NEW DATA Is the D750 still the low ISO champion???

[JKoerner007]

preponderance of evidence may apply in legal matters but it's not compelling in a scientific discussion.

It isn't?

The entire concept of "peer reviewed" findings is predicated on the concept of "a preponderance of evidence," is it not?

The more samplings, the more consistent the results, and the more peers who agree, the quicker a hypothesis is accepted as fact.

By contrast, the fewer samplings, the more disparity in results, and the greater number of disagreement among peers, the less likely a hypothesis is going to be accepted as fact

[bclaff]

It isn't?

The entire concept of "peer reviewed" findings is predicated on the concept of "a preponderance of evidence," is it not?

The more samplings, the more consistent the results, and the more peers who agree, the quicker a hypothesis is accepted as fact.

By contrast, the fewer samplings, more disparity in results, and the greater number of disagreement among peers, the less likely the hypothesis is correct

Science is not a democracy. The preponderance of peer opinion doesn't determine fact.
Peer review is important and you might be surprised at how many of my peers have reviewed and agree with my work.
(I actively seek these people out in private.)
Regarding this subject you are not a peer, nor is the Nikon marketing department.
In back channels that I cannot disclose I have no negative feedback from any camera company engineering department regarding my results.

I accept that you think my approach is wrong. Further discussion here will not be fruitful. Respond to my PM if you really want to go further.

[Wannabebetter]

It isn't?

The entire concept of "peer reviewed" findings is predicated on the concept of "a preponderance of evidence," is it not?

The more samplings, the more consistent the results, and the more peers who agree, the quicker a hypothesis is accepted as fact.

By contrast, the fewer samplings, the more disparity in results, and the greater number of disagreement among peers, the less likely a hypothesis is going to be accepted as fact

How many noes do you want? There is much in natural philosophy to which there is no legitimate remedy despite [any] "preponderance of evidence", contra proferentem. For one, I find the Scottish conclusion, verdict "not proven" delightfully refreshing as it is infuriating. Cheers!

[David H. Hartman]

May I insert my foot into my mouth?

It seems to me that the difference in DR is small and RAW files pliable enough in post processing that differences can be accommodated provided a single channel is not blow out or clipped hard.

I've seen red flowers where the red channel is blown out and obvious on my computer display but not camera LCD and a while back I was shooting bees on lavender and decided I should check the blue channel and yes it was blow out. The blinkies in white don't hint of either problem. This was a D800.

I would be dissatisfied with the control set of a D750 so I'd prefer a D8XXx, any one, to the D750. A slight low ISO advantage cannot change this.

The viewfinder, AF and tilting LCD of the D850 has me foaming at the mouth.

Dave Hartman

[bclaff]

... that differences can be accommodated provided a single channel is not blow out or clipped hard.

I've seen red flowers where the red channel is blown out and obvious on my computer display but not camera LCD and a while back I was shooting bees on lavender and decided I should check the blue channel and yes it was blow out. The blinkies in white don't hint of either problem. This was a D800.
...

Yeah, Highlights is unreliable which is why I have my Nikons set for RGB histogram.
Flowers (don't forget yellow ) often need negative exposure compensation.
Regards,

[Frank Fremerey]

Yeah, Highlights is unreliable which is why I have my Nikons set for RGB histogram.
Flowers (don't forget yellow ) often need negative exposure compensation.
Regards,

I suggested an overflow counter or a rate counter long ago. Then there is no maxing out, the PDR has no limit. I wanted to patent this ten years ago, but Kodak was faster: http://www.freepatentsonline.com/6069377.html

[OCD]

May I insert my foot into my mouth?

It seems to me that the difference in DR is small and RAW files pliable enough in post processing that differences can be accommodated provided a single channel is not blow out or clipped hard.

I've seen red flowers where the red channel is blown out and obvious on my computer display but not camera LCD and a while back I was shooting bees on lavender and decided I should check the blue channel and yes it was blow out. The blinkies in white don't hint of either problem. This was a D800.

I would be dissatisfied with the control set of a D750 so I'd prefer a D8XXx, any one, to the D750. A slight low ISO advantage cannot change this.

The viewfinder, AF and tilting LCD of the D850 has me foaming at the mouth.

Dave Hartman

I agree with David.  The sensors are all terrific and in my opinion have been since around 2012.  The good news as a result of these fabulous sensors is we can rather choose the camera we have the most comfort and fun with and which has the features that cause us to foam at the mouth.  And good point regarding the processing of the files since whenever the image rendered by the camera somehow "falls short" of our objective we can most likely square things away on the computer.  I'm primarily a JPEG guy (i.e. happy snapper retired hobbyist guy) and for the most part am happy with the rendering of the camera's JPEG files, although I shoot RAW/JPEG just in case I need the pliability that David alludes to, a rare occurrence, but kind of cool to have the RAW file, it's like an insurance policy of sorts.  And finally, the histogram doesn't lie, I love that feature.

[rs]

I understand your reasoning and it's a very common misconception ("common man" language versus technical distinction).
The fact is you can measure a signal that is below read noise; so it is distinguishable.
In any case, any non-zero signal will have photon noise in addition to read noise therefore noise at SNR = 1 is greater than read noise.
However, as I said, in practice it is of no consequence.

A nice illustration of this concept, by Jim Kasson, can be seen here:

http://blog.kasson.com/the-last-word/detectability-of-visual-signals-below-the-noise-part-2/

Regards,

Richard

[JKoerner007]

Science is not a democracy. The preponderance of peer opinion doesn't determine fact.

Very disappointed in the word twisting, Bill (either that, or I wasn't clear):

I never said science was a democracy nor did I say facts were a preponderance of peer 'opinion.'

What I said was, a preponderance of evidence, consistently obtaining, when tested amongst qualified peers, is what determine accepted scientific fact.

Peer review is important and you might be surprised at how many of my peers have reviewed and agree with my work.
(I actively seek these people out in private.)

This is expected and important, and is pretty much what I said.

Essentially, you're agreeing with my point: scientific fact is based on a preponderance of evidence reliably and consistently obtaining, after being independently-measured/confirmed by qualified peers.

By contrast, 'one guy' claiming results, with no disclosure as to method, and with no accepted peer review, does not translate to scientific fact. It begs for controlled scrutiny.

My point was, only after accepted peer review do hypotheses/theories become accepted as fact. In essence you're agreeing with me.

Another difference is, true science offers published results, not 'private, undisclosed affirmations.'

Bill, I know several scientists, and my photos have been used in published 'new species' descriptions.

When that happens, when new species are described, and accepted, it does not occur via 'private discussion/approval,' but through published, publicly-documented, peer-reviewed disclosure.

Regarding this subject you are not a peer, nor is the Nikon marketing department.

True. However, I stated my agnosticism came (not from my own expertise) but from the fact there *are* peers of yours who reach different conclusions (DxO), which produce differing results.

And, while Nikon's marketing department may not be your peers, they likely get their marketing info from Nikon's engineers, who are your peers, and likely quite a bit more, in various important areas.

I looked up your CV; you're a highly-educated software developer, not a sensor developer.  My brother is a software developer (financial loan calculations, to be exact.) So I understand "proof of concept," etc.

I also understand that two different software developers can create different solutions to the same problem, and one developer will likely develop a cleaner solution than the other.
(This applies to antivirus software, or whatever.)

This brings us back to the point: differing results among peers.

In back channels that I cannot disclose I have no negative feedback from any camera company engineering department regarding my results.

Fair enough, I have no way to know.

In the end, though, published, affirming feedback is what's important ... not a claimed 'private, lack of negative' feedback.

I accept that you think my approach is wrong.

Another misrepresentation.

I am not qualified to debate your approach. I said that back in the beginning.

What I said was, I am unconvinced of "a finding" ... a single graph result ... not of you (or of your whole approach).

That you are the only entity (of whom I am aware) placing the D750 at 'the DR pinnacle,' is what prompted me to question that single graph result, nothing more.

It's very simple: if Nikon corporation fronted the D750 as its finest DR entry, and if DxO Mark concurred with your findings, lauding the D750 in its #1 DR spot ... then we wouldn't be having this discussion ... as your graph result would be universally-accepted.

However, when Nikon downplays its own D750, offering the D850 as its finest DR entry, and with DxO lauding the D850 as "the first, ever" DSLR to reach a new milestone ... pushing the D750 down to #11 in their ranking ...  then I think it is reasonable for me (or anyone) to doubt your 'stand alone' result. I do not have to be your mathematical or software engineering peer to notice this discrepancy. Anymore than I have to be an accident recon specialist to accept, or reject, a potential expert witness as a candidate.

You admitted long ago sampling err is possible and that a re-test would be reasonable.

Your results, in this instance, are not universally-understood or accepted by all your peers.

Further discussion here will not be fruitful. Respond to my PM if you really want to go further.

I agree and I don't.

The issue is ultimately trivial, and my apologies if anything said above was unclear.

[bclaff]

A nice illustration of this concept, by Jim Kasson, can be seen here:
http://blog.kasson.com/the-last-word/detectability-of-visual-signals-below-the-noise-part-2/

Jim does great work and is one of the peers I consult.
He has also replicated my PDR work particularly for Sony cameras.

Regards,

[Frank Fremerey]

@Bill: Very helpful thread for people interested more deeply in these questions. Thank you!

[bclaff]

What I said was, a preponderance of evidence, consistently obtaining, when tested amongst qualified peers, is what determine accepted scientific fact.
...
And, while Nikon's marketing department may not be your peers, they likely get their marketing info from Nikon's engineers, who are your peers, and likely quite a bit more, in various important areas.

I don't accept your "chain of evidence" in assuming that Nikon marketing correctly conveys information from Nikon engineering.
FWIW, I used to write published articles and I can't tell you how many times editors tried to improve my wording but changed the meaning so I would have to explain to them why they needed to retain the original wording.
With marketing material there are multiple points at which things can get lost in the translation.

I looked up your CV; you're a highly-educated software developer, not a sensor developer...

I never claim to be a sensor developer; but I have been analyzing sensors for over 10 years (longer than DxOMark has existed).

...
I am not qualified to debate your approach. I said that back in the beginning.
What I said was, I am unconvinced of "a finding" ... a single graph result ... not of you (or of your whole approach).
That you are the only entity (of whom I am aware) placing the D750 at 'the DR pinnacle,' is what prompted me to question that single graph result, nothing more.
It's very simple: if Nikon corporation fronted the D750 as its finest DR entry, and if DxO Mark concurred with your findings, lauding the D750 in its #1 DR spot ... then we wouldn't be having this discussion ... as your graph result would be universally-accepted.
...

I take this to be "only" one of two since I don't accept your Nikon marketing "evidence" but only actual measurements as evidence.

You admitted long ago sampling err is possible and that a re-test would be reasonable.
...
The issue is ultimately trivial, and my apologies if anything said above was unclear.

The issue is becoming clearer and the underlying data issue is not trivial.
Regarding the D750 I am collecting additional data.
Initial indications are that the D750 PDR (as currently published) is in fact too high (!)
I see an anomaly in the initial data collection that went undetected.
I'm surprised because normally bad input data produces obviously wrong (usually quite low) results.
If the new results hold up then it will be the first time ever that bad input data produced unnaturally high PDR values.

So, with respect to the initial topic of the thread, your observation may be born out.
However, I stand by my assertion that PDR is a better measure than DxOMark Landscape score.
And I stand by my methodology regardless of any human error that might affect a specific result.

[bclaff]

Here is the comparison D750 vs D850 when it comes to PDR:

New D750 Photographic Dynamic Range (PDR) data has been collected and published which I discuss at length elsewhere.
As it turns out the D750 comes up 0.14 stops short of retaining the crown of "low ISO champion".
I'm attaching a new PDR comparison.
Although no longer the "champ" the D750 still acquits itself quite well; particularly below ISO 400 before dual conversion gain kicks in for the D850.

[golunvolo]

I have been using a d750 this weekend for still and video. Very very good low light performance from this camera is no surprise. In Spain you can get one used for little more than a 1000€ and that is an amazing IQ for the price. 24 mpx it is also a sweet spot for my type of photography. Not the perfect tool but in that light -pun intended- 0.1 steps less at low iso is negligible; and as good if not better in the higher sensitivities. Just to put this thread in some personal perspective.

  I do lust for a d850 for speed, focus and silent shutter applications. 

[Frank Fremerey]

Very interesting unexpected turn of events.

As concluded before there is only infinitesimal advancement   in the field of PDR since the D600/D800/800E appeared.

To remind everyone: there are major advancements in color fidelity and consistency over a wide range of lighting and ISO and in the field of autofocus.