Do sensors outresolve lenses - or vice versa

[arclight]

The thread title is from an article in AP, dated 10 March, by Professor Bob Newman.

It is particualry topical given the high pixel count sensors that are appearing, most recently out of the Nikon stable.

Well worth a read for those contemplating purchase of the D800 etc., especially if they are having doubts as to whether they will see any benefit.

I am not in the group of people who could claim to have an in depth grasp of the science relating to the article and cannot sensibly comment on that aspect. Consequently, I await with interest publication of the results obtained by users of the D800.

The article concludes with a statement to the effect that users of the new high resolution cameras will see a benefit as their cameras will give sharper results with all their lenses.

 

[Digifrog]

Do you have a link?

 

[GHP]

The article concludes with a statement to the effect that users of the new high resolution cameras will see a benefit as their cameras will give sharper results with all their lenses.

As I see it, you are bound to see an improvement with a higher resolution sensor. I know I noticed a difference when I went from D70 to D7000, on all my lenses, but the difference was more noticeable with the better primes I have.
The one caveat to this MIGHT BE if you put a particularly soft lens on the front of it, would the higher resolution sensor "amplify" the softness?
I'm not saying I know the answer by the way, and I don't pretend to know or understand the physics behind it either. WAY above my head.

 

[specialman]

Do you have a link?

Think it might be print-only but could be worth checking the AP website...

 

[HoppyUK]

The thread title is from an article in AP, dated 10 March, by Professor Bob Newman.

It is particualry topical given the high pixel count sensors that are appearing, most recently out of the Nikon stable.

Well worth a read for those contemplating purchase of the D800 etc., especially if they are having doubts as to whether they will see any benefit.

I am not in the group of people who could claim to have an in depth grasp of the science relating to the article and cannot sensibly comment on that aspect. Consequently, I await with interest publication of the results obtained by users of the D800.

The article concludes with a statement to the effect that users of the new high resolution cameras will see a benefit as their cameras will give sharper results with all their lenses.

It's not a question that has an answer as posed. But all the answers to rather more complicated questions are on the Nikon site, particularly this D800 Technical Guide http://www.nikonusa.com/en_US/o/Y6wrkA9OU_z04IreazIXl_22UII/PDF/D800_TechnicalGuide_En.pdf

Sharpness is not just about resolution, but about contrast as well - basis of lens MTF theory, which is the inescapable fact that as resolution goes up, so contrast goes down. You have to define both to get a meaningful measure, and of the two aspects, contrast is the more important as far as a visual impression goes.

What Nikon shows in the link is that, most of the time, the lens will be the limiting factor and they only list 16 lenses capable of doing justice to the D800. Even then, it must be at f/8 or lower to avoid diffraction, and technique especially in relation to camera-shake and subject movement, must be perfect.

When you've done all that, if you make huge enlargements and look very closely, you will be able to see a smidge more very fine detail. But overall the general impression of sharpness will be the same as a camera with far less pixels.

Nobody ever complained much about the D700 with only a third of the pixels, and Canon seems to be saying quite clearly with the 5D3 that the practical limit is around 20mp-ish. If you want to realise the sharpness potential of 30-40mp, then you need a bigger sensor like medium format which allows the lenses to deliver more of the kind of sharpness that you can actually see. But even then, it begs the question about where it is sensible the draw the line in real terms.

 

[arclight]

Do you have a link?

Could not find the article on the web so, as far as I know, you'll need to get a copy of the mag. Should be on the shelves around Tues unless you are subbed.

 

[arclight]

It's not a question that has an answer as posed. But all the answers to rather more complicated questions are on the Nikon site, particularly this D800 Technical Guide http://www.nikonusa.com/en_US/o/Y6wrkA9OU_z04IreazIXl_22UII/PDF/D800_TechnicalGuide_En.pdf

Sharpness is not just about resolution, but about contrast as well - basis of lens MTF theory, which is the inescapable fact that as resolution goes up, so contrast goes down. You have to define both to get a meaningful measure.

What Nikon shows in the link is that, most of the time, the lens will be the limiting factor and they only list 16 lenses capable of doing justice to the D800. Even then, it must be at f/8 or lower to avoid diffraction, and technique especially in relation to camera-shake and subject movement, must be perfect.

When you've done all that, if you make huge enlargements and look very closely, you will be able to see a smidge more very fine detail. But overall the general impression of sharpness will be the same as a camera with far less pixels.

Nobody ever complained much about the D700 with only a third of the pixels, and Canon seems to be saying quite clearly with the 5D3 that the practical limit is around 20mp-ish. If you want to realise the sharpness potential of 30-40mp, then you need a bigger sensor like medium format which allows the lenses to deliver more of the kind of sharpness that you can actually see. But even then, it begs the question about where it is sensible the draw the line in real terms.

That is the question that nags me. Have Nikon gone overboard with the pixels, or not.

I think it boils down to the theory being fine, but let's wait and see the results in the field after a good old hands on test.

 

[HoppyUK]

That is the question that nags me. Have Nikon gone overboard with the pixels, or not.

I think it boils down to the theory being fine, but let's wait and see the results in the field after a good old hands on test.

Exactly right Dougie

But sticking my neck out, I think there will be some disappointment with the D800, and especially the D800E where folks have spent a bit more money purely to absolutely maximise sharpness.

What they will get, if they put in the work and play strictly by Nikon's rules, is a bit more resolution of very fine details, so small that you won't normally be able to see them without pixel peeping at 100%.

I say disappointment becuase I think most people believe sharpness is all about resolution, when it's not. Contrast is far more important, and the kind of things we see as small details are more around the 30lpmm area than 100lpmm. It's good lenses that give you that, and even more so, good lenses on bigger sensors.

It's going to be interesting

 

[fracster]

It's going to be interesting

I do believe you are correct Richard.

Good question as well from the OP, will existing glass be able to render all these megapixels, or will new owners have to buy a new range of glass?

 

[arclight]

Exactly right Dougie

But sticking my neck out, I think there will be some disappointment with the D800, and especially the D800E where folks have spent a bit more money purely to absolutely maximise sharpness.

What they will get, if they put in the work and play strictly by Nikon's rules, is a bit more resolution of very fine details, so small that you won't normally be able to see them without pixel peeping at 100%.

I say disappointment becuase I think most people believe sharpness is all about resolution, when it's not. Contrast is far more important, and the kind of things we see as small details are more around the 30lpmm area than 100lpmm. It's good lenses that give you that, and even more so, good lenses on bigger sensors.

Yes, we have all seen how sharpness is apparently better in strong lighting where there is a sudden transition from light to shade.

 

[arclight]

I do believe you are correct Richard.

Good question as well from the OP, will existing glass be able to render all these megapixels, or will new owners have to buy a new range of glass?

That would be a bridge too far for me.

 

[Nawty]

I do believe you are correct Richard.

Good question as well from the OP, will existing glass be able to render all these megapixels, or will new owners have to buy a new range of glass?

I was contemplating upgrading my D700 to a D800 (well why not?) but then figured my glass was no way near good enough so decided to upgrade that in the meantime and wait for the dust to settle on the new bodies.

 

[s162216]

If you want something that definitely out resolved lenses then Kodak Technical Pan 25 film used to easily, it resolved 400 lp/mm! (assuming a perfect lens of course)

When it was released in the 70's, even Leica admitted that their lenses were outresolved by it. You can probably count the amount of lenses that can resolve that much detail on one hand (and prices in 5 or 6 digits!)

 

[HoppyUK]

If you want something that definitely out resolved lenses then Kodak Technical Pan 25 film used to easily, it resolved 400 lp/mm! (assuming a perfect lens of course)

When it was released in the 70's, even Leica admitted that their lenses were outresolved by it. You can probably count the amount of lenses that can resolve that much detail on one hand (and prices in 5 or 6 digits!)

With respect, that makes the same mistake about resolution with film as it does with lenses.

Film has MTF as well, so does printing paper, and the enlarging lens, and ink-jet paper, the printer, computer screen etc etc. They all have to be cascaded together, and you always lose something at each stage.

Stating resolution without also stating MTF is pretty meaningless, and a fairly common standard is 50% MTF. That's the stated cut-off point and below that, resolution is said to stop, but of course it doesn't.

If you shoot at test chart with very fine black and white lines, even relatively modest lenses will go on resolving detail (in the centre) to a very high level, in as much as you can just about make out the black and white lines. The point is, you can't see them clearly because they turn into increasingly similar shades of grey. MTF defines that shade of grey.

There is no way Technical Pan film, which was commonly used for lens tests, resolves 400lpmm at any meaningful level of MTF - more like 10-20% at best. It was still an exceptionally sharp film though, which was why it was used of course, but at the same time fairly irrelevant. Apart from being monochrome and very low ISO, its very limited tone range made it unsuitable for general photography. As the name suggests, it was a specialist scientific emulsion.

 

[paulcamcas]

Exactly right Dougie

But sticking my neck out, I think there will be some disappointment with the D800, and especially the D800E where folks have spent a bit more money purely to absolutely maximise sharpness.

Richard, do you reckon that if I were to crop a D700 size image (4256 x 2832) from a D800 image (7360 x4912), the IQ of the resulting D800 image would differ in any meaningful way from the D700 image? For the sake of this question, let's say I use the same very good lens and shoot both images at ISO 200 in good light.

Would the answer be any different if I were to use a lesser lens?

 

[s162216]

Apart from being monochrome and very low ISO, its very limited tone range made it unsuitable for general photography. As the name suggests, it was a specialist scientific emulsion.

I know its slow! I've got some in my camera at the moment, most of them which I've taken so far have ended up being 1 sec at f8 with an orange filter! (I know the filter will degrade the quality slightly). As the low contrast Technidol dev is now gone I'm having to get it specially processed by a lab who have an alternative (or I would have deved it myself).

Perhaps the source which I got the 400 lp/mm from slightly exaggerated it, but according to Kodak's own curves in the datasheet, 50% contrast corresponds to almost exactly 100 cycles/mm (not quite sure what cycles per mm means though!) in Technidol developer and about 190 at 30% (The curve stops there). Another technical paper claims an estimated PPI value of 7214 at 30% contrast with no lens in the path which is fairly big as if scanned with a perfect scanner then thats the same as about a 70 megapixel image from 35mm! I do realise that this only an estimation though and is unlikely/impossible in the 'real world'.

 

[HoppyUK]

Richard, do you reckon that if I were to crop a D700 size image (4256 x 2832) from a D800 image (7360 x4912), the IQ of the resulting D800 image would differ in any meaningful way from the D700 image? For the sake of this question, let's say I use the same very good lens and shoot both images at ISO 200 in good light.

Yes, a lot. In round numbers, you're only using half the frame, so the lens has to work twice as hard, so MTF is lower.

It's roughly the same sort of comparison as a D300 vs D700, and we know which one wins there.

We have been sold the 'more pixels is better' line for so long now that too many folks believe it's the key to sharp pictures. It used to be a big factor in the early days of digital when we were dealing with just a handful of megapixels, but we are way past that now.

Would the answer be any different if I were to use a lesser lens?

If you use the same lens on both, you get the same answer regardless of lens quality, just the levels change. If you're asking whether an exceptionally good lens on the D800 can make up the difference, the answer is no. The gap is too big.

It's like expecting a fast car to accellerate from 60-120mph in the same time as it takes a slightly less fast car to go from 0-60. The more you ask from a lens, the harder it has to work. Much harder.

What you tend to get with higher quality lenses is not so much more ultimate sharpness. That is usually pretty high with any decent lens if you shoot at optimum aperture and only look at the centre. Very generally, better lenses will maintain a higher level of performance at all apertures, right to the edges and corners of the frame, at all focal lengths etc. That's the main optical difference you're paying for, plus higher build quality, lower f/numbers, IS/VR etc.

 

[HoppyUK]

I know its slow! I've got some in my camera at the moment, most of them which I've taken so far have ended up being 1 sec at f8 with an orange filter! (I know the filter will degrade the quality slightly). As the low contrast Technidol dev is now gone I'm having to get it specially processed by a lab who have an alternative (or I would have deved it myself).

Perhaps the source which I got the 400 lp/mm from slightly exaggerated it, but according to Kodak's own curves in the datasheet, 50% contrast corresponds to almost exactly 100 cycles/mm (not quite sure what cycles per mm means though!) in Technidol developer and about 190 at 30% (The curve stops there). Another technical paper claims an estimated PPI value of 7214 at 30% contrast with no lens in the path which is fairly big as if scanned with a perfect scanner then thats the same as about a 70 megapixel image from 35mm! I do realise that this only an estimation though and is unlikely/impossible in the 'real world'.

Cycles per mm is the same as line-pairs per mm - one black and one white.

 

[Martyn...]

That is the question that nags me. Have Nikon gone overboard with the pixels, or not.

I think it boils down to the theory being fine, but let's wait and see the results in the field after a good old hands on test.

My thoughts exactly Dougie ... the proof of the pudding is in the eating, as they say

I seem to remember similar comments about the D3x doing the rounds when it was released.

 

[POAH]

The 5d2 would out resolve quite a few lenses and nikons line up aint that great. The 70-200 II struggled with the 24 mp if the D3X.

 

[simonblue]

The best thing is to alway try to buy the best glass,you can addford.

And yes Nikon lens are very good same as Canon

 

[vizzair]

but if the problems with the bigger sensor will be noticable, why havn't we already noticed the problems with the D7000 ?, its sensor is the same or slightly higher density that the D800 , its just cropped..

 

[arclight]

uote:
Originally Posted by paulcamcas View Post
Richard, do you reckon that if I were to crop a D700 size image (4256 x 2832) from a D800 image (7360 x4912), the IQ of the resulting D800 image would differ in any meaningful way from the D700 image? For the sake of this question, let's say I use the same very good lens and shoot both images at ISO 200 in good light.

Hoppy wrote:- Yes, a lot. In round numbers, you're only using half the frame, so the lens has to work twice as hard, so MTF is lower.

It's roughly the same sort of comparison as a D300 vs D700, and we know which one wins there.

We have been sold the 'more pixels is better' line for so long now that too many folks believe it's the key to sharp pictures. It used to be a big factor in the early days of digital when we were dealing with just a handful of megapixels, but we are way past that now.

I am particularly interested to see if the D800 when set at 15 MP produces images as good as the D700 set at 12 MP.

 

[simonblue]

The other thing that come into play,is camera shake,the greater the MP count,the more you will have to be careful about it

 

[paulcamcas]

Thanks Richard, very helpful. The more I read about the D800 the more I'm looking forward to saving £2,300 by not buying one.

 

[HoppyUK]

Thanks Richard, very helpful. The more I read about the D800 the more I'm looking forward to saving £2,300 by not buying one.

Really? Ignore 36mp for a moment, and you're still left with a sensationally good camera.

It's a bit of a win-win. Nikon has been quite canny, and the D800 can be configured to deliver different file sizes and formats that should beat any other comparable Nikon (because of new sensor tech, new processing engine) with the 36mp option always available.

 

[LongLensPhotography]

The 5d2 would out resolve quite a few lenses and nikons line up aint that great. The 70-200 II struggled with the 24 mp if the D3X.

it's true - many lenses wide open will be outresolved, like the canon 50mm f/1.4. 24-105 is just on the edge I believe, while 50D out resolved it according to many reports. 100-400 is another one. 17-40 is quite significantly outresolved on the edges, but something like 70-200mm is holding it's own really well.