Monday, January 4, 2010

Dynamic Range Test -- Does the Fuji F70EXR Exceed the Dynamic Range of a dSLR?

A badly botched test on the Fuji Talk Forum compared the dynamic range of the F70EXR and a Canon dSLR and declared the Fuji the easy winner, saying that the Fuji crushed the Canon.

This, of course, is just silly ... the Canon was shot so poorly that the highlights were blown to kingdom come, and of course we know that Canon dSLRs don't make a habit of destroying their images. I proposed photographer error immediately.

There was no EXIF available for the images, so we could not tell what mistakes were made, but we were later told that the image was shot in RAW. And that fact now makes it two serious photographer errors ... first, when the exposure was not compensated and therefore the highlights were totally blown out. And second, when the RAW converter was left at the default dials instead of trying to recover the highlights.

The perpetrator staunchly defended his test until the weight of many dissenters finally caused him to admit it was flawed. There was one lone voice of support, but he has also been savaged for many an invalid test, so it was expected ...

This supporter did manage to attack us in the 3rd person with these words:
Your detractors will drive your thread to 150, shouting aloud your incompetence and, yet, without producing a similar test.
So here is that test.

The F70EXR's big strength is its ability to hold shadows above black while protecting highlights. It does this with clever EXR hardware that realigns the bayer mask (the colored filters on each photosite that allow the camera to see in color) to allow it to do tricks.

What you will notice here is that each pixel comes in pairs. And Fuji in fact has two tricks up its sleeve with these pixels:
  1. It can combine each pair into one larger pixel, thus improving noise characteristics dramatically.
  2. It can treat the second pixel in each pair as a separate sensor, stopping their exposure early and then use those pixels in place of the blown out pixels on the first half sensor, thus protecting the highlights very dramatically.
This works so well that one can shoot in almost any light with very little adjustment to the exposures. I still use compensation, but it is necessary less often with these cameras.

So how could a dSLR keep up with such clever technology? Well, the main reason why you need such technology with these little sensors is that really bright sections of the image would force you to compensate the meter down quite a bit (less light coming in) and with such tiny photosites there would be very few photons collected in the dark parts and thus no data. No data in the shadows means they all look black. This is called "blocked shadows."

The bottom line is that, with small sensor cameras, the choise is generally between blowing the highlights or blocking the shadows. That is, until Fuji created the EXR sensor.

On the larger sensors in dSLRs, the photosites are many times larger and even with very short exposures a lot of photons are collected, providing quite a bit of detail in the dark parts of the image. Thus, their inherent dynamic range is much higher. And it remains much higher as ISO rises.

So ... the experiment:

Shoot a window frame with half covered and half open to the sky. I had a cloudy day, but the difference in light between image halves is quite obvious. I will repeat it one day in a more dramatic light, but this will suffice to make the main point.

The F70EXR is shot at DR400 and DR800 settings ... which are the two I recommend to be used at all times.

The D700 is shot in RAW, so the jpeg engine is not performing any function at all. I shot 6 images from -3ev through +2ev. This gives us a range of images to look at. Here is the image, and my analysis follows.

Remember to click on this image to see the larger version. You can also right click and select "open in new tab" ...

My analysis:

The DR800 image is slightly darker than the DR400 image ... I think it is in fact darker across the board, but that may be just a trick of the light. Anyway, that's the one I processed for the bottom version (best ACR conversion.)

The highlight side of the DR800 image looks like it is about half way between the D700's -1ev and -2ev images. Call it -1.5ev. The shadow side looks like the D700's -1ev image, almost exactly the same tones.

It therefore appears that the Fuji is underexposing the image by 1 stop on the shadow side and 1.5 stops on the highlight side. And this with the compensation dial set to 0ev.

That is the trick that makes the difference ... underexposure on both sides, and more underexposure on the highlight side than the shadow side. It works, and we appreciate the work they have done to get it this good.

Of course, the shadow side is still a bit dark, so I would open it a bit and I did just that in my "best ACR" conversion. Note that there was a bit of noise in there because it was shadow and this is a tiny sensor. But ACR handled that easily.

Now, since the highlights are at about -1.5 stops, I could start with either D700 image (-1ev or -2ev) to try to duplicate the Fuji image. In fact, I reprocessed both images in ACR to see how they turned out. And the result was nearly identical. The D700 has massive shadow range, so I could have started with the -3ev image if I wanted to really drop the highlights down. And there was no appreciable noise in the shadows, of course.

This result would have been essentially the same with the D300, whose sensor is APS-C sized, smaller than the D700 sensor, but still massively larger than the F70EXR.

A note about the detail of the trees in the far distance ... they are obviously sharper on the FUji image. Why? Because the Fuji at f/4 is approximately equivalent to the D700 at around f/22 to f/25 and I shot the D700 at f/11 ... hence the background was out of focus, as is expected on really large sensors.

And my final conclusion ...

Shot properly, a dSLR can easily match the EXR technology in dynamic range ... and most likely exceed it with less noisy images. But for a compact camera, the DR tricks the Fujis pull off in jpeg are really amazing.


Hards80 said...

The F70EXR did better than I actually expected in this test. It is actually quite amazing how close it performed to the excellent D700 with ACR best.

thanks for sharing. i like the systematic way you tackle issues on here. i was trying to figure out a dynamic range test of my own. i will use this as a starting point, and test out the f70 versus my 50d when i have a free morning or afternoon.

cheers mate,

Pedro said...

i know nothing about photography.

i noticed tough that the f200 exr has some issues when it comes to exposure , i believe.

look at some pictures in my flickr profile.

the pictures were not shot by me. i just did post process them.

click here to acess my flickr page

Kim Letkeman said...

Pedro, you did a nice job of opening the shadows for the images in that stream, but I really cannot tell how the originals were shot. Some of these should have been exposed differently, with skies brighter and shadows more open. I cannot tell if provia was used (correct) or velvia (incorrect) ... it is obvious in some cases that the flash should have been used for fill, and finally, I cannot tell whether DR400 or DR800 were used for some of these, as it should have been.

The bottom line is that these cams are pretty easy to shoot badly ... which is why I have so many articles on this blog as to how to shoot them well.

Pedro said...

thanks for the fast answer.
exactly - the fuji exr cameras " are pretty easy to shoot badly ".

the f200 had a blue tint problem.

look here please.

and here

on another note :

an overlooked good camera. pentax optio s 12.
i like the colours of the s12 pictures.
the problem is that the camera has no image stabilizer and when it comes to low light its capabilities are very por.
definetely not a low light camera.

Pedro said...

for some strange reason i did try to post html links but it did not work this time

Kim Letkeman said...

Pedro, the blue tint issue of the F200 has been addressed with a firmware update. And shooting the F200 is as easy as it is hard ... i.e. once you know how, great images are easy. I have no interest in cameras that have severe weaknesses in some areas ... thus, I look at Fuji for stills and Panny for videos.

Pedro said...

thank you for your answer

Jon C said...

Kim, would you please consider a basic tutorial on how you post process your images?

I use your recommendations on how to properly shoot the F70EXR (THANKS!!!) and would like to also be able to have some idea of how to post process some of them.

Kim Letkeman said...

Jon, I've been thinking of that myself recently. I especially would like to show what is possible when using ACR as a front end for jpegs. I will try to write something up when I get a chance.

Anonymous said...

Amiable brief and this post helped me alot in my college assignement. Thanks you seeking your information.