Sensors outresolving lenses

[john_f._lemke]

I hope I haven't missed this discussion anywhere in the archives.

 

This from a response Bob Atkins made this morning in the new article

about the Canon 5D:

 

"Right now the goal should be (and I assume is) getting yield on

12-16MP sensors up to the point where cost can come down enough to

make cameras using them affordable to the mass market. $3300 is still

too high. If they can cut that in half (and I'm sure they will),

they'll have a much larger market, which will in itself drive costs

down further. I'm pretty sure that in maybe 3 years we'll see a full

frame sensor DSLR priced where the 20D is today with a 12-16mp pixel

count."

 

In a way this is music to my ears because I've held off on "going

digital" until the prices come down and the technology matures

further. I'm definitely in the "mass market" category.

 

So the question here is this. I've read elsewhere that at around 12.8

megapixels sensors start outresolving Canon's fine middle of the line

glass, exposing "defects" in their design. Specifically the 50mm F1.4

and the 85 F 1.8, for example, both of which I own.

 

How large an issue is this? Given the parameters Mr. Atkins speaks of

above, how should we best invest in glass for the future?

 

Thanks in advance.............

[bellavance]

Even though the 20D probably can't use all the resolution of a great lens, it still can show the difference between a good and a not as good lens, or between an L series lens and a consumer grade lens.

 

I'm investing in good glass, mostly L series, and will also continue to upgrade my cameras, but probably at a slower pace than 10D then 20D... I'll probably wait for the 40D... unless the 30D has Eye Controlled Focus: I hope Mr Canon reads this. ;-)

 

Pierre

[eddie g]

I Love Mr. Canon !

[dave_hachey]

I'm going to second Pierre's comments above. But in addition I feel that a 50% increase in pixel count doesn't merit a 3-fold increase in price. IMHO it isn't cost effective. My trusty 20D is already a better camera than I am a photographer. I plan to fix that problem with more practice. Don't get me wrong, the 5D will most likely prove to be an outstanding camera, but right now I plan to invest in more L glass. If the new 24-105/4L lives up to the early reports, one will be mine by Christmas.

[beauh44]

I have a 20D and a 1DsMKII. A few years ago I bought the Canon 28-135mm IS lens and it does OK stopped down around f/8 or so. But I noticed then that I could tell a difference between it and my 70-200 f/2.8L IS lens even on a 10D. No big surprise there. But when I put the 28-135mm lens on the 1DsMKII, well it just does *not* do well at all. The lack of resolving power of the lens really does stand out. It's a bummer because it's a very convenient focal length and doesn't require working out with weights to carry it around all day. Maybe the new Canon lens(es) about to be released will do better. And while I'm yapping, I still would like to see Canon do better in the wide-angle department. The 17-40 f/4L does pretty good but it's slow. I'd love to see a 17-40mm f/2.8L IS from Canon that can kick @$$. I was hoping Canon would announce something similar but so far, no luck.

[aljaz_.]

50/f1.4 and 85/f1.8 are middle of the range lenses?

 

Is then 50/f1.8 the low range and 28-105/f3.5-4.5 the very-low range? Where is 28-70/f4-5.6, uncharted? And what is the high range - 200/f1.8 on its own?

[bobatkins]

Sensors will not outresolve lenses.

 

Think of film (remember film?). A good lens was better than a bad lens, whether you shot ISO 25 (high resolution) or ISO 800 (low resolution) film.

 

As resolution increases you may see more image improvement with really good lenses than with mediocre lenses, but you'll see some improvement with both.

 

You don't "invest" in glass - you buy the lens you need to do the job you have to do! All higher resolution sensors will give you are better images than you get now, so if your lens is good enough now, your results will only get better!

[jbq]

Like Bob said, there's a big difference between "outresolving" and exposing "defects".

 

I wouldn't be surprised to learn that good Canon primes like the one you mention are diffraction-limited in the center at f/4 or f/5.6 - which implies a resloving power much higher than what a sensor like the 5D's can record.

 

But it's also no secret that most lenses aren't as good in the corners as they are in the center, especially the shorter focal lengths. So obviously the more you try to record what the lens can capture in the center, the more you can expose the differences between the center and the corners.

 

What the 5D gives you when exploring lens defects isn't the ability to see every little problem, it's the convenience when making side-by-side comparisons. But if you take your 50/1.4, shoot test targets close to wide open on slow film and look at the result under a good loupe you will see the difference between the center and the corners, you won't need a 5D for that.

[markci]

<i>I've read elsewhere that at around 12.8 megapixels sensors start outresolving Canon's fine middle of the line glass, exposing "defects" in their design.</i>

<p>

This doesn't make a lot of sense. The 12mp+ cameras are all full-frame and have LARGER pixels than the 6/8mp cameras. If there is a concern it is with corner and edge performance.

[kelly_flanigan1]

A sensor or film often "out resolves" a lens when it is wide open; abit off axis. An F1.4 50mm lens of any brand is not a most behaved thing for resolution at the edges wide open. I think folks worry too much about this stuff. Both Canon, Nikon, Leica, Minolta, etc make fine cameras. The limits in a lab situation are often alot higher than the average Joe's handheld shots; and the resolution issues are moot. The old 300mm F2.8 Canon FL-F lens of 3 decades ago was sharp wide open at F2.8 on center and at the edges; so doent always assume all lensee require a couple of stops stopping down to be great.

[steve_dunn2]

<p>Whether or not a sensor is better than a lens deserves is not determined solely by how many megapixels it has. A 12 megapixel 1.6-crop sensor, for instance, would be a more severe test* than a 12 megapixel full-frame sensor for essentially the same reason why <a href="http://www.photo.net/equipment/digital/sensorsize/index">sensor size matters</a>. The more pixels you pack into a given space, the higher up the MTF curve you need to go, but spread the same number of pixels across a larger sensor and you're back into the good part of the MTF curve.</p>

 

<p>Also, are we talking about 100% <a href="http://www.photo.net/learn/optics/pixel_peep/pixel_peep_part2.html">pixel-peeping</a>, or about viewing an actual print? I have plenty of images which are disappointingly soft at 100% but look fine when printed at reasonable sizes and viewed from reasonable distances. If my sensor went up from 8 MP to 12 MP (without changing the overall size of the sensor), the same image would look even worse at 100%, but my reasonably-sized prints would still look fine.</p>

 

<p>Not to say that I don't want better lenses; sure, I do. Even though most of the photos from my 8 MP camera get posted on the Web at something like 800x600, or printed at 4x6", and neither use requires anywhere near the camera's full resolution, it's better to have higher-quality images for the times I <em>do</em> want a bigger print, or need to crop ruthlessly. So buying better glass is still worthwhile. But I don't think there's any need to obsess about whether or not some of Canon's finest primes are good enough.</p>

 

<p>*: Well, a more severe test of the middle of the lens' image circle; of course, a lens which is good in the middle but lousy at the edges would have this flaw become more apparent on a full-frame sensor.</p>

[gareth_harper]

"I've read elsewhere that at around 12.8 megapixels sensors start outresolving Canon's fine middle of the line glass, exposing "defects" in their design. Specifically the 50mm F1.4 and the 85 F 1.8, for example, both of which I own. "

 

Those are probably two of the finest EF lenses you can get your hands on. I have the 50mmf1.8, which at only 70 quid is one of the finest lenses I have. I also have an EF100f2 which is also superb. I shoot film, and as digital is only now getting up to around film resolution your statement doesn't make sense.

 

I just wish my 24mmf1.4L performed as well as my 50mmf1.8! But some lenses are easier to manufactuer than others.

 

Always buy the best glass you can to suit your needs. Nor does more expensive automatically mean better. For example Canon's EF35f2 and 50mmf1.8 are superb lenses for not a lot of money. Their basic portrait lenses the 85f1.8 and 100f2 are also superb. Whether you shoot film or digital it's the glass that really counts, whatever the sensor or film being used.

 

"How large an issue is this?"

 

In short it's not an issue at all, nothing has changed when it comes to selecting lenses.

[bellavance]

Well... now that the 50/1.8 is a superb lens, I wonder how we should qualify the 50/1.4 lens, which we know has better resolution and contrast... superber?

 

Pierre

[lamoine_einspahr]

Bob Atkins is correct in claiming that film will outresolve the sensors available, at least for the present. My view is this. When film processing is tossed into the mix, especially the processing available to the hinderlands, the comparision fades. My lowly Rebel produces very nice images and when printed are still very good. My EOS 3 with the same lenses and using a variety of film material, negative, positive, and several different brands, emulsions, and ISO's within those brands, and the processing available here in the center of the nation does not match the Rebel's images. Images that were printed from a CD burned from images processed from RAW in Photoshop CS. Ignoring the cost of film and processing, I still prefer the Rebel over the "3", although I yearn for a better "light box" like a 20D or a 5D. I do not print super large prints and if the images scanned on a Nikon V scanner and seen on my 21 inch monitor are anywhere near correct, the film is too grainy to do it either.

[SCL]

I have read on several forums that a number of Canon DSLR users are using Leica lenses on their cameras, as the resolving power and contrast give them better results than the Canon glass. Erwin Puts' tests on the Leica glass indicate that current films still can't capture the full potential of the glass. So, from my perspective, the glass still outperforms the sensors, and a current purchase makes sense. What I believe will occur is that more features will be added to new glass (lighter weight, better or faster focusing mechanisms, better colorimetric match to sensors, etc) which might cause one to hold off purchase. I'm quite happy using top notch current lenses for DSLRs and I also am delighted to be able to use my (some) older manual focus (Leica rangefinder) lenses on my DSLR.

[gareth_harper]

"Well... now that the 50/1.8 is a superb lens, I wonder how we should qualify the 50/1.4 lens, which we know has better resolution and contrast... superber? "

 

It does? News to me. The only optical advantages that I'm aware of are ; slightly better out of focus blurr (the f1.4 has an extra blade or two) and of course the slight advantage it has in low light. Certainly nothing to get too excited about. The 50mmf1.8 will hold it's own against any of Canon's other lenses, quite amazing considering it's price!

 

Oh, while many of Lecia's lenses are superb (albet at a price), one should be aware that some of the advantge of the rangefinder glass is lost on an SLR. The reason for this is that the closer the lens is to the film (or sensor) the better the result (please don't ask me to explain why). The design of an SLR, with it's mirror pushes the lens out away from the film or sensor. For fun try an old Canonet QL17 rangefinder, be prepared to be shocked by the sharpness.

[bellavance]

I,ve once seen a comparison between the 50/1.4 and the 50/1.8 and the 50/1.4 was very clearly sharper, which is only normal.

 

Pierre

[gareth_harper]

"I,ve once seen a comparison between the 50/1.4 and the 50/1.8 and the 50/1.4 was very clearly sharper, which is only normal. "

 

Many people have reported that the 50mmf1.8 is slightly sharper, the two are after all almost optically identicle. The extra cost in the f1.4 is, well it's f1.4, has extra blades and much better build quality (around the optics) The softest of Canon's 50mm lenses is their (discontinued) EF50mmf1.0

 

As for direct comparisons, the 50mmf1.8 is sharper than my 24mmf1.4L, my Tokina 28-80 ATX and my 70-200f4L for example, that's impressive!

 

Trying to judge the optical performance of the 50mm f1.8 and f1.4 is like trying to split hairs.

[ray robertson]

We should ask what is meant by statements such as "film/sensor outresolving a lens" or "a lens outresolving a fim or sensor".

 

One cannot determine the resolution limits of a film type or sensor by using a lens. The lens will degrade the result.

 

One cannot determine the resolution limits of a lens by using a film or sensor. The film or sensor will degrade the results.

 

Essentially, resolution of film is tested by exposing the film to finely etched lines on a glass plate in contact fashion, bypassing any intermediary lens. The resolution of lenses is tested similarly by bypassing film, using a microscope to count lines, whatever.

 

Whether we're using a superb lens in conjunction with a low resolution film or sensor, or a high resolution film or sensor in conjunction with a miserable lens, the result will always be worse than the lower performing component.

[ray robertson]

The Canon 50/1.4 is a slightly higher performing lens than the 50/1.8 by a notch or so. However, there is always some degree of quality variation amongst lenses of the same model. It's quite likely that there do exist some 50/1.8 lenses that actually are sharper than some 50/1.4 lenses. The lenses are so close there is bound to be some overlap, but on average the 50/1.4 should be a tad sharper, especially between f1.8 and f5.6.

 

The question is, if image sharpness is your main concern and you are on a budget, is the extra cost of the 50/1.4 justified? The answer is probably 'no'.

 

I believe the Canon 50/2.8 macro is a notch sharper than the 50/1.4. I already have a 50/1.8. If Canon had given us a 12MP 20D instead of the FF 5D, the 50/2.8 would have been my next purchase.

[ky2]

Ok--- here's the punch line. The 20D has a much better chance of outresolving its lenses, simply because its sensor has higher density. Yes, we're also cropping much of the projected circle, but still-- for the center part that is left there, the 20D would outresolve the 5D, as it already does for the 1DsMk2. The D2x, for example, goes even further, since it crams even more photosites into a slightly larger sensor.

 

The thing is, most lenses perform better in the center than in the corners. And yet, as Bob noted above, the better lenses are very capable of exceeding full-frame sensor resolution. Just to please your ear, John-- my 85/1.8 and 50/1.8 both outresolved my 1Ds *AND* my 20D, and with still room for more. The zooms are much more suspectible of being outresolved by the denser sensors-- and yet, my 24-70L and my 100-300/4 Sigma, still do a great job.

[mark u]

Never mind any EOS DSLR - their sensors are all outresolved by cheap digicams! Sensor resolution is principally determined by pixel size, which is smaller (just over 2 microns for quite a few P&S digicam models), compared with 8.2 microns on the 5D.

 

I often feel that far too much emphasis is placed on lens sharpness at the expense of other qualities that are important to creating pleasing images. One lesson of digital photography that applies even more strongly than it does with film is that the only truly valid comparisons take account of the entire image train - lens, sensor/sensor optics, and all subsequent image processing, both in camera and later. R&D looks at all stages of the train, as testified by the work Canon has done in re-designing sensor optics, on-sensor noise reduction, and image processing both in their DIGIC chips and in software such as DPP. Efforts will doubtless continue in all these areas.

[ray robertson]

Yaron Kidron writes:-

 

The 20D has a much better chance of outresolving its lenses, simply because its sensor has higher density.

 

No chance whatsoever. The 20D resolution is limited to around 70-75 lp/mm, and that's theoretical, assuming a perfect lens. You need at least 2 pixels per line pair. You can't expect one row of pixels to define more than one line, can you?

 

A good lens that is diffraction limited at f8 is capable of resolving, according to Rayleigh's formulas which astronomers use, 1500/8= 187 lp/mm. The MTF of such lines is around 10%, so we're talking about high contrast lines to start off with. Average real world scenes will contain little of such high contrast information (but nevertheless, some).

 

When describing the resolution limits of sensors, around 48 lp/mm for the 1Ds, 55 lp/mm for the 10D and 65 lp/mm for the 20D (give or take a few), the MTF is assumed to be around 10% which is a reasonable practical limit for visibility.

 

That's a long way off 187 lp/mm for a good lens at f8. Where does this notion that sensors can outresolve lenses originate. Beats me!

[ky2]

Raymond, you're right-- just as I wrote above. But you're refering to a "GOOD LENS". Have you ever taken a look at Canons marvelous 28~90 kit lens?

[lester_wareham]

The lens seems to be the defining factor is picture quality for 6Mp and above, but this is becase the most important thing is the contrast at line frequencies of 5-26 lp/mm depending sensor crop factor.

 

I think the main advantage of more pixels (like 30-50Mp) is no need for capture sharpening to compensate for the anti-alias filter. Sharpening, even using edge masks and Blend If to protect the shaddows and highlights results in artifacts.

 

Many good lens will have some contrast up at 100-150 lp/mm but the effect on image sharpness is limited.