This may be true for specific lens/sensor combinations, but there's something not quite right about this argument in general - otherwise we wouldn't be seeing the steady improvement in res. chart results as manufacturers have moved to higher resolution sensors in the P&S cameras.
Isn't that because they use good lenses and limit the aperture to a relatively large value?
This doesn't address my point.
No. I'm not forgetting the anti-alias filter. And I'm at a loss as to why you think I might be. My comments were based on the estimated limit of resolution of about 72 lp/mm for the D2X. That should be the resolution limit with the AA sensor in place.
All the various tests on Dpreview.com that show color moire at the limit of resolution pretty much demonstrate that htis is not consistently the case.
While I love Norman Korens site, I think his wording is suspect here. I think the paragraph is more correct if you substitute "diffraction limited" with "noticably affected by diffraction". I think that beyond f/4- f5.6, you don't necessarily lose resolution, though you will be losing MTF response.
Ron, you trouble maker you!I guess I'd say the opposite: Anybody who doesn't want more
resolution probably doesn't know what he's doing. (There - I said
it.)
Which is pretty much why I felt the opposite is true given these very real impediments to ever increasing advances in pixel density and resolution.
IF it were technically possible sure I am with you on that, but until that breakthrough in getting around the known physical limits of high resolution digital sensors occurs, practical considerations like noise, reduced DR,reduced sensititivity range, and increased cost(in particular production cost for the manufacturer) are going to be like lead walls. The present reality is there comes a point where it simply won't be profitable for the manufacturers to persue the dreams of a tiny minority of consumers, we can disagree over where that point is but my view based on current trends is that point is much closer to 12mp on an APS-C sensor than it is to say 24mp.
--
I am referring to image pixels per printed inch not sensor sites, I shouldn't have used dpi as that is usually associated with the final print device. Open a full size image in photoshop and view the image size parameter after setting the ppi to 300. Of course there is the subtelty of the final print device and weather it can render faithfully each pixel...but the D2x provides enough resolution so that if the print device can render each pixel in the image without under sampling it could represent 300 of the image pixels per printed inch at a print size of 9.493 x 14.293". I should have been more careful with my use of dpi.
Again, I am only referring to the resolution as expressed in the final print as rendered using the pixels in the image. If your printer can resolve each pixel provided by a full size D2x image, at 300 pixels per printed inch you would get a physical print that is 9.493 x 14.293" in area. How the print device uses it's resolution (dpi) to achieve that is indeed a separate matter.
Now you are confusing the pixels provided by the image and dots used to render the final print. (whatever a "dot" is for an inkjet, dye sub or color laser..etc.)
That depends on weather or not the device can actually render the pixels using whatever it has for a "dot". An inkjet will give you a great print given a source providing 300 ppi of image data over a 8 x 10" print but it will fail to provide as much detail as a dye sub will given the same source, and a color laser will give you a better print still. Each device is more capable of mapping each pixel of image data to a "dot" (whatever that is for the device) so what is "excellent" will vary with the device, print size and media chosen.
I never said they had such a chip, just that it is likely in the works.
A very important one as it determines if the camera will be resigned to a niche market or a general one...very real cost constraints on the part of the manufacturer and the unstoppable consequences of high density sensors (namely reduced sensitivity range and increased noise) will make it all the more difficult to carve out production expense just to satisfy the niche. I am with you that if it can be done why not, but unfortunately physical limits and cost will increasingly be in the way.
I never said there weren't, the point is they may not be practical for the manufacturer to produce and may be as well impractical for consumers to purchase given the reduced use such bodies will have. [low volume> high price]
Well in the example given, image pixels and sensor sites have the same count. The example strongly implies a 1:1 relationship between dpi and image detail. The discussion of pixels in the absence of a correlation to image detail seems generally irrelevant to me. But maybe that's just me.
I put the term "dot" in quotes for a reason. I am simply trying to correlate the smallest bit of resolvable data in the image to a print image that can can show 300 of these "smallest bits" in one inch. Most good modern inkjets can deliver a bit over 300 bits of real image detail (using lots more tiny ink dots to accomplish the task).
Just about any modern inkjet can render 150 lp/inch of actual image detail or 300 "bits" of image detail per inch. Of course excellent prints require more than simply good resolution, but I'd think that most people understand that you can only get excellent prints from printers capable of making excellent prints. The point is and was that the 300dpi criteria that is often quoted is often overkill - even for images intended to be viewed up close.
You are absolutely right. I misread what you wrote. My mistake. Sorry.
Yes they will and eventually diffraction does set a limit. Costs, however seem largely influenced by the size of the silicon real estate. So this seems to argue in favor of a high density DX chip since it would be cheaper to get the extra resolution on that size sensor than on a 36x24 or medium format one. If you consider the very real possibility for future developments like focus depth bracketing, such designs make even more sense.
As I said, you shoot your 25Mp DSLR in 6MP or 12Mp mode and the noise will be reduced. Much smaller pixel densities, higher noise and greater aperture limitations are already being dealt with by many millions of people cameras using smaller chips. It doesn't seem impractical at all to me to put 4 micron sensor spacing on a DX sensor when there are current sensors using 2.3 micron spacing. Heck, 2.8 micron spacing gets you almost 50Mp on a DX sensor and the ability to use all of that resolution up to around f/5.6. But hey - Nikon isn't hiring me as a consultant either. Lets check back in around three to five years and we'll see.
And this fact has nothing to do with how those image pixels are mapped to printed dots which was my original point as clarified in the last post.
Agreed, and to clarify yet again. ppi is provided by the image and dpi is tied to the print device in the first post (corrected in the second) I used dpi where I should have used ppi. Obviously, weather or not the print device can map the ppi of the image given the desired output dimension, using it's native print dpi (how ever many print "dots" it takes to render each pixel) without undersampling is a different issue tied to the device used.
Since I have not heard from Mike, I am answering my own question. In his example, he required a resolution of 49 lp/mm for a resolution of 4 lp/mm at a magnification of 12.25. Resolution at the Rayleigh limit (MTF = 9%) would be 98.8 lp/mm according to his reference. The resolution of 49 lp/mm would be about right for MTF of 50%.
Since sharpening affects the MTF response of an image. I'd be curious if judicious sharpening affects the MTF50 point in your test images. I would guess it would. The sensor will record detail at levels below MTF50, and if some of that detail can be brought up, then more enlargement could be tolerated for the same perceived image quality. This is the prime benefit of sharpening digital images.
Jay,
Hello Bill,
