If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below. |
|
|
Thread Tools | Display Modes |
#41
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
"Robert Coe" wrote in message
... On Fri, 16 Oct 2009 10:21:42 -0400, "Peter" wrote: : Differences: : Although I have a D300, with my 200 as a backup, I am seriously thinking of : a D5000 because the articulated LCD should give my back more mileage. Am I the only Canon user in this elderly crowd? Maybe there's more than one reason not to shoot flies with a Canon duck -- Peter |
#42
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
"Paul Furman" wrote in message
... Peter wrote: Floyd L. Davidson wrote Peter wrote: John Sheehy wrote DRS wrote Why do Nikons seem to have a base ISO of 200? Is it a trade off of some kind? The D3 uses a low silicon fill factor, and makes up for it with large microlenses, which, of course, fill the wells faster. snip big snip Read noise has, like shot noise, a Gaussian distribution so it is essentially random, and will mostly be seen in the shadows. To continue the analogy used above, it looks very much like a paper covered with mixture of black pepper and grey pepper. Below a certain amount of exposure, the SNR of an image is very likely to be determined by read noise instead of shot noise. (Generally that is not a good thing.) Because read noise will be seen in the shadows it is generally more annoying to the human eye. Thanks for that explanation. Yes, thanks Floyd. And JPS! My interest as a photo artist is simply to make a nice picture. When Nikon says I can shoot at a higher ISO with less noise, I am skeptical. I have a friend who claims he shoot a D300 at ISO 1,600 without little perceptible noise. Yet I have not found that to be the case on my D300. This may be the usefulness of understanding this stuff. How do you get the best performance from whatever camera you have? ETTR (Expose To The Right) is one example of how to apply this knowledge, another is helping to decide whether to buy the latest mumbo jumbo technology and what it's worth for your particular needs. Good stuff. Yes indeed, it is good information. I now understand why many of my night harness racing images shot, at 3,200 and 1,600 were so noisy as to be almost unusable. -- Peter |
#43
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
On Sat, 17 Oct 2009 10:01:26 -0400, Robert Coe wrote:
: At 72 I'm ahead of both. Lots of fun years left, hopefully for all of us. I've always assumed I was one of the oldest in the group, but I won't be 72 until next week! Never take anything for granted. I'm 75 and not far to go to 76. Any advance on 75? Going .... Going .... (not yet, I hope). Eric Stevens |
#44
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
"Peter" wrote in message
... [] Yes indeed, it is good information. I now understand why many of my night harness racing images shot, at 3,200 and 1,600 were so noisy as to be almost unusable. -- Peter Although with some recent DSLRs ISO 1600 and perhaps ISO 3200 provide quite usable images. David |
#45
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
Read noise has, like shot noise, a Gaussian distribution so it is essentially random, and will mostly be seen in the shadows. To continue the analogy used above, it looks very much like a paper covered with mixture of black pepper and grey pepper. Not quite, in many cameras. Yes, read noise has a Gaussian distribution. But much of it is 1/f noise in amplifiers, especially in CMOS sensors. That is, it has a lot of noise at low frequencies. This means that it is correlated between pixels in streaks. That is much more obnoxious than noise uncorrelated between adjacent pixels. Doug McDonald |
#46
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
Doug McDonald wrote:
Read noise has, like shot noise, a Gaussian distribution so it is essentially random, and will mostly be seen in the shadows. To continue the analogy used above, it looks very much like a paper covered with mixture of black pepper and grey pepper. Not quite, in many cameras. Especially for older cameras, though much less commonly a significant problem with newer technology. I should have said "sensor read noise". "Read noise" is an ambiguous term in that it can reference the sensor only, or everything from the sensor through the ADC digital output. Yes, read noise has a Gaussian distribution. Sensor read noise has a Gaussian distribution. Other components to the overall "read noise" do not necessarily. But much of it is 1/f noise in amplifiers, especially in CMOS sensors. That is, it has a lot of noise at low frequencies. Pattern noise... which may or may not be the same from one image to the next. It is somewhat difficult to characterize, as it depends on both the "signal" and amplifier parameters. It probably isn't correct to say "much of it", because it is probably a minor portion of the actual noise. However... This means that it is correlated between pixels in streaks. That is much more obnoxious than noise uncorrelated between adjacent pixels. The human eye is adapted to seeing patterns. Hence pattern noise, even at lower levels than other noise, can be more annoying. There are other types of noise too; but the purpose of that particular article was *only* to compare shot noise to read noise on relatively simple terms without scaring anyhone half to death with technical detail. -- Floyd L. Davidson http://www.apaflo.com/floyd_davidson Ukpeagvik (Barrow, Alaska) |
#47
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
"Peter" wrote in
: Would you please explain this concept in English? I know about SNR and why I get more noise at a higher ISO, but I don't understand the difference between read noise and shot noise. Sorry, I forget that I post in usenet sometimes. I don't have time to read the other responses right now, so I don't know what is covered and what I agree or disagree with, but I'll just give a summary. Shot noise is the only truly *necessary* noise in digital imaging. It is noise that is always in light because light itself is counted in whole particles. So, if you have a theoretically smooth, flat subject, you will never collect the same number of photons in all the pixels. Read noise is noise generated by the camera, reading the sensor and digitizing the results. It only exists in current technology, and may not exist in future ones, if ways are found to digitally count photon strikes. There are few main differences between read noise and shot noise: 1) Shot noise in an image is the same, as long as the absolute exposure is the same. IOW, if you take a shot of the same scene and lighting at ISO 1600 with f/8 and 1/100, you have the same shot noise if you had the camera set to ISO 100 and f/8 & 1/100. Read noise, with some cameras, especially CMOS DSLRs, changes when you change the ISO. With such cameras, there is almost no difference in read noise, relative to normal exposure, between the first 3 stops of ISO from base, and then well above that, they tend to have read noise proportional to ISO. Looking at it relative to absolute signal, read noise on such cameras is greatest at the lowest ISOs, and lowest at the higher ISOs. 2) In any given camera, at any given ISO, shot noise, relative to signal, doubles as you quarter the signal (2 stops less exposure), while read noise quadruples. So, the deeper you go into the shadows, the faster read noise increases compared to shot noise. Running out of photons is the least of your problems, when you decrease exposure, as read noise takes a greater toll than shot noise. 3) Shot noise and read noise have completely different relationships to real, detailed signal. Say you had a black-and-white checkerboard as a subject. As you decrease exposure, it gets to the point where the white checks don't look much different than the black checks, because they are faint, compared to the read noise "floor", which blankets over both check colors. They barely show through. If there were no read noise, and only shot noise, the black would be totally black, and the white checks would simply have a more speckled texture. 4) Only read noise has patterns in it. All those streaks and bands you see in images come from read noise - sloppy electronics without correction (most line patterns are correctable; manufacturers that have them are just apathetic about quality). Shot noise is kind of nice looking; like the actual texture of the image, especially at high pixel densities. |
#48
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
John Sheehy wrote:
Read noise is noise generated by the camera, reading the sensor and digitizing the results. It only exists in current technology, and may not exist in future ones, if ways are found to digitally count photon strikes. Such methods exist. I have devices using them. They work well. They have existed and worked well since, in some cases, the late 1940s. Solid state ones have existed for years. The problem is that they generate a bit too much heat: without active cooling am imaging device would overheat and the thermal electrons would ruin the noise performance, not to mention the device. Doug McDonald |
#49
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
Bob Larter wrote in
: Read noise is related to the gain in the sensor amp, If you break the read noise down into two components, pre-gain and post- gain, the pre-gain part is the same and fixed for any absolute exposure, regardless of gain. For the post-gain part, the lower the gain, the more read noise there can be, relative to absolute signal. There is no general history of higher gains causing more absolute read noise! I do have one camera that seems to be the exception; my Panasonic FZ50. It has the lowest absolute read noise at ISO 200, and the highest at 800 (but not a huge difference). while shot noise is relatively constant in relation to the original signal. |
#50
|
|||
|
|||
Nikon D3S (12.1 Mpix, FF, very high ISO, HD video)
|
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Nikon D90 PRO announced. 12 Mpix 20D killer | Walt Hanks | Digital SLR Cameras | 56 | April 12th 05 08:43 AM |
Nikon D90 PRO announced. 12 Mpix 20D killer | Walt Hanks | Digital Photography | 89 | April 2nd 05 09:27 AM |
Nikon D90 PRO announced. 12 Mpix 20D killer | Walt Hanks | 35mm Photo Equipment | 79 | April 2nd 05 09:27 AM |
Nikon D90 PRO announced. 12 Mpix 20D killer | Alan Browne | 35mm Photo Equipment | 0 | April 1st 05 06:22 AM |
Is 4 Mpix camera just as good as 5 Mpix when available light is the limiting factor? | Woody | Digital Photography | 17 | September 26th 04 06:44 PM |