Reading mode: Light Dark

D4 Full Well Capacity and Gain Estimates

Go to last post
Marianne Oelund said:
rhlpetrus said:
Marianne Oelund said:
It compares very closely to the D3s, by my results - close enough that you will not see the difference photographically.
Click to expand...
Is FWC measured per pixel (sensel is more appropriate I think) or per unit area?
Click to expand...
That comparison is per unit area. On a per-photosite basis, D3s is still well ahead.
Click to expand...
Of course the only way comparisons would make sense and it confirms my impression (based on info you and others have posted here) that sensor technology is reaching its practical limits.

Maybe DR was the remaining thing to be improved on the Nikon designs, compared to the Sony's sensors. There's still a chance, if Bill's graphs are confirmed, that Nikon has been able to get closer with D4 (even though the D7000's sensor tech in FF size will probably show another DR jump compared to the D3x's, in normalized terms).

Thank you and Bill and Iliah for the great job you've been doing, given the present limitations re available D4's info.
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)
 
Joofa said:
Marianne Oelund said:
No, pattern noise and other types of noise have symmetrical distributions and do not increase the average signal value.
Click to expand...
A symmetrical distribution will still increase the calculated average signal value if it is not zero mean.
Click to expand...
I meant symmetrical about zero, i.e., zero mean. Even if it were not zero-mean, the contribution to the signal would be considered part of the signal, from a practical perspective. Since it's a very small contributor either way (magnitude around 1% of signal), its impact on the analysis is insignificant. I really don't understand why you're pursuing the point. Do you also worry about dark current at 1/100sec exposure times?
Joofa said:
Do you think that pattern noise (including banding, etc.) is zero mean?
Click to expand...
The pattern noise under consideration here is zero-mean; it arises from gain and offset variation from photosite to photosite.

Regarding banding, it depends on what type you're referring to. Banding from differences in channel gains is still considered zero-mean (and part of the pattern noise); if one wants to insist that it has a non-zero mean, then that portion would be treated as part of the signal. Faint banding visible in deep shadows isn't zero-mean, but neither is it a measurable contributor at these signal levels.
 
Marianne Oelund said:
I really don't understand why you're pursuing the point.
Click to expand...
I'm trying to figure out why is there a discrepancy in BClaff's results at ISO 200:

http://home.comcast.net/~NikonD70/GeneralTopics/Sensors_&_Raw/D4_Full_Well_Capacity_and_Gain_Estimates.htm

where at ISO 200 the figure for FWC is 53,300 and the rest are closer to your estimate of 68K at ISO 12800. I noticed that both of you have used a simple mean of pixel values (though you used two images and BClaff one image). That is why I'm wondering if both yours and BClaff's calculations are more accurate at a higher ISO than at a lower ISO, and if there is some something interfering with the calculation of mean value at a lower ISO.

Sincerely,

Joofa
--
Dj Joofa
 
Joofa said:
I'm trying to figure out why is there a discrepancy in BClaff's results at ISO 200:

http://home.comcast.net/~NikonD70/GeneralTopics/Sensors_&_Raw/D4_Full_Well_Capacity_and_Gain_Estimates.htm

where at ISO 200 the figure for FWC is 53,300 and the rest are closer to your estimate of 68K at ISO 12800. I noticed that both of you have used a simple mean of pixel values (though you used two images and BClaff one image).
Click to expand...
My article might not be clear enough in this area.
The results are presented with respect to the test data sets.
I have much less test data at ISO 200 than other ISOs.
I have a great deal of ISO 12800 data.
I would not read anything into the "discrepancy" with the ISO 200 result.
I presented all data in an attempt to be transparent.

If I had withheld the ISO 200 result as an "outlier" you would never have known :)

Regards
--
Bill (visit me at http://home.comcast.net/~NikonD70/ )
 
Marianne,

I did get around to trying the pair analysis.
It worked out quite well and I revised my article.

FWIW, my window is a little smaller than yours and I located a point that computed a higher value than your original post.

The computed value from that point was almost exactly what I got from my linear fit (66,400 electrons).
Not surprising since the data was pretty linear in the range I examined.

For fun I may bother to examine the other channels.

Regards
--
Bill (visit me at http://home.comcast.net/~NikonD70/ )
 
jjnik said:
what are your overall expectations versus the D3s at this point?
Click to expand...
I don't think I'm the best qualified to answer overall.

But in the limited context of these sensor numbers I prefer to look at the "Photographic Dynamic Range".
I've used these numbers to model the D4 and predict it's curve.
You can compare with other Nikon cameras at my site or in this post:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=40356718
--
Bill (visit me at http://home.comcast.net/~NikonD70/ )
 
bclaff said:
bclaff said:
what are your overall expectations versus the D3s at this point?
Click to expand...
I don't think I'm the best qualified to answer overall.

But in the limited context of these sensor numbers I prefer to look at the "Photographic Dynamic Range".
I've used these numbers to model the D4 and predict it's curve.
You can compare with other Nikon cameras at my site or in this post:
http://forums.dpreview.com/forums/read.asp?forum=1021&message=40356718
--
Bill (visit me at http://home.comcast.net/~NikonD70/ )
Click to expand...
Thanks - that curve looks promising.
 
jjnik said:
Thanks - that curve looks promising.
Click to expand...
It is probably based on low-value compression of the read channel before the ADC. While this is useful for near-black values, I doubt that there is a 1 stop quality improvement at all ISOs. If that was the case, Nikon would be bragging about it. I have a feeling that the black noise DR test is not useful on the D4, though I would prefer to be proven wrong.
 
victor china said:
so would you still say you prefer the d3s at the higher isos?
or has your opinion changed?
Click to expand...
From all that has surfaced in the last couple of days it looks like they'll be quite close, maybe D4 slightly better at normalized output, D3s slightly better both full res. It seems Nikon is using very similar tech on both. (But my opinion is based on second hand observation of data and from a few ACR conversions, Iliah may have good reasons to disagree ;) ).
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)
 
bclaff said:
I find you comment confusing since the curve he refers to does not show a 1 stop improvement at all ISOs.
Click to expand...
Sorry about that. I'm a d3 owner. So, my comment needs to mention that I'm comparing the D3 to the D4. In reviewing the past comments, the other fellow compared D3s to D4.

Oh, and thank you for publishing your results and your efforts. As you can see, they spur much discussion.
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top