What causes blur? It seems a simple enough question, but in pursuing the answer to this, I've ended up becoming a firm believer in quantum efficiency. So where did this journey begin?
When we talk about blur in images our immediate thought is usually related to the lens system. We might want blur for artistic effect, for example with bokeh, but when it's unwanted it's probably because we haven't focused the lens correctly. That problem is compounded when we have a small depth of field and can be particularly noticeable in portraits.
Is anyone in focus?
A second source of blur is from diffraction, when we stop the lens down. Whilst the depth of field might widen, diffraction causes the spot size formed on the focal plane to increase. This might not be as severe as blurring, but it does lead to a softening of the image.
The first two reasons are lens based, however the second two are related to movement. Most obviously, if we move the camera when we are taking an image then we are likely to blur the resulting photo. Photographers use this creatively with intentional camera movement (ICM), but usually it causes us problems! Unless the shutter speed is fast enough relative to the movement of the camera, blur occurs. This is the reason that many manufacturers incorporate in-body image stabilization (IBIS), with the class-leading cameras offering up to 6-stops of compensation. A tripod is the other solution.
The last problem is when the subject is moving relative to the camera. No amount of IBIS will solve the problem of a person moving when you are taking their photo. In this scenario a faster shutter speed is needed. The special case here, common in sports photography, is where you are trying to track a subject through your frame, such as a racing car. It's here that you need good panning technique which will allow you to keep the subject sharp (even though it's moving) whilst blurring the background.
It's a real problem when the camera and the subject is moving!
Technical Setup
So what techniques can you use to get tack sharp focus? Firstly, to get correct focus many pros will set their camera to a single focus point with continuous focusing, combined with back button focus. This ensures focus is where you want it and gives you ultimate control by decoupling focus from the shutter release. It's not ideal for all scenarios, but works well in most. Secondly, watch your f-stop. Your lens is likely sharpest at mid-apertures and, if you don't need to use them, avoid small apertures.
The two movement related problems both reduce to using a faster shutter speed. If you can't get correct exposure at the aperture and shutter speed of your choice, then boosting the ISO is the best option. I switch on Nikon's AutoISO, limiting it to 3200 with a minimum shutter speed based upon the lens I'm using.
Quantum Efficiency
What are the toughest situations to shoot in? Given that the currency of photography is light, then it's when this is in short supply. Low light is the bane of every photographer from gigs, to weddings, to astrophotography. They all involve scenarios where light can be limited, but I'd like to widen that to anything that is light limited for a given shutter speed. That can involve portraiture, sports, and street to name a few. Heck any genre can be light limited. Common solutions are to use faster lenses and higher ISOs and these work well. It pays to have that 50mm f/1.8 lens permanently in your bag for just such an occasion.
On an overcast day at f/9 and ISO1600 and the shutter speed drops to 1/60s.
However there is one more variable you can choose to target: quantum efficiency or, more simply, the sensitivity of your camera's sensor. It's measured as the percentage of photons hitting the sensor that produce an electron (electrons per photon). You only need to look at smartphones to see that, with every iteration of a sensor, quantum efficiency improves. When I qualitatively compared the noise levels of Nikon's 24MP D610 with my aging 12MP D700 they were about the same. The higher resolution of the D610 means that the sensor photosites are smaller, yet the improved quantum efficiency results in noise levels that are similar.
The alternative to upgrading to the latest and greatest camera is more simple. Increase the size of the photosites, which means reducing the resolution of the sensor relative to contemporary cameras. This is the approach Nikon takes with its top level pro cameras such as the D5. Contrast it's 21MP sensor (6.5 micron photosites) with that of the 42MP (4.3 micron photosites) D850 where the photosites are half the area. However it's Sony's Alpha 7S M2 that epitomizes this approach - its 12MP sensor has 8.4 micron photosites which makes it ideally suited to light limited photography.
Marginal Gains
Photography is often about marginal gains — a method or technology is improved that enables you to perform slightly better. For example, each iteration of IBIS has increased the ability to hand hold shots. As photographers we take gains in quantum efficiency for granted and it's impact is only reported indirectly through better noise levels and dynamic range. However it's benefits are felt much more widely and no more so than in those areas of photography that dip in to light limiting situations. These occur far more often than we might think, so the next time you are considering upgrading you camera body pause to think about quantum efficiency and the types of scenes you photograph. Might it actually be better to decrease the resolution of your new purchase? Remember, the number of pixels required for a job is more nuanced. You need less than you might think. Better quantum efficiency might well let you nail the shot and that's a whole lot easier than going back and shooting your job all over again.
Of course, there remains a fifth reason for blur: resolution! It's salutary to remember that you need enough pixels to cover your subject. When shooting a job that means trying to achieve it, where possible, in-camera. Where you can't, extra resolution will allow you crop in.
Lead image courtesy of Alexander Andrews via Unsplash, used under Creative Commons.
Isn't this the site that took Squarespace to task for their use of CC images? And yet they do it themselves? Maybe I've got that wrong.
Bigger photosites have less noise - that's true - but it's only meaningful if you examine photos on a pixel-by-pixel basis. In every real world application of photography, smaller photosites are printed smaller, too. It just means that low-resolution sensors don't necessarily have less noise than high resolution sensors. For example, in our tests and real world experience, photos from the 45 megapixel D850 have less visible than the D5. Objective tests support this - DxOMark gives the D850 a low-light ISO score of 2660, whereas the D5 gets a 2434.
This! Thanks for typing it up, Tony
So a comparison between a camera which was announced Jan 2016 and the other announced Sept 2017 invalidates the proposition that quantum efficiency is improving, how?
Funnily enough, shooting at base sensitivity is precisely what I do; however, his 'I tested it, so you're wrong" would appear to fail on its face.
1) This is a strawman (Tony didn't make any argument regarding quantum efficiency)
2) The D800 was announced in 2012 and has a similar low-light ISO score of 2853 on DXOMark (which is similar still to the D600 at 24 MP and the D4 at 16 MP - all released in 2012).
Whether you think DXOMark is the end-all be-all doesn't really matter - the same can be seen using DPReview's studio comparison tool and viewing everything at the same output size. If you aren't normalizing the images, you aren't doing a fair comparison.
Precisely, which given that is the thesis of the article, I have no real idea how that constitites valid rebuttal.
Now, if you claim that we have reached the point of diminishing returns, that is valid.
That's definitely the point... if you compared cameras with equivalent silicon inside them (which Chris does with the D800, D600 and D4) how would they fare? Similarly I suspect.
Yup, my 5DsR (often criticized for noise) is far better than my 5DmkIV at equal ISOs when the 5DsR is re-sized to the same size as the mkIV.
DxO Mark ISO score is really flawed tho, the D5 is definitely better than the D850 at high ISO.
https://www.dpreview.com/reviews/image-comparison?attr18=lowlight&attr13...
the only quantum that interests me is the nuka cola quantum
At some point in time one must look into the vibrations that could be caused by many sources not limiting to just shaky hands, the tripod, wind, non stable foundations but even vibrations in camera.
A simple test is to use a pen light laser taped to the camera and pointed at a wall about 15 feet away. Then trying to release the shutter by any means while watching the little dot of light one can see the vibrations. In testing you will be surprised at how long it takes between shots to eliminate vibrations. Or say your tripod is on wooden floor and you shuffle your feet, the longer the lens or the closer in at macro range you are the more significant the vibrating issues become.
2x over shutter speed won’t eliminate vibrations on a 50 mp sensor it’s more like 3-4x.
Back in the 70's I decided (out of sheer perversity) to shoot entirely from a tripod with my Minolta X-GM.
The images were dramatically sharper. So much so that some experienced photographers wondered if I was using MF.
The informations in the article are not really exact.
Pixel size is not directly related to quantum efficiency at all, and as an example i can pick some astrophotography CMOS (some of em are actually b&w version of commercial APS-C and M43 sensors) that easily surpass any FF camera QE despite having waaay smaller pixels.
Edit: and no, not because of the lack of bayer filter, even accounting the bayer filter they would still perform better.
The D850 being a BSI sensor almost surely has higer QE than both the D5 and the A7s, yet it's performance is worse.
The real deal when talking about low light sensor noise is Read Noise, that is how precisely the sensor is able to convert the electrical charge in digital units (or ADU). Sensor with higer megapixel count tend to have higer read noise because they have more pixels to read, however a really good high mpx sensor can outperform older low mpx sensors because they might have higer read noise (ie D800 destroying the D700 in low light despite having 3x more mpx).
For detailed article, check this, in particular the "bottom line" section: http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html
- Quantum Efficiency Might Be More Important Than Image Resolution
No shit Sherlock.