Full Frame Lens and FF Sensors!

[Gerry Winterbourne]

Sorry Gerry ...

No, it's "sorry James ..."

I read the same words that you did but you have misunderstood them, as Redteg94 pointed out in a couple of posts.

Here is DXO problem

"“We have been very surprised,” explained Frédéric Guichard, chief scientist at DxO Labs, “to find out that some of the gain from wider lens openings seems to be offset by the present state of sensor technology. Our measurements all point in the same direction: as you go further than f/4 – to f/2 and wider, the accrued quantity of light falls marginally onto the sensor. A stronger and stronger part of this additional light is blocked or lost. I am therefore inclined to question the real benefit of faster lenses.”"

So the above says with "present sensor technology" there seems to be a loss of light.

So far, so good - there's no doubt that some light is lost. Moreover, this loss is explained (partly for sure, if not completely) by obliquity of light.

“We can suspect,” Guichard continued, “that sensors collect the incoming light all the more improperly, in that this light comes from a more oblique angle. Since faster lens have, by definition, a wider opening, they raise the proportion of oblique light, hence the proportion of lost energy which never lands on the pixels.”"

Yes. But this doesn't mean what you think. Look at this diagram. As near as I can manage with a quick sketch it's the 50/1.2 lens I discussed in my previous post (50mm from aperture to sensor; 42mm diameter aperture; 43mm sensor diagonal). The pixels are, of course, drawn at a larger scale.



As you see, even the centre pixel receives a lot of its light obliquely (light from both sides of the aperture) and this creates shadow areas so that part of the pixel doesn't receive its full share of light. This is what causes the loss of light from obliquity noted by DxO. Move to the corner sensor and light from one side is less shadowed, light from the other side is more shadowed: again, loss of light from obliquity. But although my diagram doesn't lay claim to perfect accuracy you can see that the two shadows on the centre pixel and the single one on the edge pixel are roughly equal.

In other words, the loss of light is roughly the same across the sensor. This is, if you think about it, the only way it makes sense to boost ISO. If the edge pixels lost a lot more light than the centre ones, boosting ISO across the whole frame would cause overexposure in mid-frame, which no one is suggesting happens.

So we go back to Luminous Landscape whos author had a meeting with these guys about the data presented. I hope everyone knows oblique means not perpendicular! So the above says that non-perpendicular light that comes out the back of a lens will create more light loss. There can be no question of that interpretation! If you understand?

Of course I understand. The statement is perfectly simple and straightforward. What isn't at all obvious is the notion that outer pixels suffer more from obliquity than central ones.

Important point : whatever the medium, all lenses suffer from fall-off to some extent. This is, indeed, caused by obliquity of light (one of many good explanations is here http://toothwalker.org/optics/vignetting.html ) but it's vital to separate the two quite different obliquity effects. I think it's this distinction that you are missing.

Now Mark Dubovoy of Luminous Landscape wrote in the link I gave, which links to the above article from DXO the following.

[Snipped to keep in word limit]

All that Dubovy is doing is repeat the DxO opinion in slightly different words. As I've said, I have no dispute with what either of them says. My dispute is with your mistaken interpretation of it.

This shows digital sensor do not like light coming at them oblique. Now my argument. The larger the sensor used for older film lens the more oblique the light will hit the photo-diodes at the edges and in fact this is a well known problem because there is a light loss at the edges. All though DXO spoke of t-stops they say current digital sensors are not very sensitive even with microlens to light coming in at an angle.

This is where you go wrong. You are taking the DxO sense of overall light loss due to obliquity in pixel wells to be the same as the universal fall-off due to lense geomtry.

Natural one can concur an APS sensor would be better for older film lens then a FF digital sensor as far as IQ goes.

As I pointed out in an earlier post, this conclusion doesn't follow from what you've said: http://forums.dpreview.com/forums/read.asp?forum=1036&message=39666290

From the position of using lens-body combinations to take pictures (as opposed to simply comparing individual measurements) there's every reason to suppose that even film-era lenses on FF digital can beat APS-C for equal FOV and DOF when used at wider angles. As it happens, I personally prefer to stick with the size, cost and convenience of APS-C and its advantages at longer lengths. But my preferences don't wipe out the advantages that FF can offer.

Digital lens are designed to make the light strike as perpendicular as possible! Hence the very good performance we get today with digital lens. Now some old lens on a larger than APS format may not be so good!

Modern digital FF lenses may be better than older FF lenses. But that doesn't mean that at the final image stage older FF lenses are worse than APS-C. It's just as likely that old lenses on FF are better than new lenses on APS-C while new lenses on FF are even better still.

--
---

Gerry


First camera 1953, first Pentax 1983, first DSLR 2006
http://www.pbase.com/gerrywinterbourne

 

[Gerry Winterbourne]

Oh something else on L landscapes article is said

"By the way, this type of light loss is the main reason that modern lenses designed for digital photography place great emphasis on the direction of light rays behind the lens. It is also an important reason why Medium Format backs (often used with camera movements such as tilt, rise/fall and shift) and the Leica M9 (with lenses that produce mostly oblique light rays) use CCD sensors with a very different structure, rather than CMOS sensors."

If you read the link I showed in my other post http://toothwalker.org/optics/vignetting.html you'll see that rangefinder cameras like the Leica are much more prone to fall-off even with film; so just because they suffer from strongly oblique light doesn't mean SLRs (either FF or APS-C) suffer to the same extent.

This is why the Pentax 645D uses CCD sensors. They have larger light gathering diodes. CMOS are just cheaper to make. Although you can argue one is not better than another (from my research) in performance, cost aside.

Does anyone make a CMOS sensor that would fit the 645D? If not, Pentax uses CCD because it's what's available.

Mark Dubovoy flat out says digital sensors need digital lens because of the problem of light at an angle hitting any digital sensor. My simple argument is the claim of all those FF pentax lens laying around will be back to their full glory with a full frame sensor is false. They will suffer in some cases a loss of IQ. Better to use a smaller format.

What aspects of IQ are you considering? Let's look at a typical moderately wide pre-digital lens, the Nikkor 35/2D. Resolution isn't the only measure of IQ but it's pretty important. According to PZ its peak centre resolution on APS-C (D7000) is at f/5.6, where it achieves 2870 lp/ph. Going out it gives 2170 at worst.

On FF (D3x) these numbers go up to 3820 and 2540: in what possible way are these values worse?

Perhaps vignetting spoils the party: 0.28 stops on FF is, I accept, worse than 0.19 on crop - but is anyone truly worried about such small values?

CAs then: not so easy to compare because PZ measures in pixels; but 0.97 on FF v 0.98 on crop looks pretty close. Barrel distortion, like vignetting, is slightly less on crop but insignificant on both. These two things are easily corrected in PP anyway.

What I've done here is look at the same lens on two form factors. This, of course, yields different FOV. However, on a straight comparison of conventional IQ measures I don't see a win for APS-C on this lens.

The truth is that old FF lenses will do pretty well the same job on FF that they do on film, which is what most people seeking FF want. Newer, digitally-optimised FF lenses might do even better; but I can find nthing to support your contention that old lenses will do worse on digital than they do on film.

--
---

Gerry


First camera 1953, first Pentax 1983, first DSLR 2006
http://www.pbase.com/gerrywinterbourne

 

[jamesm007]

No sorry Gerry...

So we go back to Luminous Landscape whos author had a meeting with these guys about the data presented. I hope everyone knows

So we go back to Luminous Landscape whos author had a meeting with these guys about the data presented. I hope everyone knows oblique means not perpendicular! So the above says that non-perpendicular light that comes out the back of a lens will create more light loss. There can be no question of that interpretation! If you understand?

Of course I understand. The statement is perfectly simple and straightforward. What isn't at all obvious is the notion that outer pixels suffer more from obliquity than central ones.

means not perpendicular! So the above says that non-perpendicular light that comes out the back of a lens will create more light loss. There can be no question of that interpretation! If you understand?

Of course I understand. The statement is perfectly simple and straightforward. What isn't at all obvious is the notion that outer pixels suffer more from obliquity than central ones.

This is where we break. I do fully understand the article and what you are saying. But here is where I tie outer pixel wells into the argument and where you say they are not related.

I think its self evident that when light is more oblique the more the loss of light. So where on a sensor is the most oblique light? The edges! Scientist have spent a lot of time and money on this very problem. Lens had to be designed for digital. Yes I know the DXO article is saying the wide aperture tends to spread light in a oblique pattern so even center photo-sites don't perform 100%. But you want to ignore the correlation of edge photo-sites and focus on strictly what the article says. The article is as you say, but at the same time it proves my point. Digital sensors do not like light that's not perpendicular. Film lens did not care about being perpendicular to the film. The further you go away from the center the more oblique the light will strike, no question about it. But you want to ignore this fact looking right at you?

Now again a quote from the K20D sensors IQ features page -

"To accurately lead the light reaching the CMOS image sensor diagonally to the light-sensitive area (or photo diode), a micro-lens is installed in front of each pixel. Theoretically speaking, the shorter the distance between the micro-lenses and the photo diodes, the higher the light-reception efficiency, and a short-gap design contributes to a considerable improvement in image quality, particularly at the edges of the image field."

http://www.pentax.jp/english/products/digital/k20d/feature.html

Now notice what it says and what we all should know

"Theoretically speaking, the shorter the distance between the micro-lenses and the photo diodes, the higher the light-reception efficiency"

You may want to slowly read the article again. DXO says the more oblique the light the more the loss. Well the further away from the lens the more oblique the light. You say you dont see a connection is hard to believe?

Don't you know how much time and effort has been done to make lens and sensor better matched in the digital age. Film lens, not all, but some film lens will perform poor on a FF digital sensor; that's a fact.

The article is child's play, its easy to understand. But if you understand why and what you will know what I say.

--
jamesm007,

http://www.flickr.com/photos/jamesm007/
http://s195.photobucket.com/albums/z77/jamesm700/

 

[jamesm007]

I see why DPR gives you one chance to edit within a short time span. After you write it you own it. This forces one to live with what they wrote and take it like a man.

So I apologize to you Gerry for the negative sounding ending of my post. Its OK for two gentleman to disagree and write and argue. However I want you to know I feel could have worded the last five or so paragraphs with more respect. They are snappy, and take cheap shots. Your a long time online Pentax fellow member and I should have worded my paragraphs with more respect than I did. I don't feel good about them!

James M

--
jamesm007,

http://www.flickr.com/photos/jamesm007/
http://s195.photobucket.com/albums/z77/jamesm700/

 

[hikenhi]

I agree, BUT, I would still like to see a PENTAX Full Frame Camera. The whole point of a Professional FF DSLR is to have a Professional Grade Camera that has Awesome Autofocuss and very High ISO Performance. The Pro's are not going to complain about having to buy New Era FF Digital Lenses if the New FF DSLR camera body costs $5-7k by itself. The true Pros will buy into the system if it's as good or better than the competition--they'll fork out the money for it and the lenses for it(.) At any rate the people with Modern Era DSLRs like the K-10D or newer are not shooting those cameras with Old-School film lenses that often don't have the nice Coatings on the glass like the new DAs. Even if they are using the older lenses, they're only using one or two at most for their specific Needs that aren't worthy of a DA upgrade. I don't use any of my film lenses. As a mater of fact I know that the Digital ones are more compatible and produce better IQ, just like the the FF Digital Ones would for a FF PENTAX DSLR would too.

 

[Gerry Winterbourne]

So I apologize to you Gerry for the negative sounding ending of my post. Its OK for two gentleman to disagree and write and argue. However I want you to know I feel could have worded the last five or so paragraphs with more respect. They are snappy, and take cheap shots. Your a long time online Pentax fellow member and I should have worded my paragraphs with more respect than I did. I don't feel good about them!

No need to aplogise, James. Over the years you and I have rarely sparred because, by and large, our opinions coincide. However, on this one I think you have the emphasis wrong (more below).

The problem with discussing complicated technical matters on this (and other forums) is that restriction of space and the copying of text in relies make it hard to hold a coherent line of reasoning. That makes it harder to get one's point across and leads to frustration. Also, having to be brief leads to terseness of expression which can be misinterpreted as bad temper (we see this often). However, I know your style well enough that nothing you have said here causes me the slightest offence.

I can guess that you might notice a similar frustration in my comments, again brought about by the structure and mot personal. And, although it seems as if we are miles apart the real differences are few. I'll try to summarise the way it all looks:

All lenses suffer fall-off because of the geometry of their makeup. Wider angle lenses suffer more. In the days of film lenses wider than about 35mm suffered quite significant fall-off. Fortunately for SLR users their WA lenses were retrofocus and didn't suffer as badly as rangefinder lenses.

The light hits film as a flat surface so all the light hitting it can be used. The light hitting a sensor has to go into pixel wells so light coming from an angle can be shadowed out, leading to loss of light captured. So far I think we are in total agreement.

The loss of light from an angle can affect digital capture in two ways. First, an exagerration of fall-off caused by the corners losing more light because the incident light is more oblique. Second, an overall loss of light at wide angles because even central pixels suffer shading of light from the edge of the aperture. The core of our debate is not whether these things are true but only about their relative significance.

The DxO article you cite shows the second loss to be siognificant at wide apertures . It also shows it to be stronger on small pixels, which tend to be used on APS-C sensors. The article doesn't try to separate overall light loss and pixek fall-off.

For a given lens, as long as it can cover the FF image circle it will show more optical fall-off on a FF sensor than on a crop sensor. At wide angles and large apertures this can be quite severe (but for old film users was a fact of life): digital pixel fall-off can make it worse but we don't know by how much. Howevr you conclude tht the total effect bad enough to adise against using FF.

But here's my problem: the small pixels of crop sensors make them suffer relative to film too, so in fact both form factors lose a bit and it's very hard to see which loses more. To get over this I go (as you have) to Photozone.

PZ does, indeed, show that WA FF lenses suffer a lot of fall-off at large apertures on FF sensors. However, when stopped down to the sorts of aperture that these lenses are typically used at they don't show significantly more than on crop sensors; and they offer better resolution. This leads me to the conclusion that there is no generic reason why FF lenses on FF sensors should give poor results.

There are, of course, some poor lenses from all eras and all form factors that are dogs. But that's not what this discussion is about.

--
---

Gerry


First camera 1953, first Pentax 1983, first DSLR 2006
http://www.pbase.com/gerrywinterbourne

 

[dlacoutu]

The loss of light from an angle can affect digital capture in two ways. [...] Second, an overall loss of light at wide angles because even central pixels suffer shading of light from the edge of the aperture. [...]

The DxO article you cite shows the second loss to be siognificant at wide apertures . It also shows it to be stronger on small pixels, which tend to be used on APS-C sensors. The article doesn't try to separate overall light loss and pixek fall-off.

What I don't understand is how you can still get two different pictures (say, between f/1.2 and f/1.6) if those extra rays are blocked? OOF area should be identical if the extra rays of light are unused...

 

[Gerry Winterbourne]

The loss of light from an angle can affect digital capture in two ways. [...] Second, an overall loss of light at wide angles because even central pixels suffer shading of light from the edge of the aperture. [...]

The DxO article you cite shows the second loss to be siognificant at wide apertures . It also shows it to be stronger on small pixels, which tend to be used on APS-C sensors. The article doesn't try to separate overall light loss and pixek fall-off.

What I don't understand is how you can still get two different pictures (say, between f/1.2 and f/1.6) if those extra rays are blocked? OOF area should be identical if the extra rays of light are unused...

Look at this diagram that I showed in another post. The blue lines show the light path from the whole aperture, whatever it happns to be (f/1.2 in this example.

The horizontal green line is the emulsion plane if using film. All the light hits the film and is recorded.

The light path is identical on digital - all parts of the cone of light reach the bottom of the pixel well (shown as a green rectangle) so optically the performance is identical to being used on film. However, part of the pixel well base is shadowed so although the image is optically identical less of the light is recorded and the image is a bit darker.



It's the same story at the centre of the image (red lines): again, all the light paths reach the bottom of the pixel well but part of it is shaded so the image is also a bit darker there than on film.

This diagram is pretty well to scale. It shows a 50mm lens and, as you see, even at f/1.2 the light paths from the edges all reach the bottom of the well. If the angle got flatter a point could be reached (in theory) when light from the edge can't reach the bottom of the well, in which case your suggestion "OOF area should be identical if the extra rays of light are unused..." would be valid. However, for the outer rays to be flatter the aperture needs to be closer to the sensor, and for (D)SLRs that can't happen. The lens has to be far enough away to accommodate the mirror etc: on K-mount bodies the back of the lens is 44mm away from the sensor and then there's the depth to the exit pupil too, so the 50mm I've drawn is about the minimum.

Shorter focal lengths are achieved by putting in more lens elements to "retrofocus" the light rays. In effect, although I've drawn a 50mm lens the same lens-sensor distance applies to all lenses shorter than 50mm. As most WA lenses have smaller maximum apertures (the odd 30/1.4 but mostly being f/2 of 2.8) the shadowing of pixel wells is less than I've drawn.
--
---

Gerry


First camera 1953, first Pentax 1983, first DSLR 2006
http://www.pbase.com/gerrywinterbourne

 

[dlacoutu]

Hummm, I understand what you say, but in this case, rays from "slower" apertures (those rays not impacted by the "well effetc") should still register more in the image...

The differences between f/1.2 and f/2 images would be exactly as if the f/1.2 extra rays were gradually underexposed and layered upon the f/2 image... Thus, OOF specular lights would have a strange, feathered profile at f/1.2, whereas they would have sharp outlines with smaller apertures...

But that's not what I experienced with my old 55/1.2 lens: specular highlights were sharp-contoured, even at f/1.2...

Strange...

EDIT: or... maybe it would rather exagerate the famous "cat's eye effect", thus being unnoticed...

 

[Roland Karlsson]

This is a well known ... myth.

The most annoying problem with FF and some lenses is that its not sharp in the corners. And thats because the lenses are cr@p, at least compared to the resolution of digital cameras.

The two most important reasons for not making FF are ...

1. Its expensive to make larger than APS-C sensors as you need exotic equipment.

2. Its expensive for a company to have several sizes. At least if you want to make lenses for all sizes.

You can of course do like Canon. They almost only have FF lenses. Just some few 1.6 crop ones. And then they have 1.0, 1.3 and 1.6 crop cameras, and claim its the same system. Clever, but you get non optimal lenses.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html

 

[Roland Karlsson]

As silicon has a very high refractive index, there is no well effect.

What may happen though is that rays that come at an angel to the sensor misses the micro lens.

--
Roland

support http://www.openraw.org/
(Sleeping - so the need to support it is even higher)

X3F tools : http://www.proxel.se/x3f.html

 

[geoson]

Comparing Leica sales to nearly any of the Japanese or Korean companies is like comparing the ratings of a TV show on cable to one on a broadcast network. 3 million viewers for a cable show is a big hit. 3 million on a broadcast network is a ticket to cancellation.

And Leica have arranged for a sensor that minimises all the problems inherent with digital sensors + old lenses.

And, now, just look at that: Leica has made an FF camera that sells like hotcakes! They can't make enough of them! So what's the real problem here?

Any argument presented by dr Klaus, mrs Mom, army of Pentax fans, etc. trying to rationalise something that cannot be rationalised (APS-C only design) makes no sense.

An FF digital Nikon has virtually resurrected the Nikon company from shambles and in a few years it has regained MUCH of its user-base. Pentax never did.

Pentax did not do it because a) it was already late to digital game, b) did not have enough $$, c) was looking for investors, d) investor Hoya couldn't care less.

It was a wrong business decision, one that cannot be hidden behind the weak technology reason. By pure luck and also by ingenious design of their APS-C cameras they have managed to create a loyal user-base.

To keep them all under the same flag, they do need an FF — badly — and they do need to spend some quality time to design those offset micro-lenses and polish the software within the FF camera to make those FF sensors work perfectly with old glass.

--
Zvonimir Tosic