Tuesday 7 June 2016

Getting Real About ISO Settings


Getting Real About Comparing ISO Settings


800 ISO f4 @ 1/50 sec with 18mm with OSS,  NEX 5n, 
a very useable combination, why would I need to push the ISO higher?


Here's a newsflash for you, raising the ISO on your cameras always trades off image quality....always...no exceptions. Raise the ISO and you underexpose the image to some degree at the sensor level and maximum image quality is always optimal when the exposure is maximised...right before the highlights you want actually clip.

And why then dear friends do we then raise the ISO? 

To access faster shutter speeds to minimise subject movement.

To access a more adequate level of depth of field via a smaller aperture without going too slow on the exposure. 

Or to minimise camera movement so we can shoot hand held under very low light or with a longer focal length lens.

With me so far, great, now answer me this. What's fundamentally wrong when testers start bagging 1" or M4/3 sensors saying they look mushy and noisy at 3200 ISO and above and claiming they don't compare well to full frame sensors?  Simple, they are not comparing apples to apples! 

Accepted wisdom is that FF cameras are way better at high ISOs, they are, but…., they dammed well better be, because you are much more likely to need “those high ISOs’. 

You see it’s never just about the ISO, shutter speed and the aperture you need, it's about the interaction between the three of these, the mechanics of the camera in question and of course what you intend to do with the image. 

Let me run some numbers for you, I love numbers they can be so edifying, and so confusing as well when mis-used by politicians, anyhow let's get some edification happening.  

Say you are shooting a group of people in a restaurant at night, you know the sort of thing, a bunch of friends sitting at a table, you want everyone sharp, so no-ones nose gets out of joint or focus.  So there you stand 1.5 m from the nearest person and the furthermost bod is 4.5 meters away. Flash won't help, the inverse square law will see that,  Mr 1.5 m will be rendered as the ghost that walks and Miss 4.5 m as a woman of the dark shadows.

Nope you probably need to run with available darkness to get a natural looking result or radically change your subjects relative locations.  Now let's say you have a lovely 24 mp FF DSLR fitted with a 35mm lens which is just wide enough to fit everyone in without producing any “Humpty Dumpty” egg shaped head distortion or that horrid empty foreground syndrome.  And let's face it, no one wants a head like Humpty Dumpty! 

So here we go, according to the DOF scales on my computer, you would need to focus at about 2.2 metres with an aperture of f10 to cover your needs.  

In practice you should get from 1.43m through to 4.7m adequately sharp, with these settings, with the total DOF being 329 cm.

But wait what if I instead use APSC, now the lens will be set at 23.5 mm for the same angle of view and using the same focus point we can run with f6.3 for basically the same result according to the app.

The figures come out at 1.46 to 4.4m, 

But, and this is important the above figure is allowing for a tighter circle of confusion for the APSC sensor, which assumes you have to blow the image up a lot more to get a print, however that’s not normally the case as most folk don’t print bigger than A4 and often only shoot for small prints or web, so in practice you could usually go with the same size CoC or circle of confusion (larger) as the FF camera.

If you keep the CoCs the same then the APSC will offer a total DOF of 7.57m extending from 125cm through to 882 cm.

In udder wordz we can easily use a whole F stop and a bit wider or an ISO step and a bit lower or perhaps helpfully a shutter speed faster for much the same look/result. In fact according to the DOF scales with f6.3 at 2.2m I can get from 1.3 to 8.1 metres of DOF.  Anyhow in real world use we can easily say that f10 on FF and f6.3 on APSC are of equal utility.

(Geeky Point:  One thing that some photographers may note is that as the focal length decreases you get more gain in acceptable focus behind the point of focus than in front, this makes perfect sense of course, but in a practical sense it means that on the smaller format, using a shorter focal length for the same angle of coverage you ideally would bring the point of focus forward a bit.  In this case instead of focusing at 2.2 metres on the APSC format you might move the focus forward to say 2 m, as the extended acceptable focus behind the focus point using the original 2.2 metre setting puts extra DOF beyond where you need it.  Moving the focus forward a little balances it out somewhat and gives extra DOF in front of the point of focus which could allow you to get away with an even wider aperture, strange but true.)

Ah but what if we have a lovely little 16mp M4/3 pocket rocket, the focal length will now be 17.5 mm and the required aperture according to the DOF Master, just f5, but again that is assuming I use a much tighter CoC.  If I relax the CoC just a little to give me something similar to the standard APSC CoC I can get away with f4 for a total DOF of 3.8 m extending from 1.4 m through to 5.2m, which of course still handily exceeds what we need for this shot.

Ah but move the focus just a little further out to 2.4 m  Ah yes we can almost get away with f3.2, now that is near enough a whole 3 f stops or 3 ISO steps lower than the full frame...like we can use 1600 iso instead of 12800 ISO! Now seriously do you really think that 12800 iso on a FF camera is always cleaner than 1600 on a state of the art M4/3 camera. It might be  but generally I would say no, it's probably about the same.

Again all of this assumes that we area trying to make roughly the same moderate sized print from each camera.

Ah but let's dig a little deeper, let's dig right down to a 1" sensor like the one in the sony RX 100. Our crop factor is now 2.7x, so 35mm equals 13mm.  Guess what aperture you will need, allowing for a sensible CoC...f 2.8  Basically we don't have that option on all the Sony devices at 13mm but we sure could in the future and Nikon has a lens that will comply with f2.8, but heck, f2.8…..that’s 3.3 stops wider than the FF DSLR or 3.3 ISO steps less. In other words its like saying iso 400 on the 1” jobbie has about the same utility as instead of 4000 iso on FF.  And trust me, that award winning x100 series performs pretty excellently at that ISO level, in fact it’s quite brilliant at 800 ISO.  

That’s a big deal fellow photographers and it provides a whole bunch of other options when the light packs up and leaves.  You of course have the option to also shoot much faster shutter speeds or go up a bit on the ISO for example. 

NOTE AGAIN...look, I know people will claim the differences in the needed aperture are not as big as I am claiming for the smaller sensor sizes but trust me in practice they are unless you are really planning to try and print files from your Sony RX100 to very large sizes.  Additionally the fall off into “unsharpness” is far less steep with smaller formats so objects close to theoretical DOF limit often don’t look as soft and unfocused as they do with larger formats….meaning basically you have more wiggle room with smaller formats.  In fact you could probably get a way with wider apertures than I have indicated here.

But there's more, well there usually is...DSLRs have mirrors and going low on the shutter speeds usually incurs a sharpness penalty. Trust me, I am a pretty steady shooter but there is no way on earth I can hand hold my old Sony A900 with its inbuilt IS and say a 100 mm lens anywhere near as slow as I can go on my old Nex 5n with the Electronic first curtain shutter enabled and a 55-210 OSS lens set at say the equivalent 65mm or so. 

I always need about an extra 2 shutter speeds higher on the A900 to get the same results clarity wise due to the greater camera movement. That folks gives the NEX 5n about a 3 stop advantage for real world low light work....or I can use lets say 400 iso instead of 3200, yes I know that sounds extreme but I have to do this stuff for a living and it really does work out that way in practice.
  
There is more, if you go to a camera with a full electronic shutter like an Olympus em5 mk2 (with 5 axis stabilisation ) then god knows how slow you can go on the shutter speed, especially considering it is far easier to properly support such a light package!

I have certainly taken plenty of sharp pics with my EM5 mk 2 in the 1/4 to 1/2 sec range.

Actually out of interest I tried shooting my old Panasonic GH2 with a 20mm stabilised lens at f5.6 and 1/2 sec mounted on a small selfie stick which I used as a grip, I got an easy 60-70% success rate, and I imagine the Oly em5 mk 2 would be even better!  Oh and I should add the GH2 does not have an electronic first curtain so there is a fair amount of shutter vibration.  

As I said I regularly have to shoot jobs at very slow shutter speeds without flash and I know from painful editing experience what works and what is just wishful full frame thinking. 

And still more! On almost all cameras, noise reduction or to put it more accurately detail reduction starts to kick in around 800 to 1600 iso, even with RAW files in many cases. Any camera that let’s you stay under say 1000 ISO is going to have some obvious
advantages, regardless of format. 

And on it goes, for a great proportion of photos the ISO you will need will be closely related to the efficiency of the cameras image stabilisation and it is here that many DSLRs start to drop back because the stabilisation is normally in the lens....or not at all. Non-stabilised glass will ultimately force you to a tripod 2 to 3 stops earlier, or cause you raise the ISO by an equivalent number of steps. 

Currently the Olympus OM models are probably king of the hill where practical shooting under low light is concerned, the 5 axis image stabilisation offers 4 to 5 stops of compensation with pretty much any lens, but maybe a bit less with some. On top of that remember that m4/3 is 2 stops ahead on DOF to start with, but maybe even 3 if you are not printing big. So up front the advantage compared to a regular DSLR with non - IS lenses in marginal light could be a massive 7 stops where the subject itself is still!   Or lets put this another way, you could use 200 iso instead of about 25000 iso. Yes I know that sounds implausible but I have hundreds of hand held images taken with my Em5 mk 2 in low light at 200 ISO on my computer, very few are lacking in critical sharpness.  

Ah but I hear the DSLR fanboys yelling, "yeah but man, we got da fast glass and the super clean high ISOs, and I respond..so? 

Just read back a few lines, I said “practical shooting” not DXO lab test king...practical.....like actually holding the camera in your hands, you know, without a tripod.  

The m4/3 cameras give you access to some equally fast native glass of top quality and they can use all your fast FF glass via adapters with full IS too in the case of the Olympus cameras and even some newer Panasonics.

Hell, you could even use a Metabones speed booster for another f stop boost with fast Canon or Nikon glass and IBIS on the Olympus!  What is that sound I hear…could it be perhaps the deflation of the wind out of someones full frame balloon. 

Now you can argue all you like, and I know those Big Boy Canon and Nikon users will, but using say a 85 mm 1.4 lens wide open under low light for anything even mildly close to the camera is never going to give you a reliable real world usable DOF for any practical use other than very low res web images. 

To hell with the current shallow DOF fashion I say, it’s just not practical when dollars and useable shots are on the line.  

Like an example, let’s say we take a shot at 5m, 85mm, f1.4, and what pray tell is your total DOF....oh about 17 cm or half a head! Your focus better be totally spot on 100% of the time otherwise it’s a crap shoot. What’s the chance of you reliably nailing a singer in a nightclub actually singing and moving around against those DOF odds? 

Now.....just hold on a bit now as I have to go into the bedroom and pop my Ultra Flameproof suit on.........dum de dum dumm dum....Ok here I am all back. 

I often run low light night photography workshops and one little aspect that became really obvious early on was that 400 ISO on one camera is not necessarily 400 ISO on another. Many Canon DSLRs in particular seem to be, ah, how can I put this...ah ....using ISO ratings that are fantasy compared to say Nikon or Sony. 

I am talking about people shooting on full manual at fixed ISO with identical apertures and shutter speeds and the good old Canons ending up around 2/3 to 1 stops under exposed compared to their peers. No it's not all Canons, the 6D for example seems to be fairly honest but frankly there's more than a fair share of porkies been told. 

Again, don’t bother arguing with me, I am not going to be convinced, remember this is when taking groups of people under real world conditions with various camera models and brands and looking at the resulting images side by side, it's not an isolated case, it’s happened in every single workshop, I don’t make this stuff up. 

Moving beyond the settings and brands, camera style and ergonomics has an enormous role to play in what you can get away with. Currently Sony has the A7r and the RX1/2 in their catalogue, one has a smooth as silk leaf shutter and no IS, the other has a rattle gun shutter ( sorry, I mean a slightly loud and little bit clunky) ....oh forget it....the shutter is crap.  Anyhow the RX1 can be shot at very low speeds despite the lack of IS, the A7r is by all accounts a bit hit and miss dependent upon the shutter speed and lens fitted.   The A7 is far more likely to actually need those marvellously clean high ISOs. (Yes I have used an A7r and confirm that I thought it was bloody hard to get reliably sharp results, but I readily admit the A7R2 is a whole different box of tricks)

Granted there are some very specific circumstances where a FF DSLRs better high ISO performance will translate into better, or at least lower noise images. Mainly these are situations where DOF really doesn't matter, like astronomy or shooting really distant landscapes under very dim light, or perhaps arty super shallow DOF stuff, of if most times you shoot using a tripod. 

Let's say you shoot a moonlit landscape at F2 using a 50mm lens, so long as the nearest element you want in focus is at least 20 metres away you're good to go....or not.  There is a fly in that ointment, we are assuming you have a lens that is able to actually perform well at that aperture under low level but high contrast light conditions, don't assume that can done well at any sane price point for a FF format lens. 

On the other hand, your puny little m4/3 camera can probably access an almost reasonably priced 24mm f1.4 lens that really will deliver at f2 (maybe even f1.4)....it's just easier to achieve this with smaller image circles, period.  Realistically with full frame you will at least need to stop down to f2.8 to clear up the residual deficiencies in most 50mm FF lenses, barring of course the new $4000.00 Zeiss Otus or perhaps that new super duper Sigma Art 50 and 35mm lenses.

So we have arrived at an end, I think those with open minds have perhaps got my drift but just to be sure, what is my take away point?  

ISO ratings and high ISO performance by themselves are quite meaningless, unless of course you're in an argument in the Pub and the next shout is at stake, then tell whatever porkies are required. 

But.... 

What matters is how the system as a whole works because that's what determines what you or I can actually get away with.  In other words unless you can tell me about, shutter action smoothness and shutter type, DOF requirements, real ISO rating, lens choice, camera ergonomics, IS or no IS, actual focus distance, intended print size, all you’re doing is sprouting useless DXO numbers that may or may not translate into improved real world shooting results. 

So the next time you hear some knowledgeable camera tester berating an m4/3 or 1 inch sensor camera for not having stellar 6400 ISO performance, think carefully before you start nodding in agreement at their infinite but badly flawed wisdom.

No comments:

Post a Comment