Reading mode: Light Dark

D4S versus D4 Photograpgic Dynamic Range (PDR)

Go to last post
B

bclaff

Forum Pro
Messages
14,415
Solutions
24
Reaction score
13,409
Location
Metro-West Boston, MA, US
I have enough data on the D4S to make PDR measurements but I'm waiting for a cleaner set of data before posting results to my site. (For those who aren't familiar with PDR it's a measurement I make that is quite similar to DxoMark Landscape DR.)

It looks like the D4S NEF files do not have any more dynamic range than the D4; in fact low ISO PDR might be slightly less.

This would seem to contradict anecdotal reports of greatly improved high ISO images.


Since the D4S has the newer Expeed 4 processor and improved signal processing I would not be surprised if JPG files look much better than the D4, particularly at high ISO; and that is significant to those who don't post process every raw file themselves.
 
Looking forward to your results, Bill.

Any chance of adding the Df to finally solve the mystery about if it is really the same sensor as the D4?

--
It's more important how an image looks as a thumbnail than how it looks at 100%.
http://inthemistphoto.com
 
Last edited:
I think results suggest that Df has an upgraded sensor (same low iso DR, marginally better high iso noise) and we know the D4s sensor needed an upgrade to support 1080p60 video (something we can never check on the video-less Df -- but if true, the lack of video may be due to a missing Expeed 4 rather than design in the Df ...).

Therefore, I would assume D4s = Df != D4
 
falconeyes said:
I think results suggest that Df has an upgraded sensor (same low iso DR, marginally better high iso noise) and we know the D4s sensor needed an upgrade to support 1080p60 video (something we can never check on the video-less Df -- but if true, the lack of video may be due to a missing Expeed 4 rather than design in the Df ...).

Therefore, I would assume D4s = Df != D4

--
Falk Lumo
Click to expand...
D4s = Df ≈ D4 ;)
 
According to my tests (link) I see slightly better SNR on the D4s RAW files when I pull details from the shadows at 100 ISO. Also the noise from the D4 tends to be more pink-ish.
 
Last edited:
crosstype said:
According to my tests (link) I see slightly better SNR on the D4s RAW files when I pull details from the shadows at 100 ISO.
Click to expand...
I sent you a PM re seeing those files.
crosstype said:
Also the noise from the D4 tends to be more pink-ish.
Click to expand...
Quite possible, I don;t measure the character of noise only total noise; but "pink-ish" sounds more like white balance.
 
DxO data is up and it's very interesting. The D4s shows a significant drop in DR for ISO 800 and 1600 vs the Df, then inverts starting at ISO 6400, peaking at ISO 12,800 where the D4s shows a 1/2 stop advantage. For SNR, the D4s is identical to the Df up to around ISO 51,200 and then takes a noticeable lead starting at ISO 102,400. This explains why the D4s scores lower for the composite score vs the Df - the composite is based on a specific cutoff of the SNR, DR, and color depth, and that usually occurs between ISO 1600 and 3200 on the Nikon FF sensors.
 
Last edited:
Horshack said:
DxO data is up and it's very interesting. The D4s shows a significant drop in DR for ISO 800 and 1600 vs the Df, then inverts starting at ISO 6400, peaking at ISO 12,800 where the D4s shows a 1/2 stop advantage. For SNR, the D4s is identical to the Df up to around ISO 51,200 and then takes a noticeable lead starting at ISO 102,400. This explains why the D4s scores lower for the composite score vs the Df - the composite is based on a specific cutoff of the SNR, DR, and color depth, and that usually occurs between ISO 1600 and 3200 on the Nikon FF sensors.
Click to expand...
It would be interesting to test a second body (DxO), since these are actually pretty small variations.

Here is the DR for the D4, D4s and Df. It seems that the D4s shows a smoother behavior than the other two, and the 6400 change is within margin of error, since it's the only point in the whole graph outside the very smooth behavior in the whole range, and very slightly so.

There is a slight high ISO advantage in favor of the D4s, which we are seeing in some photo tests, about 1/3-1/2 stop in the 3200+ range.

77696301dd904df7b11e3d033f07b8ec.jpg

--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/
Good shooting and good luck
(after Ed Murrow)
 
Last edited:
Horshack said:
DxO data is up and it's very interesting. The D4s shows a significant drop in DR for ISO 800 and 1600 vs the Df, then inverts starting at ISO 6400, peaking at ISO 12,800 where the D4s shows a 1/2 stop advantage. For SNR, the D4s is identical to the Df up to around ISO 51,200 and then takes a noticeable lead starting at ISO 102,400. This explains why the D4s scores lower for the composite score vs the Df - the composite is based on a specific cutoff of the SNR, DR, and color depth, and that usually occurs between ISO 1600 and 3200 on the Nikon FF sensors.
Click to expand...
The DR curve is one of the most "linear" for the Nikon designs, as per DxO.
 
Horshack said:
The D4s shows a significant drop in DR for ISO 800 and 1600 vs the Df, then inverts starting at ISO 6400, peaking at ISO 12,800 where the D4s shows a 1/2 stop advantage. For SNR, the D4s is identical to the Df up to around ISO 51,200 and then takes a noticeable lead starting at ISO 102,400. This explains why the D4s scores lower for the composite score vs the Df - the composite is based on a specific cutoff of the SNR, DR, and color depth, and that usually occurs between ISO 1600 and 3200 on the Nikon FF sensors.
Click to expand...
I think they need to redo the test.
 
Iliah Borg said:
Iliah Borg said:
The D4s shows a significant drop in DR for ISO 800 and 1600 vs the Df, then inverts starting at ISO 6400, peaking at ISO 12,800 where the D4s shows a 1/2 stop advantage. For SNR, the D4s is identical to the Df up to around ISO 51,200 and then takes a noticeable lead starting at ISO 102,400. This explains why the D4s scores lower for the composite score vs the Df - the composite is based on a specific cutoff of the SNR, DR, and color depth, and that usually occurs between ISO 1600 and 3200 on the Nikon FF sensors.
Click to expand...
I think they need to redo the test.
Click to expand...
Agreed and, as an engineer, I'm always amazed that people put so much emphasis on a site that tests 1 sample and then draws definitive conclusions on the model (talk about bad statistics!). Especially since the differences measured are quite small given the 5% measurement error I've seen quoted, let alone the likely sample variation?
 
jjnik said:
Iliah Borg said:
jjnik said:
The D4s shows a significant drop in DR for ISO 800 and 1600 vs the Df, then inverts starting at ISO 6400, peaking at ISO 12,800 where the D4s shows a 1/2 stop advantage. For SNR, the D4s is identical to the Df up to around ISO 51,200 and then takes a noticeable lead starting at ISO 102,400. This explains why the D4s scores lower for the composite score vs the Df - the composite is based on a specific cutoff of the SNR, DR, and color depth, and that usually occurs between ISO 1600 and 3200 on the Nikon FF sensors.
Click to expand...
I think they need to redo the test.
Click to expand...
Agreed and, as an engineer, I'm always amazed that people put so much emphasis on a site that tests 1 sample and then draws definitive conclusions on the model (talk about bad statistics!). Especially since the differences measured are quite small given the 5% measurement error I've seen quoted, let alone the likely sample variation?
Click to expand...
How were you able to establish that they performed a single sample from a single body?
 
Last edited:
Horshack said:
jjnik said:
Iliah Borg said:
Horshack said:
The D4s shows a significant drop in DR for ISO 800 and 1600 vs the Df, then inverts starting at ISO 6400, peaking at ISO 12,800 where the D4s shows a 1/2 stop advantage. For SNR, the D4s is identical to the Df up to around ISO 51,200 and then takes a noticeable lead starting at ISO 102,400. This explains why the D4s scores lower for the composite score vs the Df - the composite is based on a specific cutoff of the SNR, DR, and color depth, and that usually occurs between ISO 1600 and 3200 on the Nikon FF sensors.
Click to expand...
I think they need to redo the test.
Click to expand...
Agreed and, as an engineer, I'm always amazed that people put so much emphasis on a site that tests 1 sample and then draws definitive conclusions on the model (talk about bad statistics!). Especially since the differences measured are quite small given the 5% measurement error I've seen quoted, let alone the likely sample variation?
Click to expand...
How were you able to establish that they performed a single sample from a single body?
Click to expand...
I can't so I could very well be wrong - my bad and I shouldn't have thrown that out as fact. However, I have read many times that they only test 1 lens - there was much discussion on how they rated the new Canon 70-200 as worse than the old when the new one is far superior - so they likely had a bad copy. In any event, I doubt they have access to a large number of bodies to test and it may very well be only 1? It would be nice if they published that as without it, we can't fully understand if the extremely small differences shown are real or within expectations for sample variation and measurement error.
 
Horshack said:
Is smoothness a proxy for accuracy when measuring sensor perforomance?
Click to expand...
I agree that smoothness, per se, doesn't indicate accuracy.

But given what we know about what is being measured and my experience with about 60 or so PDR measurements I have come to expect smoothness in my PDR data (Canon intermediate ISOs not withstanding, for good reason :-) )
Horshack said:
Do sensors ever exhibit non-linear noise characteristics?
Click to expand...
Yes, of course; but they are almost always monotonically increasing/decreasing; the DxoMark data often have what appears to be obvious outliers.

Regards
 
Horshack said:
jjnik said:
Iliah Borg said:
Horshack said:
The D4s shows a significant drop in DR for ISO 800 and 1600 vs the Df, then inverts starting at ISO 6400, peaking at ISO 12,800 where the D4s shows a 1/2 stop advantage. For SNR, the D4s is identical to the Df up to around ISO 51,200 and then takes a noticeable lead starting at ISO 102,400. This explains why the D4s scores lower for the composite score vs the Df - the composite is based on a specific cutoff of the SNR, DR, and color depth, and that usually occurs between ISO 1600 and 3200 on the Nikon FF sensors.
Click to expand...
I think they need to redo the test.
Click to expand...
Agreed and, as an engineer, I'm always amazed that people put so much emphasis on a site that tests 1 sample and then draws definitive conclusions on the model (talk about bad statistics!). Especially since the differences measured are quite small given the 5% measurement error I've seen quoted, let alone the likely sample variation?
Click to expand...
How were you able to establish that they performed a single sample from a single body?
Click to expand...
Don't you think they would report the average and differences if testing several samples?
 
Iliah Borg said:
Horshack said:
How were you able to establish that they performed a single sample from a single body?
Click to expand...
Don't you think they would report the average and differences if testing several samples?
Click to expand...
Actually if it does turn out they do multiple tests and/or samples I would expect their results to represent the average. And for them not to report the standard deviation, esp. if it falls within what they typically see across samples. That said, I have no idea if they do multiple tests or samples - I would guess no.
 
Last edited:
bclaff said:
bclaff said:
Is smoothness a proxy for accuracy when measuring sensor perforomance?
Click to expand...
I agree that smoothness, per se, doesn't indicate accuracy.

But given what we know about what is being measured and my experience with about 60 or so PDR measurements I have come to expect smoothness in my PDR data (Canon intermediate ISOs not withstanding, for good reason :-) )
bclaff said:
Do sensors ever exhibit non-linear noise characteristics?
Click to expand...
Yes, of course; but they are almost always monotonically increasing/decreasing; the DxoMark data often have what appears to be obvious outliers.

Regards
Click to expand...
 
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