Nikon D850: Announcement of an Announcement of Development

[Jim Kasson]

You can't argue with measured numbers:

Landscape/studio cameras at low ISO: http://photonstophotos.net/Charts/PDR.htm#Nikon%20D810,Sony%20ILCE-7R,Sony%20ILCE-7RM2

D810 at ISO 64 - 11.6 stops (setting the floor at 20:1 SNR - DxO uses a 1:1 SNR, which is why the numbers are different but interconvertible)
A7r at ISO 100 - 11.71
A7r2 at ISO 100 - 11.42

Bill's PDR SNR depends on the resolution of the camera. It's 16000 divided by the image height in pixels. So for the D810 and the a7R, it's 3.26, and for the a7RII, it's 3.02. There is a question of how relevant this definition will be as sensor pitch decreases. When it reaches about 400 MP, PDR will equal EDR, and that doesn't seem right.

You mentioned that you can convert PDR to EDR and vice-versa. At least that what I think you said. I don't know how to do that. OTOH, given FWC and RN, I can compute both.

Jim

[shadowblade]

Bill's PDR SNR depends on the resolution of the camera. It's 16000 divided by the image height in pixels. So for the D810 and the a7R, it's 3.26, and for the a7RII, it's 3.02. There is a question of how relevant this definition will be as sensor pitch decreases. When it reaches about 400 MP, PDR will equal EDR, and that doesn't seem right.

You mentioned that you can convert PDR to EDR and vice-versa. At least that what I think you said. I don't know how to do that. OTOH, given FWC and RN, I can compute both.

Jim

His PDR values are normalised to a fixed resolution, as if the image were downsampled (or upsampled, for a low-resolution sensor). This way, output size is equalised and two sensors of differing resolutions can be fairly compared at the same output size.

DxO treats DR in the same way, normalising images to 8MP. Hence, DxO numbers can be converted to PDR numbers, just by taking the stated noise floor (1:1 vs 20:1, or whatever ratio is chosen) and compensating for it.

[Jim Kasson]

His PDR values are normalised to a fixed resolution, as if the image were downsampled (or upsampled, for a low-resolution sensor). This way, output size is equalised and two sensors of differing resolutions can be fairly compared at the same output size.

Yes, the normalization is accomplished by varying the target SNR as I explained above. I now think we probably agree on this point, but your phrasing confused me earlier.

DxO treats DR in the same way, normalising images to 8MP. Hence, DxO numbers can be converted to PDR numbers, just by taking the stated noise floor (1:1 vs 20:1, or whatever ratio is chosen) and compensating for it.

Please explain how you'd do that, since EDR has only RN, and PDR conflates RN and photon noise  [and PRNU, per Bill's advice below]in a quadrature summation. Knowing only RN or EDR won't tell you what the PDR is without additional information that lets you compute the shot noise, which is actually the point of having PDR in the first place. An exception is a very large image; the number is 384 MP.

{Added. A bit more detail on the PDR calculations. Bill and I both compute PDR by direct sample search, but our methods are quite different. I capture pairs and flat-field them to correct for lighting differences. Bill has another way of dealing with that. Nevertheless, we usually are in substantial agreement about PDR numbers for the cameras that we both test. He has in some cases taken samples from my cameras, and agreement is even better in those cases, since we are looking at the same serial number cameras, although not the same samples, since the capture protocols are different.

I did have a conversation with Bill about a model-based PDR computation, and he said that he sometimes uses one as a sanity check, but he cautioned me not to neglect the contribution of PRNU in the model, which surprised me;' i thought it would be negligible. ]

Jim

[JKoerner007]

More than you, probably. Background in the hard sciences - which, among other things, calls for the ability to evaluate and interpret data and methodology - with further research qualifications and a career based on applied science.

So? I'm a scientific/technical person. Wouldn't know the first thing about insurance laws or policies, and I'd fall asleep trying to read the first page.

That's about as much of a non sequitur as judging the skill of your brain surgeon based on his ability to play the guitar.

Notice how you're the only person who cops it, and only after you throw the first punch by launching a tirade of abuse at me and/or others?

More than me? Depends on what subject.

I threw the first punch? I don't think so. A quick re-read will prove the opposite.

You derided me, claiming I am not 'a technical person,' as if 'sensor specs' are the only way a person may be technically-proficient.

FYI, I have been a casualty (crime/loss) investigator since 1988. There are a lot of 'technical' laws, regulations, and insurance policy provisions that people need to navigate through ... which I have been handling for (probably) longer than you have been doing any 'one' thing in your life.

You seem to live in a world of fantasy; whereas I live in a world of brutal reality.

While you prognosticate about 'what Sony may do with sensors,' I deal on a daily basis with how people (like you) have to adjust to recovering from their devastating losses.

I am glad you admit you, "Wouldn't know the first thing about insurance laws or policies, and (that you'd) fall asleep trying to read the first page."

This means I know 1000x more about insurance law than you (think) you know about cameras/lenses.

If I recall correctly, you initiated a boo-hoo thread topic about your getting beat-up, having your gear stolen, and being stabbed ... after which you blamed 'your insurance company' for not covering your losses.

The truth is, your 'insurance company' isn't to blame, you are. Your lack of 'technical knowledge' regarding insurance policy provisions, applicable coverages (and your self-admitted) 'falling asleep' reading the fine print, all translated into your failure to adequately cover yourself with your own insurance choices. And, yes, what coverages YOU select are your choices.

I strongly suggest you learn to navigate the differences between RCV (Replacement Cost Value) and ACV (Actual Cash Value), as the monetary reimbursement differences can be huge. There are also Endorsement/Exclusion differences, which mean you are either covered when you travel abroad (or not), again based on your choices. Your lack of 'technical knowledge' in these respects was your own undoing. If you ever have any questions in these regards, please feel free to PM me, and I will be happy to guide you in the right direction, out of basic goodwill.

In the meantime, one more thing regarding your self-delusion demands being addressed. Your name, 'Shadowblade,' conjures-up images of Ninja-warriors ... of athleticism, stealth, sword-play, power, and evasion. I think the recent theft of all your gear, leaving you sullied and bereft, proves that the perp was the true 'Shadowblade' (who went in-and-out and escaped with the goods) ... while you were left curled-up in ball, bleeding, only to find your other choices translated to no coverage: nothing.

Perhaps some humility would suit you ... and, perhaps, you should consider a change in your chosen handle?

I think 'Mark' would be more appropriate.

Jack

PS: I am not going to bother with your other posts. All they prove is the differences in sensor performances are becoming more-and-more negligible these days, and are not worth going back-and-forth over ad infinitum, any longer.

The real differences these days now lie in lens selection and personal preferences.

Allow me to enjoy my preferences, and I will allow you to enjoy yours. Thanks.

[Michael Erlewine]

As the OP, I suggest we stop all of this and return to the main topic, the forthcoming Nikon D850.

[Jim Kasson]

As the OP, I suggest we stop all of this and return to the main topic, the forthcoming Nikon D750.

Isn't it true that anyone who knows any more than what's in the non-announcement is prohibited by employment contract or NDA from posting what they know, and that therefore anything posted here about what the camera will be is pure speculation?

Jim

[JKoerner007]

As the OP, I suggest we stop all of this and return to the main topic, the forthcoming Nikon D750.

D850 ... but I agree.

[Bernard ODonovan]

I will speculate that this will not happen:

https://nikonrumors.com/wp-content/uploads/2017/07/Nikon-Museum-in-Tokyo-89.jpg

IE they will deliver the D850...

[hogloff]

As the OP, I suggest we stop all of this and return to the main topic, the forthcoming Nikon D850.

At the very least let's focus back to photography rather than laying out personal resumes.

[shadowblade]

Yes, the normalization is accomplished by varying the target SNR as I explained above. I now think we probably agree on this point, but your phrasing confused me earlier.

Please explain how you'd do that, since EDR has only RN, and PDR conflates RN and photon noise  [and PRNU, per Bill's advice below]in a quadrature summation. Knowing only RN or EDR won't tell you what the PDR is without additional information that lets you compute the shot noise, which is actually the point of having PDR in the first place. An exception is a very large image; the number is 384 MP.

DxO's published DR results aren't an EDR either - they're also a PDR, based on sensor size and resolution. The main difference between DxO and the other website is the level they chose to use as the noise floor cutoff (1:1 vs 20:1).

[shadowblade]

More than me? Depends on what subject.

I threw the first punch? I don't think so. A quick re-read will prove the opposite.

This was your first post after my very first post in this thread:

Everything you wrote is based on this one sentence, which is not a fact, but 100% pure speculation on your part.

This was your very next post after that:

Your anti-Nikon posture makes you project this "wishful thinking" belief, so you can feel your brand-preference "is in a class by itself," but the fact is Nikon sensors are already best in class at the DX, low-end, and same at the high-end. The reason Sony doesn't take your posture is they are not that small-minded. The truth is, Sony is too big to be affected by the camera market. They have their own unique niche there, but they also have a much bigger role as a commercial supplier of sensors.

And your third:

Your posts belie this claim. You are a Canon fanboy converted to a Sony fanboy.

You derided me, claiming I am not 'a technical person,' as if 'sensor specs' are the only way a person may be technically-proficient.

Nothing to do with sensor specs, but data interpretation. Something that's common to all scientific fields - the ability to read data tables and charts, to understand and critique the methodology by which the data was derived and to infer conclusions from the data (including the lack of any conclusions, if the data is insufficient or the methodology flaky). It was applicable to hard sciences (of which physics and mathematics - the sciences behind optics - are two) and is equally applicable to optics, medical science, biostatistics and even military science.

So far, you've never produced any data that backs your claims, or any data interpretation beyond 'X is bigger than Y' (without any understanding of what X and Y are actually measuring).

All you've done is screech rhetoric and brandish Lenscore overall scores, like a partisan lawyer or politician, without any understanding of what the scores actually mean (an understanding which no-one has, since Lenscore themselves don't publish how they got their results or what each of their scores/categories actually measures).

You seem to live in a world of fantasy; whereas I live in a world of brutal reality.

No, you live in a world of man-made clauses, contracts and words. By their very nature, these are mutable, subjective and open to interpretation.

I deal with objective realities. When someone has a blood pressure of 55/30, they have a blood pressure of 55/30, no matter how you want to spin it. When someone has a penetrating injury to their common femoral artery, they have an immediately life-threatening vascular injury, no matter how you want to spin it. The exact words you use to describe the circumstances don't change the situation, the necessary actions or the outcome, and aren't particularly helpful to either the patient or the treating team.

While you prognosticate about 'what Sony may do with sensors,' I deal on a daily basis with how people (like you) have to adjust to recovering from their devastating losses.

I am glad you admit you, "Wouldn't know the first thing about insurance laws or policies, and (that you'd) fall asleep trying to read the first page."

This means I know 1000x more about insurance law than you (think) you know about cameras/lenses.

If I recall correctly, you initiated a boo-hoo thread topic about your getting beat-up, having your gear stolen, and being stabbed ... after which you blamed 'your insurance company' for not covering your losses.

The truth is, your 'insurance company' isn't to blame, you are. Your lack of 'technical knowledge' regarding insurance policy provisions, applicable coverages (and your self-admitted) 'falling asleep' reading the fine print, all translated into your failure to adequately cover yourself with your own insurance choices. And, yes, what coverages YOU select are your choices.

I strongly suggest you learn to navigate the differences between RCV (Replacement Cost Value) and ACV (Actual Cash Value), as the monetary reimbursement differences can be huge. There are also Endorsement/Exclusion differences, which mean you are either covered when you travel abroad (or not), again based on your choices. Your lack of 'technical knowledge' in these respects was your own undoing. If you ever have any questions in these regards, please feel free to PM me, and I will be happy to guide you in the right direction, out of basic goodwill.

I don't need to know any of that. I pay an agent to do it. Just like you probably pay someone else to perform surgery on you (before you ask, yes, I have performed surgery on myself before).

They screwed up. They got me a policy that didn't do the job.

My travel insurance was meant to cover everything, including cameras. That's why it cost so much. No point having dedicated camera insurance when, unless I'm travelling, it's sitting at home and covered by home and contents insurance anyway.

In the meantime, one more thing regarding your self-delusion demands being addressed. Your name, 'Shadowblade,' conjures-up images of Ninja-warriors ... of athleticism, stealth, sword-play, power, and evasion. I think the recent theft of all your gear, leaving you sullied and bereft, proves that the perp was the true 'Shadowblade' (who went in-and-out and escaped with the goods) ... while you were left curled-up in ball, bleeding, only to find your other choices translated to no coverage: nothing.

Perhaps some humility would suit you ... and, perhaps, you should consider a change in your chosen handle?

I think 'Mark' would be more appropriate.

One part's a call-sign, the other is a sports team. And, yes, the call-sign was mine. Medical evacuation in the Himalayas, rappelling from helicopters to assess and stabilise injured and sick climbers and hikers before evacuating them to appropriate facilities.

Anyone can get bashed by a group when unarmed and alone.

And why would I choose 'Mark', of all names?

[Jim Kasson]

DxO's published DR results aren't an EDR either - they're also a PDR, based on sensor size and resolution. The main difference between DxO and the other website is the level they chose to use as the noise floor cutoff (1:1 vs 20:1).

If it's 1:1 it's virtually EDR.

https://www.dpreview.com/forums/post/55664246

But forget what you call it, my original question still stands: how do you get from one to the other without some other information?


Jim

[JKoerner007]

And why would I choose 'Mark', of all names?

Your 'book smart' cluelessness is cute

I will let you think about that for awhile ... and will respond to the rest at another time

Gotta head out for a bit.

Have a nice weekend.

[bclaff]

Bill's PDR SNR depends on the resolution of the camera. It's 16000 divided by the image height in pixels. So for the D810 and the a7R, it's 3.26, and for the a7RII, it's 3.02. There is a question of how relevant this definition will be as sensor pitch decreases. When it reaches about 400 MP, PDR will equal EDR, and that doesn't seem right.

At that point I'll simply pixel-bin differently. Should work out OK.

You mentioned that you can convert PDR to EDR and vice-versa. At least that what I think you said. I don't know how to do that. OTOH, given FWC and RN, I can compute both.

Right. There is no easy way from PDR to EDR because PDR is SNR based not read noise based.
With FWC (gain) and RN you can get "close enough for government work", but even FPN can affect the SNR-based PDR.

Regards

[bclaff]

...
I did have a conversation with Bill about a model-based PDR computation, and he said that he sometimes uses one as a sanity check, but he cautioned me not to neglect the contribution of PRNU in the model, which surprised me;' i thought it would be negligible. ]

Perhaps I misspoke; DSNU might matter, PRNU wouldn't. :-)

Regards

[shadowblade]

If it's 1:1 it's virtually EDR.

https://www.dpreview.com/forums/post/55664246

But forget what you call it, my original question still stands: how do you get from one to the other without some other information?


Jim

Well, EDR on a whole-sensor level, perhaps. But that's not photosite-level dynamic range - it's a value that's normalised to a standard resolution, which is much more applicable to actual use than the dynamic range of any individual photosite.

I'm not sure there is a formula for direct conversion using the DR value alone. DxO's DR calculation is based off read noise and FWC alone, whereas PDR uses total SNR. Converting between the two, with a simple subtraction to account for the different SNR criteria gets you a working approximation that's good enough for most purposes (a bit like saying that sqrt(2) = 1.41 - it doesn't, but it's close enough for most purposes). But DxO also publish separate charts for SNR at 18% grey results, which takes into account all noise sources. This should be more useful if you need an exact result. Not sure of the exact formula or the exact criteria used for PDR - you'll have to ask him that.

[bclaff]

Well, EDR on a whole-sensor level, perhaps. But that's not photosite-level dynamic range - it's a value that's normalised to a standard resolution, which is much more applicable to actual use than the dynamic range of any individual photosite.

Engineering DYnamic Range (EDR) is based on pixel level and is not normalized.

I'm not sure there is a formula for direct conversion using the DR value alone.

There isn't.

DxO's DR calculation is based off read noise and FWC alone,

No, just read noise normalized for resolution; FWC is not involved.

whereas PDR uses total SNR. Converting between the two, with a simple subtraction to account for the different SNR criteria gets you a working approximation that's good enough for most purposes (a bit like saying that sqrt(2) = 1.41 - it doesn't, but it's close enough for most purposes).

Often not so close as you think.

But DxO also publish separate charts for SNR at 18% grey results, which takes into account all noise sources. This should be more useful if you need an exact result. Not sure of the exact formula or the exact criteria used for PDR - you'll have to ask him that.

Yeah, I use DxOMark SNR data to compute a DxOMark derived PDR value which I publish on PhotonsToPhotos as a fall back for cameras that I have not tested.

BTW, I think we're pretty far afield; maybe we should return to the D850?

[shadowblade]

Saw that you had replied while I was typing up my previous reply.

Engineering DYnamic Range (EDR) is based on pixel level and is not normalized.

That's what I've been trying to say - DxO doesn't provide the EDR, but basically normalises it around an 8MP sensor, which (for most sensors) gives a higher 'whole image' DR value than the actual EDR.

There isn't.

I think the confusion was over whether it was possible to calculate PDR using DxO data (which it seems to be) vs doing it from DxO's DR value alone.

Yeah, I use DxOMark SNR data to compute a DxOMark derived PDR value which I publish on PhotonsToPhotos as a fall back for cameras that I have not tested.

Thought you might. I didn't know whether you performed your own tests or used the data from DxO to calculate PDR values.

[jeremyrh]

BTW, I think we're pretty far afield; maybe we should return to the D850?

Please. Pretty please.

[bjanes]

BTW, I think we're pretty far afield; maybe we should return to the D850?

+1. However, the information that you and Jim have provided was interesting. Furthermore, since no one who can post here has any real data on the D850 it might be best to defer further discussion on the D850 until further data are available.

Bill