D800e Imatest Stepchart Results

[bjanes]

Imatest Stepchart is a tool used to measure the tonal response, noise, and dynamic range of a camera system using a Stouffer stepwedge. On finally obtaining my D800e, I performed an analysis on the camera. As expected, the dynamic range is very good and the camera was able to capture 39 of the 41 steps of the wedge. This is much better than the results I previously obtained with my D3.

The first step is to expose to the right so as to capture the full dynamic range of the camera. Shown below is the best exposure analyzed by Rawdigger. The exposure at 1/13 sec f/8 places step 1 slightly less than 0.3 EV from clipping, but 1/10 sec clips the green channels. The full DR was not quite captured because of the limitations imposed by the granularity of the shutter speed.



The most important results are shown:



The lower steps are completely black on the screen as shown:



Rendering with ACR 7.1 PV2012 with +5 EV exposure and increasing the blacks to maximum does show the lower steps. Pattern noise is minimal, but I do see slight banding. Altogether, an impressive performance.



Regards,

Bill

[Bart_van_der_Wolf]

On finally obtaining my D800e, I performed an analysis on the camera.

Hi Bill,

Congratulations with the camera, the long wait has been rewarded.

The tonal response curves look very good. There's a lot to be gained by testing one's equipment. Knowing the weaknesses, if any, will allow to find workarounds before being confronted with difficult shooting situations. Of course knowing the strengths also allows to exploit those better.

The slight increase in highlight noise, and the slight flattening of the noise spectrum near Nyquist might suggest some preprocessing influence, but that could be attributable to the Raw converter. When you do a direct (DC)Raw conversion with Imatest you may be able to pinpoint the source (Camera firmware/ASIC or Rawconverter). But it is nothing to worry about either way, the output looks great.

Enjoy you new tool,
Bart

[LKaven]

Bill, have your experiences given you any insight into the issue of thermal noise with this camera at high gain, and how to treat that?

[bjanes]

The slight increase in highlight noise, and the slight flattening of the noise spectrum near Nyquist might suggest some preprocessing influence, but that could be attributable to the Raw converter. When you do a direct (DC)Raw conversion with Imatest you may be able to pinpoint the source (Camera firmware/ASIC or Rawconverter). But it is nothing to worry about either way, the output looks great.

Bart, Thanks for your comments. The conversion was with DCRaw, using the facility built into Imatest. The settings are shown. I don't really know much about interpreting the noise spectrum or how the source of the flattening could be determined. I uploaded the raw file to Yousendit in case you want to look at it.



Thanks,

Bill

Yousendit

[bjanes]

Bill, have your experiences given you any insight into the issue of thermal noise with this camera at high gain, and how to treat that?

Luke,

Unfortunately I have no data on thermal noise. You might search the DPReview D1-D800 forum for posts by Marianne Oelund. She is an expert in this area and frequently posts there.

Regards,

Bill

[LKaven]

Luke,

Unfortunately I have no data on thermal noise. You might search the DPReview D1-D800 forum for posts by Marianne Oelund. She is an expert in this area and frequently posts there.

Marianne is off on projects these days.  I have a couple of black frames at ISO6400 1/80th if you're interested in the NEFs.  One is made with the camera cold, and the second is made after 1 minute of live view. 

It is a new thing to see this much thermal noise at handheld shutter speeds at this sensitivity.  Though I suspect that the smaller Exmors also do this. 

Anyone who shoots at ISO6400 and up on the D800 needs to understand thermal noise and how to mitigate it.  And it seems that a black frame subtraction (matching shutter speed of course, as well as WB, gain, etc) does quite well -- well enough that one cannot ignore the difference.  But some of us are curious about what would be the best way to handle this noise, and whether it means a simple black frame from a stored library, or stacked frames, or gaussian-blurred frames, or what-have-you. 

As a very knowledgeable contributor, your thoughts are being solicited.

[EricWHiss]

Bill,
Thanks for posting. That's a very interesting result.
Eric