Camera ISO Settings

The Truth About ISO

Back in the days of ‘wet photography’, we had rolls and sheets of film that carried various ISO/ASA/DIN numbers.

ISO stands for International Standards Organisation

ASA stands for American Standards Association

DIN – well, that’s ‘Deutsches Institut für Normung’ or German Institute for Standardisation

ISO and ASA were basically identical values, and DIN = (log10)ISO x10 +1, so ASA/ISO 100 equated to DIN 21….nope, I’m not going to say anything!

These numbers were the film ‘speed’ values.  Film speed was critical to exposure metering as it specified the film sensitivity to light.  Metering a scene properly at the correct ISO/ASA/DIN gave us an overall exposure value that ensured the film got the correct ‘dose’ of light from the shutter speed and aperture combination.

Low ISO/ASA/DIN values meant the film was LESS sensitive to light (SLOW FILM) and high values meant MORE sensitivity to light (FAST FILM).

Ilford Pan F was a very slow mono negative film at ASA 50, while Ilford HP5 was a fast 400 ASA mono negative film.

The other characteristic of film speed was ‘grain’.  Correctly exposed, Pan F was extremely fine grained, whereas correctly exposed HP5 was ‘visibly grainy’ on an 8×10 print.

Another Ilford mono negative film I used a lot was FP4.  The stated ASA for this film was 125ASA/ISO, but I always rated it (set the meter ASA speed dial) to 100ASA on my 35mm Canon A1 and F1 (yup, you read that right!) because they both slightly over-metered most scenes.

If we needed to shoot at 1/1000th and f8 but 100ASA only gave us 1/250th at f8 we would switch to 400ASA film – two stops greater sensitivity to light means we can take a shutter speed two stops shorter for the same aperture and thus get our required 1/1000th sec.

But, what if we were already set up with 400ASA film, but the meter (set at 400ASA) was only giving us 1/250th?

Prior to the release of films like Delta 1600/3200 we would put a fresh roll of 400ASA film in the camera and set the meter to a whopping 1600ASA! We would deliberately UNDER EXPOSE Ilford HP5 or Kodak Tri-X by 2 stops to give us our required 1/1000th at f8.

The two stops underexposed film would then be ‘push processed’, which basically meant it was given a longer time in the developer.  This ‘push processing’ always gave us a grainy image, because of the manner in which photographic chemistry worked.

And just to confuse you even more, very occasionally a situation might arise where we would over expose film and ‘pull process’ it – but that’s another story.

We are not here for a history lesson, but the point you need to understand is this – we had a camera body into which we inserted various sensitivities of film, and that sometimes those sensitivities were chemically manipulated in processing.

That Was Then, This Is Now!

ISO/ASA/DIN was SENSITIVITY of FILM.

It is NOT SENSITIVITY of your DSLR SENSOR….!!! Understand that once and for all!

The sensitivity of your sensor IS FIXED.

It is set in Silicon when the sensor is manufactured.  Just like the sensitivity of Kodak Tri-X Pan was ‘fixed’ at 400ASA/ISO when it was made at the factory.

How is the sensitivity of a digital sensor fixed?  By the SIZE of the individual PHOTOSITES on the sensor.

Larger photosites will gather more photons from a given exposure than small ones – it’s that simple.

The greater the number of photons captured means that the output signal from a larger photosite is GREATER than the output signal from a smaller photosite for the same exposure value (EV being a combination shutter speed and aperture/f number).

All sensors have a base level of noise – we can refer to this as the sensor ‘noise floor’.

This noise floor is an amalgamation of the noise floors of each photosite on the sensor.

But the noise floor of each photosite on the sensor is masked/obscured by the photosite signal output; therefore the greater the signal, the larger the signal to noise (S/N) ratio is said to be.

In general, larger photosites yield a higher S/N ratio than smaller ones given the same exposure.

This is why the Nikon D3 had such success being full frame but just over 12 megapixels, and it’s the reason that some of us don’t get overly excited about seeing more megapixels being crammed into our 36mm x 24mm sensors.

Anyway, the total output from a photosite contains both signal and noise floor, and the signal component can be thought of as ‘gain’ over the noise floor – natural gain.

As manufacturers put more megapixels on our sensors this natural gain DECREASES because the photosites get SMALLER – they have to in order to fit more of them into the finite sensor area.

Natural gain CAN be brought back in certain sensor designs by manipulating the design of the micro lenses that sit on top of the individual photosites. Re-design of these micro lenses to ‘suck in’ more tangential photons – rather like putting a funnel in a bottle to make filling it easier and more efficient.

There is a brilliantly simple illustration of how a sensor fits into the general scheme of things, courtesy of digital camera world:

%name Camera ISO Settings

The main item of note in this image is perhaps not quite so obvious, but it’s the boundary between the analogue and digital parts of the system.

We have 3 component arrays forward of this boundary:

  1. Mosaic Filter including Micro Lenses & Moire filter if fitted.
  2. Sensor Array of Photosites – these suck in photons and release proportional electrons/charge.
  3. Analogue Electronics – this holds the charge record of the photosite output.

Everything forward of the Analogue/Digital Converter – ADC – is just that, analogue! And the variety of attributes that a manufacturer puts on the sensor forward of this boundary can be thought of mostly as modifying/enhancing natural gain.

So What About My ISO Control Settings Andy?

All sensors have a BASE ISO. In other words they have an ISO sensitivity/speed rating just like film!  And as I said before THIS IS A FIXED VALUE.

The base ISO of a sensor photosite array can be defined as that ISO setting that yields the best dynamic range across the whole array, and it is the ISO setting that carries NO internal amplification.

Your chosen ISO setting has absolutely ZERO effect on what happens forward of the Analogue/Digital boundary – NONE.

So, all those idiots who tell you that ISO effects/governs exposure are WRONG – it has nothing to do with it for the simple reason that ISO effecting sensor sensitivity is a total misconception….end of!

Now I’ll bet that’s going to set off a whole raft of negative comments and arguments – and they will all be wrong, because they don’t know what they’re talking about!

The ‘digital side’ of the boundary is where all the ‘voodoo’ happens, and it’s where your ISO settings come into play.

At the end of an exposure the Analogue Digital Converter, or ADC, comes along and makes a ‘count’ of the contents of the ‘analogue electronics’ mosaic (as Digital Camera World like to call it – nice and unambiguous!).

Remember, it’s counting/measuring TOTAL OUTPUT from each photosite – and that comprises both signal and noise floor outputs.

iso1 900x900 Camera ISO Settings

If the exposure has been carried out at ‘base ISO’ then we have the maximum S/N ratio, as in column 1.

However, if we increase our ISO setting above ‘base’ then the total sensor array output looks like column 2.  We have in effect UNDER EXPOSED the shot, resulting in a reduced signal.  But we have the same value for the noise floor, so we have a lower S/N ratio.

In principal, the ADC cannot discriminate between noise floor and signal outputs, and so all it sees in one output value for each photosite.

At base ISO this isn’t a problem, but once we begin to shoot at ISO settings above base, under exposing in other words, the cameras internal image processors apply gain to boost the output values handed to it by the ADC.

Yes, this boosts the signal output, but it also amplifies the noise floor component of the signal at the same time – hence that perennial problem we all like to call ‘high ISO noise’.

So your ISO control behaves in exactly the same way as the ‘gain switch’ on a CB or long wave radio, or indeed the db gain on a microphone – ISO is just applied gain.

Things You Should Know

My first digital camera had a CCD (charge coupled device) sensor, it was made by Fuji and it cost a bloody fortune.

Cameras today for the most part use CMOS (complimentary metal oxide semi-conductor) sensors.

  • CCD sensors create high-quality, low-noise images.
  • CMOS sensors, traditionally, are more susceptible to noise.
  • Because each photosite on a CMOS sensor has a series of transistors located next to it, the light sensitivity of a CMOS chip tends to be lower. Many of the photons striking the sensory photosite array hit the transistors instead of the photosites.  This is where the newer micro lens designs come in handy.
  • A CMOS sensor consumes less power. CCD sensors can consume up to 100 times more power than an equivalent CMOS sensor.
  • CMOS chips can be produced easily, making them cheaper to manufacture than CCD sensors.

Basic CMOS tech has changed very little over the years – by that I’m referring to the actual ‘sensing’ bit of the sensor.  Yes, the individual photosites are now manufactured with more precision and consistency, but the basic methodology is pretty much ‘same as it ever was’.

But what HAS changed are the bits they stick in front of it – most notably micro-lens design; and the stuff that goes behind it, the ADC and image processors (IPs).

The ADC used to be 12 bit, now they are 14 bit on most digital cameras, and even 16 bit on some.  Increasing the bit depth accuracy in the ADC means it can detect smaller variations in output signal values between adjacent photosites.

As long as the ‘bits’ that come after the ADC can handle these extended values then the result can extend the cameras dynamic range.

But the ADC and IPs are firmware based in their operation, and so when you turn your ISO above base you are relying on a set of algorithms to handle the business of compensating for your under exposure.

All this takes place AFTER the shutter has closed – so again, ISO settings have less than nothing to do with the exposure of the image; said exposure has been made and finished with before any ISO applied gain occurs.

For a camera to be revolutionary in terms of high ISO image quality it must deliver a lower noise floor than its predecessor whilst maintaining or bettering its predecessors low ISO performance in terms of noise and dynamic range.

This where Nikon have screwed their own pooch with the D5. At ISOs below 3200 it has poorer IQ and narrower dynamic range than either the D4 or 4S.  Perhaps some of this problem could be due to the sensor photosite pitch (diameter) of 6.45 microns compared to the D4/4S of 7.30 microns – but I think it’s mostly due to poor ADC and S/N firmware; which of course can be corrected in the future.

Can I Get More Photons Onto My Sensor Andy?

You can get more photons onto your sensor by changing to a lens that lets in more light.

You might now by thinking that I mean switching glass based on a lower f-number or f-stop.

If so you’re half right.  I’m actually talking about t-stops.

The f-number of a lens is basically an expression of the relationship between maximum aperture diameter and focal length, and is an indication of the amount of light the lens lets in.

T-stops are slightly different. They are a direct indicator of how much light is transmitted by the lens – in other words how much light is actually being allowed to leave the rear element.

We could have two lenses of identical focal length and f-number, but one contains 17 lens elements and the other only 13. Assuming the glass and any coatings are of equal quality then the lens with fewer elements will have a higher transmission value and therefore lower T-number.

As an example, the Canon 85mm f1.2 actually has a t-number of 1.4, and so it’s letting in pretty much HALF a stop less light than you might think it is.

In Conclusion

I’ve deliberately not embellished this post with lots of images taken at high ISO – I’ve posted and published enough of those in the past.

I’ve given you this information so that you can digest it and hopefully understand more about how your camera works and what’s going on.  Only by understanding how something works can you deploy or use it to your best advantage.

I regularly take, market and sell images taken at ISO speeds that a lot of folk wouldn’t go anywhere near – even when they are using the same camera as me.

The sole reason I opt for high ISO settings is to obtain very fast shutter speeds with big glass in order to freeze action, especially of subjects close to the camera.  You can freeze very little action with a 500mm lens using speeds in the hundredths of a second.

Picture buyers love frozen high speed action and they don’t mind some noise if the shot is a bit special. Noise doesn’t look anywhere near as severe in a print as it does on your monitor either, so high ISO values are nothing to shy away from – especially if to do so would be at the expense of the ‘shot of a lifetime’.

The ND Filter

Long Exposure & ND Filters

D4R3875 Edit Edit 598x900 The ND Filter

A view of the stunning rock formations at Porth Y Post on the Welsh island of Anglesey. The image is a long exposure of very rough sea, giving the impression of smoke and fog.  30 seconds @f13 ISO 100. B&W 10stop ND – unfiltered exposure would have been 1/30th.

The reason for this particular post began last week when I was “cruising” a forum on a PoD site I’m a member of, and I came across a thread started by someone about heavy ND filters and very long exposures.

Then, a couple of days later a Facebook conversation cropped up where someone I know rather well seemed to be losing the plot over things totally by purchasing a 16 stop ND.

The poor bugger got a right mauling from “yours truly” for the simple reason that he doesn’t understand the SCIENCE behind the art of photography.  This is what pisses me off about digital photography – it readily provides “instant gratification” to folk who know bugger all about what they are doing with their equipment.  They then spend money on “pushing the envelope” only to find their ivory tower comes tumbling down around them because they THOUGHT they knew what they were doing………..stop ranting Andy before you have a coronary!

OK, I’ll stop “ranting”, but seriously folks, it doesn’t matter if you are on a 5DMkIII or a D800E, a D4 or a 1Dx – you have to realise that your camera works within a certain set of fixed parameters; and if you wander outside these boundaries for reasons of either stupidity or ignorance, then you’ll soon be up to your ass in Alligators!

Avid readers of this blog of mine (seemingly there are a few) will know that I’ve gone to great lengths in the past to explain how sensors are limited in different ways by things such as diffraction and that certain lens/sensor combinations are said to be “diffraction limited; well here’s something new to run up your flag pole – sensors can be thought of as being “photon limited” too!

I’ll explain what I mean in a minute…..

SENSOR TYPE

Most folk who own a camera of modern design by Nikon or Canon FAIL at the first hurdle by not understanding their sensor type.

Sensors generally fall into two basic types – CCD and CMOS.

Most of us use cameras fitted with CMOS sensors, because we demand accurate fast phase detection AF AND we demand high levels of ADC/BUFFER speed.  In VERY simplistic terms, CCD sensors cannot operate at the levels of speed and efficiency demanded by the general camera-buying public.

So, it’s CMOS to the rescue.  But CMOS sensors are generally noisier than CCDs.

When I say “noise” I’m NOT referring to the normal under exposure luminance noise that a some of you might be thinking of. I’m talking about the “background noise” of the sensor itself – see post HERE .

Now I’m going to over simplify things for you here – I need to because there are a lot of variables to take into account.

  • A Sensor is an ARRAY of PHOTOSITES or PHOTODIODES
  • A photodiode exists to do one thing – react to being struck by PHOTONS of light by producing electrons.
  • To produce electrons PROPORTIONAL to the number of photons that strike it.

Now in theory, a photodiode that sees ZERO photons during the exposure should release NO ELECTRONS.

At the end of the exposure the ADC comes along and counts the electrons for each photodiode – an ANALOGUE VALUE – and converts it to a DIGITAL VALUE and stores that digital value as a point of information in the RAW file.

A RAW converter such as Lightroom then reads all these individual points of information and using its own in-built algorithms it normalises and demosaics them into an RGB image that we can see on our monitor.

Sounds simple doesn’t it, and theoretically it is.  But in practice there’s a lot of places in the process where things can go sideways rapidly……..!

We make a lot of assumptions about our pride and joy – our newly purchased DSLR – and most of these assumptions are just plain wrong.  One that most folk get wrong is presuming ALL the photodiodes on their shiny new sensor BEHAVE IN THE SAME WAY and are 100% identical in response.  WRONG – even though, in theory, it should be true.

Some sensors are built to a budget, some to a standard of quality and bugger the budget.

Think of the above statement as a scale running left to right with crap sensors like a 7D or D5000 on the left, and the staggering Phase IQ260 on the right.  There isn’t, despite what sales bumph says, any 35mm format sensor that can come even close to residing on the right hand end of the scale, but perhaps a D800E might sit somewhere between 65 and 70%.

The thing I’m trying to get at here is that “quality control” and “budget” are opposites in the manufacturing process, and that linearity and uniformity of photodiode performance costs MONEY – and lots of it.

All our 35mm format sensors suffer from a lack of that expensive quality control in some form or other, but what manufacturers try to do is place the resulting poor performance “outside the envelope of normal expected operation” as a Nikon technician once told me.

In other words, during normal exposures and camera usage (is there such a thing?) the errors don’t show themselves – so you are oblivious to them. But move outside of that “envelope of normal expected operation” and as I said before, the Alligators are soon chomping on your butt cheeks.

REALITY

Long exposures in low light levels – those longer than 30 to 90 seconds – present us with one of those “outside the envelope” situations that can highlight some major discrepancies in individual photodiode performance and sensor uniformity.

Earlier, I said that a photodiode, in a perfect world, would always react proportionally to the number of photons striking it, and that if it had no photon strikes during the exposure then it would have ZERO output in terms of electrons produced.

Think of the “perfect” photodiode/photosite as being a child brought up by nuns, well mannered and perfectly behaved.

Then think of a child brought up in the Gallagher household a la “Shameless” – zero patience, no sense of right or wrong, rebellious and down right misbehaved.  We can compare this kid with some of the photodiodes on our sensor.

These odd photodiodes usually show a random distribution across the sensor surface, but you only ever see evidence of their existence when you shoot in the dark, or when executing very long exposures from behind a heavy ND filter.

These “naughty” photodiodes behave badly in numerous ways:

  • They can release a larger number of electrons than is proportional to their photon count.
  • They can go to the extreme of releasing electrons when the have a ZERO photon count.
  • They can mimic the output of their nearest neighbors.
  • They can be clustered together and produce random spurious specks of colour.

And the list goes on!

It’s a Question of Time

These errant little buggers basically misbehave because the combination of low photon count and overly long exposure time allow them to, if you like, run out of patience and start misbehaving.

It is quite common for a single photodiode or cluster of them to behave in a perfect manner for any shutter speed up to between 30 seconds and 2 minutes. But if we expose that same photodiode or cluster for 3 minutes it can show abnormal behavior in its electron output.  Expose it for 5 minutes and its output could be the same, or amplified, or even totally different.

IMPORTANT – do not confuse these with so-called “hot pixels” which show up in all exposures irrespective of shutter duration.

Putting an ND filter in front of your lens is the same as shooting under less light.  Its effect is even-handed across all exposure values in the scenes brightness range, and therein lies the problem.  Cutting 10 stops worth of photons from the highlights in the scene will still leave plenty to make the sensor work effectively in those areas of the image.

But cutting 10 stops worth of photons from the shadow areas – where there was perhaps 12 stops less to begin with – might well leave an insufficient number of photons in the very darkest areas to make those particular photodiodes function correctly.

Exposure is basically a function of Intensity and Time, back in my college days we used to say that Ex = I x T !

Our ND filter CUTS intensity across the board, so Time has to increase to avoid under exposure in general.  But because we are working with far fewer photons as a whole, we have to curb the length of the Time component BECAUSE OF the level of intensity reduction – we become caught in a “Catch 22” situation, trying to avoid the “time triggered” malfunction of those errant diodes.

Below is an 4 minute exposure from behind a Lee Big Stopper on a 1Dx – click on both images to open at full resolution in a new window.

Canon1Dx 900x560 The ND Filter

Canon 1Dx
4 minutes @ f13
ISO 200 Lee 10stop

Canon1Dx2 900x560 The ND Filter

The beastly Nikon D800E fairs a lot better under similar exposure parameters, but there are still a lot of repairs to be done:

D800E 900x558 The ND Filter

A 4 minute exposure on a D800, f11 at 200ISO

Most people use heavy ND filters for the same reason I do – smoothing out water.

NDuse2 708x900 The ND Filter

The texture of the water in the top shot clutters the image and adds nothing – so get rid of it! D4,ISO 50, 30secs f11 Lee Big Stopper

Then we change the camera orientation and get a commercial shot:

D4D9874 Edit 581x900 The ND Filter

Cemlyn Bay on the northwest coast of Anglesey, North Wales, Approximately 2.5 km to the east is Wylfa nuclear power station. Same exposure as above.

In this next shot all I’m interested in is the jetty, neither water surface texture or horizon land add anything – the land is easy to dump in PShop but the water would be impossible:

NDuse1 900x511 The ND Filter

I see the bottom image in my head when I look at the scene top left. Again, the 10 stop ND fixes the water, which adds precisely nothing to the image. D4 ISO 50, 60 secs, f14 B&W 10 stop

The mistake folk make is this, 30 seconds is usually enough time to get the effect on the water you want, and 90 to 120 seconds is truly the maximum you should ever really need.  Any longer and you’ll get at best no more effect, and at worst the effect will not look as visually appealing – that’s my opinion anyway.

This time requirement dovetails nicely with the “operating inside the design envelope” physics of the average 35mm format sensor.

So, as I said before, we could go out on a bit of a limb and say that our sensors are all “photon limited”; all diodes on the sensor must be struck by x number of photons.

And we can regard them as being exposure length limited; all diodes on the sensor must be struck by x photons in y seconds in order to avoid the pitfalls mentioned.

So next time you have the idea of obtaining something really daft, such as the 16 stop ND filter my friend ordered, try engaging your brain.  An unfiltered exposure that meters out at 1/30th sec will be 30 seconds behind a 10 stop ND filter, and a whopping 32 minutes behind a 16 stop ND filter.  Now at that sort of exposure time the sensor noise in the image will be astonishing in both presence and variety!

As I posted on my Book of Face page the other day, just for kicks I shot this last Wednesday night:

D4D0710 Edit Edit Edit 900x598 The ND Filter

Penmon Lighthouse in North Wales at twilight.
Sky is 90 secs, foreground is 4 minutes, D4, f16, ISO 50 B&W 10 stop ND filter

The image truly gives the wrong impression of reality – the wind was cold and gusting to 30mph, and the sea looked very lumpy and just plain ugly.

I spent at least 45 minutes just taking the bloody speckled colour read noise out of the 4 minute foreground exposure – I have to wonder if the image was truly worth the effort in processing.

When you take into account everything I’ve mentioned so far plus the following:

  • Long exposures are prone to ground vibration and the effects of wind on the tripod etc
  • Hanging around in places like the last shot above is plain dangerous, especially when it’s dark.

you must now see that keeping the exposures as short as possible is the sensible course of action, and that for doing this sort of work a 6 stop ND filter is a more sensible addition to your armoury than a 16 stop ND filter!

Just keep away from exposures above 2 minutes.

And before anyone asks, NO – you don’t shoot star trails in one frame over 4 hours unless you’re a complete numpty!  And for anyone who thinks you can cancel noise by shooting a black frame think on this – the black frame has to be shot immediately after the image, and has to be the same exposure duration as the main image.  That means a 4 hour single frame star trail plus black frame to go with it will take at least 8 hours – will your camera battery last that long?  If it dies before the black frame is finished then you lose BOTH frames……………

 

Please consider supporting this blog.

This blog really does need your support. All the information I put on these pages I do freely, but it does involve costs in both time and money.

If you find this post useful and informative please could you help by making a small donation – it would really help me out a lot – whatever you can afford would be gratefully received.

Your donation will help offset the costs of running this blog and so help me to bring you lots more useful and informative content.

Many thanks in advance.