Why go full frame?

The Ran wrote in post #13172736
Crop doesn't magically have a 50-60% lower resolution, that is dependant on the sensor resolution. Take for example a 20mp crop sensor and a 20mp full frame sensor, with no AA filter and combined with a lens that outresolves both of them they will capture the same amount of detail. Now, if the lens gets soft enough at the edges that it then no longer out resolves the sensors then the crop will capture more detail.

Ok, maybe you can explain to me why when using an extremely sharp lens like the 70-200 2.8L MKII at it's sharpest settings (70mm @ F4 from what I can tell) the 5Dmkii produces roughly 40% more centre sharpness, despite only a 15% mega pixel advantage?
How could that possibly be possible, if like you say, the 5Dii is only producing substantially higher MTF figures due to a 15% mega pixel advantage?

HTTP response: NOT FOUND | MIME changed to 'image/png'

rhys216 wrote in post #13172909
Ok, maybe you can explain to me why when using an extremely sharp lens like the 70-200 2.8L MKII at it's sharpest settings (70mm @ F4 from what I can tell) the 5Dmkii produces roughly 40% more centre sharpness, despite only a 15% mega pixel advantage?
How could that possibly be possible, if like you say, the 5Dii is only producing substantially higher MTF figures due to a 15% mega pixel advantage?

Those figures are presented as linewidths per picture height (not linewidths/mm on the sensor). The 5D2 has more picture height, so it can have more linewidths.

rhys216 wrote in post #13172909
"Cropped vs. Full Frame Sensors. One needs to be extra careful when comparing MTF charts amongst cameras with different sensor sizes. For example, an MTF curve at 30 LP/mm on a full frame camera is not equivalent to a different 30 LP/mm MTF curve on a 1.6X cropped sensor. The cropped sensor would instead need to show a curve at 48 LP/mm for a fair comparison, because the cropped sensor gets enlarged more when being made into the same size print."

The above is saying that if a lens produces 30 Lp/mm then to get the equivalent sharpness on a crop sensor, you would have to find a lens that equals 30lp/mm multiplied by the crop factor.

i.e. If a lens was 30 Lp/mm on a 5Dmkii, a 7D would need a lens that delivered 48 Lp/mm to produce the same level of sharpness.

i.e. 30 x 1.6 = 48

when being made into the same size print."

Also the 5D2 likely has a rather weak AA filter just like the 5D classic.

AJSJones wrote in post #13172947
Those figures are presented as linewidths per picture height (not linewidths/mm on the sensor). The 5D2 has more picture height, so it can have more linewidths.

Can you clarify what you mean here.

AJSJones wrote in post #13172947
when being made into the same size print."

Yes, if your printing an image at the same size, then the crop will need a sharper lens to reach the same sharpness.

The Ran wrote in post #13172969
Also the 5D2 likely has a rather weak AA filter just like the 5D classic.

Any particular data (or statement from Canon) that show that, or is this more internet wisdom ? The extent of blurring by the 5D2 camera would be less than the extent of blurring by the 5D if Canon stays with the only published info on the design of their AA filters (blurring extent is related to pixel pitch - see my post above). So are we meaning weaker in terms of pixel widths of blurring, or total blurring?

Heh! This discussion sure does take a lot of twists and turns!

Hey, I have an idea -- let's compare two sensors:

The first is the 7D sensor: 18MP 1.6 crop!

The second: a ff sensor that would have the same pixel density as the 7D sensor, the same AA filter, and so in the center would have the same resolution and IQ. But, because it is a full-frame sensor, it woud have 18x2.56 MPs, that is, 46 MegaPixels! Hey, which would you choose!?

Of course, in the "real world" we don't have that -- the 21MP 5D2 has made use of the advanced sensor technology to get better IQ/cleaner pixels, although it is 3 years old. But it still outresolves 18MPs on a crop sensor, and with cleaner pixels at that!

Now who knows what the next generation will bring!?

AJSJones wrote in post #13173002
Any particular data (or statement from Canon) that show that, or is this more internet wisdom ? The extent of blurring by the 5D2 camera would be less than the extent of blurring by the 5D if Canon stays with the only published info on the design of their AA filters (blurring extent is related to pixel pitch - see my post above). So are we meaning weaker in terms of pixel widths of blurring, or total blurring?

Apparently each camera usually has a different strength AA filter, with the average being 30%, so the average camera with 10 Mp, will effectively only have effective 7 Mp's.


"Only currently available on select DSLR cameras, the ICF and AA assembly is removed. A new ICF is installed because the stock ICF is cemented to the AA. The new ICF blocks IR light. By getting rid of the AA filter, also known as the Blur Filter, camera sharpness is enhanced by about 30% typically. The drawback on an HR camera is greater potential for moire patterns. The details on why the manufacturers put in a blur filter are technical, but, basically it has to do with digitally sampling an analog signal (Nyquist Theorem). The manufacturers are not willing to completely eliminate moire patterns because this would require a minimum 50% blur filter. So instead, they make a compromise which varies by camera model, but is often about a 30% blur."


http://www.maxmax.com/​digital_still_cameras.​htm

AJSJones wrote in post #13172947
Those figures are presented as linewidths per picture height (not linewidths/mm on the sensor). The 5D2 has more picture height, so it can have more linewidths.

rhys216 wrote in post #13172984
Can you clarify what you mean here.

At the value of MTF 50 for saying "we defined a linewidth" (numbers close but not super-precise )

7D
15 mm picture height; 1500 LW/picture height = 100 LW/mm; 3456 px/15mm; 230 px/mm - 2.3 px/LW

5D2
24 mm picture height; 1900 LW/picture height = 79 LW/mm; 3744 px /24 mm; 156 px/mm ~2 px/LW

This suggests that the 7D takes 2.3 (4.4 µm) pixels to define one "linewidth" while the 5D2 takes 2 (6 µm) pixels.

Thus it takes 10.2 µm of a 7D sensor and 12 µm of a 5D2 sensor to define a LW.

7D's measured lineal resolution is better than the 5D2, as expected (but only 87% of what might be expected solely by decreased pixel dimensions i.e. 2/2.3). (but confounded by any testing variables I ignored )

Is that clearer now ?

rhys216 wrote in post #13173054
Apparently each camera usually has a different strength AA filter, with the average being 30%, so the average camera with 10 Mp, will effectively only have effective 7 Mp's.


"Only currently available on select DSLR cameras, the ICF and AA assembly is removed. A new ICF is installed because the stock ICF is cemented to the AA. The new ICF blocks IR light. By getting rid of the AA filter, also known as the Blur Filter, camera sharpness is enhanced by about 30% typically. The drawback on an HR camera is greater potential for moire patterns. The details on why the manufacturers put in a blur filter are technical, but, basically it has to do with digitally sampling an analog signal (Nyquist Theorem). The manufacturers are not willing to completely eliminate moire patterns because this would require a minimum 50% blur filter. So instead, they make a compromise which varies by camera model, but is often about a 30% blur."


http://www.maxmax.com/​digital_still_cameras.​htm

Thanks - that confirms a generality of the desire to blur to a fairly similar extent based on pixel dimensions, as Canon's published diagram indicates. (I wish I knew what the numerator and denominator were in that "30% blur" number) By splitting the image of one ray into 4 rays, 4 adjacent pixels each get 25% of the photons from that ray, so colours should be quite well sampled. MaxMax could actually do the measurements on all the cameras from which they extract the AA filter - that would settle all these speculations and possibly (or not) drum up business from the sharpness freaks (but they know who they are anyway). How much it varies by camera model is hardly something that can be measured without tests as well controlled as theirs! The difference between before and after on those comparisons never struck me as really worth it, compared to what can be done by (Canon's recommended) 0.3 radius, 300% USM as a starting point, or something like Topaz InFocus.

AJSJones wrote in post #13173104
At the value of MTF 50 for saying "we defined a linewidth" (numbers close but not super-precise )

7D
15 mm picture height; 1500 LW/picture height = 100 LW/mm; 3456 px/15mm; 230 px/mm - 2.3 px/LW

5D2
24 mm picture height; 1900 LW/picture height = 79 LW/mm; 3744 px /24 mm; 156 px/mm ~2 px/LW

This suggests that the 7D takes 2.3 (4.4 µm) pixels to define one "linewidth" while the 5D2 takes 2 (6 µm) pixels.

Thus it takes 10.2 µm of a 7D sensor and 12 µm of a 5D2 sensor to define a LW.

7D's measured lineal resolution is better than the 5D2, as expected (but only 87% of what might be expected solely by decreased pixel dimensions i.e. 2/2.3). (but confounded by any testing variables I ignored )

Is that clearer now ?

Yes it is clearer.

Take for instance the figures I'v made bold. The 7D is squeezing approx 50% more pixels into the same picture (sensor) height that needs to be resolved in the same surface area of the lens as the 5dii. Now according to CIC the lens is often the limiting factor in terms of resolution/sharpness.

"Frequently, the resolution of your digital photos is actually limited by the camera's lens — and not by the resolution of the camera itself."

Now as the 7D and 5Dii have rather small pixel densities, my guess is true resolution is nearly always limited by the resolution of the glass, as that would corroborate with the frequent 50% sharpness advantage the 5Dii often has over the 7D in terms of MTF figures.

Put simply, the figures suggest that the lens is the limiting factor, therefore the MP count (between the two bodies) has no effect on sharpness, and the reason that the 5Dii delivers a sharper image is because each pixel on the sensor, is resolved by a larger surface area of the lens, thus effectively reducing the bottleneck of the lens by approximately 50% in most cases (the 70-200 at it's sharpest setting showed this percentage can be lower at approximately 40%).

I'm not going to make any comment on that until I know what you mean by "sharpness" - not a cop-out I simply don't want to think things through only to find you meant something else!
Think in terms of the following variables for comparisons of sharpness between cameras : 100% crop viewing? Same size print viewing? same degree of enlargement (FF print bigger than crop but viewed from same distance)? target occupies same fraction of sensor area? same absolute area? etc.

I dig full frame

Why go full frame?

Because the DOF alone is worth it, most of the time like tonight, I want all the bokeh I can get.

MIME changed to 'text/html' | Content warning: script

What was the depth of field there, about a meter? Would an extra 60cm have made much difference? I doubt it. Or if it really did then you could've have used the money saved by getting a crop body to get a 35mm f/1.8 (assuming you used the 50mm, if it was the 85mm then just use the 50mm on crop) and get the same depth of field and a better lens.

The Ran wrote in post #13177149
What was the depth of field there, about a meter? Would an extra 60cm have made much difference? I doubt it. Or if it really did then you could've have used the money saved by getting a crop body to get a 35mm f/1.8 (assuming you used the 50mm, if it was the 85mm then just use the 50mm on crop) and get the same depth of field and a better lens.

Uh, I think your mistaken, which is also probably the reason you struggle to understand why FF is so appealing to most togs with a few specialised exceptions.

35mm 1.8 on a crop would achieve about the same FOV, however it does not equal the same DOF as a 50mm 1.8 on FF.