InitiallyDifferences between DX and MFT sensors abound, becoming all the more interesting as underlying sensor technology improves. Indeed, some have suggested that the century old exposure triangle relationship of ISO/Aperture/Shutter speed will in a handful of years be no more. Indeed, we can see how modern cameras are overcoming ISO albeit unofficially.
We all know that if we change one of either ISO, Aperture or Exposure (shutter speed) it must be balanced by a change in the opposite direction by one of the other elements in the group. This has been our training and understanding with this table is an illustration:
ISO 100 200 400… 1600 3200 6400
Aperture f8 f11 f11… f32 f32 f64
Shutter 1/500 1/500 1/1000 1/1000 1/2000 1/2000
Digital sensors generate noise that is usually seen in underexposed or images at very high ISOs. But what if we imagine that ISO is removed from the equation, allowing the camera to set ISO automatically? – a claim that is now abound across many forums.
Hot on the heels of a presentation by a night sky and low light specialist speaker at a camera club meeting I got home, sat down in the front room and did a quick experiment sat in my armchair. I first took a shot at ISO 2000, 1/20 sec f 4.8 – no manipulation, no correction of auto WB, just a basic DxO conversion from a RAW (NEF) file.
Then, in manual mode, I adjusted to ISO 100 without changing the aperture setting, but also changing the exposure to 1/50 sec to test the speaker’s claim that 6 stops underexposure is possible and then adjusting the RAW converter to correct. The shot unsurprisingly turned out to be pretty dark. Then I re-converted the RAW file in DxO but then adjusted the ISO by 4 stops, and bent the gamma curve a mite upwards. This got the third result:
WOW! To me this was unexpected. Having switched totally to digital since 1999, and used to avoiding noise and rarely looking beyond anything other than the native ISO setting, let alone the enormously high ISOs like 2000, I was truly amazed. In order to have best results we have been encouraged to ‘expose to the right’, in other words ensure there is as much information/data as possible of tones to the right of the histogram though being careful of clipping. If we are more confident in lifting shadow areas it seems now we should ensure there is sufficient margin to the right in anticipation of raising ISO level.
I went to sleep thinking why, in my ‘normal’ photography, need I expose to the right when the information is there!
The following morning I decided to be a little more methodical, though I’m not a ‘techy’ or ‘geek’ but I needed to know more.
I like the old oak tree above my home and decided to replicate the previous night’s indoor image but this time outdoors in the sunlight. With manual settings, I took a shot as my reference at ISO 3200 with exposure of 1/2000 at f8. Then, without adjusting exposure I changed the ISO to ISO 200 and retook the shot and got the expected dark grossly underexposed result. In the DxO RAW converter I boosted ISO by 4 stops (back to ISO 3200) with the third result:
For curiosity I undertook the same procedure with my Olympus OMD EM-5 with these results, the first taken at 1/2000 sec, the second at 1/160 sec:
This is all well and good, but one might now ask what is the point or incentive to explore this facet? And it is true that for the average daylight shot there is nothing appreciable to be gained down this route, but the technique holds more potential when photographing in low light, night or even in the studio where there is a wide dynamic range. The reason in these instances, is that software can isolate the dynamic range extremes and treat different areas of the image more appropriately.
This shot in my studio includes continuous lighting, but the lights are not shining on the dapper model; thus most of the scene is pretty dark but with very bright areas – 1/3 sec f4.8. I then lifted the exposure in the PS levels dialogue to gain a better response. The second version shown was taken at ISO 100 (underexposed at 1/30 sec f5.6) and adjusted in DxO to an appropriate ISO that was undecided at the time of the shot – 4 stops and reduced contrast significantly. In PS levels I adjusted exposure, though not as radically, and gave back some of the contrast. The outcome was an image that is less sharp but tonally more balanced:
Before adjusting levels there was strong evidence of banding and one can see a colour balance shift in the last image. Indeed, the image is confused as far as colour is concerned. A conversion to mono seems quite satisfactory.
Note that I have not looked to getting the best outcome in each image, and have applied processing to the whole rather than selective parts of the images in order to demonstrate and illustrate my thoughts. Also note that manipulation of ISO must take place in the RAW conversion, too much data is lost if an image has been converted to another file.
We can make an analogy with digital audio recording. Simply, sounds are initially analogue and are converted to digital but the electronics create background hiss/rumble/hum/noise. It is a choice where to amplify the audio spectrum to make a balanced recording. It is a choice at what stage amplification is undertaken in order to keep audible sound above the noise floor. Similarly with digital images – there is an initial analogue to digital conversion that takes place before further image processing in the camera. The challenge is to make the sensor as noise-free as possible, but aside from that the RAW file being created can be described as ‘on-chip’ or ‘after-chip’. The Sony stable of sensors (they appear in many Sony, Olympus, Nikon, Fuji cameras) seem to be the former, and seem to have led the way in terms of noise. Canon has its own sensors and approach, but it has been suggested on some forums that the most recent model is now ‘on-chip’.
The DxO website plots the dynamic ranges of the cameras it reviews, and it may be noteworthy that those cameras that can be described as having ISO Invariance have a straight logarithmic curve but others an otherwise adjusted curve. This is a graph of dynamic range at various ISO levels of three cameras – the D7100 (APS-C), the renowned ‘noiseless’ D810 (full frame), and Canon’s 7D (APS-C). The latter’s curve is not as straight, having a shoulder at the lower ISOs before dropping. There is obviously a difference, and indicated how well Nikon’s D7100 fares provided it is not ‘pushed’.
There is a good article at http://improvephotography.com/34818/iso-invariance/ and discussions in the forums at dpreview.com which can explain the technology and engineering a little more clearly and from an engineering perspective than perhaps I can.
What matters in the end are the approaches towards low light image making that can be taken, and these can vary depending on the way the particular camera sensor works… it is not necessarily one way is better than another.
Fundamentally, reducing the noise floor of a sensor is paramount to giving the photographer wider options in his creativity – using a narrower aperture in low light, or by raising shutter speed to catch action.
For my own photography I think I have more work to do in order to succeed in broadening the dynamic range even more. (I’m not comfortable using multiple image HDR whether in good technique or the production of an image that may be ethically suspect. I have no problem with single shot HDR since this is manipulation of data that is captured in the single image.)
Initially, from these initial experiments ISO Invariance does not seem to impart advantages for normal daylight photography, except the old adage of expose to the right loses its value. But there is one feature of modern DSLRs that can make use – AutoISO .
Until now I have never thought about using this feature because, like all auto features it implies a loss of control over the image.
To explain I must first explain that I often shoot in aperture priority but when I’m on my own and with time available I often default to manual mode. When using cameras that are ISO Invariant the effects of loss of dynamic range is forecasteable, freeing the photographer to concentrate on aperture/shutter speed. It is now quite feasible to make use of this feature successfully, but I would caution regarding increase in noise, and would suggest setting a ceiling for ISO according to taste (say ISO 1600). Testing will also identify whether very long shutter speeds add to noise in the same way as increasing ISO (turning up the gain/volume control).
But watch this space – more work utilising the various facets Invariance suggests may change my mind!
To go to other Gallery pages click from the menu
Contact Phil at firstname.lastname@example.org
If you would like to leave a post, or contact Phil please email or use the COMMENT page for your response. Your comments/views are invited! There are hundreds of Phil’s images at alamy – clicking the ALAMY logo take you to that site