Photon wrote:
But a 24 MP sensor would come nowhere near the diffraction limited performance of a lens at f/2.8. There are a few lenses out there now that can resolve more than any current full frame digital sensor can fully utilize, and the total resolution of the system is a product of both factors, so more pixels can only help. Increasing pixel density can also eliminate moire issues and reduce Bayer demosaicing artifacts without an anti-aliasing filter.
From a practical standpoint, it can certainly be argued that the output from a current top of the line 12 MP camera might give the overall best results for high ISO shooting, but I see no reason to decide a priori that a higher MP camera will be at a disadvantage. Whatever Nikon does with its next FF body series, I would expect a higher pixel count and similar or even lower noise levels taking the image as a whole. It's a win-win situation....Show more →
An excellent point regarding diffraction @ f/2.8. To see diffraction limitations at this setting would rquire a HUGE print and would also require more than 24MP too. So more MP would help but only if you plan on making prints large enough to see it. You certainly WILL NOT see it on a 24X36" or smaller prints.
As a personal preference, I'd make a 12MP file out of that one by cropping off the top to make it a faux pano. Which shows another benefit of the high-MP camera, of course: if I tried that on my beloved D300, I'd end up with a 6MB file... good enough to print at any normal size, but certainly with less "enlarge-ability" than the same image from a 20+ MP sensor.
Thanks Rodolfo. Yes - that's right - higher rez sensors are great for cropping. If you know you only need a 12 megapixel file, and your lenses are good enough, then you can take fewer lenses: more megapixels save weight when travelling! I often hear it said around here, i.e. in the Alt, Nikon, and Canon forums, that people don't want more megapixels. I don't understand that. If there's little or no high ISO concession for more pixels, then where's the downside? Pixel-binning raw modes like the 50D's sRaw and mRaw mean that the number of pixels is only a ceiling.
Now, see , that's a very good shot (the crop doesn't do anything for me). But then, we're talking about resolution. That's a different subject.
I wonder how many remember how the Canon folk would jump on and make trolly comments like why they would "never use something like the D2x because of the pixel pitch, it was hard to get a sharp image. You had to be even more aware of your shutter speed".
No , some people don't need more MP. I've got a D300, and am more than happy with what I've got. The lenses and other accessories is what I'm chasing. There are some that will benefit more MP.
nikt wrote:
Now, see , that's a very good shot (the crop doesn't do anything for me).
The crop is just there to show the level of detail available in the full rez version, and illustrate that the D3x is going to be able to make use of Nikon's excellent lenses without diffraction always spoiling the party.
> I wonder how many remember how the Canon folk would jump on and make trolly comments like why they would "never use something like the D2x because of the pixel pitch, it was hard to get a sharp image. You had to be even more aware of your shutter speed".
Well, funnily enough the trolly comments were absolutely right. To get the best out of a 21 megapixel file (5.7 micron pitch), I do need to use a higher shutter speed and a better lens and technique than with my previous workhorse (8.2 microns). You also need to be sure to focus really accurately. The 'reach' of high density sensors comes with a higher lens sharpness, shutter speed, focus accuracy and technique requirement. IMO that is why some 1Ds3 users have claimed (falsely, in my experience) that the camera's results need more sharpening. They were really just seeing the sharpness limits of their lens/shutter-speed/focus-accuracy.
Yes, the comments were correct. What I'm getting at is that I remember some of those people. They are the same people that bang on about the resolution of the 50D for instance.
I do understand why you posted the the crop. No one is disputing what extra resolution gives you in detail. And my point is pretty much what you've said. Many people will have to improve , among other things, their technique if they are to use a high MP / sq.cm chip.
Most people simply won't get the benefit out of these cameras.
To be honest, the iso 12800 results from the two cameras are so close after noise reduction that I'm not sure I could reliably tell them apart. In my view this vindicates not only my suggestion that the 1Ds3 might snap at the heels of the D700 for high iso performance, but it also vindicates my argument that sites like Luminous Landscape and DPReview are not correctly assessing noise and high ISO performance because they compare camera crops at different magnifications. Each of those sites suggests that there is a significant difference in high iso performance between these cameras. That assessment is based, as far as I can see, on prejudice, and the failure to do fair tests. You don't have to believe my test here, but I advise you to heed my warning that those sites have a serious flaw in their method of comparison and you would be better to do your own comparisons, carefully, using equal magnification, as I have, than to believe what you read at such trusted review sites.
To be totally clear, by saying "equal magnification" you really mean same printed picture size (as in physical size) with the same DPI (absolute printed resolution) when comparing image quality between cameras?
On screen, that would mean downsizing the DsIII images to the same amount of pixels as the D700, and then comparing.
This is a very fair test and I totally agree on the logics behind it. This is how it should be done, if you're interested in the final result rather than the "per pixel" technochratic oriented quality.
It also concludes that the DsIII would walk all over the D3/D700 in low light situations if it had the same "per pixel" quality, which it apparently hasn't... But it makes up for it (part of the way at least) by being able to use more pixels to cover the same print area (as I understand it, the way the word magnification is used here)
I do understand that for pro's, more is almost always better, as long as it doesn't affect your shooting speed, ISO-quality (at the same "magnification") or time spent in PP. For me, with the print-sizes I use privately (mostly up to 20x30.... in cm - that's ~8x12 in inches), any printer requiring a rip (final resize before the printing engine) larger than ~6MP would cost me, or the company that does the printing a couple of 100k$. It would have to have usable real-world full-colour resolution noticeably beyond 300DPI, and very few printshops can do that.
I used to work for a company that had this kind of equipment They were one of the smaller plants of the European division to be granted with pre-press proofing printers this expensive. They had a yearly turnaround of roughly 150M€ - that's around 200 million USD...
I've been interesting in seeing such a comparison, and this was well done. Thanks for sharing!
I agree with your main conclusion, which is that DPR and others are doing the comparisons poorly. I don't agree with you that they are so close after noise reduction that they are hard to tell apart. Even in the web-sized versions you posted in your last post, the Canon is struggling more, particularly with the shadows. On the other hand, the Nikon advantage may simply be due to the software processing. NX2 is very good at handling chroma noise in the shadows. If you processed both files in the same RAW developer and then applied noise reduction as needed, I bet the results would be closer.
AndreasE wrote:
In case you are interested.
These 2 shots were taken with ISO 200 and ISO 12800 (with a D3).
regards,
Andy
Andy, can I ask why you (or anyone else for that matter) would EVER shoot a picture like that, with plenty of available light, at ISO 12800?
Seems like such shots do nothing to help determine a camera's low light capabilities. What I'm interested in are shots that would otherwise be lost "without" ISO 12800, not shots that could just as easily be captured at ISO 200.
Sheesh you can see the softness of the 12800 shot even in these web sized images
Mattbtn wrote:
Andy, can I ask why you (or anyone else for that matter) would EVER shoot a picture like that, with plenty of available light, at ISO 12800?
,
to be fair it does demonstrate the softness that comes with high ISO...or is it NR?
EDIT: oops sorry didn't realise this was an old thread, this link has been posted in other threads
http://cyberphotographer.com/d700v1ds3/12800a_sansNR.jpg
BTW, I think that the non-processed crops above illustrate that Nikon does apply NR in camera before saving raw files. There is chroma noise on the 10-20 pixel scale, as you can see in the monitor stand, but none whatsoever on the 1 pixel scale. Local small scale chroma noise doesn't appear, whereas local small scale luma noise does. How can that be, bearing in mind that it's a standard bayer sensor? Some people are going to claim that the Nikon sensor has a mysterious ability to see noise in black and white, despite being a 3 colour device, but I think there is a much easier explanation: Nikon cameras have excellent in-built chroma noise reduction which blends out high frequency chroma noise with a blending radius appropriate to the voltage levels coming off the sensor. It's an excellent way of reducing noise, and it's precisely the method I normally use in Photoshop to noise-reduce noisy files, as well as being the method used by all 3rd party noise reduction software. Some astro-photographers aren't too keen on it, I understand, for obvious reasons, but generally it's a very useful feature. The D700 essentially has NeatImage/NoiseNinja built in, and impressively, it can do it at 8 frames per second, which makes it a low-light sports shooter's wet dream come true.
Equally, one can see easily from this crop that desktop chroma noise reduction is going to reap big rewards in the Canon file, whereas there's not much more to be gained in the Nikon file, unless you use a larger radius, and then the edges will go water-colourish. So my view is that the D700 iso 12800 raw file has already had some very useful noise reduction, but the Canon file seems to catch up once it has similar treatment on the desktop.
Finally, a couple of warnings to those considering using a 1Ds3 for iso 12800 work (as I do): desktop processing is time consuming, and a D700 is going to pay for itself really quickly if you shoot at those isos often. Moreover, the 1Ds3 has amp noise in the bottom right corner in very low light at very high iso, and correcting that isn't always easy, and will waste even more of your time.
,