I guess we're all thankful. It just took us (me) awhile to figure out what was being tested.
I'm still not sure I understand the usefulness/meaning of this test. What are native Q lenses like at 100% crop? If I saw that then I could comment more intelligently. Oh wait, no, this is me. Revision: ...If I saw that then I could comment.
Bifurcator wrote:
I guess we're all thankful. It just took us (me) awhile to figure out what was being tested.
I'm still not sure I understand the usefulness/meaning of this test. What are native Q lenses like at 100% crop? If I saw that then I could comment more intelligently. Oh wait, no, this is me. Revision: ...If I saw that then I could comment.
The context for me is to see how much resolution it possible to get with 36x24 mm sensors. the Q is only a test tool, and the performance of a Q lens is irrelevant to me. I print large, and would make use of 100-200 megapixels if it was available. Does it make sense to collect expensive lenses for this format, or is larger format the only way to go?
We have not previously been able to see with our own eyes how much of the resolution is the camera/sensor related and how much of it is the lens. Now we can see what the lenses actually deliver, in the context of large sensor cameras.
Between FF and APS-C sensors, we often do not see the full resolution advantage we expect, even with the best lenses. It is often interpreted as a sign that we are close to the resolution limit.
When new cameras with more megapixels are released, and the actual increase in detail is found to be modest, some people conclude that it is pointless. I have said the opposite, we need much higher increase in megapixel count to get more detail.
This kind of tests may provide som novel insight to these questions.
But how do we know the sensor isn't just the soft on it's own? Do you have any native lenses for the Q? Post up a 100% crop it if you do. That should round out this lill' test.
interesting, thanks!
I'd like to see the same crop with a Q lens if possible? I assume it would be better than this, which means that the only reason why the Zeiss lens is better and more expensive is because of the bigger image circle? (if I understood the whole test concept properly)
I print large, and would make use of 100-200 megapixels if it was available. Does it make sense to collect expensive lenses for this format, or is larger format the only way to go?
I am not 100% sure my calculations are correct, but if they are, the diffraction won't let you resolve more that 236MP at f/2.8, 116MP at f/4 and 59MP at f/5.6.
With 200MP you will start seeing diffraction impact at f/3.3 and with 100MP at f/4.7.
That is for 24x36mm sensor, no AA filter and 560nm light. So you probably would not be able to take advantage of the resolution if you are shooting three dimensional objects and need some depth of field.
Bifurcator wrote:
Well! When you put it that way...
Actually it sounds like a neat test.
But how do we know the sensor isn't just the soft on it's own? Do you have any native lenses for the Q? Post up a 100% crop it if you do. That should round out this lill' test.
Spyro P. wrote:
interesting, thanks!
I'd like to see the same crop with a Q lens if possible? I assume it would be better than this, which means that the only reason why the Zeiss lens is better and more expensive is because of the bigger image circle? (if I understood the whole test concept properly)
Ok, I will toss in similar image with the 8 mm prime (the only Q lens I have) later today. The perspective and DoF will be different, so the comparison will have to be reviewd with caution.
Mirek Elsner wrote:
I am not 100% sure my calculations are correct, but if they are, the diffraction won't let you resolve more that 236MP at f/2.8, 116MP at f/4 and 59MP at f/5.6.
With 200MP you will start seeing diffraction impact at f/3.3 and with 100MP at f/4.7.
That is for 24x36mm sensor, no AA filter and 560nm light. So you probably would not be able to take advantage of the resolution if you are shooting three dimensional objects and need some depth of field.
Your calculations sound about right, and the trade-off between DoF and diffraction will be painful (as it will for large format film)
However, we need to factor in that the effect of diffraction is not a clean cut. We can at least separate the effect into these levels:
a) Where diffraction softening starts to get visible
b) Rayleigh criterion for 560 nm and nominal pixel spacing.
c) Absolute cutoff for 560 nm and nominal pixel spacing.
d) Rayleigh criterion for correct sampling of blue light (470 nm) with a Bayer CFA
e) An oversampled image may "look better" than an interpolated image even without more fine detail.
This test also does provide some subjective answers. It is the equivalent of 360 MP. I would say that f/2.8 looks fine for that. So I would adjust your numbers slightly (for my preference) and say that for 100 MP I would go to f/5.6 and for 200 Mp to f/4. These numbers are not meant as a universal standard of course.
Bifurcator wrote:
But how do we know the sensor isn't just the soft on it's own?
How can a sensor be soft? The pixels are discrete aren't they? I think the only way it could be soft is if there is an AA filter that is too strong to give sharp results.
AhamB wrote:
How can a sensor be soft? The pixels are discrete aren't they? I think the only way it could be soft is if there is an AA filter that is too strong to give sharp results.
The camera may also soften the raw file. IIRC, dxomark detected smear in their noise analysis of the Q.
Mirek Elsner wrote:
I am not 100% sure my calculations are correct, but if they are, the diffraction won't let you resolve more that 236MP at f/2.8, 116MP at f/4 and 59MP at f/5.6.
With 200MP you will start seeing diffraction impact at f/3.3 and with 100MP at f/4.7.
That is for 24x36mm sensor, no AA filter and 560nm light. So you probably would not be able to take advantage of the resolution if you are shooting three dimensional objects and need some depth of field.
Your calculations seem about right for 550nm light, but your assumptions are all wrong... :-)
The distance between two pixels of the same color can never be "1 pixel width", since they do not populate every pixel instance. The absolute minimum distance between two green pixels is 1.41 x pixel size, sqrt(2) since the distance is a diagonal 45º distance.
Then you need to add in some optical losses that mask the visibility of the diffraction, say that in an almost perfect lens like the CO60 or an APO-Elmarit 180/2.8 you need to add in at least 20% more to this. 1.2 x sqrt(2) = 1.7.
This means that in reality, in a real image taken with a Bayer based sensor and with a real lens, your figures are about 1.7^2 = 3 times to low.
You can see the diffraction hitting in hard at F5.6 at a "300MP FF sensor", but you just CANT see the diffraction at F4.0 on the same sensor.
In the smallest pixel FX sensor I have access to today, the D800, I can still sharpen an F11 image so that it looks absolutely identical to the F5.6 image in the plane of best focus. I can't do that with F16, so I'd say that the real diffraction limit in real use with a D800 with 4.8µm pixels is somewhere between F11 and F16. On the D3x, F16 was never a real problem as long as you have more light than you need (either sunshine, or studio strobes).
The F-stop where you can actually SEE that the pixel-to-pixel contrast has started to drop (by inspecting the image at bigger-than-life magnifications) by a few percent should not be called a "limit", it should be called a "visibility threshold". Those two ways of saying it means two totally different things to most people.
When I first read this post (when it first appeared), I was thinking "What the hell?" Only after seeing the 5d2 image did I see what was to be seen: a minuscule portion of an image. I knew my ZE21 was good, I didn't know it was that good.
Come on Sony, give us the 36mp full frame NEX, I have the lens already!
The test with the Pentax 8.5 mm lens was a little difficult because the perspective was so different, I had to walk much closer. The 21 mm lens becomes a tele lens on the Q, while the 8.5 mm is a normal. I tried to make a compromise, with the flowers just minimally larger in the frame, at the cost of the background magnification a little less. The light was also very different.
So Bif, do you still think the sensor is soft?
I think this shows that it is much easier to produce a lens with very high resolution when the image circle is very small.
AhamB wrote:
So Bif, do you still think the sensor is soft?
I think this shows that it is much easier to produce a lens with very high resolution when the image circle is very small.
That is true, but in terms of resolution a lens cannot be better than a diffraction-limited lens. Thus ... how can this lens be so much better than the supposedly (close to) diffraction-limited Canon lens?
Toothwalker wrote:
That is true, but in terms of resolution a lens cannot be better than a diffraction-limited lens. Thus ... how can this lens be so much better than the supposedly (close to) diffraction-limited Canon lens?
All close-to-diffraction-limited lenses are equal, but some are more equal than others.
I have made some initial tests with the TS-E 24 II and the results are encouraging. 
of the day! 
