MitchAlsup
Veteran Member
While it is within the definition to relegate 'noise' to random signal varriations; it is ALSO within the definition of 'noise' for anything other than the desired image itself to be considered 'noise'.Ok, let's get this straight. Noise is entirely random. Lines are
not random. Lines are not noise. Noise does not produce lines.
Actually, if you bother to look at the images with 'banding' you will find that the pattern is 2 dimensional, both horizontal and vertical. That is it is NOT banding, but a 2 dimensional pattern noise. In fact, my 20D has almost the exact pattern as does Christian Buil (google him). He shows an image that contains the backdrop of this noise in an easy to view manner. And he goes further and BOTHERS to subtract another image with no signal and shows the ressulting flat rield result. 95%+ of the pattern nosie is gone (ahem::banding). If banding bothers you learn to conquer it.
Then learn how to get rid of it.So what we are talking about is a camera that generates very low
contrast lines in all of it's images at all ISOs, in any region of
the image that is sufficiently dark. You think that's ok. I don't.
Pure wordsmithing. Everything IS in the image data, the image data itself, The possion noise data, the thermal noise data, the pattern noise data, and the 1/f noise data.Again, these lines are not lines in the noise, they are lines in
the image data
.
So, any automobile that doesn not actively prevent accidents is defective?An explanation is not justification. The camera
should be designed to eliminate this sort of interference. If it
isn't, it's a defect.
Fine, then go and use it.My G3 doesn't have lines in the image data,
and it cost a hell of a lot less.
Nope:: the different amplifiers read different quadrents of the sensor. If the amplifiers were out of calibration a whole quadrant would be higher in noise.The 20D uses four ciruits to improve the speed
of data processing, and slight differences in their electrical
calibration will result in - you guess it - stripes in the picture.
I think the pattern is partially in the 1 transistor amplifiers in the sensor cells themselves, and another is in the amplifier at the 'top' of a column. Still another results from minor varriations of timing due to the construction of the row decoders in the sensor.
It would have been, but As seen above, these are not the suspected components.If it is so obvious to you that 'slight differences in their
calibration' can lead to visible stripes in the picture, shouldn't
it be all the more obvious the the folks that built the camera?
A weak maybe at best.Shouldn't they have been able to fix it so that such things don't
happen? Don't they have a responsibility to fix it now?
Just a few years ago, to get ISO 3200 we had to bake our films in a hydrogen enriched gas mixture. Now all you have to do is dial up the intermediate amplifier gain. And you believe ou have the right to complain it doesn't work well enough.I doubt you'd even have to do the level correction.I love the quality of my 20D at ISO 1600, and I am sure that I
could find some so-called "banding" in my underexposed shots if I
cared to crop dark parts and do a levels correction to lighten them
up.
You just don't get it. These lines will be viewable even inPeople, get a life - or better: Shoot pictures, print them, hang
them into your living-rooms and enjoy them. Just don't print those
100%-crops of shadows with auto-levels and especially don't hang
them into your living-room![]()
prints. If not in a 4x6, then an 8x10, and it does not require
underexposure or autolevels. You could easily take a perfectly
exposed picture at 3200 ISO, print an 8x10 and see the lines. This
could also happen at 1600 ISO, though it's less likely.
Do you know that at ISO 3200 you get 2 A/D counts for each e- in a sensor cell! Why not complain that the A/D can't distingush between 1 e- and 1.3 e- in a cell? Know what, you can't get a fraction of an e- in a cell!
Case in point:: I have many images that COULD have shown banding if I didn't bother to set the black point ever so slightly high after pushing the shadow data up with levels. So, at least for my own images, when I see banding, I can get rid of it.
In addition (after reading Christian Buils web site) and for the last 300 images (5 days), I have been using automatic noise reduction. This does a commendable job of eliminating the pattern noise.
So, once again, we see that if it bothers you, YOU CAN get rid of it. And you can get rid of it a lot faster and for a lot less grief than by simply complaining (forever). After you get used to the added load in the workflow, its just a couple more seconds per image.
The noise performance of my 20D (even with my pattern noise) is so good, I don't bother with resetting the ISO most of the time. I shoot at 100 and push the image in DPP and PS. The only difference is the amount of noise from the intermedate amplifier.And I do shoot, I shoot a lot. The banding issue forces me too
avoid 3200 ISO almost entirely, an advertised capability of my
camera, one of the primary reasons I bought it. It also forces me
to take a lot more shots at lower ISOs because I know that exposure
had better be dead on, or else I risk finding banding in the
shadows during post processing.
--
Mitch