X-Pro 1 reviewed at DPReview

[semilog]

Really?...

Many opinions here, relatively few facts. You're entitled to the opinions.

 

[Aristophanes]

Many opinions here, relatively few facts. You're entitled to the opinions.

The DP Review tests are designed to be factual data points. Their motive is journalistic and edited as such.

Do you not consider the findings of their sample shots to be factual?

The Fuji sensor at pixel level smears reds into whites, and generally mixes colours.

The Sony sensor does not, and in cameras far less expensive.

The price of the camera is also objective, but the cost to each individual is relative.

Each person can therefore make a subjective decision whether or not this is an $1,800 sensor.

Then balance all the facts about VF, bulk, lenses, etc.

That's the raisond'etre of DP Review whose editorial opinion is under scrutiny here.

 

[semilog]

Each person can therefore make a subjective decision whether or not this is an $1,800 sensor.

At work, I have paid $30,000 for a sensor -- a digital camera without optics or image processing pipeline or finder or autofucus or illumination systems. Those items can easily add another $100,000.

For pictorial photography, I have never purchased a sensor -- only complete cameras with lens mounts, finders, imaging pipelines, and other integrated systems. I use them to take photographs. A 5DIII is no more a $3000 sensor than an M6 is a roll of Tri-X.

The motive is largely artistic, not technical, and I evaluate cameras as whole systems, not as collections of component parts. The question is whether they are good tools for my photography. Currently I use an M6, an X-Pro1, and a several others. They all have strengths and weaknesses.

Perhaps your avocation is different than mine.

 

[semilog]

The Fuji sensor at pixel level smears reds into whites, and generally mixes colours.

How do you know that? Have you taken the trouble to look at the sensor-level data without an intermediary RAW developer or JPEG pipeline? If not, you have no idea whether you're looking at the characteristics of a sensor, or the characteristics of the downstream image processing pipeline (firmware + RAW development software).

The X-Pro sensor uses a new mosaic pattern, and hence the de-mosaicing algorithms for this sensor are still in development. In contrast the comparable algorithms for Beyer arrays are rather mature and settled, since the Beyer array has been around for many years now. It will be at least a couple of years before one can reasonably draw the sorts of conclusions that you seem to think you are drawing for the new array that Fuji is using.

 

[back alley]

seems like an endless argument...

 

[Aristophanes]

How do you know that? Have you taken the trouble to look at the sensor-level data without an intermediary RAW developer or JPEG pipeline? If not, you have no idea whether you're looking at the characteristics of a sensor, or the characteristics of the downstream image processing pipeline (firmware + RAW development software).

The X-Pro sensor uses a new mosaic pattern, and hence the de-mosaicing algorithms for this sensor are still in development. In contrast the comparable algorithms for Beyer arrays are rather mature and settled, since the Beyer array has been around for many years now. It will be at least a couple of years before one can reasonably draw the sorts of conclusions that you seem to think you are drawing for the new array that Fuji is using.

I simply downloaded Fuji's RAW data which all of us work with, including DP Review (and all current owners of the camera). The JPEG's show the same lack of colour fidelity as the RAW. So at least one level of processing error has been eliminated.

"Still in development" is getting close to being an apologist. Do you have access to Fuji's binary code? If not, your argument? That the colours are smearing because an algorithm was not ready for retail? Fuji's SuperCCD's demonstrated similar effects. Yes, I am inferring it's a sensor issue, but I would think that for a flagship product Fuji would have not let orange become grey, and white pink if there was a software fix, firmware or RAW. DPR used what Fuji gave them.

It's Fuji's complete product and their hardware/software chain towards IQ can't tell red from white as posted on the biggest online review site. I simply pointed to the evidence via journalistic testing and consumer evaluation. The market can decide through analyzing the samples provided.

The discussion at point is not whether Fuji will upgrade their firmware, it's about the DPR findings. If you don't pixel-peep because you are happy with the artistry, great. But if you do, then it looks like this sensor compromises colour where others do not. My concern is how this smearing is magnified in post. It sounds like you see the smearing as well and it is not an issue for you, and you hope some future firmware will correct it. We'll see if people's trust in Fuji prevails.

I really wanted to like this camera, but I think it's got issues, particularly at the price point. AF and sensor fidelity are concerns, both of which lag far less expensive cameras. The pent-up demand for an RF-like digital should not overwhelm competent analysis with the product as it is now at the price offered (and I think DPR gave it more credit than due in part because to its novel approach; they underweighted AF and IQ issues and overweighted the niche market idiosyncrasies). If you are personally willing to overlook these flaws for other reasons, or willing for Fuji to beta test on its early adopters, then that's your market choice.

 

[semilog]

"Still in development" is getting close to being an apologist. Do you have access to Fuji's binary code? If not, your argument? That the colours are smearing because an algorithm was not ready for retail? Fuji's SuperCCD's demonstrated similar effects. Yes, I am inferring it's a sensor issue, but I would think that for a flagship product Fuji would have not let orange become grey, and white pink if there was a software fix, firmware or RAW. DPR used what Fuji gave them.

And Fuji may not yet know how best to demosaic this pattern.

Beyer demosaicing is better now — much better — than it was five years ago, or ten. Beyer files developed with LR4 look different and considerably better than the exact same files did with LR1. And not everyone takes the same approach. RPP64 files look different yet again. We are looking at the VERY FIRST commercial product with the new Fuji array pattern, and your experience with SuperCCDs (these were Beyer CCD's, not random-array CMOS devices) is irrelevant to the present discussion, since we are dealing with both a wholly different electronic design and a different array pattern. What is relevant is that we're dealing with 1.0 versions of the demosaicing algorithms and there is clear and obvious room for improvement.

If you've been following the rapid progress with the RPP64 algorithm, this is evident. The current RPP64 algorithm for the X-Pro sensor is far from perfect but worlds better than the other engines available — particularly ACR, which for the X-pro gives genuinely poor, really disappointing results on many files, and was clearly rushed to market. And the great progress with RPP64 has happened just over the last few weeks, driven by a single small software developer.

Unless you are looking at the raw data stream at the pixel level (NOT demosaiced files) you simply cannot say much about the inherent color discrimination capabilities of the sensor, particularly at the sub-array level of resolution. When you do this you are totally dependent on the demosaicing algorithm used. As I suspect you realize.

In terms of color accuracy over larger areas of the sensor, it appears to be reasonably competent, at least with the camera's in-built JPEG engine. This suggests that the sensor's intrinsic capabilities are satisfactory (note that the tests at Imaging Resource provide a lot more quantitative detail than those at DPR).

How good? Time will tell.

 

[GaryLH]

Semilog

I have been using jpg from the camera, is RPP64 now better than jpgs straight from the camera?

Thanks
Gary

 

[semilog]

Depends on what you're looking for. I like the JPEGs from the camera for most purposes, but RPP64 provides good control and nice rendering of fine detail. It's particularly useful when exposure or WB is not bang-on or a lot of tonal adjustment is required. For these files I'm developing in RPP64 to yield a LAB .tiff and then sharpening and doing final tonal adjustment in LR4.


DSCF0257-2 by Semilog, on Flickr

 

[GaryLH]

Thanks for the update Semilog. Looks like I need to give RPP64 a try.

Cheers
Gary

 

[semilog]

Thanks for the update Semilog. Looks like I need to give RPP64 a try.

Note that I've been using beta versions. Links for download are in this thread.

 

[icaro2007]

The most recent beta version is available here
http://www.raw-photo-processor.com/RPP/RPP64_1576Beta.zip

 

[GaryLH]

Thanks for the links.

Gary

 

[gavinlg]

Really?

Sensor: ISO 800, for middle of the spectrum on the referenced DP Review samples:

Fuji X-Pro 1

Nikon D7000

Pentax K-5 (for good measure)

The Fuji sensor has great resolution and low overall chroma, lower than the other 2 models by a hair, but that proprietary mosaic demonstrates issues with colour fidelity, especially along narrow boundaries. Look at the word "Fujitsu". The Fuji sensor clearly bleeds into whites whereas the Sony sensor does not; those lines are discrete as they should be. Where the text should be bright white, it is pink on the Fuji samples.

There are other places on the DP Review shots where the Fuji sensor similarly bleeds, such as where it says "Tin Light" and "Mr. Robot" on the little Robot. The reds are muddy with blues on the Fuji, not at all on the Sony sensor.

In fact, the Fuji bleeds everywhere. Look at the transition along the perpendicular axis of the battery. On the Fuji the orange line distinctly changes from orange to grey. Not on the Sony sensor. The Fuji sensor has proximity issues.

But that's the way it has always been with Fuji sensors. The S5 was designed to smooth skin tones, and for all the low-light performance of the SuperCD series, they achieved that with some smearing effects, just as seen here.

The colour fidelity of the Sony sensor looks to be better at retaining the DR. Go up ISO and the difference increases in favour of the Sony chip. Look anywhere there is text or fine lines and the Fuji bleeds. I'd say that's a function of the sensor design itself.

Overall I'd also say the Sony sensor in either Pentax or Nikon is a modest 1/4 stop better in the shadows. But that's where the inferior chroma arises. Every sensor has its compromises.

Both are excellent sensors, but I value having whites white, not pink, because overall chroma can be handled beer in PP, but one a white has gone pink at the pixel level, I'm not getting it back unless I edit each pixel 😡 I also don't like line edge colour shifts. No amount of PP is getting those back.

I'm a big time Fuji fanboy currently sporting 3 models of their product line, but I'm not looking at a sensor with the X-Pro 1 that outpaces the competition, certainly not at $1,800. I'll leave the FF or 1" sensor comparisons out of this. It's not a class-leading sensor, but one that keeps pace with other APS-C cameras if you can deal with non-correctable colour bleeds and shifts.

So, no, nothing wrong with my eyes.

The major knock against the Fuji is poor AF for an AF-biased $1,800 camera (its MF is worse than some entry-level DSLR's). AF-biased cameras at less than 1/3 the price focus much faster be it DSLR or the V1. The sensor doesn't make up the difference with the compromises noted above, and pretty much every manufacturer puts out super-sharp primes. So the Fuji X-Pro 1 premium appears primarily to be for a unique VF experience, lots of old school control, and a modestly smaller kit.

Just an honest criticism using the same DP Review material and on track with the OP.

I can barely even see those 'color bleeds' in the crops you posted at 100% view, let alone in a web print or an actual print where you will never ever see them - I guarantee that much. The RRP raw processor has no color bleeding which shows that it's just the algorithm, not the sensor. It's basically a non-issue anyway unless you print little squares of 100% crops and look at them with a magnifying glass. The k-5 and d7000 are definitely noisier at every iso in both luminance and chroma.

 

[jsrockit]

As a X-Pro1 user (and X-1000 user), the AF is not fast with the 35mm 1.4 lens. If I try to use AF on the streets with moving people, it IS too slow at times. Additionally, I have had times where the AF has missed even with the corrected frame. The AF in both cams is easily tricked by reflections (in windows for example)...more so than any other camera I have or have used. I'm just being honest. I still love both cameras though...overall my best cameras.

 

[gdi]

I am glad the OP posted this. I haven't visited DPReview in a while; I have spent time checking it out over the last couple of days. I had been considering getting either an XPRo-1 or OM-D, so it was interesting to do the side by side compare of image quality.

I am very impressed with the Fuji at at the really high ISOs; it looks like by about ISO 1600 the chroma noise of the Oly is reaching a point where the Fuji looks better, even with less detail. After 1600 the Fuji shots are clearly usable and the Oly is probably too noisy. But it seems the Fuji may have gone a little heavy handed with the NR for low ISOs; maybe there is a better raw converter out there?

So, I decided to get the OM-D and ordered one yesterday. The image quality looks to be so close that the deciding factors were my existing m4/3s lenses (and the growing options to choose from), the AF speed, and the significant price difference. Now I'll see how impressed I am when the Oly arrives...

 

[gavinlg]

I am very impressed with the Fuji at at the really high ISOs; it looks like by about ISO 1600 the chroma noise of the Oly is reaching a point where the Fuji looks better, even with less detail. After 1600 the Fuji shots are clearly usable and the Oly is probably too noisy. But it seems the Fuji may have gone a little heavy handed with the NR for low ISOs; maybe there is a better raw converter out there?

It's not noise reduction that makes them look so smooth - it's a characteristic of the x-trans sensor with its random color array. Despite the low noise characteristic, it resolves the same detail as a 25ish MP full frame camera, so it's not 'smoothening' as it were.

 

[gdi]

It's not noise reduction that makes them look so smooth - it's a characteristic of the x-trans sensor with its random color array. Despite the low noise characteristic, it resolves the same detail as a 25ish MP full frame camera, so it's not 'smoothening' as it were.

Humm... where do you see that it matches the detail of a 25MP FF camera? Is there a good comparison review out there? (There may be, I don't spend a huge amount of time looking at reviews and probably wouldn't know about it.)

I do see that it doesn't have as much detail as a lot of other cameras in the DPReview shots, and they do indicate that "the X-Trans CMOS sensor includes a degree of chroma NR".

Thanks

 

[retow]

It's not noise reduction that makes them look so smooth - it's a characteristic of the x-trans sensor with its random color array. Despite the low noise characteristic, it resolves the same detail as a 25ish MP full frame camera, so it's not 'smoothening' as it were.

My screen an eyes tell me that it resolves less details than a crop sensor 24MP NEX 7.

 

[semilog]

My screen an eyes tell me that it resolves less details than a crop sensor 24MP NEX 7.

In theory the difference should be ~22% in linear resolution — a rather minor difference. However, the folks at DP review say that in practice, they're about the same. The images I've seen corroborate that. Bottom line is that they're both superior APS-C sensors, albeit with slightly different strengths and weaknesses. I'd have no problem with an X-Pro1 body incorporating the NEX-7, or vice versa.

The differences in imaging performance are much smaller than operational differences between these cameras that actually make a difference for practical photography. One might say the same of the comparison between the NEX-7 and the Olympus EM-5, or for that matter between any of these cameras and a D700.