Camera Sensor Dynamic Range (continued)

A few emails were sent so I’ll delve a little bit deeper on the last post on dynamic range of camera sensors.

It was a typical “all other things kept constant” type of statistic.  In reality, there are so many variables that there is no way to weight the cameras the same way for all people.  For example:

1.  Auto AEB defaults are  all over the map.  From 3 to 9 AEB and the D40x doesn’t even have AEB.

2.  Total megapixels vary greatly.

3.  Some Full Frame sensors, one APS-H sized sensor, various APS-C sensors.

4.  Various maximum Frames Per Second for continuous shooting.

5.  Huge difference in low light performance.

The test performed by DXOmark.com was to measure the values from the first recorded pixel of light to the brightest recorded pixel.  That is over simplified, but the idea is to measure total potential Dynamic Range of the sensor.  Obviously with any camera there will be noisy shadows and difficult to recover highlights.  With that in mind, look at the extremes in the chart from the last post.

A Nikon D3x has a 13.7 rating versus a Panasonic G1 at 10.3.  That is a huge difference.  You can miss a shot with a D3x and still have a lot of latitude in the middle to mess with a RAW file compared to a G1.  Someone out there may be able to equate this to the full range / latitude of film (Slide/Print).  The concept here is the same.  The measuring of Dynamic Range of the sensor’s floor and ceiling spread is what the DXOMark testing gives us.

I can attest that when I started using a D3 over my 5D the 1.1EV expanded dynamic range (according to DXOMark) of the D3 was apparent to me when I would try to edit a single shot.  I could bring down highlights and pull up shadows more dramatically.  (I’m talking shots taken at base ISO in good light, because each of these cameras perform differently in low light).

These stats by DXOMark were not around when I made the switch to the D3, but nobody had to tell me the D3 had a greater dynamic range.  It was easy to figure out during my shoot off.  You see, I was evaluating the Nikon system as my new HDR pipeline.  What I did was take shots from a tripod with the D3, then swap out the 5D.  I was actually testing the sharpness of the D3 + 14-24mm f/2.8G versus the 5D + 16-35mm f/2.8L II.  But in the process of testing I also figured out that the D3 was capturing more light per image.  The main reasons I went to the D3 were to get 9 AEB and 9 fps in those bursts.  The fact that the 14-24mm was sharper and the D3 had a larger DRange were all bonus reasons to switch my main capture system to Nikon.

As we all know.  Not all those pixels in the full measured dynamic range are clean.  Every camera has noise in shadows and as you approach clipping saturation does weird things in the highlights.  Even in post using RAW converters to fill light or recover highlights will not always give you exactly what you want.  There have been shots I barely got away with using one shot using a D3, but I know if I had been using an old 5D (or any other camera lower ranking on that DRange chart) that it would not have cut the mustard.

[ [ [ After seeing where the D3x falls on that chart I am seriously considering selling my D3 on eBay and upgrading to the D3x ] ] ]

When you consider that the sweet spot of each shot is the “mountain in the middle” per se, it becomes critical when you think in terms of HDR.  The truly sweetly exposed pixels is more narrow than you probably realize.  The true range of pure pixels is a fraction of that number on that dynamic range chart, so starting with a large dynamic range to begin with makes it even more critical for when you start spreading out your EV steps on an AEB sequence.

This is where a camera that only shoots 3 AEB and using a wide 2.0 EV step to grab a high contrast scene becomes an issue.  The tonemapper will only use data it has in the HDR file from the merge.  So if you are merging say 3 images from a high contrast scene (where 5 or more shots would be more appropriate for coverage), then this is where the DR of the sensor becomes an issue.  If the camera that is taking those 3 shots has a much shallower dynamic range than another camera taking 3 AEB, then the tonemapper will inevitibly be pulling up shadows with more noise than the camera with a wider Dynamic Range.  The dynamic range becomes the limiting factor of clean pixels.

If however, you are going to shoot many shots at say 1EV jumps between frames (or as small as 2/3rds or 1/3rd EV steps), then the DR of any of the cameras becomes a moot point.  Because then the tonemappers will have plenty of clean pixel data at each small EV jump and overlap ranges greatly at that.

The DR of a camera sensor matters much more to single shots and/or HDR with only 3 AEB.

I hope this clarifies a wee bit.

Like I said, there are so many differences in total megapixels, crop sensors, high iso performance, frames per second shooting and lens differences… that it is impossible to really and truly compare all these cameras in totality.

Tomorrow I’ll show a chart with their default AEB capabilities.  It will also show the max EV steps between each frame which will give you an idea of which cameras are capable of capturing very large dynamic range scenes using the camera defaults alone.