When Fuji launched the X PRO 1 in January the main excitement was for the new sensor. But any time you see some innovation you really need to test it to see if it delivers what it promises. Sandy McGuffog from Chromasoft wrote a couple of articles about the Fuji X PRO 1 sensor. Click here to read part one, part two and part three.
The conclusions are pretty surprising: “The hard truth is that any interpolation engine is going to have trouble with pixels that are spread out, not aligned, and unevenly spaced.” and “So my conclusion is, sorry to say, that the Fuji X-Pro1 X-Trans sensor doesn’t deliver the Fuji promise of outperforming similarly sized sensors. In fact, it underperforms similar DX sensored cameras – with the official SILKPIX raw developer, the underperformance is too slight to be noticeable under normal circumstances, but is still there if you look closely“
Also ReidReviews tested the X PRO 1 sensor. His article can be viewed only by subscribers but in sort he says that at High ISO the files are very clean but also have less color detail.
I am not posting this to denigrate Fuji. Quite the opposite. The Fuji X PRO 1 is a great camera but still…not perfect!
38 Responses
Basically, the higher IQ is from the lack of AA filter? Interesting.
With all due respect considering the whole of work you have done, the 400% crop is not relevant for printing @ 100% or viewing pictures @<100%.
What is relevant is the ability of the sensor to capture both light and shadows. The best raw processor is the one that will convert the data closer to the way that the incamera processor does in terms of contrast, colour curve, sharpness and better noise handling.
For bigger prints, 400% crops or pixel peeping buy a medium format.
Bravo Fuji for the new sensor approach, the hardware will become more powerfull in time and will process faster the information from the sensor.
Fuji needs to protect their jpeg engine ("Fuji colors") and will not share it to anyone.
Again, this is not a medium format camera. It's just a small, convenient one that delivers great image quality.
Admin: Unfortunately you are a couple of weeks late in your reporting on this subject!
Later developments on the subject which is a lot about RAW converter development to demosaic the data from the new sensor. Most converters have problems here, partly because Fuji not being willing to cooperate. ReidReviews and Sandy McGuffoc has lately worked together on this and moved things forward. There is now also a Mac only RAW converter that seems to deliver better.
The basic problem is that there is a longer way between red and blue sensitive pixels with the Fuji sensor, compared to the common Bayer pattern. New algorithms need to be found.
Having said that, I felt that, for a $1700 camera you can expect things like this should be sorted out before sales start. Fuji is the culprit here and camera owners + RAW converter developer are the victims….
I decided NOT to buy Fuji – X100 was a concept camera rather than a production camera, X10 similar and now the X-pro1 is a concept camera.
I decided for a Nikon D800 which is a mature camera and the images from it are so beutiful! My m43 system is at the moment shelved.
I tend to agree with the fact that Fuji should get its act straight, they have great features, IQ and “gestalt” in their cameras, but they fail at delivering a well rounded product – particularly the xpro1 which can be thought as the second iteration of the x100 concept, and even more at the current price tag. They have a great opportunity, the handling is great, the tactile controls are a breadth of fresh air, IQ is there, size and weight is almost perfect…so they are *this* close to a home run, but not quite.
Kyleberg properly observed the problem. Let me to describe it in details.
RAW data is the stream of red, green and blue pixels. The demosaic should recover each one into RGB using neighbour dots. There are 3 choice:
1) We are on green and should recover red and blue components. For some greens (that are on center 2×2 green patch) red and blue are not uniformly distributed and tend to place at one side of the green pixel, so we are obviously losing color information here compared to bayer sensor with uniformly distributed 4 blues and 4 reds around.
2) We are on red and trying to recover green and blue. In this case we have 5 greens and 3 blues, which although distributed uniformly, but they are too rare, compared to 4 greens and 4 reds of bayer color filter.
3) We are on blue recovering green and red. The same as with red.
So, the conclusion is: we lose red and blue color accuracy with fuji sensor. The possible accuracy drop is quite dramatic at the center of 2×2 green patch. We have slightly better green accuracy than with bayer filter (the difference isn’t high due thanks to better uniformity of bayer filter). It’s quite observable, fuji tends to have problems with reds and blues, with high noise and low tonality in these colors. It is also lead to the drop of effective resolution compared to the bayer filter. But due to the fact the green has the highest weight in luminosity formation (60%) we have lower luma noise.
Unfortunately, these problems cannot be resolved with “better algorithms”, we just lose color information (other than green) with this sensor. Again, with 24Mp sony sensor we will have even higher green accuracy than fuji as well as much more precise reds and blues. Due to higher SNR of todays sony 24Mp compared to Sony 16Mp technology (Fuji uses it) we’ll have higher luma noise with 24Mp sensor, but it’s not a problem at base iso.
So, if you want better colors: don’t by Fuji. The so called “fuji colors” are just profiles they used. If you want better details: don’t buy fuji, the 24Mp sensor better preserves them. If you want better low light capability: yes, the fuji is your choice (if you don’t afraid of poor AF in low light with this camera :D).
I think, and seen examples showing that RAW converters can/are improving as developers understand the problem better and better. I am not as pessimistic as Denis. Coming from the IT-world I know that impossible today can be very doable tomorrow! 🙂
BTW: For the price of a X-pro1 camera housing you get a NEX-7 + the Zeiss 24/1.8
@Admin, I think it’s worth looking at the new Canon Rebel T4i / EOS 650D particularly its AF system. It may give hints to the future canon mirrorless and possible ability to work with legacy glass
i wrote that before and all here told me to shut up.
i have testet it and i have not seen any resolution benefits.
So, it’s seems, that IQ-wise Nex-7 is the best enthusiast mirrorless-camera at the moment.
It’s interesting to see what Canon can bring to mirrorless-segment…
I wouldn’t hold my breath though. Their sensor development haven’t been very impressive lately.
The Nex-7 always is tempting, but still the issue with lack of lenses from Sony. Also, on the good lenses even, a question of corner performance with the 24mpx sensor.
I guess it is still an immature market… look forward to the next gen on all of these products.
A web sample from the low resolution and poor color quality X-Pro 1 camera, jpg from raw by RPP 64 (no Chromasoft inside):
http://farm9.staticflickr.com/8157/7164285193_d99ed7ae7e_o.jpg
Another, less aggressive, poor color and resolution shot:
http://farm9.staticflickr.com/8022/7341727386_7aa1fb8d04_b.jpg
I’m pretty sure that this rage against the Fujifilm X-Pro1 is an orchestrated move. Bet it is comes from the DSLR-world, anxious to discover that the mirror less evolution will fade out their range of too expensive, heavy camera’s with even more unaffordable lens systems that still don’t match up with f.i. Leica or Zeiss, and just say it: Fujinon. Fujifilm is dangerous, they have the money to make it even more difficult to those big ones. Certainly when they would collaborate with or even acquire Olympus, there are still a lot of strange rumors about this. It would hurt them all.
Having after a lot of DSLR-stuff the X-Pro1 myself, I must still discover the first proof of chromatic smearing and water-colour effects. When you start zooming into pictures until the pixel-level and would compare the available RAW-convertors, I’m pretty sure other sensors will also show up with odd differences. it’s at a level that doesn’t even interest me.
In reality, this camera delivers an outstanding IQ and color-setting that can easily match up with the best FF-camera’s. Even with a certain German company I will not name. And even more easy, with all the other recent mirrorless concepts. Precisely this is the problem, not the sensor.
That being said, I’ve never seen so much nonsense and troll-writing about a camera as with the X-Pro1.
A toy is a toy is a toy. Is the price a joke?
Go to a camera store with a blank SD card and take a few shots with the Fuji, with the Sony, with the OMD. Go home and look at the files. Look at the jpgs from the FUJI, vs raw files from the other two. If you’re really interested, use the RPP beta and look at the raw Fuji files.
The Fuji files will totally have more resolution that the other two cameras. To be fair, this is as much a result of the lens quality as the sensor quality. RPP has been doing weekly updates to its conversion algorithms, and they keep getting better. This is not a trivial problem (new pixel array conversion) and while I would have loved Fuji to share their algorithm, I can understand their wantig to keep it proprietary.
I’m not saying the Fuji is a better camera than either the Sony or the Olympus. All are great cameras, and each has its own strengths. But to say that the OMD out resolves the Fuji is nonsense, as anyone who looks at the files they take will see.
I have read so much data about the processing of Fuji x pro 1 files and have studied my images and tried just about every combination. Even after looking at Reidreviews & Sandy McGuffog review and blog, I am a bit dismayed. I bought into the whole xpro1 system because the x100 performed so well.
I blame Fuji for not making sure when they designed the sensor that it was going to be able to be addressed by software. It obviously cannot. It is an excellent camera by my standard and for the type of shooting I like to do, but it also an expensive major disappointment because of the problem with software support.
I am fortunate because I also have the new Olympus OM-d system and Pens system.
interesting… the details on the jpeg are really good. i am wondering if the built-in fuji processing engine is a better algorithm? why is it so harder for other companies to come up with an algorithm?
Silkypix development is showing significantly more details than internal JPGs.
They also shows a large amount of artifacts. So, their effective resolution is still lower than even APS-C with weak AA.
I have been saying this for some time now. When you look at controlled test samples (e.g DPReview or Imaging Resource), the X-Pro 1 is very clearly softer and smearing fine detail relative to other cameras with the same nominal resolution. It also still has moiré, just in different places to bayer cameras (and IMHO more objectionable). It does have the lowest noise at high ISO for its pixel density, but this is almost certainly due to RAW NR anyway. All in all, I consider Fuji’s new sensor layout a failure.
Well that comes down to the problem being with the interpolation engine and not the sensor. So you can reinterpret what she is saying.
I work in a camera repair shop, and I see all cameras all the time, and I see very quickly which ones stand out from the crowd. I can tell you difinitively that the Fuji X-Pro 1 images rival that of anything out there on the market today 35 full frame or even medium format, -the only thing it fails against is slow speed film.
These technical hesitations about the sensor akin to masturbation of the brain. 99% of photographers will not see the differences reported because they may not show much interest in terms of practical results in an average scenario. These considerations are relevant only for those interested in the pure technology of photography, but those who say: “I will not buy the X-Pro 1 for this reason” seem to suffer from a kind of obsessive compulsive disorder. If the choice of a purchase is made on such criteria, then do not buy anything! Because no gear is perfect! Or, buy an Alpa 12, 80MP digital back and Rodenstock / Schneider lenses (and even then, I guess, well there a way to find something …). I may shock some, but I taught the technology of photography for years and I think ultimately, after work, both with 35mm as Sinar p2, with film as a digital camera, that the best gear is always the one that makes you hard, the one you kick for, the one that will you leave your home in bad weather, the one that will open your eyes to the world … And often, this is a device of good quality, reasonnable range of lenses, settings well placed and intuitive, and excellent image quality for the use you make of it … Following on what I know and what I’ve read and seen, the X-Pro 1 is one of these wonderful devices.
Pot calling the kettle black, Jean-Michel.
You old “pro(phet)s” seem to have lots of obsessive clinging to analog mechanical film era of 50 years ago.
This kind differences in sensor become fast fully irrelevant for choosing camera when comparing other aspects of X-Pro 1 to other camera systems around.
For starters something entirely else than number of pixels or any tech stuffed inside covers, comfort in use: For its bigger size than other mirrorless bodies I was expecting it to feel somewhat better than others but when trying in hand it’s as comfortable as brick/piece of wooden plank.
But of course horse wagon is horse wagon and you can’t expect features of car from it. Then again I can’t claim any of mirrorless systems to be balanced on that aspect, differently decorated microcars don’t form full range of cars.
For controls analog fixed function dials are far from good way to get the most effective overall control layout for digital camera. And that WW2 era positioning of them was caused by limitations of that analog mechanical tech.
And three primes reasonable range of lenses? You’ve got one narrowminded view of photography.
But then non-TTL viewfinders don’t even work well with flexible range of various type lenses.
I am totally with Jean Michel. To me, the fact that an expert is not able to create the right algorithm for a camera (in this case the Fuji X-pro 1) is just an indication that he should try harder.
I used to have quite good Canon gear. I now own a Fuji system. The difference is stunning. Really.
It’s impossible. Their filter makes it impossible. They force to make guesses, due to lack of information, “right” algorithm just doesn’t exist. These guesses are not always right. So there are detail and color rendition artifacts. And it looks like their “interpolation” uses kind of pattern recognition and that’s why it doesn’t look as shitty as raw. But man — with such a fake approach they can apply this to conventional bayer filter and they’ll get “stunning difference” too.
Hey Guys…it is not quite that straight forward. Try using individual cells of R OR B OR G to determine luminosity and then run a color add/subtract on the whole array for color values. Then combine.
Hint for R&B: Think triangle, then rectangle, then two diagonal passes for luminosity of Red and for Blue. Green is just two diagonal passes. Math many of you have never seen before, btw.
Jeez…
It’s obvious there is no perfect camera.
It’s just plain stupidity to post something just to prove that.