Fuji X-T2 announced

[Les Olson]

That's an interesting statement, but I don't understand what you mean.
Would you care to expand the concept?

The weaknesses of the mirrorless design in its current incarnations are not intrinsic to the design but come from the obsession with keeping the cameras small and light. 

One drawback compared to the SLR design is higher power consumption, so that for a given battery size mirrorless cameras get fewer shots per charge than SLRs (the XT2, eg, gets 340 shots per 1250 mAh while the D7200 gets 1110 shots out of 2400mAh).  So mirrorless cameras ought to have bigger batteries than SLRs, but to keep them small and light they have been given smaller batteries.  Fuji has given the XT2 an optional battery grip with two extra batteries that allows you to get 1000 shots per charge - still less than an SLR and the camera is no longer small and light. 

Another weakness of the mirrorless design is the short flange focal distance that mirrorless cameras have in order to remain small.  There are two aspects to that.  The first is that digital sensors are unhappy when light strikes them obliquely.  Purely by chance, the long flange focal distance imposed by the mirror of the SLR design is much better for digital sensors.  The second is that the short flange focal distance means that the angle between the periphery of the sensor and the back of the lens is large - especially with larger sensors.  The result is severe peripheral light fall off, which is dealt with by automatic correction and in Sony's case but not (apparently) Fuji's by designing barrel distortion into the lenses, which then also has to be automatically corrected.  This is not ideal for image quality.

A third issue that is not really a weakness of the mirrorless concept but is a contradiction of the design philosophy is sensor-based image stabilisation.

Of those, the only one that Fuji has addressed is stabilisation.   

[Frank Fremerey]

I wish more cameras followed Fuji's classic button design with all the dedicated dials clearly marked on top.
One look is enough to see which lens is being used, at which aperture, shutter speed, ISO, etc.
I like the button programmability of my Sony's but basic functions should just have dedicated buttons or dials like on the Fuji or Df cameras IMHO.

I could not agree more with you, JA!
Here is my new old kiddo, following the same philosophy (ISO, Time, Aperture, Distance):

[Eddie Draaisma]

It can be argued that the weakest spot in the Fujifilm X cameras is the X-Trans colour filter layout. Very nice results (wrt colour rendition) can be obtained (eg with PhotoNinja) but not so perfect results ("colour starved") with ACR/LR, which today simply is the industry standard for raw conversion. It means that most Fujifilm X users will use that, they might not even realise that better results can be obtained, and they will never switch to another converter. In fact, they might have been better off with a Bayer sensor.

[stenrasmussen]

It can be argued that the weakest spot in the Fujifilm X cameras is the X-Trans colour filter layout. Very nice results (wrt colour rendition) can be obtained (eg with PhotoNinja) but not so perfect results ("colour starved") with ACR/LR, which today simply is the industry standard for raw conversion. It means that most Fujifilm X users will use that, they might not even realise that better results can be obtained, and they will never switch to another converter. In fact, they might have been better off with a Bayer sensor.

I agree that ACR/LR is poor wrt. conversion of RAF's. I have tested most raw conversion softwares and found Raw Therapee to be superior in demosaic'ing the X-trans pattern.

[Eddie Draaisma]

I agree that ACR/LR is poor wrt. conversion of RAF's. I have tested most raw conversion softwares and found Raw Therapee to be superior in demosaic'ing the X-trans pattern.

Thanks for the tip Sten, I will give Raw Therapee a try too

[Bjørn Rørslett]

Raw Therapee is a mixed bag of surprises. Some good, some not so good. The speed is glacial and stability of the software is not the aspect the developers have paid most attention to in my experience (Win7/10). However, on the other hand, new versions are rapidly forthcoming thus there is always hope one of them works well enough. If one materialises, I'll set up a Linux box for it.

Apparently Raw Therapee 'understands' X-type sensors as it has its own RAW module for X with lots of parameters to experiment with. I'm using RT in order to extract the most from my S3/S5 Pro RAFs, and hope lingers on still.

[pluton]

Another weakness of the mirrorless design is the short flange focal distance that mirrorless cameras have in order to remain small.  There are two aspects to that.  The first is that digital sensors are unhappy when light strikes them obliquely.  Purely by chance, the long flange focal distance imposed by the mirror of the SLR design is much better for digital sensors.  The second is that the short flange focal distance means that the angle between the periphery of the sensor and the back of the lens is large - especially with larger sensors.  The result is severe peripheral light fall off, which is dealt with by automatic correction and in Sony's case but not (apparently) Fuji's by designing barrel distortion into the lenses, which then also has to be automatically corrected.  This is not ideal for image quality.

Based on the Fuji XR 14mm/2.8 and 23mm/1.4 lenses I've got, it is clear to me that the very short flange-to-focus distance that Fuji uses has not impeded them from designing distortion-free wide prime lenses featuring normal/fully acceptable amounts of light falloff at the periphery.  So with the APSC-sized target at least, it can be done.

[pluton]

One look is enough to see which lens is being used, at which aperture, shutter speed, ISO, etc.

This is the key idea.

[bjornthun]

Based on the Fuji XR 14mm/2.8 and 23mm/1.4 lenses I've got, it is clear to me that the very short flange-to-focus distance that Fuji uses has not impeded them from designing distortion-free wide prime lenses featuring normal/fully acceptable amounts of light falloff at the periphery.  So with the APSC-sized target at least, it can be done.

I use Sony and have the Sony/Zeiss 16-35/4 and the Zeiss Batis 25/2, and both function very well.

Indeed, all of the wide angle lenses designed natively for mirrorless (Fuji, Sony, Olympus, Panasonic) are retrofocus, but with short back focal distances. Since they are retrofocus, problems with periphersl vignetting and smearing are avoided. Though the natively designed wide angles have a short back focal distance, they are designed with the rear exit pupil suffficiently far away from the sensor, thus avoiding the problem of light hitting the sensor at a too oblique angle. This means that they are designed differently from Leica M compatible rangefinder wide angles.

By looking through a Zeiss Batis 25/2 Distagon, we can see that diameter of the rear exit pupil is much larger than the diameter of the front entry pupil. This is the tell tale sign that we are dealing with a retrofocus design and not a symmetrical wide angle design. Yet the rear lens element is closer to the sensor than what is possible with a DSLR. "Distagon" by the way is Zeiss' label for retrofocus wide angles. The Zeiss Touit 12/2.8 Distagon is made for both Sony and Fuji APS-C mirrorless, and is also a retrofocus type with a short back focal distance.

Specifically, a long back focal distance (distance from the rear lens element to the sensor) is NOT required to make a lens retrofocus. It's only required to make the lens fit onto a DSLR.

[Hugh_3170]

There is an aftermarket sensor filter for the 135/24x36mm format Sony's that is specifically designed to enable the Leica M and Contax lenses to better function on Sony cameras so equipped - the peripheral vignetting and smearing issues of such lenses on Sony cameras are significantly reduced when this sensor filter is used.


eras

I use Sony and have the Sony/Zeiss 16-35/4 and the Zeiss Batis 25/2, and both function very well.

Indeed, all of the wide angle lenses designed natively for mirrorless (Fuji, Sony, Olympus, Panasonic) are retrofocus, but with short back focal distances. Since they are retrofocus, problems with periphersl vignetting and smearing are avoided. Though the natively designed wide angles have a short back focal distance, they are designed with the rear exit pupil suffficiently far away from the sensor, thus avoiding the problem of light hitting the sensor at a too oblique angle. This means that they are designed differently from Leica M compatible rangefinder wide angles.

By looking through a Zeiss Batis 25/2 Distagon, we can see that diameter of the rear exit pupil is much larger than the diameter of the front entry pupil. This is the tell tale sign that we are dealing with a retrofocus design and not a symmetrical wide angle design. Yet the rear lens element is closer to the sensor than what is possible with a DSLR. "Distagon" by the way is Zeiss' label for retrofocus wide angles. The Zeiss Touit 12/2.8 Distagon is made for both Sony and Fuji APS-C mirrorless, and is also a retrofocus type with a short back focal distance.

Specifically, a long back focal distance (distance from the rear lens element to the sensor) is NOT required to make a lens retrofocus. It's only required to make the lens fit onto a DSLR.

[bjornthun]

There is an aftermarket sensor filter for the 135/24x36mm format Sony's that is specifically designed to enable the Leica M and Contax lenses to better function on Sony cameras so equipped - the peripheral vignetting and smearing issues of such lenses on Sony cameras are significantly reduced when this sensor filter is used.


eras

Yes, that's right. It's offered by Kolari Vision.

[Les Olson]

Based on the Fuji XR 14mm/2.8 and 23mm/1.4 lenses I've got, it is clear to me that the very short flange-to-focus distance that Fuji uses has not impeded them from designing distortion-free wide prime lenses featuring normal/fully acceptable amounts of light falloff at the periphery.  So with the APSC-sized target at least, it can be done.

Maybe.  A retrofocus lens must have barrel distortion (and a telephoto lens must have pincushion distortion).  Correction cannot be complete in an asymmetrical lens, and prime lenses in the 20mm (35mm equivalent) range typically have around 2% barrel distortion (Zeiss 12/2.8 2%, Zeiss 21/2.8 1.7%, Sigma 20/1.4 2.4%, Canon 20/2.8 1.9%, Nikon 20/1.8 1.6%, Nikon 20/2.8 2.3%, Pentax 21/3.2 2%).  According to Photozone the Fuji 14/2.8 has 0.4% measured distortion and the 23/1.4 less than 0.3%, in both cases the same in RAW and JPEG.  Those are incredibly (literally) low values.  There has to be a suspicion that they are correcting the distortion in RAW. 

The 14/2.8 has 2.4 stops of light fall-off at f/2.8 and 1.8 stops at f/4. There are no direct comparisons but the Nikon 20/1.8, eg, has, on FX, 1.2 stops of light fall-off at f/2.8 and the Nikon 12-24/4 has 1 stop of light fall-off at 12mm and f/4.  The Fuji 23/1.4 has just over 2 stops of light fall-off at f/1.4.  The Nikon 24/1.4 has 0.9 stops of light fall-off on DX at f/1.4. 

Of course, everyone decides for themselves what is acceptable, but it is clear that the mirrorless user pays a penalty.   

[Eddie Draaisma]

According to Photozone the Fuji 14/2.8 has 0.4% measured distortion and the 23/1.4 less than 0.3%, in both cases the same in RAW and JPEG.  Those are incredibly (literally) low values.  There has to be a suspicion that they are correcting the distortion in RAW.   

The XF 35/2WR has significant barrel distortion; it is NOT corrected in RAW.

[bjornthun]

Maybe.  A retrofocus lens must have barrel distortion (and a telephoto lens must have pincushion distortion).  Correction cannot be complete in an asymmetrical lens, and prime lenses in the 20mm (35mm equivalent) range typically have around 2% barrel distortion (Zeiss 12/2.8 2%, Zeiss 21/2.8 1.7%, Sigma 20/1.4 2.4%, Canon 20/2.8 1.9%, Nikon 20/1.8 1.6%, Nikon 20/2.8 2.3%, Pentax 21/3.2 2%).  According to Photozone the Fuji 14/2.8 has 0.4% measured distortion and the 23/1.4 less than 0.3%, in both cases the same in RAW and JPEG.  Those are incredibly (literally) low values.  There has to be a suspicion that they are correcting the distortion in RAW. 

The 14/2.8 has 2.4 stops of light fall-off at f/2.8 and 1.8 stops at f/4. There are no direct comparisons but the Nikon 20/1.8, eg, has, on FX, 1.2 stops of light fall-off at f/2.8 and the Nikon 12-24/4 has 1 stop of light fall-off at 12mm and f/4.  The Fuji 23/1.4 has just over 2 stops of light fall-off at f/1.4.  The Nikon 24/1.4 has 0.9 stops of light fall-off on DX at f/1.4. 

Of course, everyone decides for themselves what is acceptable, but it is clear that the mirrorless user pays a penalty.   

Let's compare these figures to two Nikon mount 20/21mm lenses. The Nikon AF Nikkor 20/2.8 has a vignetting of 2.26 EV at f/2.8 and the Zeiss ZF.2 21/2.8 has a vignetting of 1.99 EV at f/2.8 which is less than 0.5 EV difference from the Fuji 14/2.8 at f/2.8. It is well known that all lenses will vignette less as you stop them down, so comparing a lens that starts at f/1.8 to one that starts at f/2.8 may be a skewed comparison.

Let's consider distortion, and compare the Zeiss Loxia 21/2.8 for mirrorless Sony and the Zeiss Milvus/ZF.2/ZE 21/2.8 for DSLRs. Both are retrofocus Distagons.

Zeiss Loxia 21/2.8: http://www.zeiss.de/content/dam/Photography/new/pdf/en/downloadcenter/datasheets_loxia/loxia_2821.pdf

Zeiss Milvus/ZF.2/ZE 21/2.8: http://www.zeiss.de/content/dam/Photography/new/pdf/en/downloadcenter/datasheets_milvus/milvus2821.pdf

From the graphs for distortion, we see both lenses have very similar, about 2% barrel distortion without any digital correction applied.

The graph for vignetting indicates approx. -2 EV for the Milvus/ZE/ZF.2 (DSLR) lens. The graph for the Loxia (mirrorless) indicates less, but I think that one is digitally corrected.

Both the mirrorless and the DSLR Zeiss 21/2.8 show very similar quality MTF graphs, though the mirrorless lens has a more gentle curve drop off toward the corners, where the DSLR lens drops more abruptly near the edge.

The takeaway for Zeiss is that distortion is comparable for the DSLR and mirrorless lenses.For the Fuji 14/2.8 and corresponding full frame DSLR Nikon 20/2.8 and DSLR Zeiss 21/2.8 the vignetting is actually comparable rangin from 1.99 EV to 2.4 EV. My guess for the Zeiss Loxia 21/2.8 is an uncorrected vignetting between 2-2.5 EV.

In summary it appears that little, if anything is given up with mirrorless wide angles. In the case of Zeiss, the mirrorless Loxia 21/2.8 weighs 398 grams, about half of it's DSLR sibling, so some weight is lost.

[Les Olson]

The XF 35/2WR has significant barrel distortion; it is NOT corrected in RAW.

Digging further, it appears Fuji includes lens correction parameters in all its RAW files.  Lightroom and ACR use that information to automatically correct the RAW files, and as I understand it you can't turn that off (which would account for Photozone's results).  Some other converters do allow you to turn the correction off. 

According to Imaging Resource the Fuji 35/2 does not have significant distortion in the uncorrected files (http://www.imaging-resource.com/lenses/fujinon/xf-35mm-f2-r-wr/review/). 

The 35/2 and the 35/1.4 are both quasi-symmetrical designs (http://www.fujifilm.com/products/digital_cameras/x/fujinon_lens_xf35mmf2_r_wr/specifications/ and http://www.fujifilm.com/products/digital_cameras/x/fujinon_lens_xf35mmf14_r/specifications/) - as you would expect a "normal" lens to be - so it would be surprising if they did have much distortion.