Imaging qualities between X-Trans sensor and Bayer sensor

[PeterZheng]

Imaging qualities between X-Trans sensor and Bayer sensor

Besides holding the 35mm and APS-c DSLR cameras, I also wanted a small-sized type's system camera, so I pay attention to Fuji X-e1, I have no interest in Sony NEX-7 with a sensor 3.8μm. Within next 3 years I will still not like a 3.8μm sensor camera.

I never believed the so-called "Overall Scores", that were some tricks for the amusement of the children. But I do a double clicks on button "screen mode" through them webpage, (pixel by pixel), to watch under "screen mode" the measured chart and data, and not A4 size the print mode and an 8MP Overall Score.

For some camera reviews, I would not swallow the puffery words. Meantime, I find some questions, as you can see pictures below.

Please notice, here is not talking what one software would be better. Most of the puffery words of the camera reviews used the lenses and software are like below, not else.

Some camera review articles and fans have claimed the third party softwares cannot improve X-Trans sensor camera's imaging quality and is likely worse, even more claimed the X-Trans sensor camera's JPG quality is much better.

Most of the puffery words about the X-Trans sensor are based on the Resolution Chart shooting. Sounds like very good, but, I seen some reversed results, like still life photos are below. This is inexplicable.

For easier to see the problems, photos taken cropped and magnified 150 - 600% as follows.

The photos taken ISO 200, aperture f/8

Colour details lost

1. Sony NEX-5n and Fuji X-e1 @ACR vs Camera JPG

2. Sony NEX-5N and Fuji X-e1 @ ACR vs ACR

3. Nikon d5100 and Fuji X-e1 @ ARC vs Camera JPG

5. Fuji X-e1 and Fuji X-pro1

6. Snoy NEX-5n and Fuji X-e1 @ ACR vs Camera JPG adj

DPR testing of the E-mount cameras did not use with the E-Mount 50mm f1.8 OSS lens, but is with the lowest-end A-Mount 50mm f1.8 DT. For other brand's a camera in testing processes DPR has ever had replaced with another lens model and switched of the published photos.

Resolution and Sharpness

7. Sony NEX-5n and Fuji X-e1 @ Res and Sharp

8. Nikon d5100 and Fuji X-e1 @ Res and Sharp

Nikon d5100

In the imaging quality that the d5100 is slightly better than the d7000. The d5100 is 4 fps for better (faster fps to be worse in low-end camera), and slower to run the shutter blades and mirror flipping, all the vibrations are lower than the d7000, the imaging sharpness better.

Some cons in the d5100:
No providing the electronic first shutter curtain mode, (it can be Zero internal vibration, and switchable turned on or off),
A poor OVF is too small,
No AF fine adjustment and storage,
No build-in AF motor,
much more plastic and entry-level built in,
The mirror lock must take the Live-View mode, or open the menu controlling, and the 2sec delay mode must buy the ML-L3 remote device, although cheap.

If Nikon d5300 gets back with a 16MP sensor, and has the electronic first shutter curtain mode, 4 fps, a large OVF, AF fine adjustment and storage, the build-in AF motor, that is better.

9. Sony NEX-6 and Fuji X-e1 @ Res and Sharpness

Sony NEX-6
The imaging quality is very excellent,
The electronic first shutter curtain mode, can be Zero internal vibration, and switchable turned on or off,
The hybrid AF system, despite its on-sensor the built-in PD-AF is not as fast as the mid-level DSLR camera, but it has better performance in all the mirrorless cameras,
22ms shutter lag and was based on the Zero internal vibration,
An OLED XGA EVF built in the body,
Small size and easier carrying.

Cons in Sony NEX-6
The AF fine adjustment storage is merely usable LA-EA2 adaptor cases for A-mount lenses.
The AF fine adjustment storage is neither in the E-mount lenses usable, nor in the hybrid-AF available. Certainly, in Live-View mode all of the DSLR cameras are also not available, but it is a problem existing.
Running the PD-AF 99-points area is not flashing prompt with every one point of the 99-points, the PD-AF 99-points are merely with 9-patchs to do selectable and is in hybrid AF mode, but Sony a99 is flashing prompt every one point in the 102 points.

Sony NEX-6 is most worthy buying one with all the mirrorless cameras at 2012.

The NEX lenses are too few models and have acceptable optical performance with a modest price. The E50mm f1.8 OSS and the E35mm f1.8 OSS are with acceptable optical performance limited, but the E35mm f1.8 OSS is slightly high price, not so desirable. The Sonnar T lenses are unacceptable price on the E-mount camera. The E10-18mm F4 OSS is obviously a too high price. Other Sony E-mount lenses lose the worthy optical performance.

The E-mount lens models are too less and some models in the price and optical performance are also not desirable, they are far not matching. If you used an adapter and other lenses, the AF and AE exposure will be lost, merely the Alpha mount lenses and Sony adapter excepted.

In E-mount cameras the selectable lens models, optical performance and price, the three key points seriously weaken the competitive power on E-mount camera.

But the E-mount lens market situation will get changed. In 2012 already have 3 major camera makers joined up with the 18mm flange focal distance's APS-c lens mount, maybe third party lens maker will be actively in future.
Who is coming in fourthly? Another one major camera maker would join in the 18mm flange focal distance's APS-c lens mount?

Also, if you have old E-mount lens and wanted to suit the NEX-6's built-in PD-AF, the lens firmware needs updated.

P.S.
about the AF-errors and Lens residual spherical aberrations
Move the mouse cursor between marks f1.8 and f4 below to observe the respective AF-errors.
http://www.photozone.de/sony_nex/733-heliar75f18?start=1

The Lens Voigtlander Heliar 75mm f1.8 VM on Sony Nex-7 and takes MF and the CD-AF assist denoted. The AF-errors came from the lenses design conditionality, at any one lens design that is always existing.

Many time actually the true AF-errors are not in a camera body, but actually is since the lenses design conditionality. In a lens design the residual spherical aberrations impossible to become a Zero. Also, most of the lenses designed the spherical aberrations compensating are not in the shortest target distance, except macro lens.
If all of the residual spherical aberrations are compensating became Zero aberration, you will merely get a very bad lens, despite the AF-errors could be inclined Zero, in fact impossible to do.
Must more notice the residual spherical aberrations with every aperture stop to be different, while is the target distance changed then the residual spherical aberrations will be also changing, these are real causation in AF-errors.
Usually, a larger aperture lens easier get AF-errors, the spherical aberrations compensating are more difficult for some larger glass diameter, especially a large aperture wide-angle lens. The zoom lens is much more complicated cases.
What are residual spherical aberrations?
http://www.photozone.de/lens-test-faq

In a PD-AF system's minimum AF-errors is with PD-AF sensor's designed value, which is like the equivalent aperture f5.6 stop, or f4 stop, or f2.8 stop, except the PD-AF related an equivalent aperture value but seems like other aperture values to upward move the AF-errors, actually is lens in the residual spherical aberrations with every aperture stop to be different.

Seems like the CD-AF in the maximal aperture stop of lens reduced AF-error, since it works through maximal aperture of the lens.
However, relative to the maximal aperture stop, the CD-AF system does not ensure other aperture values to also get AF-errors as lesser as maximal aperture stop.
If a lens is maximal aperture f1.4, seems like the CD-AF can f1.4 stop have more accurate AF, but is likely larger AF-error in the f5.6 stop or f2.8 stop.

So, Whether it is PD-AF or CD-AF, the AF fine adjustment and storage to be needed.

If you use Nikon DSLR camera and Nikkor lens, usually you should take with lens aperture f5.6 stop to check the PD-AF-error, not more note f1.4 stop or f1.8 stop.
The target distance should be 3m or 5m or far distance (it based on lens focal length differences), with sun lighting or Incandescent light bulb, not using with fluorescent lamp (the near 650 - 780 nm red spectrums lost), not related to camera brands.

The eos 5d2 PD-AF system has an equivalent f2.8 detection cells, so, you observed the EF 50mm f1.4 USM in f2.8 stop is right AF accuracy, but the f1.4 stop and f5.6 stop have some the observable AF-errors.
http://www.photozone.de/canon_eos_ff/564-canon50f14ff?start=1

Some lens designs also have a little larger of the residual field curvature, so, when shooting a 2D plane target that can cause the focusing discrepancies in between center and side. Usually, people are very less to shoot a 2D plane target, except the chart test's shooting.
Read more,
http://www.photozone.de/canon_eos_ff/528-canon2470f28ff?start=1
What is the field curvature ?
http://www.photozone.de/lens-test-faq

 

[Detail Man]

Regarding the Fuji claims here:

http://www.fujifilm.com/products/digital_cameras/x/fujifilm_x_e1/features/

... note that they are comparing their 6x6 array sensor (with no optical low-pass filter) with a an unspecified Bayer-arrayed 2x2 array sensor of unknown photo-site size ("pixel-pitch") that does have an (unspecified, not characterized) optical low-pass filter. Evidently playing specification-number games ?


Not surprising to see (any) 6x6 arrayed image-sensor show such differences relative to a 2x2 arrayed image-sensor. I would tend to believe my own eyes before believing those Fuji resolution claims ...

 

[PeterZheng]

Yes,

We need to believe our own eyes before believing those Fuji claims

In yesterday, We heard "CaptureOne Fujifilm X-Trans Raw support tested".
Well, we to observe the photos.

do a view original size, clicks on picture below,

CaptureOne7x and Fuji X-pro1

 

[The Davinator]

...you do understand that if you need to use 600% crops to see a difference, you're wasting your time. Even 100% rarely shows in all but the largest prints.

Kind of like telling us how well a Honda Civic corners at 400mp/h.

 

[PeterZheng]

Some crops 600% are only for easier to observe the problems, a NOTE.
Also, the aforesaid pictures are including with 150%, 200%, and cropped 600%, not all with 600%..
some pictures crops 100% can also be seen of the problems. The problem can induce more in depth discuss.

"wasting time" might be merely with you, not others.

Dave Luttmann wrote:

...you do understand that if you need to use 600% crops to see a difference, you're wasting your time. Even 100% rarely shows in all but the largest prints.

Kind of like telling us how well a Honda Civic corners at 400mp/h.

 

[Erik Magnuson]

PeterZheng wrote:

We need to believe our own eyes before believing those Fuji claims

Well to be fair, Fuji's claims would only apply to Fuji's processing so anything C1 v7 does is irrelevant. Anyway, a high-contrast but aliased pattern does result in a higher MTF-50 score so Fuji's claim is not disproven by your example.

 

[The Davinator]

PeterZheng wrote:

Some crops 600% are only for easier to observe the problems, a NOTE.
Also, the aforesaid pictures are including with 150%, 200%, and cropped 600%, not all with 600%..
some pictures crops 100% can also be seen of the problems. The problem can induce more in depth discuss.

"wasting time" might be merely with you, not others.

Wasting time is staring at differences that would never manifest themselves in a print. At this resolution, and even 150% views, we are talking prints well over 60" wide when viewing on screen. So yes, I'd say for the overwhelming majority of people here, what shows in a 60" to 70" print from the Fuji or the Nex is a waste of time....as these differences vanish in prints below 30"....which is where most reside in their printing.

Do you regularly print at 40" x 60" and larger?

Dave Luttmann wrote:

...you do understand that if you need to use 600% crops to see a difference, you're wasting your time. Even 100% rarely shows in all but the largest prints.

Kind of like telling us how well a Honda Civic corners at 400mp/h.

 

[The Davinator]

Erik Magnuson wrote:

PeterZheng wrote:

We need to believe our own eyes before believing those Fuji claims

Well to be fair, Fuji's claims would only apply to Fuji's processing so anything C1 v7 does is irrelevant. Anyway, a high-contrast but aliased pattern does result in a higher MTF-50 score so Fuji's claim is not disproven by your example.

 

[PeterZheng]

Erik Magnuson wrote:

PeterZheng wrote:

We need to believe our own eyes before believing those Fuji claims

Well to be fair, Fuji's claims would only apply to Fuji's processing so anything C1 v7 does is irrelevant. Anyway, a high-contrast but aliased pattern does result in a higher MTF-50 score so Fuji's claim is not disproven by your example.

Thanks, you are right,

In Imatest measurement, that an aliased pattern does a result in a higher MTF-50 score, but it is not same as the general result in colour photos.

 

[PeterZheng]

In Imatest measurement, that an aliased pattern does a result in a higher MTF-50 score, but it is not same as the general result in colour photos.

This is saying Fuji's MTF-50 score, but, the X-Trans sensor itself is not same as the good results in general colour photos, and is unlike others. So, you seen aforesaid examples .

There picture-sized are fair on per camera.

We needed to believe our own eyes before believing those Fuji claims.

 

[The Davinator]

PeterZheng wrote:

In Imatest measurement, that an aliased pattern does a result in a higher MTF-50 score, but it is not same as the general result in colour photos.

This is saying Fuji's MTF-50 score, but, the X-Trans sensor itself is not same as the good results in general colour photos, and is unlike others. So, you seen aforesaid examples .

There picture-sized are fair on per camera.

We needed to believe our own eyes before believing those Fuji claims.

The problem with believing our eyes based upon your samples, is that they are of no relevance in the real world. As I said, even with your screen view of a 150% shot...it's the equivalent of looking at a print well over 60". Your 600% view is like staring up close at a print 25 FEET wide. Fuji I'm sure is more interested in the real world issues of users as opposed to people "testing" what a 25 foot wide print would show....

 

[PeterZheng]

Dave Luttmann wrote:

PeterZheng wrote:

In Imatest measurement, that an aliased pattern does a result in a higher MTF-50 score, but it is not same as the general result in colour photos.

This is saying Fuji's MTF-50 score, but, the X-Trans sensor itself is not same as the good results in general colour photos, and is unlike others. So, you seen aforesaid examples .

There picture-sized are fair on per camera.

We needed to believe our own eyes before believing those Fuji claims.

The problem with believing our eyes based upon your samples, is that they are of no relevance in the real world. As I said, even with your screen view of a 150% shot...it's the equivalent of looking at a print well over 60". Your 600% view is like staring up close at a print 25 FEET wide. Fuji I'm sure is more interested in the real world issues of users as opposed to people "testing" what a 25 foot wide print would show...

.Do you regularly print at 40" x 60" and larger?

you are wrong

Yes, I can now print a photo 60x40-inch with a 16MP NEX-6, and is an acceptable quality, no resizes in software, no digital zoom. In 200 % sizes.

The X-Trans does not have possible, must others also not? I would tell you if needed

I have doubt that why do you not ask Fuji to move up?

you are no wasting time.

 

[Draek]

You're pixel-peeping, at 600%, to JPEGs, converted for the web. There's so many things wrong with that it'd take me the rest of the day to list them all.

Worst part? it's not the only mistake you make; listing lens price as if it were relevant when the cheaper lens in question is a 50mm, a focal length whose optical formula has been refined so much and for so long even used models going for $20 at a thrift shop put many "pro" lenses to shame.

Still, I do find it interesting how aliasing seems to appear on the Fuji as slight staircase artifacts, while the Sony and *specially* the Nikon have notorious amounts of color moire, instead. I'd love to play with some RAW files to test which one's easier to clean, though given the insane magnifications you're peeping at, I'd expect neither would be a problem in real-world shooting.

 

[Erik Magnuson]

PeterZheng wrote:

Yes, I can now print a photo 60x40-inch with a 16MP NEX-6, and is an acceptable quality, no resizes in software, no digital zoom. In 200 % sizes.

There would be some demosacing errors if you looked for them, e.g. there is color moire in your OP 5N/D5100 examples.

The X-Trans does not have possible, must others also not?

It's possible for X-Trans, just not for images where tiny, red, regularly spaced details are important to the photo (which is how many photos?) Different sensor organizations have different strengths/weaknesses w.r.t. pixel peeping detail. If this is what people notice about your 60" print, then you have failed in a much more important way than which

 

[CAcreeks]

I believe the white grid pattern is caused by the luminance (green) pixel pattern in the X-Trans sensor. The loss of color is caused by fewer chroma (red blue) pixels than in Bayer sensors. It's a personal choice whether you are willing to accept these flaws in exchange for wide dynamic range and minimal noise at high ISO.

PeterZheng wrote:

Imaging qualities between X-Trans sensor and Bayer sensor

Colour details lost

1. Sony NEX-5n and Fuji X-e1 @ACR vs Camera JPG

 

[The Davinator]

CAcreeks wrote:

I believe the white grid pattern is caused by the luminance (green) pixel pattern in the X-Trans sensor. The loss of color is caused by fewer chroma (red blue) pixels than in Bayer sensors. It's a personal choice whether you are willing to accept these flaws in exchange for wide dynamic range and minimal noise at high ISO.

PeterZheng wrote:

Imaging qualities between X-Trans sensor and Bayer sensor

Colour details lost

1. Sony NEX-5n and Fuji X-e1 @ACR vs Camera JPG

The color filter matrix has nothing to do with dynamic range.

 

[Jack Hogan]

Some high flying claims in that Fuji page. From what I understand, the Fuji X-Trans sensor aims to do away with the antialiasing filter and make up for it by introducing a pseudo random CFA pattern. If that's the case, the main advantage would be less moire' and a very slightly different sensor Sensitivity rating - SNR and DR really should not substantially enter the equation, all other things being equal - and they seldom are.

Data on the effect of removing an antialiasing filter is available, in these days of the D800 and D800e. PDR is virtually unaffected, SNR only minimally even in the worst of circumstances, unless one thinks they need or can measure these variables with a precision greater than .1 of a stop. Resolution is slightly improved, but since the D800's 4-dot beam splitter AA filter is well behaved, it is not too difficult to recover a portion of this difference with a simple deconvolver: so confirm reports about the D800's ability to 'take' a fair amount of sharpening well. Incidentally, the D800 apparently uses the same type and size of AA filter as the D5100/D7000 mentioned in this thread, so its results should be directly applicable to them as well.

On the other hand, the compromise of a CFA pattern of limited experience is harder/unproven color reconstruction, which can be seen in the examples in this thread and several other threads on the inability of several leading raw converters to de-mosaic its data without smearing or show results that do justice to the resolution claims. Is it a coincidence that many of the images that portend to show X-Trans' supposed resolution advantage are shown in B&W? Perhaps.

Imho the jury is still out on the claims on this sensor. I nevertheless salute Fuji for its technological boldness - it may be just a matter of giving the software writers time to come down the de-mosaicing learning curve.

Jack

 

[CAcreeks]

Dave Luttmann wrote:

The color filter matrix has nothing to do with dynamic range.

Thank you for your very useful contribution to this discussion.

Have you looked at the DPreview test of the X-Pro1, dynamic range page? Do you have a theory why DR is off the charts?