D3, 1Ds3 noise comparo

[226178]

For the life of me, I don't know why everyone gets so worked up.  I agree with others that it is a terrific time to have access to cameras with the kind of technology and sophistication that allows us to move our craft into areas not previously (or perhaps easily) accessible, no matter whether one previously shot film and/or digital.  And I am appreciative to those such as Michael or Jay or Jeff who are willing to test these cameras and share with us their experiences.  I mean, how many of us could consider purchasing multiple cameras to post our observations on our blogs?

My appreciation notwithstanding, my experiences will never be Michael's, my priorities never Jeff's.  No one forces me to read their opinions nor to take workshops from them.  For example, did I put myself in the path of a wayward seal?  No, but I did enjoy the images of the "wolfpack" seals.  And of course, I thoroughly enjoyed the commentary.

What a boring site this would be if we all agreed!  I say, celebrate our differences in perspective and our tendencies to compete.  Because (as I started) without all of this, we would still be shooting deguerrotypes.

Rick Allen

[John Sheehy]

Unless your photography involves long exposures of distant point light sources, them being stars or whatever, I do think that this issue will not affect the quality of your pictures.
[a href=\"index.php?act=findpost&pid=160614\"][{POST_SNAPBACK}][/a]

Well, there's a little more to it than that.  Canon has traditionally been the astrophotographer's DSLR of choice not just because of the lower overall noise, but because Canon does not clip the RAW data anywhere near black, but far below it.  Unlike the shot noise due to photon collection (and leakage noise, too), which can never be less than photonic black, the noises incurred in reading the sensor have both positive and negative elements.  Reading out the sensor so that the true black level is zero (as most manufacturers do) in the RAW data clips away half of the noise at black.  That is not a good thing because the noise also pushes the signal up and down, so signal near black is lost in the process, to a greater degree than noise is lost.  To clip above black (as is done in the D3) not only reduces the strength of signals and SNRs near black, but loses the very lowest levels of signal almost completely, because even piggy-backed on top of the noise, they still don't reach the clipping point.  For the lowest ISOs, no signal may survive at all, because there is not enough noise for anything to reach above the clipping point.

None of this matters much in a well-exposed image, but if you really need to get down in the shadows, and stack or bin multiple images, the Canon style of RAW data is the way to go, as you can stack and bin in such a way that you don't lose any signal, and the near black means are totally linear right up until the point that black is finally clipped, and much more linear than they would otherwise be, after the clipping.

[bjanes]

Well, there's a little more to it than that.  Canon has traditionally been the astrophotographer's DSLR of choice not just because of the lower overall noise, but because Canon does not clip the RAW data anywhere near black, but far below it.  Unlike the shot noise due to photon collection (and leakage noise, too), which can never be less than photonic black, the noises incurred in reading the sensor have both positive and negative elements.  Reading out the sensor so that the true black level is zero (as most manufacturers do) in the RAW data clips away half of the noise at black.  That is not a good thing because the noise also pushes the signal up and down, so signal near black is lost in the process, to a greater degree than noise is lost.  To clip above black (as is done in the D3) not only reduces the strength of signals and SNRs near black, but loses the very lowest levels of signal almost completely, because even piggy-backed on top of the noise, they still don't reach the clipping point.  For the lowest ISOs, no signal may survive at all, because there is not enough noise for anything to reach above the clipping point.

None of this matters much in a well-exposed image, but if you really need to get down in the shadows, and stack or bin multiple images, the Canon style of RAW data is the way to go, as you can stack and bin in such a way that you don't lose any signal, and the near black means are totally linear right up until the point that black is finally clipped, and much more linear than they would otherwise be, after the clipping.
[a href=\"index.php?act=findpost&pid=160635\"][{POST_SNAPBACK}][/a]

John,

An informative post. Canon applies a bias voltage so that the read noise is not clipped. Of course, this could decrease the dynamic range a bit if clipping at the high end occurs sooner. What factors do you think caused Nikon to use their current strategy and what are the advantages and disadvantages?

As for astrophotography, it seems to me that Christian Buil's main complaint was not clipping of the blacks but the way they handle thermal noise by filtering the raw data.  This would seem to eliminate the need for a second exposure for subtraction of a reference dark signal map and could speed up things for normal photography. Again, what is your take on the two approaches?

Bill