[biz-engineer]

The capacity of each photosite is much greater and that quite costly to make.

Could you provide references? I'd be interested.

From a camera user standpoint, good DR at 64 ISO is not as good as having the same DR at 100 ISO. Take BSI sensors for example: higher quantum efficiency is the small advantage of BSI over non BSI sensor technology. Higher quantum efficiency of BSI offers increased native sensitivity (ISO) with same max DR as non BSI sensors. BSI is considered an improvement over non BSI, it was not used in large sensor because it requires TSV, hence more manufacturing steps and it is more tricky to produce with high yield (yield = percentage of good sensors that pass the production tests).

Technically if you look at the K1, it can achieve the SNR+DR at ISO100 that the a D810 can achieve at ISO64. Now, in you opinion, which one is better? 14ev DR at ISO100 or 14ev DR at ISO64?

Anyway, DR is clamped by either the 14bits ADC, or the breakdown voltage of the PIN junction.
The smallest value that can be coded by the ADC is 2^14 smaller than the larger value, that's the max theoretical pixel level DR achievable.
With regards to the breakdown voltage of the PIN junction, do you realize that one the output voltage of the sensel reaches 7 volts, it needs to go up to 14 volts in order to get 1ev of extra DR, and 28 volts to get 2ev of extra DR (and think about clocking the ADC at 400MHz under 14V or 28V supply... good luck with that). No wonder why even after 20 years of efforts, no consumer sensor maker can offer more than 14ev of DR. That is why the easy way to get best possible IQ is to use a larger sensor, it's as simple as that.

Pentax pixel shift is not very practical, but from the signal processing standpoint it is a very good idea because it is increasing the DR beyond what the 14 ADC limit and how? Moving the sensor and stacking the pixel of difference silicon area averages out of the read noise (quantization noise of the ADC) increases resolution. So, in theory, if the ADC is perfect, pixel shift can get 1 extra bit over the 14bits, but that only 15ev. Sigma delta converters work the same way, it starts from a 1bit converter (comparator) into an integration loop followed by low pass filtering. Please do some research before you throw the baby in water like a bad child.

---------- Post added 08-09-17 at 08:53 ----------

I would love to have iso 64 on my D80o and is one of my biggest complaints.

Well, users of D810 say they CANT tell the difference between 100 ISO and 64 ISO...

---------- Post added 08-09-17 at 08:55 ----------

The problem in photography is that would need 20ev of dynamic range. So, whether Nikon can do 0.3ev more than Pentax, is just a marketing gimmick in comparison to lighting situations found in nature.

---

Last edited by biz-engineer; 09-08-2017 at 10:41 AM.

[normhead]

Well, users of D810 say they can tell the difference between 100 ISO and 64 ISO...

People say all kinds of s#!^.

What you see is an output device or print, none of which come anywhere close to even 10 EV. No one sees the actual 13 EV captured. It's just a bunch of 1s and 0s. There is no physical device you can look at that can display 10 EV.

The scientific unreliability of claims made without double blind testing is really good. Anecdotal evidence is again practically guaranteed to be un-reliable.

[Rondec]

Could you provide references? I'd be interested.

From a camera user standpoint, good DR at 64 ISO is not as good as having the same DR at 100 ISO. Take BSI sensors for example: higher quantum efficiency is the small advantage of BSI over non BSI sensor technology. Higher quantum efficiency of BSI offers increased native sensitivity (ISO) with same max DR as non BSI sensors. BSI is considered an improvement over non BSI, it was not used in large sensor because it requires TSV, hence more manufacturing steps and it is more tricky to produce with high yield (yield = percentage of good sensors that pass the production tests).

Technically if you look at the K1, it can achieve the SNR+DR at ISO100 that the a D810 can achieve at ISO64. Now, in you opinion, which one is better? 14ev DR at ISO100 or 14ev DR at ISO64?

Anyway, DR is clamped by either the 14bits ADC, or the breakdown voltage of the PIN junction.
The smallest value that can be coded by the ADC is 2^14 smaller than the larger value, that's the max theoretical pixel level DR achievable.
With regards to the breakdown voltage of the PIN junction, do you realize that one the output voltage of the sensel reaches 7 volts, it needs to go up to 14 volts in order to get 1ev of extra DR, and 28 volts to get 2ev of extra DR (and think about clocking the ADC at 400MHz under 14V or 28V supply... good luck with that). No wonder why even after 20 years of efforts, no consumer sensor maker can offer more than 14ev of DR. That is why the easy way to get best possible IQ is to use a larger sensor, it's as simple as that.

Pentax pixel shift is not very practical, but from the signal processing standpoint it is a very good idea because it is increasing the DR beyond what the 14 ADC limit and how? Moving the sensor and stacking the pixel of difference silicon area averages out of the read noise (quantization noise of the ADC) increases resolution. So, in theory, if the ADC is perfect, pixel shift can get 1 extra bit over the 14bits, but that only 15ev. Sigma delta converters work the same way, it starts from a 1bit converter (comparator) into an integration loop followed by low pass filtering. Please do some research before you throw the baby in water like a bad child.

---------- Post added 08-09-17 at 08:53 ----------


Well, users of D810 say they can tell the difference between 100 ISO and 64 ISO...

---------- Post added 08-09-17 at 08:55 ----------

The problem in photography is that would need 20ev of dynamic range. So, whether Nikon can do 0.3ev more than Pentax, is just a marketing gimmick in comparison to lighting situations found in nature.

I do question it a bit. There is supposed to be 0.4 EV difference in dynamic range between iso 64 and iso 100 on the Nikon. Seeing that sort of difference is questionable, but maybe a bit more of a placebo effect? Maybe a little like how every time there is a firmware update there are a group of people who are convinced that their auto focus is more reliable and faster.

The biggest thing I see that iso 64 gives you is an extra stop to play with if you want a long exposure in bright light, but you are still better going with some sort of ND filter -- 1 stop is seldom enough.

[normhead]

The biggest thing I see that iso 64 gives you is an extra stop to play with if you want a long exposure in bright light, but you are still better going with some sort of ND filter -- 1 stop is seldom enough.

If you can't fake a stop, you aren't a photographer. It's the composition that counts. for 90% of whether or not folks like an image. The technical affects the last 10%. But only if it's a compelling composition. If your image has no appeal, no one care about 50 stops, forget about 1.

[biz-engineer]

People say all kinds of s#!^.

oupppsss typo mistake (corrected). Nikon users report not to see a difference in image between shooting at ISO100 and ISO64, except the apparent shutter speed is slower.
Pixel shift is a different story, in terms of shutter speed is equivalent to 4 exposure times, so it could eventually equate shooting at ISO25.

If you realize it, the Pentax K1 in pixel shift mode is the world highest image quality for the 24x36 format. Only the Sony A7S comes close to it in terms of noise , but without the resolution of the Pentax K1.

[Rondec]

If you can't fake a stop, you aren't a photographer. It's the composition that counts. for 90% of whether or not folks like an image. The technical affects the last 10%. But only if it's a compelling composition. If your image has no appeal, no one care about 50 stops, forget about 1.

Oh, I don't know. It's harder than it sounds in bright daylight to get a 1/8 second exposure without going to f22 or just blowing out your highlights. I don't shoot waterfalls every day, but there are times that I want to drag my shutter a bit and having a lower iso, even without extra dynamic range would be handy.

I suppose that's why they make ND filters.

[biz-engineer]

Actually, with K1, I now tend to shoot at ISO200 with highlight correction On as default setting because images are so clean that 100 ISO is not a must, ISO200 gives you faster shutter speed , plus 1 stop of highlight headroom on the RAW files.

[normhead]

Oh, I don't know. It's harder than it sounds in bright daylight to get a 1/8 second exposure without going to f22 or just blowing out your highlights. I don't shoot waterfalls every day, but there are times that I want to drag my shutter a bit and having a lower iso, even without extra dynamic range would be handy.

I suppose that's why they make ND filters.

That might be the exception. There's barely noticeable image degradation from diffraction, and there is unmistakeable degradation. Often ƒ22 falls into the later category. But that is best solved with neutral density filters.

[patarok]

sry... this may sound crazy to some of you, if you remember what i posted until now.(But, ASB being honest is what will take us further... and i dont see my recent postings corrupted by the following.)

If i had to decide, if i would take a Nikon D810 or a Pentax K-1... (actually I am in doubt if DxO has the integrety people think it has.)
Man... i tell you, i would still take the K-1 even if AF-area may be smaller and not as dense as on Nikons.

And i tell you why. I already had the chance to put the K-1 to the test against a Nikon D850...
It did the trick nearly as good as the Nikon D850 ... both had mounted a 24-70 2.8

And despite the fact, that the D850 has the same AF-sensing as the D500/D5 ... its AF seems to perform nearly nowhere the D5/D500...
that may be due to the fact that there is so much data( running through the system whilst it would also have to process AF (sensing AF and EV for every shot)algorithms.

alas, both cameras performed not so good when it came to tracking movement on the Z-axis...(towards and away from the sensor -frontally)
I had set the Nikon D850 to "Dynamic Area"-Mode and the Pentax K-1 to SEL-S(or was it M? ...)
both set on AF-C resp. AF.C


Anyway, the only camera that really is good at that seems to be the Sony A9....
bottomline... if you own Pentax glass: keep it and stop thinkin about Nikon D850, or try it and realize that you will rather buy a used K-1 extra as a backup camera before you jump ship

Only case that it may be wise to jump ship (to D850) is, if you really feel in need for faster bursts with longer shooting in a row..-bigger buffer

and just because someone said, he wouldnt know what the fun-factor is. pls tell it to the marines... since i have written "xxx, xxx, good video - fun factor" u should know, that i meant having good video on board of a DSLR that costs more than 1000 bucks is maybe not needed but it makes fun because if you do video with it, you can really show off the material. Instead of asking yourself why fast moving cars rather seem to jump their way through the frame instead of driving...
(smooth movement is only achieved at a good framerate)

[normhead]

sry... this may sound crazy to some of you, if you remember what i posted until now.(But, ASB being honest is what will take us further... and i dont see my recent postings corrupted by the following.)

If i had to decide, if i would take a Nikon D810 or a Pentax K-1... (actually I am in doubt if DxO has the integrety people think it has.)
Man... i tell you, i would still take the K-1 even if AF-area may be smaller and not as dense as on Nikons.

And i tell you why. I already had the chance to put the K-1 to the test against a Nikon D850...
It did the trick nearly as good as the Nikon D850 ... both had mounted a 24-70 2.8

And despite the fact, that the D850 has the same AF-sensing as the D500/D5 ... its AF seems to perform nearly nowhere the D5/D500...
that may be due to the fact that there is so much data( running through the system whilst it would also have to process AF (sensing AF and EV for every shot)algorithms.

alas, both cameras performed not so good when it came to tracking movement on the Z-axis...(towards and away from the sensor -frontally)
I had set the Nikon D850 to "Dynamic Area"-Mode and the Pentax K-1 to SEL-S(or was it M? ...)
both set on AF-C resp. AF.C


Anyway, the only camera that really is good at that seems to be the Sony A9....
bottomline... if you own Pentax glass: keep it and stop thinkin about Nikon D850, or try it and realize that you will rather buy a used K-1 extra as a backup camera before you jump ship

Only case that it may be wise to jump ship (to D850) is, if you really feel in need for faster bursts with longer shooting in a row..-bigger buffer

and just because someone said, he wouldnt know what the fun-factor is. pls tell it to the marines... since i have written "xxx, xxx, good video - fun factor" u should know, that i meant having good video on board of a DSLR that costs more than 1000 bucks is maybe not needed but it makes fun because if you do video with it, you can really show off the material. Instead of asking yourself why fast moving cars rather seem to jump their way through the frame instead of driving...
(smooth movement is only achieved at a good framerate)

I just wasted my whole bath wondering if the D850 was the perfect camera. I want my bath time back. I been had.

[Nicolas06]

It is tough for me to see a situation where you could use iso 64 and not use pixel shift. If the point is to get better dynamic range, then you don't get much -- not enough to really impact a final image. If the whole point is to get a longer exposure without having to stop your aperture to the point that diffraction sets in then you are probably better off using an ND filter.

Regardless, I still think the K-1 meets criteria to be D810 lite. DXO Mark gives sensor scores to them both of 96 and 97 and state that a difference of less than 5 points is unlikely to be seen in real world shooting. To me that says that they are awfully close in terms of image quality, assuming similar quality glass.

Most of the photo taken in day light at wide or normal focal length could afford iso 50 just fine and a significant part of them iso 25. I often get 1/250, 1/500 sometime more in daylight... If I am not using a tele lens or trying to capture a fast moving subject most often 1/50 or 1/100 is enough meaning that I have 2-3EV of margin for many shots.

The camera that would fully leverage that would gain a significant advantage.

Pentax has pixel shift it provide more improvements but require basically a still subject and a tripod. And once you have that, you can do use various techniques like HDR or panos for even better results with significantly more work. Well HDR is quite easy once the tripod is there and give 8EV of dynamic range with a 5 shot setup.

Nikon solution bring far less improvement than pixel shift in D810, but is available for many more shots including no tripod, moving subjects. And in D850, we will see the sensor perf but the 45MP BSI is quite likely to bring significant improvements.

[Nicolas06]

Could you provide references? I'd be interested.

From a camera user standpoint, good DR at 64 ISO is not as good as having the same DR at 100 ISO. Take BSI sensors for example: higher quantum efficiency is the small advantage of BSI over non BSI sensor technology. Higher quantum efficiency of BSI offers increased native sensitivity (ISO) with same max DR as non BSI sensors. BSI is considered an improvement over non BSI, it was not used in large sensor because it requires TSV, hence more manufacturing steps and it is more tricky to produce with high yield (yield = percentage of good sensors that pass the production tests).

Technically if you look at the K1, it can achieve the SNR+DR at ISO100 that the a D810 can achieve at ISO64. Now, in you opinion, which one is better? 14ev DR at ISO100 or 14ev DR at ISO64?

Anyway, DR is clamped by either the 14bits ADC, or the breakdown voltage of the PIN junction.
The smallest value that can be coded by the ADC is 2^14 smaller than the larger value, that's the max theoretical pixel level DR achievable.
With regards to the breakdown voltage of the PIN junction, do you realize that one the output voltage of the sensel reaches 7 volts, it needs to go up to 14 volts in order to get 1ev of extra DR, and 28 volts to get 2ev of extra DR (and think about clocking the ADC at 400MHz under 14V or 28V supply... good luck with that). No wonder why even after 20 years of efforts, no consumer sensor maker can offer more than 14ev of DR. That is why the easy way to get best possible IQ is to use a larger sensor, it's as simple as that.

Pentax pixel shift is not very practical, but from the signal processing standpoint it is a very good idea because it is increasing the DR beyond what the 14 ADC limit and how? Moving the sensor and stacking the pixel of difference silicon area averages out of the read noise (quantization noise of the ADC) increases resolution. So, in theory, if the ADC is perfect, pixel shift can get 1 extra bit over the 14bits, but that only 15ev. Sigma delta converters work the same way, it starts from a 1bit converter (comparator) into an integration loop followed by low pass filtering. Please do some research before you throw the baby in water like a bad child.

---------- Post added 08-09-17 at 08:53 ----------


Well, users of D810 say they CANT tell the difference between 100 ISO and 64 ISO...

---------- Post added 08-09-17 at 08:55 ----------

The problem in photography is that would need 20ev of dynamic range. So, whether Nikon can do 0.3ev more than Pentax, is just a marketing gimmick in comparison to lighting situations found in nature.

Storing signal of 16, 32 or whatever number of bit is easy. Encoding too. Then where you have a point is that to store more photons, you need photosites with more capacity that would return a stronger signal and that costly. This is exactly what you explained in different words. Nikon paid to go to iso 64... And because it is hard is the reason why they didn't go to iso iso 10 or iso 1.

Please notice that we start to have alternatives. Newer Sony sensors bundle memory with the sensor. If there a memory cell for each photosite, then you could just as well count of many time the photosite filled up. That would give you iso as low as you want with unlimited dynamic range as long as the exposure is long enough. A multi exposure shots like pixel shift or HDR is just a variation of that idea but not automated up to the photosite level, that all.

I heard there prototypes about that. I agree we never seen one in reality but if your ask for more than 14EV of dynamic range, MF cameras offer just that and I think that the 645Z managed signficantly more than the best FF and some MF rawa re using 16bit per channel for that reason.

[Rondec]

Most of the photo taken in day light at wide or normal focal length could afford iso 50 just fine and a significant part of them iso 25. I often get 1/250, 1/500 sometime more in daylight... If I am not using a tele lens or trying to capture a fast moving subject most often 1/50 or 1/100 is enough meaning that I have 2-3EV of margin for many shots.

The camera that would fully leverage that would gain a significant advantage.

Pentax has pixel shift it provide more improvements but require basically a still subject and a tripod. And once you have that, you can do use various techniques like HDR or panos for even better results with significantly more work. Well HDR is quite easy once the tripod is there and give 8EV of dynamic range with a 5 shot setup.

Nikon solution bring far less improvement than pixel shift in D810, but is available for many more shots including no tripod, moving subjects. And in D850, we will see the sensor perf but the 45MP BSI is quite likely to bring significant improvements.

Well, I see that you have in your mind an idea. I will say that pixel shift is more useful than you think and that there are a lot fewer instances where iso 64 would make a big difference over having a base iso of 100 (understanding that there is actually only a very small increase in dynamic range with it). The only instance I could think of was in situations where I want to lower my shutter speed (mostly for flowing water) but not put my aperture at the smallest. Even there, you are just getting a stop -- not enough certainly to go from 1/30 second shutter speed at f16 to 1/8 second.

Saying that there are many images in bright daylight where you probably could use iso 64 is not the same as saying why you would use it. If your shutter speed is 1/250 at f2.8, you could drop down to iso 64, but why? We have established that the dynamic range and noise differences between your iso 100 and iso 64 images aren't going to be significant in the final image. Are you dropping your iso 64 just because 1/250 second is too fast a shutter speed?

As far as HDR goes, you get a specific look with that that I don't like very well. It is better to process a single image if possible. Panoramas are possible, but often are significantly more complicated than pixel shift and give odd shaped images that need to be cropped in order to be used.

As far as the D850 sensor goes, I am skeptical. I think the big thing it brings to the D810 platform is faster processing and 4K video. All well and good, but it looks like it isn't quite as good at high iso or with regard to dynamic range at low iso. It often works that way. You can't have 14.6 EV dynamic range and 9 frames per second pulled off the sensor. But I guess we'll see.

[biz-engineer]

Storing signal of 16, 32 or whatever number of bit is easy. Encoding too.

No, it's not "easy". It's easy in the digital domain but not easy within the analog to digital conversion. How low the small the noise of the conversion should be with 16bits... impossible to get at 400MHz, cannot go fast and low noise at the same time, there's a barrier, a tradeoff to be made. If that's easy to encode 32(true) bits depth, you should file a patent and sell your rights to Sony, you could be very rich and stop working.

Please notice that we start to have alternatives. Newer Sony sensors bundle memory with the sensor. If there a memory cell for each photosite, then you could just as well count of many time the photosite filled up.

It's not true, there is no memory for each photosite. And for your info, there isn't one ADC per pixel...you need to find space to put that ADC and that space reduces the silicon area available to capture light. Sony A9 uses a conventional CMOS sensor; with the memory stacked to acquire digital data by segment. That is to achieve more parallelism for the data transfer into memory, with the drawback that it increases the general noise level of the image sensor substrate.

---------- Post added 09-09-17 at 20:30 ----------

If there a memory cell for each photosite, then you could just as well count of many time the photosite filled up.

In theory and lab on experimental sensor that works, it's an more than 40 year old technique call folding interpolation. There is only a small problem... where to fit the folding circuitry on the sensor without reducing effective sensor area. That is why the lab experiments did not make it into a sensor for the mass market.

[Winder]

As far as HDR goes, you get a specific look with that that I don't like very well. It is better to process a single image if possible. Panoramas are possible, but often are significantly more complicated than pixel shift and give odd shaped images that need to be cropped in order to be used.

In any of your landscapes, have you tried to stack pixel shift images taken at different exposures? Does the color hold up better?