Maker Pro
Maker Pro

Interlaced ICX429 CCD camera project

I'm working on a low light still camera project using low cost Sony
ICX429AL interlaced CCD sensors.

Project architecture is : a PIC 16F877A (@20Mhz) for timing signals
generation, double correlated sampling, 16 bits ADC, data download to a
PC via USB.

In software I generate an acquisition sequence as follows :

1. Sensor reset ( via a single PhiSUB pulse )
2. Delay ( exposure )
3. Odd field start sequence ( from ICX429 datasheet )
4. Odd field download ( a loop of vert.shift and horiz.reg. read )
5. Even field start sequence
6. Even field download ( a loop of vert.shift and horiz.reg. read )

Camera works fine, but I have an important difference in ADU levels for
odd and even fields. That generate a very annoying pattern of
horizontal stripes ( sometimes called "Venetian blind" )

Even though a small amount of "venetian blind" seems to be normal for
interlaced sensors, the amount i have is not normal. Gap between even
and odd fields could be as high as 5000 ADU (in a 0-32767 scale).

I checked hardware and software sequences many times, and everything
seems correct. So, I guess, I'm missing something very basic that I may
give as granted, but that must be wrong.

Anyone with more experience could please give an advice ?

Don't be afraid to ask, if you need more details ( I'd like to get rid
of this....!!! )

Regards,

Marco Paolilli
Italy
 
R

Roger Hamlett

Jan 1, 1970
0
I'm working on a low light still camera project using low cost Sony
ICX429AL interlaced CCD sensors.

Project architecture is : a PIC 16F877A (@20Mhz) for timing signals
generation, double correlated sampling, 16 bits ADC, data download to a
PC via USB.

In software I generate an acquisition sequence as follows :

1. Sensor reset ( via a single PhiSUB pulse )
2. Delay ( exposure )
3. Odd field start sequence ( from ICX429 datasheet )
4. Odd field download ( a loop of vert.shift and horiz.reg. read )
5. Even field start sequence
6. Even field download ( a loop of vert.shift and horiz.reg. read )

Camera works fine, but I have an important difference in ADU levels for
odd and even fields. That generate a very annoying pattern of
horizontal stripes ( sometimes called "Venetian blind" )

Even though a small amount of "venetian blind" seems to be normal for
interlaced sensors, the amount i have is not normal. Gap between even
and odd fields could be as high as 5000 ADU (in a 0-32767 scale).

I checked hardware and software sequences many times, and everything
seems correct. So, I guess, I'm missing something very basic that I may
give as granted, but that must be wrong.

Anyone with more experience could please give an advice ?

Don't be afraid to ask, if you need more details ( I'd like to get rid
of this....!!! )

Regards,

Marco Paolilli
Italy
Normally, VBE, is a very few counts indeed. The main problem, comes with
moving targets, where the image from the second field is displaced
relative to the first, rather than problems with the intensity
differences.
I must admit, that my immediate suspicion, is that you are starting
exposures on both fields together, and then stopping the first field
before the second, so the second set of lines, have actually taken an
image for exposure time, plus the download time of the first set.

Best Wishes
 
Hi Roger and thanks for your replay.

You are correct : the second set of lines gets a total exposure time
that is longer than the first one while the first set is downloaded (
it may take 5 seconds on the USB port ).

But if I download both frames in total darkness (as I checked) that
difference shouldnt exist anymore (remains just a little contribution
from termal noise that cannot give that "huge" difference by itself).

It seems like some residual charge remains on the horizontal register
after each reading and adds to the reference level of the next line
(even thought i don't have any "smear like" effect)

Thanks,

Marco
 
R

Roger Hamlett

Jan 1, 1970
0
Hi Roger and thanks for your replay.

You are correct : the second set of lines gets a total exposure time
that is longer than the first one while the first set is downloaded (
it may take 5 seconds on the USB port ).

But if I download both frames in total darkness (as I checked) that
difference shouldnt exist anymore (remains just a little contribution
from termal noise that cannot give that "huge" difference by itself).

It seems like some residual charge remains on the horizontal register
after each reading and adds to the reference level of the next line
(even thought i don't have any "smear like" effect)

Thanks,

Marco
There is also the 'amplifier glow', as well as thermal noise. Not knowing
what your amplifier gain is set at, or the temperature of the CCD, it is
possible that all the components may be a lot larger than you think. The
normal method is to sequence the line timing starts, so that both sets of
lines receive equal timings. However there were a lot of tweaks in the
past to the Starlight Express driver for their cameras driver, to do with
better flushing, so obviously a simple reset, may not give a comprehensive
flush.

Best Wishes
 
Top