Fujifilm GFX 50R First Impressions Review

Thanks Richard,
I understand, but my point was that as IQ becomes a demand (not saying it should be) it takes careful shooting and processing to demonstrate what potential exists within a format.

It almost feels int his comparison you made less effort to show the differences, which is probably why you might break 1000 comments :)

Why was more sharpening applied to the small sensor?

Because base sharpening is lower on the Nikon to start with.

why is that?

Well, it kinda does at the same aperture, but most medium format lenses are slower, so it cancels out. Moreover, to deal with the shallower DOF one has to stop down more.

There are always situations in which one or other format performs better (like sport, video or low light) but at the end of the day when you need DOF, that is background separation, the bigger the format the easier to reach the solution. Of course this story can be build more complex with image processions, with lens qualities, light gathering per pixels etc, etc, but the question: is that worth it? And finally, according to my feeling the test scenario was good to equalise the two different system but not good to present how they can shine in their own fields. Recently I turned from Sony a7II to Sony A7R ( backward ?:-) ) I had to slow down taking the pictures because of the strange mechanical shutter and had to change my lens sets because of missing PDAF, but the IQ compensates me. If I had the need/sources for MF I would go for that and I would be sure that I should even more slow me down but the IQ would also compensates me.
Enjoy your own system and shoot nice photos!

@CameraMiki, this is a theoretical vs. practical conversation.

The lens aperture and subject distance are the primary determinants of DoF (and for noise performance) for a given viewing size and angle of view. Note that sensor size is not in there.

The sensor size starts to come into play when looking at lenses that are actually available for a system. For example, to get the same image from a full-frame 105mm F/1.4, you would need:
> 52.5mm F/0.7 on 4/3
> 70mm F/0.93 on APS-C
> 133mm F/1.8 on MF (44x33mm)

(These lenses all have the same aperture: 75mm. And on their respective systems, they have the same angle of view).

But not all of these lenses exist--or exist without significant lens aberrations. That's when the sensor size (and really, the system) come into play.

@Karroly - what I meant was that you want smaller aperture to get enough DOF to work with. This is not about exposure.
@beatboxa - yes, by "aperture" I meant "aperture setting" a.k.a f/stop.
@D7000ShooterUK - leave the light meter alone :) here is the working situation: your light meter says f/8, 1/125 at ISO800. you happily set it on your 35mm full-frame camera and take a portrait of three people, all in focus. to get this people in focus on a medium format camera, depending on a frame size, you might need f/11-f/16, which will demand a lower shutter speed, or higher ISO. This is the example of how MF higher light gathering capability does not work in reality.

@martindpr - it is not a myth. Rather, it's a situation with this new crop of "sub-medium-format" cameras. Try 6x4.5 at f/2 and see what happens.
In reality, these 30x45mm sensors are much more convenient to work with than real medium format, as they provide a good balance between the medium format look and relative ease of use.

@beatboxa - I really wonder what algorithm you use to read my words. I have never stated that only one parameter determines the DOF, and I never would, because, as you rightfully stated, it's a nonsense. DOF is a characteristic of a system, not a lens, and it also depends on a distance, not only f-stop and the sensor size. The very same lens at the very same f-stop will render different DOF on sensors of different sizes. Let's just stop there and move on to what my original post was about.
What you do not understand is that illusion of depth has nothing to do with either a given f-stop, or a maximum aperture of the lens. A good lens renders 3D at f/8 just as well as when wide open.

You didn't write the words to read. The only parameters you defined were: "Try 6x4.5 at f/2 and see what happens."

So I gave you one DoF for 645 at F/2. Because there are infinite DoFs for 645 at F/2. Because you are missing key info. That's the point.

Even in this last post, you've got many things wrong, and you're just parroting back to me what I just wrote to you. Except, I provided enough information for actual comparisons, such as apertures (f/f-number), distances, and sensors sizes (for enlargement).

Above, I also wrote: "The lens aperture and subject distance are the primary determinants of DoF (and for noise performance) for a given viewing size and angle of view." This is true. You didn't even know the difference between f-number and aperture--I had to explain.

You're the one who is repeatedly claiming that a sensor size & f-number only determines DoF. It doesn't. Just like how you confused f-number with aperture, you're confusing sensor size with enlargement.

@beatboxa - wow...

@Richard Butler "At the same F-number a larger sensor will experience more light (same light per square mm, more square mm). Considering this perspective tells you much more about the final image than exposure does (the same exposure gives very different performance on different formats, even though the brightness is the same)."

No, not really. It depends whether the two sensor sizes have the same number of pixels. If they do then the pixel pitch on the larger sensor will be greater. If they have the same MP then the photosites on the larger sensor will be bigger thus allowing them to make use of more of the light falling on the sensor resulting in probably better dynamic range and better high ISO performance. I say 'probably' because other factors come into play such as processing, S&R firmware, etc. This is why larger sensor sizes generally give better image quality but it is simplistic and wrong to say that this is because bigger sensors gather more light.

@D7000ShooterUK:
That's wrong, and a myth. What is MUCH more important is the total area, because all photosites are highly efficient today, in part due to a microlens array over the pixel. That's the whole point of it.

It is absolutely correct to say that larger sensors perform better because bigger sensors gather more total light. In fact, you can even correlate & quantify this to sensor size because it is so predictable.

A full-frame has 2x the area of APS-C, 4x the area of 4/3, and 7x the area of 1".
Use this fact to compare images with the same total light:
https://www.dpreview.com/reviews/image-comparison?attr18=daylight&attr13_0=nikon_d850&attr13_1=nikon_d500&attr13_2=panasonic_dcg9&attr13_3=sony_dscrx100m4&attr15_0=raw&attr15_1=raw&attr15_2=raw&attr15_3=raw&attr16_0=12800&attr16_1=6400&attr16_2=3200&attr16_3=1600&attr126_0=1&attr126_2=1&attr171_2=1&attr171_3=1&normalization=compare&widget=1&x=0.00021670278379749475&y=0.006868370448483285

Predictably, they look the same.

Richard Butler,
By writing "In practice, in the same lighting conditions working with MF equipment requires smaller apertures." without any more explanations, I understood that, because the large sensor gathers more light, one have to stop down a little bit more to get the same exposure (image brightness), which is, for sure, absolutely not the case. Thus my reply.
But I am perfectly aware that the larger sensor gathers more photons on the whole surface, for the same light flux (= same F stop), and thus produces an image with less noise for the same print size, and that the pixel size does not really matters in this case.

Richard Butler,
No need to apologize ;-)
By the way, do you know if a gapless microlens array is commonly adopted on today sensors, whatever their size, or there are patents limiting its use ?

I have yet to see a medium format camera or back, which performs really well at high ISO. Right now decent ISO12600 in full-frame segment is not unheard of, while none of the medium format manufacturers can make this claim.
Most likely, this is because larger sensors produce more heat adversely affecting the signal-to-noise ratio. However, the noise looks less critical on a print, because the size of the noise particles relative to the frame size is smaller than on FF, or APS-C, and MFT let alone smartphones and stuff.

@karroly - they use microlenses but also BSI sensorst that move everything not collecting light to the back so you have almost all of the surface for photosites. This is much more important on smaller sensors with smaller photosites (cellphones) but helps with larger sensors as well since you can now put more and better circuits as they do not affect the light gathering (more and faster ADCs, stacked RAM, etc)

Shallower depth of field and better control of it should be the first argument for medium format (IMHO). Probably because digital medium format means just a little bit bigger sensor boring sharpness and noise are the subject....
Noise is then reduced to a low-light problem while it pollutes colours in even daylight pictures.
Try this out with cloud photography (a good idea anyway) and be surprised like me.....

otto k
Thanks, I know about BSI sensors... But there are still small gaps at the silicon level between photosites. So, maybe I should rephrase my question : does a gapless microlens array may help improving efficiency of a BSI sensor in a significant manner or not. But maybe, making a gapless microlens array is now a standard technology, so it is used on both BSI or non-BSI sensors ?

It is used on all sensors with notable example being Leica that designs microlenses so that they work better with their legacy lenses (they were designed for film and are not telecentric so do not work spectacularly on new digital sensors, especially wide angle ones).

I don't have hard data, but IIRC they all use microlenses. I remember articles from at least two decades ago so it's fairly old technology. If you care I could dig them up?

Otto k,
I am specifically talking about GAPLESS microlens arrays. You do not use the word gapless in your answer, so I am wondering if we are talking about the same thing... Microlens arrays were not gapless at the beginning.

Given that the current GFX lenses have slow maximum apertures (only one F2 lens), you can always open the full frame lens 2/3 stops or more to get similar dof.

Ok, I found a bunch of articles but not by camera manufacturers so of little use here.
FWIW, ephotozine says https://www.ephotozine.com/article/digital-camera-image-sensor-technology-guide-16808#gapless-microlenses

Canon introduced gapless microlenses with the introduction of the Canon EOS 50D (2008)
...
This was further developed in the EOS 5D Mark II (2008) and EOS 7D (2009) with a reduction in the distance between the sensor and the gapless microlenses
...
Gapless microlenses can be found on a wide range of other cameras as well, including the Sony Alpha A77-II, A6000, and A7R to name a few.

@Richard Butler, you are wrong! Where there is a "2/3 EV difference", for example, between sensor sizes, it is because of different pixel pitch and other factors. To state that larger sensors (everything else being equal) gather more light is incorrect! Larger sensors do give advantages but they do not gather more light simply because of their large surface area.

D7000ShooterUK,
I am sorry, but to be more rigorous, I should have said what happens to the "N" term of the S/N ratio. As I said in my example above, "S" is 4x bigger, but "N", the noise, is random, so the noise coming from the 4x bigger area of the FF sensor is not 4x bigger. And that is why the S/N ratio is increased.
One can make a parallel between 1) using a bigger sensor and 2) stacking images taken in low light to reduce noise. When stacking, we multiply the "S" of an image pixel by the number of images while "N", the noise, being different in each image, has a tendency to cancel itself. One can call this "temporal" stacking because the informations added to create the final image were captured at different times.
Using a larger area, with a bigger sensor, to create an image pixel is like combining several points from different locations having nearly the same "S" values, but very different "N" values which cancel themselves. One can call this "spatial stacking".

OK @Karroly , but surely the signal, or "S" is dependent on the total number of pixels rather than the size of the sensor area. Does a 24 MP FF sensor produce a higher "S" than a 24 MP APS-C sensor?

@Irakly Shanidze: How should a test be carried out to demonstrate significant differences in your opinion? Do you think there is a dramatic change with shorter focus distances? In which setup do you think there would be (objectivly measurable, visible) differences? I would be honestly interested in that.

@Irakly Shanidze You're absolutely right! I think they should throw in some sport actions tests, some run and gun tests, then may be we can know the rest 80% of what we need to know ;)

@Irakly Shanidze: I understand. What I appreciate is that they did a comparison where they cancelled out the obvious size effect. In what discipline would you think the differences are more visible? Do you have any links to other comparisons where one could verify the advantage of the (slightly) bigger sensor?

Irakly - this isn't a lens test. This is a comparison of the technical image quality benefit (or not) of a current 50MP medium format sensor versus a full-frame sensor of near-equivalent resolution. I think that's made clear, but if you want some more context, we also covered this topic in a piece from last year, here:

https://www.dpreview.com/opinion/2341704755/thinking-about-buying-medium-format-read-this-first

@rationalist - Barney posted a reference to an article, which addresses many valid points. For further discussion and examples of MF photos take a look at my article here
https://www.artphotoacademy.com/medium-format-why-on-earth/

Karroly It's not pixel size that matters but pixel pitch and therefore the size and design of the micro lens array. More total light hits a larger sensor but that is irrelevant as exposure is per unit area. A larger pixel pitch allows the sensor to be designed such that it can make better use of the light falling on the sensor. A 24 MP FF sensor can be designed to give better dynamic range than a 24 MP APS-C sized sensor for these reasons due to the larger pixel pitch (not larger pixels).

D7000ShooterUK,
Where did I say that pixel size matter ? I only talked about the advantage of a large sensor versus a small sensor. I even said myself that pixel size does not matter, read more carefully.
"More total light hits a larger sensor but that is irrelevant as exposure is per unit area."
So, you really do not undertand what S/N ratio is. Same exposure by unit area, for different sensor sizes, does not mean the same S/N ratio on the final printed image. And I explained why...
" A 24 MP FF sensor can be designed to give better dynamic range than a 24 MP APS-C sized sensor for these reasons due to the larger pixel pitch (not larger pixels)."
You are wrong here too. The better DR does not come from the larger pixel pitch but mainly from the larger sensor size. You can make a large sensor and small sensor with the same pixel pitch and the large sensor will still have a better DR and S/N...

No. Full frame, meaning, it's a 35 mm that's bigger than the crop sensors. Medium formats are bigger. But the Sony medum format is actually a crop version, compared to medium format film cameras and a few digital cameras that are full frame medium format.

@Consinaphile. While crop sensors are full frame sensors in their own right. The reply is in proper context, due to the reason why the term, " full frame" and aps/crop sensors were started from the get go. So yes, in the context of the question (although I'm guessing the person asking the question was joking), "full frame" is smaller than a medium format.

crop sensor is a relatively new term .... it was not mentioned or applied to cameras before a few yrs ago when the sensor of 24x36 size became affordable or widespread and flame wars of the infallibility of so called ff sensors began

Ok, I'm back.

'Crop'-sensor is not a new term per se in relation to 35mm; 50 years ago (and for many decades) there were 'Half-frame' cameras using half of the full-frame size of 24×36mm (so 24x18mm).

But now we are too far away from the joke of the OP.

I had the same thought!

4sofnature "While the leisure class whines about the image quality..." you are doing exactly what about the homeless problem?

@Techo: I think all the magic is coming from the word "medium format", not related to sensor size ;)

You can't apply the crop factor linearly to megapixels. In this case you need to look at the physical dimensions of the sensor and figure how big one pixel is. Then you can extrapolate that size (ie how many of those can you fit horizontally and vertically in a full frame sensor) and you have your resolution. A 42mp FF sensor is roughly equivalent to an 18MP APS-C in terms of pixel density. You can cheat and apply the crop factor to the vertical number and the horizontal number to arrive at the result. For example a 24mp FF sensor at 6000 by 4000 pixels would be (6000 / 1.5) by (4000 / 1.5) which is 4000 by 2666, which is roughly 10.7mp. That math works due to the aspect ratio remaining the same.

@Christos Cloud
AND "pixel density" means absolutely nothing. Example are A7iii vs A7riii, D850/Z7 vs 50r/s etc.

@cksin
Maximum ISO ratings on cameras have nothing to do with pixel pitch. They are completely arbitrary numbers for marketing. Yes, the a7s/II has the highest ISO available of Sony cameras, simply because it is marketed as a low-light camera.

The Pentax K1 II has a max ISO up to 819,200 despite having the same sensor as the original K1 (also the same as the Sony a7r and Nikon D800/e). It simply has a new image processor and noise reduction algorithms.

Most modern cameras, though, top out at 25,600 or 51,200, entirely regardless of their sensor size or pixel count. Companies could advertise 2,000,000 max ISO if they want. Doesn't mean anything.

Also, larger pixels *will* be cleaner at the pixel level, precisely because of the higher SNR.

@MattyMustng
HI, I believe you're mixing up the concept of larger pixel with dual gain sensor. Pentax K1 II used a higher gain setting on the sensor to get to a higher max ISO, which can be archive alternatively by putting a larger pixel into a sensor. These 2 different approach are both boosting the gain of a sensor, but not necessarily providing an image with less noise.

@Erk1024 that is part of what I was trying to say, to keep the same example, the very reason I don't have my own pair of 660s is because they are twice the price, and to me at this point, that isn't worth it. I never said that MF was clearly the better choice overall, I just said it does in fact have better IQ in these examples, because slightly, it does.

All I really intended is to point out that everyone needs to stop trying to put FF(or any format) on some untouchable pedestal. This whole list of comments is filled with people making excuses for FF or MF one way or the other, but it's completely ridiculous to argue about why FF is superior in IQ to MF in this example. It's about the features you want and what you value more, and that is something that makes sense to discuss.

hey, the camera bag for this combo is called an automobile :-)

you can use 4x5 lenses on ff milc, and even get tilt/shift functionality with it, using a variety of adapters.

"Conclusion - shoot crop and forget about FF because youll get same performance for a whole lot less money."

:)))))

@MoodyRuby - if you read my OP carefully, i stated that lower base ISO makes a difference, but if there was a D7200 type sensor with base ISO 64 there would be no difference again. Btw, I did use D850 quite extensively for about 3 months, and I really couldnt see that much difference nor could anyone i presented my photos to, so whatever. I stayed with smaller and lighter.

@AstroStan - color bit depth is in [bits] and dynamic range is in EV if you really need the units.

@jaywalles - as the measurement show, no you do not have 1 stop more DR. DR is exactly equal at base ISO. You might have less DOF with FX, but i dont find that beneficiary, and yes, I stated that bigger sensors have high ISO advantage.

"color bit depth is in [bits] "

That makes no sense, especially for "Bit depth: 24.8"
please Google ""color bit depth"

@AstroStan i really dont have to gogle it. It is like I said. The number of digitaly displayed signal is in bits. No mater if that is visual signal or audio or whatever.

willful ignorance?

Just for fun, try to understand the meaning of bit.
Google it.

An assertion that a camera has "24.8 bits" is a bit crazy because bits are integer (comprised only of 1 and 0). But set that aside and compare "24.8 bits color" to a normal JPEG, which is 24 bits color (8*3). A "24.8 bit color" camera would hardly be any deeper than a simple JPEG, i.e. pretty crappy.

DSLR/ICL CMOS detectors are native 14 bits and output 16 bit raw files (true raw files are not really "color" and only have a single layer). There are no 25 bit CMOS/CCD detectors nor do 25 bit files exist.

So what are you talking about? Presumably you copied those numbers from somewhere but you profoundly misunderstood them.

@AstroStan - i really dont have to google it i know it is in bits and what you are trying to explain is wrong. So whatever man :)
Try not to overthink simple facts. You are confusing signal processing with color depth.

it was you who stated "color bit depth is in [bits]"
but enough of this.

Yes and i was right

You have been utterly incapable of explaining yourself and it has become obvious that you have no idea what you are talking about, nor even understand the most basic concepts.

But of course you insist that gibberish is "right" (in true DPR fashion)!

If you asked nicely I would explain.
But then again, why waste my time explaining as you clearly cant understand basics... so whatever man

Silly bluff. if you could rationally and scientifically explain then you would have already done so. Or at least you would have had an intelligent comment on my honest attempts to understand your concept of "bit" and "bit depth".

I suspect you took those numbers from somewhere so they probably have some sort of meaning. If you can scientifically define them then I'm listening.

Bit is 2 on power of a number and bit depth is color depth in bits.
I would have explained all if you had just asked nicely and didnt start rhis wgole conversation by being a prick. So whatever.... ask someone else to explain it to you or learn to be nice.

But anyway, Ill try to be the nice guy here and give you a quote:
"Color depth, also known as bit depth, is either the number of bits used to indicate the color of a single pixel, in a bitmapped image or video framebuffer, or the number of bits used for each color component of a single pixel."
You can follow up from there

So you Googled after all. Its too bad you didn't bother to think about it. Had you read my earlier post you would have seen that very definition and why it makes absolutely no sense in regard to your original claims. You are way out of your depth but too proud to acknowledge it.

As for those numbers, why don't you just say where you got them from?

P.S. It is almost certain that you and I are the only ones following these posts, which have gone into hiding on a stale topic. So you don't have to get upset that your reputation is in danger of being trashed by your own ignorance and foolishness.

It's OK If you are incapable of understanding these math and engineering concepts. Many people are illiterate in these domains. But less OK that you doggedly avoid learning and understanding to instead evade, obfuscate and insult. Time to to either let it go or learn something. I'd be glad to help but I suspect you are not that interested in understanding and more dedicated to being "right" (even when you don't know what that is).

But if you do want to learn then we can start with this question: what do you think "24.8 bits" means? You might want to take a look at my post above where I note that fractions are not bits and that 24 bit color is much too shallow for a real DSLR.

@AstroStan
Can you please explain it then? he probably got that information from "DxOmark".. What exactly is that "color bit depth" thing?

I did look into it. It ia not in bits. Its a plain number

Btw, why would 24bit be "too shallow for a dslr"?
And you are the first to have started with insulting so get off your high horse

24 bit color = 8 bits per color = jpeg

Most dslr output 16 bit raw file = 48 bit color
But most CMOS are 12 or 14 bit so the 48 bit file format is overkill and the true color bit depth = 36 or 42 bit. Both are considerably deeper than 24.

You are talking about theoretical values. And that just isnt correct. Real values are far less than theoretical values.

Most real raw files from a 14 bit cmos really do have 14 bits of data. After deBayering that becomes 36 bits of color data. I write professional image processiing software so I know.

But where it can get complicated is that different ISO gain results in number of electrons per DN (ADU) varying widely. Most raw images are adjusted so that the max DN is invariate. But if one calculates the minimum bit depth necessary to count electrons at a particular ISO then the 24.8 figure may describe a relatively high ISO. That's a strange way to characterize a cam and that's why I asked you to explain.

Yes, internally you do get 14bit of data per chanel but that data isnt usable data. Real values, what you get in photos (numbers that the measuring instrument can see), are much lower.
As there is no way to write formulas here, ill just link it
It is under chapter Definition
https://www.dxomark.com/glossary/color-depth/

OK now I see how that is derived. It is a secondary measure of "color depth", which is a complicated measure.

Sorry I gave you such a hard time.

No biggie. Ill trx to get to the point faster next time without acting smug.

or not

"f1.4 lenses are plenty." Yes, that is the true test.

You're right. Computational photography will render most of these topics null. And bots will take most of the landscape and street photos that I see here and everywhere else.

Since people are using phones for professional work, it's a serious point.

m43 vs 1inch = insignificant.
Let's all go get a Sony RX100V.