Opinion: Thinking about buying medium format? Read this first
The recent announcements of Fujifilm’s GFX 50S and Hasselblad's X1D have turned a lot of heads, and for good reason. To take the GFX 50S specifically (since it's more likely to represent an affordable option for DSLR shooters), we love Fujifilm cameras. It’s hard not to – they offer excellent ergonomics with a level of direct control that photographers itch for, and Fujifilm’s color science renders images that harken back to the days of film, while retaining all the advantages of digital. Meanwhile, the X-Trans color filter array (CFA) offers a number of advantages compared to traditional Bayer CFAs, showing decreased false color and a slight noise advantage due to a (relatively) greater proportion of green pixels.
Ultimately, though, the image quality of Fujifilm’s best cameras was limited by their APS-C sized sensors, which simply cannot capture as much light as similar silicon in larger sizes. And if you’ve kept up with our recent technical articles, you’ll know that the amount of total light you’ve captured is arguably the largest determinant of image quality.
That left many of us wondering when Fujifilm would step up to full-frame (35mm). But Fujifilm went one better – they skipped the arguably saturated full-frame market and went straight to medium format. In a rather compact, lightweight mirrorless form-factor at that. That made a lot of sense especially when you consider Fujifilm’s heritage in medium format film cameras, and its experience making medium-format lenses for other brands.
So, finally, here comes the GFX 50S: Fujifilm ergonomics and colors, but with all the advantages offered by larger sensors. But while heads turn, eyes widen, and colleagues fight over who gets to take the camera out for a shoot, personally I’m in need of a little convincing. And think you should be too, if you’re thinking about plopping down a fat wad of cash for this seemingly drool-worthy system.
But what’s not to like, you ask? Bear with me…
Theoretical advantages of larger sensors
The potential advantages of larger sensors can broadly be split into four areas: noise in low light, dynamic range, subject isolation (shallow depth-of-field), and resolution. But zoom into the following 36MP at 100% - are any of those lacking?
The question is: does the GFX 50S currently deliver on all, or any, of these advantages over what the best of full-frame has to offer? Let’s look at each separately.
Low light (noise) performance
For the same f-number and shutter speed (or ‘focal plane exposure’), a larger sensor is exposed to more total light. The same light per unit area is projected by the lens, but the larger sensor has more area available capturing it. An image made with more light has less relative photon shot noise (the noise that results from the fact that light arrives randomly at the imaging plane). The more light you capture, the more you ‘average’ out these fluctuations, leading to a cleaner image (that’s the laymen’s description of it anyway; read about it more in-depth here).
That’s why a full-frame camera generally gives you cleaner images than your smartphone.* So if more light means better images, that’s a clear win for the GFX 50S, right?
Not so fast...
No, literally, not so fast. The lenses available for the GFX format simply aren’t as fast as those offered by full-frame competitors. The fastest lens on Fujifilm’s GFX roadmap is F2, which in full-frame equivalent terms is F1.56** (the concept of equivalence is out of scope for this article, but you can read about it in-depth here; for now, just remember the GFX has a reverse crop factor, relative to full-frame, of 0.79x). And most of the current MF lenses hover around F2.8 and F4, or F2.2 and F3.2 equivalent, respectively. That means that if they had the exact same underlying silicon technology (or sensor performance), a full-frame camera with an F2.2 (or F3.2) lens should do just as well as the GFX 50S with its F2.8 (or F4) lens. Even if were were to think ahead to the MF 100MP sensor Sony provides in the Phase One cameras, its 0.64x crop factor at best yields an F1.3 full-frame equivalent lenses from the one F2 lens announced, still not beating out the Canon 85/1.2, and barely beating out the plethora of available F1.4 full-frame lenses. So even if the newly announced G-mount lenses cover the wider medium format image circle (which I'd sure hope they would), things still aren't so exciting.
But full-frame can do better than that: F1.4 and F1.8 lenses are routinely available for full-frame cameras, typically for less money too. An F1.4 lens projects twice as much light per unit area than an F2 lens, and 4x as much as an F2.8 lens, amply making up for the 1.7x smaller sensor surface area of full-frame.
That means full-frame cameras can capture as much, or more, light as the GFX 50S simply by offering faster lenses. But wait, it there's more...
Companies like Sony have poured a lot of R&D into their full-frame (and smaller) sensors, and the a7R II uses a backside-illuminated design that makes it more efficient than the sensor used in the 50S. It also offers a dual-gain architecture that flips the camera into a high gain mode at ISO 640, allowing it to effectively overcome any noise introduced by the camera’s own electronics. In other words, the a7R II’s sensor is better able to use the light projected onto it, relative to the MF sensor – ironically a sensor made by Sony itself - in the G50S (or Pentax 645Z, or Hasselblad X1D). This allows it to match the low light noise performance of the larger sensor in the GFX (and Pentax 645Z and Hasselblad X1D) even at the same shutter speed and f-number. See our studio scene comparison widget above.
So if we start with parity, guess what happens when you open up that aperture on the a7R II to an f-number simply unavailable to any current medium format system? You guessed it: you get better low light performance on full-frame. Whoa.
Although the same f-number and shutter speed give a larger sensor more total light, they receive the same amount of light per unit area. Most sensors of a similar generation have broadly similar tolerance for light per unit area (technically: similar full well capacity per unit area). But a larger sensor devotes more sensor area to any scene element, so can tolerate more total light per scene element before clipping. That means that for the same focal plane exposure, despite clipping highlights at a similar point, a larger sensor will render shadows (whose noise levels define the other limit of dynamic range) from more total light. And the same logic that applies to low light noise applies here as well: more total light = less relative shot noise and less impact of any noise from camera electronics. That means cleaner shadows, and more dynamic range.
So another clear win for the larger sensor GFX, no? Well, no. Because someone poured a lot of R&D into the Nikon D810 sensor (noticing a trend here?), giving it higher full-well capacity per unit area than any other sensor we’ve measured to date: its ISO 64 mode. Each pixel can hold more total charge before clipping, relative to equally-sized pixels on any other sensor in a consumer camera. That means it can tolerate a longer exposure at ISO 64, longer enough (at least 2/3 EV, or 60% more light) to capture as much total light as the 68% larger sensor in the GFX 50S exposed at its base ISO (100). Don’t believe us? Check out our real-world dynamic range comparison of the Nikon D810 vs the Pentax 645Z, which ostensibly shares the same sensor as the GFX 50S:
In this shoot-out, we exposed each camera to the right as far as possible before clipping a significant chunk of pixels in the brightest portion of the Raw (in the orange sky just above the mountains). The D810, in this case, was able to tolerate a full stop longer exposure***, which allows its (pushed) shadows to remain as clean as the 645Z. That's the (scientific, not baloney) reason we claimed the Nikon D810 to have medium format-like image quality. Because its dynamic range and overall signal:noise performance at ISO 64 rivals many current medium format cameras their base ISOs (though not the huge new 100MP MF Sony sensor in the new Phase One). Just look at its massive SNR advantage (read: image cleanliness) for all tones at ISO 64 over the Canon 5DS R at ISO 100 - we intend to plot the Fujifilm GFX 50S on the same graph, and don't expect it to show any advantage to the D810. Because science.
Read about this all more in-depth in our D810 review here, and check out Bill Claff's quantitative data that shows a 0.22 EV base ISO dynamic range difference between the D810 and 645Z - hardly noticeable, much less something to write home about.
‘OK but it’s not fair to compare ISO 64 to ISO 100!’
Fair enough, there’s a little more to the story. ISO 64 does require more exposure than ISO 100, either via a brighter lens, or longer exposure time. But one might argue that under circumstances where you care about dynamic range – i.e. high contrast scenes – you’re typically not light-limited to begin with, and can easily give the camera as much light as needed. Either because you’re shooting on a tripod, you’re using studio lights and can just crank them up, or because there’s so much light to begin with (it is a high contrast scene, right?) You’re working at or near base ISO anyway, so you shouldn’t have trouble adding 2/3 EV exposure by opening up the lens or lengthening the shutter speed a bit.
But, yes, if you’re in a light-limited situation (i.e. you’re not shooting at base ISO) and it’s high enough contrast that you care about dynamic range (have to expose for highlights then push shadows), then the GFX 50S will have the upper hand here. But dare I say, that’s quite the niche use case: keep in mind that most situations demanding higher ISOs tend to be in lower light, where you care more about general noise performance, not dynamic range (since low light scenes tend to have lower contrast). And if that’s what you care about, there’s the a7R II which, although it may clip highlights a bit earlier, can give you as good, or better, low light noise performance.
But I'll concede – if you want both the base ISO dynamic range of the D810, and the low light noise performance of an a7R II (albeit with F2 or slower lenses), then the GFX might be your ticket.
As we calculated in our ‘Low light (noise) performance’ section above, the fastest lens on Fujifilm’s roadmap is ~F1.6 full-frame equivalent, with most current available lenses being F2.2 equivalent or slower. Since full-frame routinely has F1.4 (equivalent) lenses available, you actually get more subject isolation, and blurrier backgrounds, with full-frame than with medium format.
And, no, the ‘but larger formats have more compression because you use longer focal length lenses for the same field-of-view’ argument is false. Just say no to the compression myth. For equivalent focal lengths/apertures, there's no extra compression. Compression is relative only to equivalent focal length and subject distance (or subject magnification), and its relative distance to the background. Not the format you're shooting on. Don’t believe us, have a look for yourself:
Full Frame (70mm F4.5)
APS-C (46mm F2.8)
46mm F2.8 on APS-C is roughly equivalent to 70mm F4.3 on full-frame - meaning the two shots above should be virtually identical. And they are, save for a tiny bit more DOF in the full-frame shot because F4.5 was the closest I could get to F4.3. Now, of course, you can get shallower DOF on full-frame, for example by shooting at F2.8. But that's because those faster lenses are available for full-frame.
They're not in Fujifilm's lineup, which includes two F2.8 lenses, one F2 lens, and a few F4 lenses - which are equivalent to F2.2, F1.6, and F3.2 in full-frame terms, respectively.
Without brighter lenses, there’s just no reason to get excited about medium format for subject isolation and blurry backgrounds. If you're a bokeh fanatic, full-frame’s arguably the sweet spot.
OK, finally, some good news. Well, theoretically anyway.
If you have two differently sized sensors with the same pixel count, the smaller one will be more demanding on its lens (it samples the lens at more lines per mm for the same scene frequency). Manufacturing larger lenses to resolve the same total system resolution is also slightly easier, since any lens needs lower resolving power (in lp/mm) to resolve the same amount of total megapixels on a larger sensor. Furthermore, the Fujifilm has some tricks up its sleeve - less 'gapless' microlenses mean fewer stray photons make it to any given pixel, yielding more per-pixel sharpness, at the cost of more aliasing (due to lower spatial sampling resolution) and slightly more noise (due to less light collection).
What's that mean in laymen terms? The GFX files are sharp. Tack sharp.
So if you’re looking for true 50MP of detail across the frame, you’re more likely to get it with the GFX 50S than with a comparable 50MP full-frame sensor, simply because of the realities of lens design and tolerances. That said, we’ve been told that some of the newer full-frame lens designs were designed with 80 to 100MP in mind, on full-frame sensors. And with the eye-popping performance of some of the newest full-frame lenses we’ve seen, from varied manufacturers, we’re not inclined to disagree. We’ve seen some 50MP files from the 5DS R paired with truly stellar lenses where we simply can’t imagine anything better, resolution-wise. In fact, at ~F5.6-6.2 equivalent, I'm not seeing a major resolution advantage of the medium format cameras over the full-frame cameras in our studio scene comparison tool, and the 50MP full-frame image below isn't exactly starved for resolution, is it?
Put another way: if you’re seeing eye-popping resolution at F2 above and here and here (and even at F1.4 on some new lenses) when viewing a Canon 5DS R 50MP full-frame file at 100% (do click on the above image and view at 100%), do you want or need a truer 50MP? Or do you want even more than 50MP, particularly if it'll come at the cost of more depth-of-field, since there are hardly any F2 equivalent lenses that'll give you the subject isolation and background bokeh you see in the full-frame shot above?
Only you can answer that question, but it is true that physics being physics, larger sensors will always tend to out-resolve smaller sensors with equivalent glass. And so this is the area where we most expect to see an advantage to the Fujifilm system, especially over time as we approach 100MP, and beyond. It’s probably easier for an F1.8 prime paired with the GFX 50S to out-resolve an F1.4 prime on a 5DS R when both systems are shot wide open, but whether that will be the case (or if Fujifilm will even make an F1.8 or brighter prime for the system) remains to be seen. I certainly don’t think it would be a cheap combination.
Thanks, DPR, for saving me money / killing my hopes and dreams
Still excited about the Fujifilm GFX 50S and Hasselblad X1D? Perhaps you still should be. You get Fujifilm ergonomics and color science in a body capable of far better image quality than Fujifilm's APS-C offerings. But remember you can emulate much of that color science in Raw converters with proper profiles (we're looking into a separate article on this). More importantly, remember that equivalence tells us that an F1.8 medium format prime is what the GFX 50S actually needs to at least match the performance from modern full-frames paired with F1.4 lenses, from the perspective of noise and shallow depth-of-field. And that’s before you consider the advanced silicon technologies we’ve seen in different full-frame (and smaller) sensors that we haven’t yet seen in any medium format sensor. These advances have, for example, allowed a Nikon D810 to catch up to the dynamic range of the Pentax 645Z at base ISO, and the BSI, dual-gain a7R II sensor to catch up to the GFX 50S in low light noise performance.
Still, as I've said, physics is physics. For equivalent apertures and final output resolutions, we do expect medium format to yield a slight resolution advantage, thanks to its lower demands on resolving power of lenses. But the extent of this advantage, especially given some of the tremendous progress we’ve seen in recent lens designs, remains to be seen: I'm not starving for eye-popping detail at 1:1 viewing of 50 and 42MP files when pairing a 5DS R or a7R II with stellar modern prime lenses.
Of course, as medium format evolves, the same gains we see in full-frame over smaller sensors might find their ways into the format. But this will require both the silicon to keep up, and for the development of faster lenses. At least as fast as the fastest lenses full-frame offers. One thing does make us hopeful - recent conversations with some forum members alerted us to the fact that certain full-frame lenses, like the Zeiss Otus primes, actually project an image circle large enough for at least a square crop on Fujifilm's new MF format. That would essentially get you high quality F1.1 equivalent glass on the GFX 50S. Cool, if you can focus it, anyway... (the GFX focus even with native lenses is anything but fast or intelligent, by the way). But if we see more and more fast full-frame lenses able to cover the image circle of the GFX G50S, then we're more likely to actually experience the benefits of the larger sensor format, though native fast lenses (that aren't slow unit focus, please) are really what we need.
Else, the potential advantages may be outweighed by the disadvantages: the extra weight, heft, price and severely lacking autofocus. And the GFX 50S has given up some of the noise and false color advantages their X-Trans cameras show...
For now, we hope that looking at the problem through the lens of equivalence at least gives you an idea of how big (or small) you can reasonably expect the differences to be. Maybe it even saves you a dime or two. Or makes you want to yell at us for bringing up equivalence, again.
But at the end of the day, equivalence has left me rather equivocal about medium format digital. What about you? Let us know in the comments below.
Editorial note: The headline of this opinion article has been updated to make it clearer that the points expressed are not intended to be taken as being specific to a single product, but represent discussion of the pros and cons of the emerging enthusiast medium-format camera class as a whole.
* It’s also why ‘multi-shot’ modes yield cleaner images than single shots: these modes essentially capture more total light, averaging out shot noise. It’s also why brighter scenes generally look cleaner than low light scenes: more light = more photons captured = less relative shot noise = higher signal:noise ratio (SNR, or ‘cleanliness’ in laymen terms).
** The GFX 50S’ 44x33mm sensor has an effective 0.78x crop factor, so you can multiply the MF lens’ f-number by 0.78 to get the equivalent full-frame f-number.
*** We don’t control for T-stop, which may partially explain the larger-than-expected exposure difference between ISO 64 and ISO 100 on the D810 and 645Z, respectively. This doesn’t affect our experiment though, as we applied well-vetted 'Expose to the Right' (ETTR) principles for a fair comparison.
**** Blind test: our ISO 12,800 studio scene shots of the GFX 50S and the a7R II have both been resized to 42MP, and a 576px wide 1:1 crop has been taken. Can you tell which is which?
a7R II or GFX?
GFX or a7R II?