-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
Rumours say X-H2 is Xtrans
- Thread starter HatWearingFool
- Start date
GreatOceanSoftware
Senior Member
I think Fujifilm has carved themselves a nice niche and they’re not going to abandon X-Trans to be like other manufacturers or be (better) compatible with Adobe’s interpretation of its raw files.
From Imaging Resource, referring to the 4th generation…
”X-Trans technology is fundamental to Fujifilm's strategy. They view image quality as a key differentiator for their cameras and have invested a huge amount in developing their own sensor technologies in pursuit of it. X-Trans is just the latest iteration in that process, one that was planned years before the processor technology would arrive to make it feasible.”
I think the Fujifilm magic that I enjoy so much is comes from the sum of their components, of which X-Trans is a major contributor. I think their adventures into Bayer in certain models was a marketing test, good deal on sensors, or some other compelling reason to experiment. And I think those experiments have informed the decisions they’re making now. And they’re sticking with X-Trans.
Here is a link to the full article…
https://www.imaging-resource.com/news/2020/04/20/fujifilm-x-trans-is-it-really-all-that-different
More 2c from the peanut gallery. But I am super satisfied with my XT3 and XE4. I wouldn’t change too much, and certainly not the heart of the camera.
--
Randy
Last edited:
michaeladawson
Forum Pro
Yes and no. I didn't mean to be taken literally that it can't be done. Just that it's more difficult and I would think never implemented by Fujifilm. But let's look at your possible method.
So you make 6 shifts (5 actually) of one pixel, either down a column or across a row. When you combine all the images you get alternating pixels with different ratios of R, G, and B. One pixel will have 4 green, 1 blue, and 1 red. The next pixel will have 2 green, 2 blue, and 2 red. And so on.
So the raw converter needs to apply a different channel gains to the alternating pixels to get back the right color.
This contrasts with Bayer where you need only 4 shots in total and every pixel is composed of 2 green, 1 blue, and 1 red. So the same gain can be applied to all pixels.
Is this a big deal? I don't know. But it takes 50% more captures, 6 vs. 4, and the algorithm is a little more complicated to demosaic the raw image. Does the 4:1 oversampling of green in X-Trans make any material IQ difference to the 2:1 ratio of Bayer? Again, I don't know.
timwit
Leading Member
I personally can't take that article seriously, since it was sponsored by Fujifilm.
Tim C.
Last edited:
GreatOceanSoftware
Senior Member
Even better. I think it sums up a good explanation of their design thinking and intentions.
So if XTrans is the greatest thing since canned beer - why does Fuji uses Bayer in the GFX? Hmmmmmm…..?
GreatOceanSoftware
Senior Member
I don’t know. I think you’d have to be a fly on the wall at the Fujifilm design meetings to know all of the factors in that decision. This video seems to support the “it’s not necessary to prevent moire” reasoning…
Fujifilm Manager Explains Why Medium Format and Not Full Frame, Talks About Fixed Lens GFX, X-Trans MF Camera and More - Fuji Rumors
But I suspect it was also affected by the desire to keep the price down and attract those considering full frame. Or maybe such large X-Trans raw files would have indeed been harder to process on the software of the day. Remember that those decisions were probably made several years ago, based on technology that was available at the time or certain that it would exit the pipeline.
As for APSC, it’s clear that X-Trans is still on the table, at least for the next generation of cameras. I hope they follow the same pattern of identical sensor/processor/software in all of the next generation APSC cameras. That takes IQ out of the equation when deciding on bodies.
--
Randy
Last edited:
michaeladawson
Forum Pro
Well, yes. Just look at the marketing back then. They pushed 16 MP X-Trans cameras as having the resolution of 24 MP sensors.
Without an AA filter, and when talking about OOC jpeg files, this was a valid claim. It gets a little murkier with RAW files and knowledgeable post-processing.
The D800 and D800E were perfect cameras to test out the impact of an AA filter. One had an AA and one had a self-canceling filter. There were several comparisons of resolutions between two models, where appropriate sharpening was added to images from each of the cameras. At 100% it was difficult to see any resolution differences.
I can't say I buy the technology pipeline argument. Nothing has really changed other than ever increasing processor speeds. Anticipating increasing processor speeds doesn't seem like something that would, or should have been a significant risk factor during planning and design stage.
Whether Bayer, or X-Trans, the point of pixel shifting is to capture full color information at every pixel so you no longer need to demosaic the image. Since the exposures are identical for each shift, I would assume duplicates can just be averaged.
Here is where it gets a bit messy. The phase detect pixels have no color filter. The standard Bayer moves you describe above can potentially leave you with incomplete color information on some positions because of the phase detect pixels. With the X-Trans moves I described above, the every position would still capture full color information.
Yes, X-Trans would require 6 exposures instead of 4, but in some ways, the processing would be easier as you don't need to "guess" the colors at the phase detect pixels.
I would guess it is simply a matter of sensor cost. Fuji sells far, far more crop cameras than medium format. Given the volume, a custom color filter array on the crop sensors would have little to no financial consequences. The same is not true for the volume of medium format sensors.
Assumptions/perceptions of the people buying crop versus medium format could be playing a role too.
Sampled systems can suffer from aliasing. It is a mathematical fact easily provable using simple Fourier analysis and follows directly from the Paley-Wiener theorem and the uniqueness of analytic continuation. In audio, digital signal processing and digital RF systems, an antialiasing filter is used in the processing chain to eliminate the aliasing or more precisely to attenuate the energy above the Nyquist threshold for aliasing.
Back in the early days of digital imaging the resolution was fairly low so strong AA filters were used. There of course was always a criticism in that people saw such images as soft and that softness was confused will loss of "detail." In reality the AA filters did not destroy any information and what was often confused as "detail" in digital images without an AA filter were actually aliased artifacts rather than valid information.
But the holy grail for digital photography became to eliminate the AA filter. Add to this the fact that a CFA sensor is not one but three digital sensors - with different color filters. However, unlike say the Foveon that captures one color at each pixel location the Bayer has a red sensor and blue sensor sampled at 1/2 the rate each of a green sensor. Hence the green sensor has 2X the resolution of the Red and Blue. It is pretty much a lie to call a 20 MP CFA a 20 MP color sensor. It is more like a 10 MP when it comes to spacial frequency response. Back when Sigma was using the original Foveon X3, one would find this on going debate. Sigma actually listed the resolution of their cameras as the number of individual detectors which is three times the number of pixels. Of course that was a lie but no bigger lie than called the Bayer resolution as the number of detectors. In reality the spatial frequency response was 1/2 the number of detectors which is the resolution of the red and green channels. You could see this especially if one looked at color charts instead of B&W. The Foveon's resolution was constant across color - the Bayer was not.
The XTrans is a CFA - not Bayer but still a CFA. It has about 1/2 the number of red and blue detectors as green detectors. The Bayer is based on a repeating 4x4 pattern where as the XTrans is based on a 6x6 pattern. These patterns are periodically repeated. In the 36 detector XTrans tile - there are 20 Green, 8 Red and 8 Blue detectors which gives slight more Green detectors than in a Bayer of the same size. The distribution for a Bayer is 1/2 Green, 1/4 Blue and 1/4 Red. The XTrans 5/9 Green, 2/9 Red and 2/9 Blue.
As an aside when Sigma when to the "Foveon Quartto" chip it the top layer was 4 times the number of detectors than the other two channels. The detail was captured by the two layer and the bottom two layers were used for color information. The estimate was a 20 MP Quartto (20 million detectors) had the same resolution as a 39 MP Bayer.
Loading…
www.camerastuffreview.com
So digital camera companies have been playing marketing speak abusing the term resolution for years. Fact the highest spatial frequency a digital sensor can represent is determined by the underlying detector. When one tries to turn a device that measures light energy into one that also records color information - the highest spatial frequency is reduced by 1/2 in Bayer and maybe one could argue by 5/9 in XTrans of the total number of detectors. So one might argue that a 20 MP XTrans would have about 11% more resolution that the same 20 MP Bayer. But since there is an interpolation process involved in both - that would need to be established.
As Mike pointed out when the D800 and D800E came out - same sensor, one with and one without an AA filter in practical sense it was difficult to see any difference in resolution and when one did it was at 100% and was very image specific. It is also unclear how the aliasing presents itself in the XTrans. Without a AA filter the energy higher than the Nyquist spatial frequency cutoff will be recorded as lower spacial frequency - that is details that are not actually there. We know that Bayer tends to produce color artifacts because the Nyquist cut offs of the red and blue channel are 1/2 the green. We know the Foveon X3 produced aliasing artifacts that could mask as "details." Because of that they didn't look as bad but they were faux detail.
It would be an interesting analysis to see just how the aliased information presents itself in the XTrans. I am sure Fuji has a pretty good understanding. I also suspect they are not going to share the details publicly.
Loading…
www.camerastuffreview.com
I can't really fault Fuji for its marketing hype on resolution. The entire digital camera industry has been doing it for years. As an aside the first color image from space was taken on NASA's AT3S space space craft in 1967. The output of 3 channels (red/blue/green) was transmitted back to earth and the color image reconstructed on the ground on film.
The first operational color sensor packages flown in space where on the Landsat program. At that time the sensor technology was a lot lower than the commercial sensors of today. A three way beam splitter was used - with each channel sent through a color filter to a separate sensor. The three sensor captures were then used to create a color image. That generated a true color image.
Carter Mead had a vision of putting that concept onto a single chip - the Foveon X3. The CMOS technology was not up to implementing his vision. Maybe some day.
Sounds like a fair assumption. ;-) Big difference in size between the sensors though. Maybe the small APSC really benefits from the XTrans whilst the much bigger GFX doesn’t need it.
The interesting thing is Fuji has always been pushing "alternative" concepts. They have pushed them with a lot of fanfare and hype. The Super CCD. Remember that one.
Super CCD - Wikipedia
The first iteration was the detectors laid our in a diagonal pattern instead of a linear horizontal/vertical configuration with claims of superior resolution. Horse pucky as my old Grandpappy would say. They they added a secondary small detector - claiming class leading dynamic range. Of course it was never verified. But that was the advertising blitz. Amazing enough those scams and claims disappeared in short order. But that scam lasted for 10 years before it became obvious that the Super CCD was in fact a scam and provided no benefit over the sensors of the time and people weren't buying it.
Is the XTrans the follow on scam. Hmmmm... the last Super CCD was announced in 2010, the XTrans XPro1 was announced in 2012. That is about the right timing to spin up their new CMOS sensor scam. But like the Super CCD, the XTrans only managed to reach a niche level in the market. Well a few years later Fuji introduced the GFX with a - wait for it - Bayer CFA. Maybe they knew they could not scam the medium format market when competing with Pentax, Hasselblad, etc.
The Super CCD was not the greatest thing since canned beer and solve all problems. It was really no better than the Bayer CCD cameras on the market. The XTrans - the follow on scam. That's for everyone to decide. Alternative concepts are good. The Foveon was a great concept - a much better concept than XTrans a real three color array. But it didn't pan out. But at some point alternative concepts if not significantly better border on marketing scams. Even Sigma has thrown in the towel on the Foveon and just came out with a high density Bayer sensor.
However, every camera today is the sum of its parts - the film days are over we buy a camera we get the sensor that comes with it. So it gets down to trade offs. Welcome to the world of digital cameras.
MrALLCAPS
Senior Member
Have you ever thought that if they covered a Medium format sensor with an X-Trans layer, it would cost way more that a regular Bayer one, which would bring the price of their Medium format cameras by a lot?
MrALLCAPS
Senior Member
It's funny to read the complaints of X-Trans vs Bayer, the "well I guess the X-_________" is going to be my last Fujifilm camera...", and so on and so on...
Well, the good thing is Nothing is keeping you to continue to buy Fuji. Don't like X-Trans, upset that the X-H2 will be X-Trans? -Fine...
You don't have to buy it.
You would buy Leica, but it's too expensive, you would go Full Frame, but it may be to big and heavy for certain situations, but why don't the complainers realize that All Cameras are a Compromise?
It's too easy to look at a competitor's camera and say, "I want dog eye AF too!". Well, if you really want it, then go get it. If financials is an issue, then shoot with what you have. Haven't you shot dogs before Without "Dog eye AF"? Didn't you get by?
It's also amusing how the most of us come from Bayer sensor cameras, but really liked the output from X-Trans. How some of us has/had issues with PP in say Lightroom, but we get by and still enjoy the cameras, lens selection and output for what it is.
Have y'all ever thought that if Fujifilm hasn't given us a major Fuji X body with a Bayer sensor, what makes y'all think they will change? There are still minor complaints with lightroom, but did that change Fuji's direction? -NO, because it's not about how their algorithm in their sensors work with a 3rd party app, it's about their Cameras, Lenses, manual controls and overall Enjoyment of the system! We are not forced to use lightroom and this is coming from one that uses it, as well as Capture one!
Some people want to complain, just to complain, I guess. "His ice cream cone has 3 scoops, while mine only has 2" mentality.
I'm totaly quite happy with 2 scoops and will be enjoying it.
Well, the good thing is Nothing is keeping you to continue to buy Fuji. Don't like X-Trans, upset that the X-H2 will be X-Trans? -Fine...
You don't have to buy it.
You would buy Leica, but it's too expensive, you would go Full Frame, but it may be to big and heavy for certain situations, but why don't the complainers realize that All Cameras are a Compromise?
It's too easy to look at a competitor's camera and say, "I want dog eye AF too!". Well, if you really want it, then go get it. If financials is an issue, then shoot with what you have. Haven't you shot dogs before Without "Dog eye AF"? Didn't you get by?
It's also amusing how the most of us come from Bayer sensor cameras, but really liked the output from X-Trans. How some of us has/had issues with PP in say Lightroom, but we get by and still enjoy the cameras, lens selection and output for what it is.
Have y'all ever thought that if Fujifilm hasn't given us a major Fuji X body with a Bayer sensor, what makes y'all think they will change? There are still minor complaints with lightroom, but did that change Fuji's direction? -NO, because it's not about how their algorithm in their sensors work with a 3rd party app, it's about their Cameras, Lenses, manual controls and overall Enjoyment of the system! We are not forced to use lightroom and this is coming from one that uses it, as well as Capture one!
Some people want to complain, just to complain, I guess. "His ice cream cone has 3 scoops, while mine only has 2" mentality.
I'm totaly quite happy with 2 scoops and will be enjoying it.
Jerry-astro
Forum Pro
Excellent post. Very much on the same page.
John Gellings
Forum Pro
- Messages
- 10,705
- Solutions
- 1
- Reaction score
- 9,345
Cheers, should have been in ALL CAPS!
--
https://www.johngellings.com
Instagram = @johngellings0 and @nycrandomjg
Last edited:
The_Davinator
Member
I like the idea of the XH2 and don’t mind which sensor option used. Rumours of a 43mp sensor render it rather moot.
michaeladawson
Forum Pro
Can you point out examples where an X-Trans APS-C is so much more expensive than a comparable Bayer sensor?
KariP
Veteran Member
Thanks - refreshing opinions Mr A !
I understand part of the theories and facts people write in this thread.
I think photography as a hobby or profession is often something about taking good or better photographs, even creating something like art (sometimes). Good equipment can help if we have something there 12 cm behind the viewfinder.
I do not understand this heated bayer vs x-trans discussion at all - perhaps it was OK when some people were troubled by worms and odd looking forest images because they could not use their LR or some other software.
Blaming X-tans if your photographs are not perfect is - childish. And X-trans causes serious headache it is extremely easy to swap to bayer sensor cameras. I like Fujifilm cameras because the size (X-E3) is right and the handling is nice (H1). Worms or artifacts have never troubled me.
I use also Canon cameras (FF today) because I started with Canon and kept some lenses. The new R6 is the first camera that feels in many ways "better" than my Fuji gear - biggest problem is the price of lenses... I have seriously tried to compare IQ of my two systems. I see some tiny differences, but I can not see anything I could think is the sensor array difference. Both systems are great for big prints u to A3 - perhaps FF is slightly better in A2 size.
I really would like to see some printed images where the X-Trans vs Bayer "problem" is visible. Pixel peeping is a separate hobby.
Anyone can buy a Sony, Nikon or Bayer and live happy with the new partner - and find some new things to complain about.
Similar threads
Keyboard shortcuts
- f
- Forum