Format equivalence, engineering and practical envelope

So which one has the biggest practical shooting envelope? They’re all the same; read on to find out why***.

Much has been written about depth of field, angle of view etc. equivalency for the various common sizes – I won’t repeat that. What I’m more interested is what consequences it has in practical terms on shooting envelope limitations, and how the apparent multitude of choices aren’t really choices at all – with a very few exceptions. To complicate things further, just because something can be done from an engineering standpoint doesn’t mean that it’s desirable from a marketing standpoint, and that’s before we even attempt to factor in how other things like haptics, controls, build quality etc. affect the overall shooting experience. Two examples: a consumer APS-C-sized camera with weather sealing and no feature or control compromises (think D5600 or 200D size); or a 1″ camera with really top class interchangeable optics (well, Nikon tried, but the market didn’t accept it). Or a rugged ‘professional’ compact, sensor size irrelevant. See what I mean?

Any camera or camera system is a complex balance of tradeoffs between sensor size, sensor resolution, lens size, lens resolution, maximum aperture, focal length range, physical body size/weight, and the big secondary impact: cost. There is no free lunch: you cannot keep everything constant and only change one variable. If you want a larger sensor, then the lenses either have to get larger or slower; if you want a longer lens range, the maximum aperture either gets slower, performance suffers, the housing gets larger – or more likely, all three. Actually, more often than not, pricing is the deciding factor: you never see different tiers of medium format lenses by performance; there’s usually only one design and they’re all as good optically as they can reasonably be made – no plastic elements or mounts to save a few bucks (consumer kit zoom variants, anybody?). Through long, expensive and painful experience, I’ve found that it’s best to prioritise the most important performance parameter and be willing to compromise on the rest – lest you land up with something that’s a compromise across the board, and thoroughly unsatisfying. It’s very rare for a Swiss Army knife to be as useful or satisfying as it is convenient.

By this, I mean that compact superzoom are great on paper and on the back, but you’re going to hate the images later, or curse when you run out of light. Every format has its strengths and weaknesses – I prefer to think of only the former – make choices on this basis and you can’t go too far wrong, providing the hardware exists. In general and taking into account both practical (manufacturing, product offering, marketing etc.) considerations and technical ones – smaller formats are better at maintaining a given (or even extended) depth of field for a given light gathering ability – determined by aperture. Lenses can also be made faster. However, the tradeoffs are that you generally lose a stop in sensor performance (dynamic range, sensitivity for a given overall noise level) every time you go down a size*. You don’t, however, gain a stop in lens performance – there are practical limits to how fast lenses can be made at sufficient cost (generally, the smaller the sensor, the lower the tolerable cost; but the more you make, the cheaper the unit cost becomes) and sufficient performance. You also start to run into diffraction very quickly: smartphone cameras are almost all diffraction limited even at f2, though peak performance is also limited by the corrections required for f2 and a real focal length of ~4mm in a tiny physical size. In practical terms: you’re not going to get much faster than f1.8, with an upper limit coming in at f0.95-f1.4 for enthusiast (read: willing to pay and carry more) formats.

*Practically common sizes: 1/3″, 1/2.3″, 2/3″, 1″, M4/3″, APSC 24×16, FF 36×24, MF 44×33. Things get a little complicated because the photosite-level technology and processing routines for each sensor-DSP combination are not of equal age or sophistication, which means that the relationships are not quite linear. Smartphones, for example, have massively more computing power than most cameras, and consequently have the ability to punch above their weight class.

Curiously, whilst you can buy dedicated f0.95 lenses for M4/3 (the Voigtlander series) and FF (Leica Noctilux) – there’s not much love for APS-C other than through mount sharing. But again, mount sharing and short flange distances have complicated things further: there are f1.4 FF lenses that also cover 44×33 (and smaller, of course – just with increasingly impractical angles of view). In accessible and practical** terms, we have the following:
[Sensor size] > [max common aperture] > [crude performance deficit to FF factoring in max aperture and sensor size]
1/3″, 1/2.3″, 2/3″ > f1.7 (in smartphones) > -4.5 stops
1″ > f1.8 (in high spec compacts, though Nikon did make a 32/1.2) > -3.5 stops
M4/3″ > f0.95 in MF; f1.2 in AF > -1 stop
APSC > f1.4 > -1 stop
FF > f1.4 > benchmark
44×33, 645 > ~f2 (though some adapted f1.4 FF lenses also work). > 0 stops

**Considering a normal lens range of say 28-135mm equivalent or thereabouts.

In short: going up a size may gain you a stop in sensor performance, but it doesn’t necessarily cost you a stop in lens performance (though size and weight are something else).

What we haven’t addressed are resolution, output size, and pixel level performance. This is a much more complicated mix: if we fix output size and pixel-level noise, the larger format will always look better: there’s less enlargement required; this is well known from the film days. What complicates things now is that for the vast majority of uses and users, even the 1/3″ stuff is capable of meeting ‘normal’ output sizes (web, small print – if that) with acceptable image quality and little strain. Everything is oversampled, and by a significant factor: going from a 12MP smartphone to 4MP Instagram (max, usually less after factoring in compression) is already 3x; let alone 24, 50 or 100MP. That 4.5 stop deficit is no longer 4.5 stops for normal output. This is undoubtedly why most consumers or inexperienced users cannot see much obvious benefit from trading up. Things are of course different if you plan to print very large.

In reality, pixels aren’t created equal, either: not only do the larger sensors have more of them, but they’re larger, too – which means significantly more light gathering ability. On most 12-16MP smartphones today we are now down to 1-1.2um pixels; compare that to 5.3um of 50MP 44×33: as much as 25x the area, and 25x the light gathering ability – per pixel. This means lower noise, higher dynamic range, more accurate color (it’s easier to precisely measure a stronger signal than a weaker one) and higher pixel-level acuity. Factoring this in, for the latest available sensors (in brackets) we now have:

[Sensor size] > [max common aperture] > [crude performance deficit to FF factoring in aperture, best in class sensor performance and resolution]
1/3″, 1/2.3″, 2/3″ > f1.7 > -4.5 stops (12MP CMOS BSI PDAF, iPhone 8 Plus)
1″ > f1.8 > -2.5 stops > (20MP CMOS BSI PDAF, RX100V)
M4/3″ > f0.95 in MF; f1.2 in AF > -1.5 stop (20MP CMOS PDAF, E-M1.2)
APSC > f1.4 > -1.5 stop (24MP CMOS, D5600, X-T2)
FF > f1.4 > benchmark (47MP CMOS, D850)
44×33, 645 > ~f2 > +1 to +3 stops (50MP, X1D; 100MP, H6D)

And there’s the computing power deficit: there’s a LOT of signal processing going on inside an iPhone, for instance; I don’t know of any camera that has anywhere close to the same graphics abilities. Even the pro level sports cameras struggle to chunk around data as quickly. Hard to quantify, because it can also produce some very strange artefacts at the pixel level that are not desirable if noticeable (think watercolour noise, edge destruction, big flat areas) – let’s call this one a wash.

What isn’t a wash (and is a hot topic) is the question of stabilisation – sensor shift, optical, or in some cases, both. The new systems are really superb – you can really leave your tripod behind, and hit silly slow shutter speeds with pixel level perfection. Before somebody asks why we don’t have it on all cameras – you need to understand that effective stabilisers must move both quickly and precisely; the higher the resolution, the more precise the movement required. And the heavier the thing being moved, the more power is required to move it – and non-linearly so, since all of these systems work with electromagnets. Basically, bigger things require more power than you might think. It is generally easier to do lens-based stabilisation because the thing being moved can be reduced to one or two fairly small and light elements, plus if the right elements are used, then large deviations can be easily corrected by small displacements – this is good for longer lenses, where sensor shift may simply run out of latitude. However, it’s possible that if the stabilising element goes significantly out of plane or alignment, then optics suffer; it isn’t something that is so common with sensor-shift mechanisms due to the way these are constructed. Bottom line: the smaller the format, the more effective stabilisation – whatever type – will be; in practice this is mostly true, except size limits once again determine how big actuators may be, and effective performance. Let’s alter that table again, remembering that M4/3 will stabilise everything including the f1.2 lenses, double-stabilise some, and there are precious few FF f1.4 stabilised options (only Pentax comes to mind, and that isn’t class leading in high ISO performance). The rest are slower – f2.8, f4 or worse – and the stabiliser doesn’t really open the envelope much.

[Sensor size] > [max common aperture] > [deficit factoring for best in class, including stabilization]
1/3″, 1/2.3″, 2/3″ > f1.7 > 3.5 stops
1″ > f1.8 > -2.5 stops
M4/3″ > f0.95 in MF; f1.2 in AF > 0 stop to +1 stop
APSC > f1.4 > -1.5 stop (same lenses as FF)
FF > f1.4 > benchmark
44×33, 645 > ~f2 > 0 to +2 stops (no stabiliser)

Note: stabilisers are not always a replacement for shutter speed, especially for moving objects.

This table looks like I’m spoiling for a fight: M4/3 has an advantage over FF? Under a surprising number of situations, yes. In all cases, I’m trying to match the final output look (read: DOF and angle of view) as closely as possible:

Example #1: a low light portrait or documentary capture, handheld (imperfect):
M4/3: E-M1.2, 42.5/1.2, f1.2, 1/30s ISO 400 – this is well within plausibility and capability of hardware; 1/30 chosen because of the subject’s ability to sit still. So, not all advantage to the camera.
FF: D850, 85/1.4G, f1.4, 1/100s ISO 1600-2000 – assuming you have steady hands. The parameters look or slightly worse for MF; your longer lens to match the same FOV will be slower, and you’ll need more shutter speed to counter hand shake. The sensor is a bit cleaner and you’ve got more pixels, so overall IQ is better but not that much so.

Yes, FF has shallower DOF; yes, double the resolution and slightly larger photo sites mean that assuming no upsizing, the FF file will probably look a bit cleaner – but not that much. Look at the ISO difference. I’ve deliberately ignored some of the subtleties, but in practical terms: I think you get my point. Let’s take another scenario: where you can maximise the potential of the sensors.

Example #2: a landscape or controlled lighting setup, tripod (perfect):
M4/3: E-M1.2, 25/1.2, f5.6, ISO 200 (base)
FF: D850, 55 Otus, f11, ISO 64 (base)
MF: H6D-100c, 80/2.8, f16, ISO 64 (base)
I think it’s pretty clear in this situation that the MF camera is going to wipe the floor with the other two – resolving power, dynamic range, color etc.

The reason I’ve chosen these two examples is because they’re at the practical limits of what we tend to photograph: it’s rarely at either end of the spectrum all the time. So whilst the larger formats have much higher potential, the further away you get from perfect conditions – the more difficult it becomes to deploy it. If you shoot at the edge of the envelope, your life is probably going to be easier with a smaller format, and the results not that different. If you live in a studio, then the bigger the better.

The one final piece of the puzzle to factor in is size and weight: whilst you can have a series of f1.4 primes for FF, you’re far more likely to be at f2.8 or f4 to keep the tradeoff between range and size (and cost) manageable – which means the 1″ compact with the f1.8-f2.8 zoom is looking mighty good about now, especially if you don’t print (and even if you do). There are curveballs with really poor shooting envelope (the Canon G1X III, for instance) or unicorns with very extended envelope and small size (Leica Q), that Hasselblad X1D-Leica 50/1.4 ASPH M combination I’m now using, but these are few and almost involve a huge cost penalty. There are far too many ‘photographers’ chasing the latest gear and trying to cheat physics: I’m not the only one to have noticed that the latest supposedly ‘small’ mirrorless cameras are now huge; a similarly kitted out A7RIII and D850 are about the same size and weight. A Panasonic G9 body is larger than an X1D. You can’t bend the rules of optics, unfortunately.

I’m aware that this article is getting both long, dry, and probably confusing. We’re nearly there.

Here’s my take on all of this: there is no free lunch, from an engineering standpoint or a commercial one. Different formats have different creative strengths, and this is generally how I choose on a given day (all other factors being equal and non-liming): the small formats are great for punchy, contrasty, all-in-focus compressed scenes and portability; the larger ones for printing, subtlety and nuance. Low light sits somewhere in the middle, with either M4/3 or FF as the best all-round compromise. Factor in size, and you start to prefer M4/3. Having used all of these formats extensively, and with output requirements perhaps a bit more demanding than most – I tend to steer towards FF or MF, with (fortunately) a unicorn as the ‘compact’. It’s definitely easier to shoot at the edge of the envelope with a smaller format, but if you can make it work, larger formats can deliver something special. We all want that magic single camera solution, and there’s nothing wrong with chasing it – but remember to keep the creative objective first, or else you’ll never be in that contented place where the hardware becomes a transparent tool and you can just shoot. MT

***First image: H6D-100c and 100/2.2; X1D and 90/3.2; GX85 and 12-32/3.5-5.6. What the H6D-100c gains over the X1D in lens speed and sensor size, it loses in weight and practicality; what the GX85 gains in weight, dual lens and sensor stabilisation over the MF gear, it loses in sensor and lens performance. We haven’t even considered subject tracking and AF. But the MF guys will have more print potential in a given light condition. I have uses for all three, depending on my creative and output objectives. Don’t underestimate opportunistic photography and carrying something just in case, either – I’ve made a lot of images I like with my phone, simply because I had it and nothing else. Composition does not and should not change with hardware.

__________________

Visit the Teaching Store to up your photographic game – including workshop videos, and the individual Email School of Photography. You can also support the site by purchasing from B&H and Amazon – thanks!

We are also on Facebook and there is a curated reader Flickr pool.

Images and content copyright Ming Thein | mingthein.com 2012 onwards unless otherwise stated. All rights reserved