[Klaus]

Surprisingly just slightly reduced borders/corners compared to the corresponding Fuji tests:

http://www.opticallimits.com/sony_nex/881-zeiss12f28e

 

[Guest]

Hum. Not sure of what to make of that ? Maybe the native corner resolution is not great? Also there was some tests by the folks at lens rental (Roger specifically) that found when these lenses show up in different mounts (fuji, sony, 4/3) they sometime benefit by having optical glass inserted infront of the lens dependent on lens design and thickness of sensor filter. Was quite surprise by the magnitude of impact this glass had on the lens resolution (mtf).

[davidmanze]

  Dollars paid  X  Flange registration distance= edge resolution,

 

 You pay a price for shallow bodied mirror-less cameras with wide angle lenses!

[Klaus]

The 24MP Sony sensor has the reputation of producing edge smearing.

This doesn't show up all that much on the Z12. However, on the Z32 it is a completely different story.

The Z32 confirms the pattern that the sensor doesn't like the light ray characteristic of ultra-large aperture lenses (just see the Sony E 50/1.8 and E 35/1.8 reviews) at large apertures.

 

The comparison between the NEX7 and A6000 will be interesting although honestly I don't expect that much here.

 

I BELIEVE as opposed to KNOW that the lenses are actually quite fine.

[Brightcolours]

The 24MP Sony sensor has the reputation of producing edge smearing.

This doesn't show up all that much on the Z12. However, on the Z32 it is a completely different story.

The Z32 confirms the pattern that the sensor doesn't like the light ray characteristic of ultra-large aperture lenses (just see the Sony E 50/1.8 and E 35/1.8 reviews) at large apertures.

 

The comparison between the NEX7 and A6000 will be interesting although honestly I don't expect that much here.

 

I BELIEVE as opposed to KNOW that the lenses are actually quite fine.

The difference in border performance with the Fuji is small, but still seems to be there. What is striking (again?) is the unreliability of the CA measurements of the software you use. More than double with a slight resolution difference.

 

By the way, both Fuji X and Sony CFKAN reviews lack the "excellent/good/fair" and so on bar to give guidance to the results...

[Klaus]

Imatest has nothing to do with the correctness of the CA figures. CAs vary rather heavily between lens samples - I suspect due to variations in centering. The CA readings are taken in all 4 border regions and then "plain" averaged. I reckon we could improve this somewhat by just taking the 2 middle figures while ignoring the extremes.

But yes, CAs are the most unreliable figure we provide.

[Brightcolours]

As I ointed out before, the CA figures give by imatest often make no sense. Sometimes in your samples one can see CA reaching 4 pixels or so, when the imatest figure only says 1 pixels or so.

 

I kinda doubt that CA can vary a lot per lens sample, decentering will have to be very severe for CA way stronger... A shame you don't have uncorrected samples for each sample (the Fuji X and this one), to see the real CA performance.

[thxbb12]

CA can also be very much sensor dependent. I remember using the exact same lens on 2 Pentax bodies: K20 with a Samsung sensor and K10 with a Sony sensor. The K20 had much lower CA with the exact same lens. The Samsung sensor was less prone to CA than the Sony in the K10. I remember reading similar statements in the dpr forum.  It was actually a nice surprise.

[Brightcolours]

CA can also be very much sensor dependent. I remember using the exact same lens on 2 Pentax bodies: K20 with a Samsung sensor and K10 with a Sony sensor. The K20 had much lower CA with the exact same lens. The Samsung sensor was less prone to CA than the Sony in the K10. I remember reading similar statements in the dpr forum.  It was actually a nice surprise.

You must have confused something else with (lateral) CA then, CA is purely lens "based". It is what the lens projects onto the image plane, different colours refracting differently through the optics. Maybe some kind of PF got you confused?

 

CA (or the tell tale sign of CA at least) always comes in colour pairs, each colour on one side of a dark area surrounded by lighter areas. Magenta/green, blue/yellow, red/blue-green. The colours depend on which part of the spectrum gets projected bigger or smaller than the rest of the spectrum. Totally uncorrected lenses can show a bigger mess.

 

The only way one camera can show lower (lateral) CA than another, is when the CA colour edges get removed in-camera, or the camera performs CA correction ((shrinking/expanding the R, G or B channel till they all match). 

[Klaus]

As I ointed out before, the CA figures give by imatest often make no sense. Sometimes in your samples one can see CA reaching 4 pixels or so, when the imatest figure only says 1 pixels or so.

 

I kinda doubt that CA can vary a lot per lens sample, decentering will have to be very severe for CA way stronger... A shame you don't have uncorrected samples for each sample (the Fuji X and this one), to see the real CA performance.

 

Well, the difference between the different test quadrants is visible in the chart images. 

 

However, honestly we aren't talking about magnitudes here. You may notice that the CAs are (just) in the 2nd lane in case of the Sony whereas they are in the 1st lane in the Fuji test. Thus the CA ratings are not worlds apart. Technically we are talking about a difference of 0.4px (0.6px non-normalized) here and please don't come up with the difference in % - that's not helpful regarding the overall significance.