EF lenses performance on DSLR cameras

[mgunduz]

A Photography magazine in Australia claims EF lenses would yield poorer results(lower

resolution and chromatic aberration increase) when used on Canon DSLRs with sensors

smaller than full frame.

They claim to have tested two popular L- lenses( 17-40mm f4L and 70-200mm f4L) to prove

and recommend them to be used on full frame DSLRs or film bodies for peak performance.

As a newcomer to Canon range of lenses I find their claims surprising and find forum

discussions(especially Bob Atkins responses) quite useful.

Any input to above will be appreciated.

[beauh44]

I've certainly not found that to be the case; I have a 1DsII and 20D and 10D. If anything a smaller sensor uses the best part of the image circle. And I'm confused because most of Canon's lenses are EF lenses. I'm not sure what they'd use if they don't use EF. Therefore my .02 cent's worth is: Hogwash! ;-)

[kendall_koning]

On the other hand, the same image projected over a larger surface area (read, full frame vs. APS crop) will need to resolve fewer lines per mm of sensor surface to get the same number of lines per frame, and should therefore be sharper where this effect is stronger than the effect of lens quality falloff.

 

So it would surely be sharper in the center, but of course depending on the lens, may be softer in the corners, and assuming the same lens is used on full frame and APS crop, definately have worse vigenetting and chromatic abberation charictaristics at the edge of the frame.

[bobatkins]

I'm guessing you're misinterpreting what they are saying.

 

They're probably saying that really good lenes give better images on full frame DSLRs than on APS-C DSLRs, but that's because under those circumstances, the full frame DSLRs give higher quality final prints of a given size.

 

Lousy lenses with really crummy edge and corner performance might give better prints of a given size on APS-C DSLRs than full frame DSLRs because you don't use the edges and corners of the 35mm frame and thus don't see just how bad they are there!

 

I'm hoping to get a 5D for review on photo.net quite soon and this is one of the things I intend to look at.

[mendel_leisk]

If a full frame lens has a certain ultimate resolving power, and you use it with a crop format dslr, aren't you compromising that resolving power, as a function of the crop factor? In other words, if you shoot a scene with a 20D, then transfer the lens to a 5D and step back and re-shoot, so as to capture the same scene, you are utilizing more of the lens' available area, and accordingly, increasing your potential resolving power.

 

On the other hand, you are increasing vignetting and the corners of the 5D image are likely more prone to show the lens' chromatic abberation and sharpness fall-off.

[mgunduz]

Bob,

Here is an excerpt on 17-40mm f4L from the article;

"A significant drop-off in resolution and increase in chromatic aberration was revealed

when the lens was tested on the EOS 300D. Here the uncorrected MTF50 figure dropped to

an average of 878.5line widths per pixel height, while chromatic aberration averaged

0.086%, which is just into the'moderate' range(visible at high print magnifications). These

results are not surprising but they confirm the fact that lenses designed for 35mm

cameras cannot perform at their best on DSLR cameras with the sensors smaller than

24*36 mm."

Thanks for everyone response.

[awindsor]

There is a definite trade-off.

 

Essentially you are just magnifying the centre section of the 35mm image. This increases any lens defects present in the region and lowers resolution. Similarly any chromatic aberration present gets magnified.

 

However most lenses performance falls off as you approach the edge of the image circle. In the case of wide angle lenses for SLRs this can be dramatic especially wide open. Numerous reviewers report that lenses they loved on their APS-C digital cameras simply didn't hold up on their full-frame 1D or 5D camera (cf Petteri, Reichmann)

 

I suspect that the magazine has drawn their conclusions based on figures from samples at the centre of the image. Done this way full frame cameras will always win. However comparing the edges of the 17-40/4 on an APS-C camera and on a full frame camera will tell a different story. I liked the 17-40/4 on my film bodies for the ultrawide view but edge performance suffered and their was some vignetting. My opinion, based on nothing more scientific than using the lens, is that it is optically better on the APS-C when the performance over the whole frame is considered.

[eos 10 fan]

Might be that the article is available online.

 

--

[steve_dunn2]

<p>The lens has no way of knowing whether it's being used with film or digital, or what crop factor the digital sensor will be; it simply projects exactly the same image onto whatever happens to be there to see the image. Same resolving power, same chromatic aberration.</p>

 

<p>Now, the results you'll get on your film/sensor depend on both the lens and the film/sensor. You'll get more fine detail out of your 20D than your D30, assuming that the lens is a reasonably good one, because the 20D has nearly twice the linear resolution of the D30. You'll probably also find CA more noticeable; CA which is one pixel wide on the 20D is only around half a pixel wide on the D30.</p>

 

<p>But those changes have nothing to do with the lens and everything to do with the sensor. If you were to take a full-frame sensor with pixels of the same size as the 20D's, you'd get virtually identical results from the two (within the 1.6-crop area, of course). And if you're adjusting zoom position to counteract the difference in crop factor, then you're using different parts of the zoom range, and of course the lens will perform different at different focal lengths.</p>

[bobatkins]

<em>...Here the uncorrected MTF50 figure dropped to an average of 878.5 line widths per pixel height...</em>

 

<p>

 

I have no idea what this is supposed to mean. "line widths per pixel height" makes absolutely no sense to me. "Lines per picture hight" would have some meaning.

[awindsor]

Bob,

 

I bet they are using Imatest. It produces figures LW/PH, which is line widths per picture height (where picture height is actually sensor height). That the can't even decode the acronyms correctly makes me a little suspicious of the rest of their conclusions.

 

Steve,

 

While you are correct that the image produced is identical you are neglecting the need to normalize to compare different formats. Comparing by mm on the sensor is pointless if the sensors are different size. Imatest normalizes by sensor height. This is the same as comparing perfect prints of the same size from the different formats. On the sensor the line pairs blur at an identical point but when you produce the print that same line pair produces fewer LP/MM on the print from the APS-C, since it is enlarged more, than on the print from the full-frame sensor.

[steve_solomon2]

Greetings, All.

How about those new Canons?! While I'd love the 1DsMIII, I also like paying a mortgage, so I'd be interested to see how the 40D tests. Also, has anyone used an APS-C DSLR with Canon's TS-E lens, particularly the 90mm? I'd be curious to see what the IQ is from that combination.

Thanks for all the great advice and tips!

~Steve