The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

Sensor Size/Image Quality

Status
Not open for further replies.
T

TheReason

Guest
Question for the forum:

With the upcoming announcement of the Nikon D800 with a rumored 36MP Full Frame 35mm Sized Sensor, what if any are the advantages of a MF sensor now that DSLR systems are making huge leaps in the MP count?
 

Godfrey

Well-known member
Question for the forum:

With the upcoming announcement of the Nikon D800 with a rumored 36MP Full Frame 35mm Sized Sensor, what if any are the advantages of a MF sensor now that DSLR systems are making huge leaps in the MP count?
Each capture format size has a different relationship of Field of View to Depth of Field. There is a corollary derivative relationship as well between light gathering power and depth of field driven by sensitivity (itself influencing format size as a function of photosite technology, size, and desired pixel resolution).

Once sensitivity, pixel density, acutance, and dynamic range are at a satisfactory quality plateau for any format and specific shooting scenario, the choice of format is all about those relationships.
 
T

TheReason

Guest
Thank you for the insightful, factual and very technical answer. I understand the relationships between all these factors will make the question almost an apples to oranges comparison, however I am really seeking a more simple answer. Will the image quality of a 36MP FF DSLR sensor be equal to a 40MP MF sensor?
 

GrahamWelland

Subscriber & Workshop Member
Teeny tiny pixels and an AA mask and mass produced DSLR bodies and tolerances aren't going to be conducive to equivalent image quality as current crop 40MP MF sensors and cameras. It's basic physics & optics at work.

I'll also wager that there will be a lot of disappointed Nikon shooters who can't get the image quality they're expecting from the lenses available today & technique. Diffraction will raise it's ugly head at ever wider f-stops, just as it did already with the D3x.

Btw, I'm a long time Nikon shooter and have no Nikon axe to grind - I hope that they can overcome these basic considerations. We'll have to wait and see.
 

jlm

Workshop Member
it will boil down to this:

almost the same pixel count, but one is crammed into a smaller format, so the pixel cells are not the same size. this can affect noise, colorcast effects, iso ratings, etc in various ways
the firmware built into the sensor can be quite different, mfg to mfg. some people hate nikon's, canon's work, some love leica and Phase and blad, for example
measuring image quality is ultimately subjective as only a few features can be quantified
 
T

TheReason

Guest
Teeny tiny pixels and an AA mask and mass produced DSLR bodies and tolerances aren't going to be conducive to equivalent image quality as current crop 40MP MF sensors and cameras. It's basic physics & optics at work.

I'll also wager that there will be a lot of disappointed Nikon shooters who can't get the image quality they're expecting from the lenses available today & technique. Diffraction will raise it's ugly head at ever wider f-stops, just as it did already with the D3x.

Btw, I'm a long time Nikon shooter and have no Nikon axe to grind - I hope that they can overcome these basic considerations. We'll have to wait and see.
Great points. THose 36 MP in the upcoming D800 will have to be tiny. Really tiny. It seems like the upcoming D4 and the upcoming Canon 1Dx have decided image quality and sensitivity are more important with the FF having an ideal amount of larger pixels and better ISO performance with very respectable MP count. Glad to know the MF still has natural advantages.
 

Stefan Steib

Active member
Not so tiny at all !

about the same size as on a 7D, which if can remember right is one of the best Cameras actually for lowlight stuff and image quality in APS-C.
double the chipsize and there you have 36MPix.

And what about the 80 Mpix IQ180 - 5.2 Micron Pixel size and leader in Dynamics and Image quality on DXO Mark ?

So..........

regards
Stefan
 

GrahamWelland

Subscriber & Workshop Member
I think that it's certainly true that Nikon know what they are doing, as witnessed by the initially maligned V1 that has been showing the world that they can squeeze excellent image quality and low noise out of a 3.3u sensor. I'm just not sure that the glass available will be up to the task.
 
T

TheReason

Guest
Not so tiny at all !

about the same size as on a 7D, which if can remember right is one of the best Cameras actually for lowlight stuff and image quality in APS-C.
double the chipsize and there you have 36MPix.

And what about the 80 Mpix IQ180 - 5.2 Micron Pixel size and leader in Dynamics and Image quality on DXO Mark ?

So..........

regards
Stefan
Wow. Great facts in this post. Thanks for the info. So in your opinion, why would the major DSLR manufactures not offer higher MP count in the flagship models? (D3x and 1DX) why are they going to be outdone by prosumer models like the D800?
 

carstenw

Active member
I would guess that Nikon and Canon sell many, many more of the D700 and 5DII than the D3x and 1DsIII, to the point where the lower profit margin still yields more money total due to higher sales. However, the sports pro cameras must be selling more, i.e. D3/D3s and 1DIV. I cannot think of any other reason for Nikon to release a D4 with 16MP and a D800 with 36MP (although neither has been confirmed, to my knowledge).
 
V

Vivek

Guest
Teeny tiny pixels and an AA mask and mass produced DSLR bodies and tolerances aren't going to be conducive to equivalent image quality as current crop 40MP MF sensors and cameras. It's basic physics & optics at work.
According to the rumor site, there will be one version with an AA filter and another without an AA filter (rumored Nikon D800).

What physics and optics are you talking about, Graham?

Is 5.2 micron (even smaller pixel than the one in a Nikon D70) limiting for a MF digital back and are there lenses that would do justice to this for the entire area (not just the center) of a medium format sensor?
 

Stefan Steib

Active member
I think it absolutely misleading to define Flagship=most MPix.
Flagship is probably technologically highest devellopment with highest system price of a brand.

MPix alone don´t do nothing.

And - it is definitely easier to build 35mm lenses with shorter focal lenght and needed sharpness than MF Glass.
As the new Schneider/Mamiya lenses show they can reach nearly Nyquist on 80 Mpix/6x4,5 why should Nikon(and Canon/Sony) not be able to do the same on 24x36mm ?

And finally - there is the method of oversampling bayer sensors (e.g. used by Canon´s new 300C ) as described here:

http://image-sensors-world.blogspot.com/2011/10/eric-fossums-lecture-on-image-sensors.html

Greetings from Munich
Stefan
 

GrahamWelland

Subscriber & Workshop Member
What physics and optics are you talking about, Graham?

Is 5.2 micron (even smaller pixel than the one in a Nikon D70) limiting for a MF digital back and are there lenses that would do justice to this for the entire area (not just the center) of a medium format sensor?
Let me give you an example from the Nikon world from my experience shooting with the D3x for several years. The only Nikon lenses I had that were truly up to the task of rendering the resolution of the D3x 24mp sensor were the 14-24, 24/1.4G, 35/1.4G, 24 PC-E, 45 PC-E, 85 PC-E, 200 micro, 200 VR and 600 VR. I had other excellent glass 24-70, 70-200 VR II, 200-400VR, 105 micro, etc that were all great lenses in their own right but ultimately struggled to provide across the frame resolution and distortion free images with the 24mp sensor. Now obviously pixel peeping capabilities and great image rendering are not the same thing and I'm not saying at all that these lenses weren't good, just that ultimately not all glass is equal even at the premium end of a manufacturers offerings.

Now with 36mp and smaller sensor pitch you are going to effectively have an even smaller CoC to work with and the onset of diffraction will occur at wider apertures than before, effective depth of field will be reduced and the ability to render any lens or shooting flaws will be ever greater. Put simply, you'll start to see flaws that previously weren't rendered clearly before.

Regarding your question about 5.2u sensors on MF backs being harder on glass, I think you'll find that there is plenty of anecdotal evidence and discussion here about that. In short, yes it's tougher on glass particularly when it comes to full frame coverage and light and sharpness drop off depending upon the lens design and focal length. I don't think that anyone is saying that the best glass is being out resolved by any of the sensors yet but it is tough to provide that same performance across the frame.
 

Bob

Administrator
Staff member
Whenever a sensor is developed with smaller cell sites over a given sensor format, the probable first effect is that (AA filters aside) is that lenses that seemed to be acceptable in the corners just look soft. This is not such a big a deal if the ultimate print size is not increased, but if it is or if you pixel-peep, the additional softness will become annoying. Using Phase as an example, I found that going from a P45+ to a P65+ to an IQ180 at each step put more demands on the lenses. This can be predicted upon careful examination of the MTF plots. Say, for example, that you consider "nice and sharp", a lens that delivers 50% contrast at your pixel pitch times by 2 (to get to line pairs per mm) and if you shoot with a 5.2 micron sensor, that converts to about 96 line pairs per mm.
Looking at the Nikon 35mm f1.4 G lens MTD plot, you can see that at 30 lines per mm and at 21.6 mm off center (the corner of a 24x36 mm frame) you see contrast at about 30%. This lens will begin to appear soft in the corners of even a 15 micron pixel pitch sensor. Of course an AA filter just makes it appear to be a touch soft all-over.
True apparent sharpness is also related to the subject frequencies and the sampling frequencies desired. Nyquist criteria dictate that 2 samples per period is the minimum sample rate to recover a frequency component. However, all that gives you is a square wave or a picket fence with lots of moire. A smoother sample may be possibly in the neighborhood of 4 samples per period which means that your 96 line pairs per mm might really help you get smooth samples of a 48 line pair per mm subject.
At least theoretically, it therefore seems that several of the nikon lenses I have browsed will go unacceptably sharp just in the corners assuming that 50% contrast is your limit.
Comparing those lenses with what I shoot on a technical camera fitted with an IQ180 shows a similar story but for a larger sensor. At F/8 with the Rodenstock 70mm Digeron-W and 40 lines per mm it gets to about 30% at the corners but stays above 50% to about a 35mm image height. There is more contrast at higher frequencies at more displacement from the image center.

So it looks to me that if the cursed AA filter is removed, the better Nikon lenses will approach but not quite equal medium format corner resolution should a 36 MPix FX sensor be produced.
BUT, I have 80MPix in an IQ 180, so for an equivalent finished print size, it will need less scaling so it will appear to produce more sharpness over the final print.
YMMV and of course, these MTF plots are theoretical and actual lenses usually are not quite as good.

OTOH a D800 is a different animal than a technical camera and it is usually held differently. Remember that a smaller cell site also requires a more stable platform or a faster shutter speed to reduce motion blur which often becomes the deciding factor in sharpness for the hand-held shooter.

-bob
 
V

Vivek

Guest
Let me give you an example from the Nikon world from my experience shooting with the D3x for several years. The only Nikon lenses I had that were truly up to the task of rendering the resolution of the D3x 24mp sensor were the 14-24, 24/1.4G, 35/1.4G, 24 PC-E, 45 PC-E, 85 PC-E, 200 micro, 200 VR and 600 VR. I had other excellent glass 24-70, 70-200 VR II, 200-400VR, 105 micro, etc that were all great lenses in their own right but ultimately struggled to provide across the frame resolution and distortion free images with the 24mp sensor. Now obviously pixel peeping capabilities and great image rendering are not the same thing and I'm not saying at all that these lenses weren't good, just that ultimately not all glass is equal even at the premium end of a manufacturers offerings.

Now with 36mp and smaller sensor pitch you are going to effectively have an even smaller CoC to work with and the onset of diffraction will occur at wider apertures than before, effective depth of field will be reduced and the ability to render any lens or shooting flaws will be ever greater. Put simply, you'll start to see flaws that previously weren't rendered clearly before.

Regarding your question about 5.2u sensors on MF backs being harder on glass, I think you'll find that there is plenty of anecdotal evidence and discussion here about that. In short, yes it's tougher on glass particularly when it comes to full frame coverage and light and sharpness drop off depending upon the lens design and focal length. I don't think that anyone is saying that the best glass is being out resolved by any of the sensors yet but it is tough to provide that same performance across the frame.
Good points on the D3x and lenses, Graham. Interestingly, it will be a mismatch of the body (rumored D800) and the glass for a large section of the users (relatively inexpensive body with high demands for resolution).

I don't think that anyone is saying that the best glass is being out resolved by any of the sensors yet but it is tough to provide that same performance across the frame.
It is a tough one isn't it?
 

Stefan Steib

Active member
Bob
I think there is one factor in the 35mm world which is not that apparent and valid for the MF backs. All the Canon, Nikon and Sony DSLR´s are using extremely powerful image processors which crunch the actual rawdata inside the camera before they are released to the user. These CPU´s are little workstations which - used 10 years ago in a Sun or HP - would have made people very happy for 35000 $. There is sharpening, chroma removal, devignetting, colorcorrection of differing transmission on differing lens constructions to match a color consistency and much more in fractions of a second.
I beg to say that probably a processor on a Phase or a Leaf or Hasselblad back is not nearly as advanced as the Bionz, Digic and Expeed counterparts in japanese mass products - the difference is plenty millions of dollars of R&D.
This is the naked truth and probably the reason why the advantage of MF will shrink further and further - hen and egg. Not enough customers - not enough R&D, not enough R&D even less customers.

Greetings from Munich
Stefan
 

Bob

Administrator
Staff member
Bob
I think there is one factor in the 35mm world which is not that apparent and valid for the MF backs. All the Canon, Nikon and Sony DSLR´s are using extremely powerful image processors which crunch the actual rawdata inside the camera before they are released to the user. These CPU´s are little workstations which - used 10 years ago in a Sun or HP - would have made people very happy for 35000 $. There is sharpening, chroma removal, devignetting, colorcorrection of differing transmission on differing lens constructions to match a color consistency and much more in fractions of a second.
I beg to say that probably a processor on a Phase or a Leaf or Hasselblad back is not nearly as advanced as the Bionz, Digic and Expeed counterparts in japanese mass products - the difference is plenty millions of dollars of R&D.
This is the naked truth and probably the reason why the advantage of MF will shrink further and further - hen and egg. Not enough customers - not enough R&D, not enough R&D even less customers.

Greetings from Munich
Stefan
Image processors cannot create data that is not there. and there is plenty of image processing available to all in post production.
I have seen the bad effects of those in-camera image processors, delivering over-sharpened files with obvious artifacts. It is one thing to offset lens vignetting, but it is quite another to reconstruct image detail beyond the theoretically possible.
THAT is one reason I shoot MF.
I want my images al dente, not over-cooked.
-bob
 
V

Vivek

Guest
Whenever a sensor is developed with smaller cell sites over a given sensor format, the probable first effect is that (AA filters aside) is that lenses that seemed to be acceptable in the corners just look soft. This is not such a big a deal if the ultimate print size is not increased, but if it is or if you pixel-peep, the additional softness will become annoying. Using Phase as an example, I found that going from a P45+ to a P65+ to an IQ180 at each step put more demands on the lenses. This can be predicted upon careful examination of the MTF plots. Say, for example, that you consider "nice and sharp", a lens that delivers 50% contrast at your pixel pitch times by 2 (to get to line pairs per mm) and if you shoot with a 5.2 micron sensor, that converts to about 96 line pairs per mm.
Looking at the Nikon 35mm f1.4 G lens MTD plot, you can see that at 30 lines per mm and at 21.6 mm off center (the corner of a 24x36 mm frame) you see contrast at about 30%. This lens will begin to appear soft in the corners of even a 15 micron pixel pitch sensor. Of course an AA filter just makes it appear to be a touch soft all-over.
True apparent sharpness is also related to the subject frequencies and the sampling frequencies desired. Nyquist criteria dictate that 2 samples per period is the minimum sample rate to recover a frequency component. However, all that gives you is a square wave or a picket fence with lots of moire. A smoother sample may be possibly in the neighborhood of 4 samples per period which means that your 96 line pairs per mm might really help you get smooth samples of a 48 line pair per mm subject.
At least theoretically, it therefore seems that several of the nikon lenses I have browsed will go unacceptably sharp just in the corners assuming that 50% contrast is your limit.
Comparing those lenses with what I shoot on a technical camera fitted with an IQ180 shows a similar story but for a larger sensor. At F/8 with the Rodenstock 70mm Digeron-W and 40 lines per mm it gets to about 30% at the corners but stays above 50% to about a 35mm image height. There is more contrast at higher frequencies at more displacement from the image center.

So it looks to me that if the cursed AA filter is removed, the better Nikon lenses will approach but not quite equal medium format corner resolution should a 36 MPix FX sensor be produced.
BUT, I have 80MPix in an IQ 180, so for an equivalent finished print size, it will need less scaling so it will appear to produce more sharpness over the final print.
YMMV and of course, these MTF plots are theoretical and actual lenses usually are not quite as good.

OTOH a D800 is a different animal than a technical camera and it is usually held differently. Remember that a smaller cell site also requires a more stable platform or a faster shutter speed to reduce motion blur which often becomes the deciding factor in sharpness for the hand-held shooter.

-bob
The handheld macro shooter (in available light) is my favorite. ;)

Nice post, Bob.:thumbup:

I wonder, though the area is small for a FF sensor, if Nikon or Sony or Canon configure it in a way there is efficient heat loss away from the sensor area as in an IQ back. Even if they attempt it, it is unrealistic with x number of frames/s captures.
 

Stefan Steib

Active member
Don´t forget we are at the beginning still, this technology is in it´s infancy. To be able to fire a gun does not mean it´s useful to do it. The knowhow of how a decent image does need to look like for a "rich" file (to use for prints or large blowups) is just going into mainstream and reaching these makers with a delay. They did not have interest for this until now, their customers where sports , press , wedding and portrait (on standard sizes) and amateurs.
The large resolution does change a lot. They are learning - and they learn so fast........

Greetings from Munich
Stefan
 

GrahamWelland

Subscriber & Workshop Member
As per Eric Fossum's quote which I think applies very well to ever greater consumer/prosumer small sensor resolutions:

"The force of marketing is greater than the force of engineering"

:thumbs:

We're getting into silk purses from sow's ear territory!
 
Status
Not open for further replies.
Top