excerpt a paragraph on interline transfer imagers from my section in Bruce Batchelor's machine vision book:
IT Limitations Like FT sensors, IT sensors suffer from vertical striping but the cause is slightly different. Because there is no direct illumination of the vertical transport register where the charge is shifted, it would seem that there should be no striping. Unfortunately, the close proximity of the photoelements to the transport registers allows some light to sneak under the opaque aluminum layer. The amount is generally small, but the shifting proceeds slowly so there is ample time for the stray light to produce a signal. In a typical IT sensor, the vertical shifting takes exactly one frame to be completed. This means that the integration time for the stray light is essentially equal to the integration time for the photodiodes acquiring the desired image. Thus, even if the leakage is only 1%, the signal produced in the transport registers by an area just at saturation will be larger than one LSB in an eight-bit system. Oversaturated points can produce much larger stray signals. Confusing matters is the problem that the stray light is acquired from the frame that will be produced after the frame currently in the transport registers. Thus, fast-moving bright objects can cause vertical striping away from the positions where they appear in the real image. This problem is exacerbated by lenses with their last element very close to the sensor because the proportion of light rays arriving at the sensor at low angles – exactly the rays that get under the aluminum - is increased.
I remember having read about frame interline transfer CCD area array. The first transfer is interline transfer followed by frame transfer. This appeared to be a good solution to striping at the cost of added complexity and silicon
This comment has been removed by the author.
ReplyDeleteEric, I have to correct you.
DeleteSmear can show up ABOVE as well as BELOW the spot. This is the case for frame-transfer and interline-transfer CCDs.
This comment has been removed by the author.
DeleteIndeed, I did learn something new. ABOVE is from the current frame and BELOW is from the previous frame (or vice-versa). Thanks Prof. AT.
DeleteYou learn fast, very fast ! Would love to have you in my group ;-)
DeleteWhy previous and current frames? Why not simply from the current frame? The leakage photo carriers are confined within the channel stops and travel vertically above and below directions and appear as white columns. Am I missing something?
DeleteHow about smear in full frame devices? Should be very bad due to low readout speeds
ReplyDeleteFull frame devices are typically used with a mechanical shutter so that there is no smear (there can still be blooming).
DeleteTo excerpt a paragraph on interline transfer imagers from my section in Bruce Batchelor's machine vision book:
ReplyDeleteIT Limitations
Like FT sensors, IT sensors suffer from vertical striping but the cause is slightly different. Because there is no direct illumination of the vertical transport register where the charge is shifted, it would seem that there should be no striping. Unfortunately, the close proximity of the photoelements to the transport registers allows some light to sneak under the opaque aluminum layer. The amount is generally small, but the shifting proceeds slowly so there is ample time for the stray light to produce a signal. In a typical IT sensor, the vertical shifting takes exactly one frame to be completed. This means that the integration time for the stray light is essentially equal to the integration time for the photodiodes acquiring the desired image. Thus, even if the leakage is only 1%, the signal produced in the transport registers by an area just at saturation will be larger than one LSB in an eight-bit system. Oversaturated points can produce much larger stray signals. Confusing matters is the problem that the stray light is acquired from the frame that will be produced after the frame currently in the transport registers. Thus, fast-moving bright objects can cause vertical striping away from the positions where they appear in the real image. This problem is exacerbated by lenses with their last element very close to the sensor because the proportion of light rays arriving at the sensor at low angles – exactly the rays that get under the aluminum - is increased.
I remember having read about frame interline transfer CCD area array. The first transfer is interline transfer followed by frame transfer. This appeared to be a good solution to striping at the cost of added complexity and silicon
ReplyDelete