You must be looking at some other DxO from the one I can see, then, because the DxO SNR figures I'm looking at give the D3x at about 5 dB better than the D300 ISO for ISO.
You are very mistakend. Look at ISO 6400. 19.8 vs 20.8db.... and throughout the curve you have what equates to linear correlation
Just try selecting the 'print' tab. DxO explains why you need to do that here:
http://www.dxomark.com/index.php/eng/Insights/More-pixels-offsets-noise !
If you don't do that then waht you are doing is comparing the two cameras at different size of output.
This is basically showing you that pixel pitch is what determines the quality excluding generational improvements.
Nope - look at the DxO reference I gave you.
Total integrated noise is dependent on total surface area but that is not what represents the image appearance.
I guess that is why SNR is used as the typical measure of signal quality, crazy engineers and their unrepresentative numbers...right, that is what you are saying?
You've cropped what I wrote there. I was talking about the read noise. Constant amplitude read noise is 'injected' into the image at a spatial frequency dependant on the pixel pitch. This makes it not diectly comparable between different cameras with different pixel pitch. SNR by itself is not a useful measure of noise, you also need to know the bandwidth over which the noise is measured. That's why engineers tend to use noise spectral density (in V/Hz) as a measure of noise in a system. Then there is also the quastion of the perceptual effect of noise at different spatial frequencies, which is a whole new kettle of fish.
What you are completely ignoring is that the SNR is not size dependent. If you have two sq ft of sensor or if you have one micrometer, the percentage of error per noise is the same if the SNR is the same.
SNR is completely size dependent. That is in the nature of white noise. Capture an infinite bandwidth, you'll have infinite noise power. See my comment above.
Indeed, and at the same f-number the 35mm lens has a smaller aperture than the 50mm lens, and so refracts less light onto the sensor. that is why the larger sensor collects more light for the same exposure.
Total...yes if you integrate the area but once again, SNR is the same whether 1000 of the pixels are used or the whole sensor.
You can's simply compare the SNR of individual pixels,because if they are a differnt size, you're comaring the noise over different bandwidths.
Just crop your Fx to 6mp and print at say 240ppi. It will look the same as if you had just printed the full image at 240ppi.
But the image will be much smaller.
The noise is the same. You will now argue about the size...that will be addressed via optical magnification to show the equivalent image which doesnot affect the SNR.
Oh yes it does. However you magnify the image, you are stretching the noise content over one set of spatial frequencies to another set. However you do it, the noise comes with the detail. The only way to lose it is lose the detail, and that can be done as easily with a high resolution camera by proper resampling to the same output resolution (in which case you'll find your full frame image is head and shoulders above the cropped one, noisewise).
If the wafer of the fx cam is masked to dx dimensions, the noise percentage stays the same.
It doesn't actually. The shot noise is as I say above, but the read noise integrated over the sensor reduces, because there are fewer pixels. However, if you now blow that cropped portion up to produce a larger image, the scale of the read noise is larger (because each pixel represents a bigger proportion of the image) so the noise will look worse.
What does that have to do with read noise especially since it is incorporated into the SNR already? Now you are arguing about 6mp vs 12mpx. Yep, 12mpx blows up better but that is not part of the SNR equation, different issue as I addressed above in my 'not resolution dependent' disclaimer.
I only raised the read noise issue because I wasn't clear which noise you were talking about. If it's only the shot noise that you are interested in, then your whole argument is misplaced. Shot noise is
in the image it isn't
produced by the sensor . The only thing making the pixels bigger does is lose detail in the shot noise along with detail in the image. If you want, you can easily lose the same detail in processing and end up with a better image (due to the higher Nyquist of the AA filter).
The OP wants D3s performance in a Dx frame at 6mp. That puts a pixel pitch between that of the D700 and D3x but with D3s SNR.
I do have an FF camera, I have tried it and what you say is wrong. When you make a crop and blow it up to the same size as the full sensor it looks much more noisy. Anyone who regularly PP's images knows that. If you do 'overexpose' by a stop or so (or alternatively, reduce the ISO so the increased exposure is no longer 'over') the magically the noise is back to where you started.
That is not a FF vs crop issue, that is over magnification.
magnification is
entirely what the FF vs crop issue hangs on.
You are mixing up your arguments.
Oh no Im not.
12mp in a Dx and a 12mp Fx with the same SNR will scale exactly the same
only if you compare them in images scaled according to sensor size.
Intl We all agree that proper framing is necessary if you wanna shoot at 6mpx and compare with a higher rez capture, different issue
Same issue exactly. The OP's proposed 6MP camera will produce the 'performance' of the D3s only if you view the images 1.5x smaller (linearly - more than a whole ISO paper size down) - but so would a 12 or 18MP camera viewed the same size, and the higher res cameras will give better microcontrast.
--
thomas
