DMR under expose and push or just shoot 1600iso

[robsteve]

Advertisement (gone after registration)

It has been mentioned a few times that some people are underexposing 400iso shots by two stops and then pushing two stops in the RAW processor, rather than just shooting the DMR at 1600iso. With the improved firmware and the newest Flexcolor (4.8.4), I decided to give it a test.

 

I just ran a test in my backyard with the DMR, shooting with the 1600iso setting and then shooting at the 400iso setting at the same aperture and shutter speed. This meant the 400iso shot was underexposed two stops. Using Flexcolor, I moved the exposure slider two stops on the 400mm shot.

 

It looks like the underexposed 400iso shot does not have any more noise than the 1600iso shot, but the under exposure seems to loose some fine detail and the pushing in post more of the highlights clip. If you look at the crop into the slide, you can see that the surface of it is more over exposed in the 400iso shot than the 1600iso shot.

 

These are my test shots. I used a tripod and the shutter speed was 1/8000th of a second, so any detail difference can't be camera shake.

 

Here is the 1600iso shot.

 

 

 

Here is the 400iso underexposed two stops and pushed two stops in Flexcolor.

 

 

Crops showing the better detail in the 1600iso shot versus 400iso

 

1600iso:

 

400iso 2 stop push:

[robsteve]

1600iso:

 

400iso 2 stop push:

 

Notice how the highlights hold better in the 1600iso shot compared to the 400iso pushed two stops.

[topoxforddoc]

Rob,

 

Is it me or is the shadow detail better in the 400 -2EV frames. that's what I would shoot when I wanted high ISOs, say when shooting concerts etc. In those circumstances retrieving information from underexposed shadows is really important. What do you think? The crops from the first frames show it best, particularly below and to the right of the yellow triangle.

 

Charlie

[jaapv]

I would say the apparent reduction of noise is cause by loss of resolution. Applying gaussian blur to the 1600 ISO shot might well produce a similar result. I had not tried 1600 ISO on the DMR yet. The quality is beyond my expectations. I was led to believe it was virtually unusable.

[robsteve]

Rob,

 

Is it me or is the shadow detail better in the 400 -2EV frames.

Charlie

 

I think you may be looking at the crop where there are trees in the background that may have been moving in the wind. When you look at the full files, there seems to be more shadow detail in the staight 1600ios shot. Look at the lower right back corner of the play set that is in the shade.

 

Here is a link to the full size files as jpegs. They are just under 2mb.

 

1600iso

http://www.robsteve.com/LUF/L%201010064.jpg

 

400iso -2ev

http://www.robsteve.com/LUF/L%201010066.jpg

[robsteve]

I would say the apparent reduction of noise is cause by loss of resolution. Applying gaussian blur to the 1600 ISO shot might well produce a similar result. I had not tried 1600 ISO on the DMR yet. The quality is beyond my expectations. I was led to believe it was virtually unusable.

 

Jaap:

 

In most areas I don't think there is much difference in the noise between the two files. With the 1.3 firmware I think there is some noise reduction done by the camera on areas of even tones, like the darkest shadows on the sample shots or well lit parts of the yellow slide.

[topoxforddoc]

Advertisement (gone after registration)

Rob,

 

I've had a look at the full files and you're correct as usual. Just goes to show we medics should stay out of this techno stuff. Either way I still rate my DMR as a piece of image taking kit.

 

Charlie

[gogopix]

thyere have been other tests that show the opposite, but were hardly controlled. From a technical point of view 'pushing' by underexposing just throws two low end bits of data away, as Robert's test show in the shadow detail.

Now, that's the theory. In practice, noise can be non linear. AT 1/8000 there is not much noise buildup. For longer exposures, the 400 may have a definite noise advantage over 1600 depending on internal processing.

 

All depends; if both camera and post amplification are linear, should be no effect. If noise builds up at 1600 for longer exposures (Robert, can you try that?) then there may be an effect on noise

 

However, if the light doesn't tickle the last two pixel bits, the detail is just plain lost.

 

Victor

[jaapv]

I found in testing this to extremes on the M8 (up to ISO 5000) that this method works very well for low contrast images. On high contrast images the loss of dynamic range results in loss of detail in the shadows and a flat,noisy image. That is in line with Viktors explanation, I think.

[robsteve]

thyere have been other tests that show the opposite, but were hardly controlled.

 

AT 1/8000 there is not much noise buildup. For longer exposures, the 400 may have a definite noise advantage over 1600 depending on internal processing.

 

Victor

 

Victor:

 

I had believed in the underexpose and push theory with my first digital, an EOS D2000. It had a Kodak chip. I later came to the conclusion that is was just better shooting at 1250iso with the proper exposure than 800iso underexposed a stop. This is why I never tested it on the DMR before now.

 

As for trying longer exposure, I will have to wait until dusk and by then, there will be very few shadows to look for shodow detail.

 

Robert