SLR Lens resolution

[peterso]

As the number of pixels on the chips are increasing the new resolution limit we are up against seem to be the current repertoire of lenses. Even Canon's L series of lenses seem to be insufficient with the highest MP cameras. I am using the Canon XSi and only the best lenses are able to justify the 12MP. This problem was not evident before. The following lenses in my inventory are able to resolve up to or above the resolution of the camera, Canon EF-S 60mm f/2.8 macro (extremely sharp), Canon 50mm F/1.4, Canon 24-105mm f/4 (third copy that is). The rest are ok, Sigma 17-35mm f/2.8-4, Canon 70-300mm f/4-5.6.

The resolution of the chip tells only half the story; more importantly is the pixel density.

Canon 5D Mark II  21MP  - 2.4 mp/sq cm
Canon XSi  12 MP              - 3.7
Canon G10  14.7 MP        - 34 (not a typo)

What puzzles me is this. How can Canon produce stunning images with the G10? Why is the glass on the G10 able to resolve more than 10 times the resolution of the glass in front of a Canon 5D Mark II. All Canons’ L glass start at $1000, the entire G10 camera can be had for $400. Of course the lens is much smaller, but they have shown that is quite possible to produce lenses that are able to resolve much higher resolutions than the current L line of lenses. Why are we running into resolution problems with our $1000 lenses when the $100 lens on the G10 can resolve 10x the L glass with no problem?

Peter

[dalethorn]

I'm no expert on lenses, since the only semi-pro cameras I've owned are the Leica M's. That said, there has been significant progress in lens design, but it would seem the major manufacturers are playing it cheap by not upgrading their big lenses.  Does upgrading SLR lenses mean trouble for them?  If they upgrade 3, will users then demand they upgrade 30?  It's a nightmare for dealers who stock them.

A few years ago Casio announced success with ceramic lenses, and used them in production consumer cameras.  Whatever happened to ceramic lenses, which were supposed to be much smaller and lighter than conventional lenses?  I hope this isn't the 200 mpg carburator question.

[BJL]

The optical designs of lenses tend to limit angular resolution (except when diffraction come into play), so that for example if the same 4x standard zoom lens design is scaled down from a 28-105/3.5-4.5 for 35mm to a 5.4-21mm f/3.4-4.5 for a compact camera of about 1/1.8" format, it will resolve about the same angular detail at equal aperture, but that means resolving five times as many line pairs per mm. Or being able to keep up with pixel densities of up to 25 times greater.

(Aside: we are talking about lens resolution here, so linear measures like l/mm or pixel widths are more relevant than that DPReview creation of "pixel density", which exaggerates differences by squaring all the numbers.)

In reality, the smaller format lens will not do as well as this, due to greater difficulty in manufacturing the smaller designs and diffraction effects at smaller apertures, but there is a solid trend of smaller format lenses having higher absolute resolution in lines per mm, though they give less image detail in the sense of lower angular resolution in "lines per degree", or less "lines per picture height".

For example, many good 35mm format lenses out-resolve the best large format lenses in l/mm.

[james_elliot]

The answer is simple: a 10MPix LX2 was not even on par with a 6MPix APS-C RD-1 regarding MTF (the test is here if you are interested).
I am pretty sure that the so-called "stunning results" of the G10 wouldn't stand such a test either, and I am not even sure the G10 would outresolve the RD-1, or any APS-C 6MPix camera.
With a lens aperture around 3.5, diffraction starts to kick in with pixels from 3.5x3.5 in the red and 1.7x1.7 in the blues (see here). The G10 has 1.7x1.7 pixels, and thus the 14MPix argument is just marketing (Panasonic has wisely decided to stick with 10MPix in the LX3).

And I am not even speaking of the real Point Spreading Function of the lens, which is, of course, much worse, even if manufacturing lenses for a small sensor is easier than manufacturing lenses for 35mm...

[Panopeeper]

How can Canon produce stunning images with the G10? Why is the glass on the G10 able to resolve more than 10 times the resolution of the glass in front of a Canon 5D Mark II

1. The linear size proportion is 3.65, not 10. The pixel sizes are: 5D2: 6.2µ G10: 1.7µ

2. Is that lens really capable of resolving so much finer? The attached captures show the raw pixel view of a 5D2 and a G10 shot.

[Ray]

Is that lens really capable of resolving so much finer? The attached captures show the raw pixel view of a 5D2 and a G10 shot.

Interesting! I think it's clear that the 5D2 at F8 is doing a better job than the G10 at F5.6 in this comparison. One might argue whether or not the G10 lens is sharpest at F5.6 and whether or not the 35mm lens used with the 5D2 is sharpest at F8, but no matter. If one were to go up one stop with both cameras, the differences would probably be similar.

The major issue here is DoF. In order to produce images with approximately equal DoF, the F stop used should vary in proportion with the diagonal of the sensors, as a rough guide. If it doesn't, DoF differences will be very apparent in real-world scenes that have significant depth.

I'm reminded here of Michael's comparison of the G10 and P45+. As I recall, the G10 was used at F3.5 and the P45+ at F11 to take shots of the same forest scene from the same position. At A3+ print size, experienced photographers could not identify which camera had been used until they realised that one of the prints had a slightly shallower DoF and that was likely the print from the P45+.

The fact is, the diagonal of the G10 sensor is only 9.5mm. The diagonal of the P45 sensor is 60mm. To get the same DoF that the G10 produces at F3.5, the P45 should be used at F22 (60/9.5 x 3.5 = 22).

Similarly, to get the same DoF that the G10 produces at F5.6, the lens with a 5D2 would have to be set at F25. I think in these circumstances the G10 might produce the sharper result.

I think it's generally true that the larger format will always produce sharper results than the smaller format when the lenses for the different formats are use at their sharpest aperture. But the trade-off is less DoF for the larger format.

[JamesA]

As the number of pixels on the chips are increasing the new resolution limit we are up against seem to be the current repertoire of lenses. Even Canon's L series of lenses seem to be insufficient with the highest MP cameras. I am using the Canon XSi and only the best lenses are able to justify the 12MP. This problem was not evident before. The following lenses in my inventory are able to resolve up to or above the resolution of the camera, Canon EF-S 60mm f/2.8 macro (extremely sharp), Canon 50mm F/1.4, Canon 24-105mm f/4 (third copy that is). The rest are ok, Sigma 17-35mm f/2.8-4, Canon 70-300mm f/4-5.6.

The resolution of the chip tells only half the story; more importantly is the pixel density.

Canon 5D Mark II  21MP  - 2.4 mp/sq cm
Canon XSi  12 MP              - 3.7
Canon G10  14.7 MP        - 34 (not a typo)

What puzzles me is this. How can Canon produce stunning images with the G10? Why is the glass on the G10 able to resolve more than 10 times the resolution of the glass in front of a Canon 5D Mark II. All Canons’ L glass start at $1000, the entire G10 camera can be had for $400. Of course the lens is much smaller, but they have shown that is quite possible to produce lenses that are able to resolve much higher resolutions than the current L line of lenses. Why are we running into resolution problems with our $1000 lenses when the $100 lens on the G10 can resolve 10x the L glass with no problem?

Peter

The G10 is like most P&S's, it is capable of pretty decent resolution, owing to the fact it is easier and cheaper to make a lens to support a tiny sensor accurately than it is to make a lens as accurate for a much larger sensor.  But having said that, the lens means little when you have ugly, smeary noise reduction even at 80 ISO and even then, still have noticeable noise.  Resolution of black lines on white backgrounds means next to nothing.  Even if you look at the resolution charts, you'll see the lines are very unsmooth, rough looking and if you look at the sample images, there is noise in all of them, even at low ISO, detail-robbing noise.
http://www.dpreview.com/reviews/canong10/page8.asp