Canon Lens Resolution

alastairb said:
a post i just saw made me think,
how much of canon's glass gets close to/beats the resolving power to be able to have enough detail to completly resolve every pixel of the 7D's sensor. as this equates to needing 233.51 lines per mm?
and is there anywhere to see a list of what resolving power different lenses have?
Click to expand...

Remember you also need to factor in the AA filter on the 7D sensor, so the maximum actual resolution is somewhat less. DPReview measured the 7D sensor at 2450 lines/picture height (absolute) vertically. This equates to 165 lines per mm (ish) - rather than 233.

Andy
 
Lines per mm (more correctly line pairs per mm, or cycles per mm) is not the same as Lines per Picture Height, so 7D's sensor is equal to about 116 lpmm. But you cannot compare lenses and sensors like that, even though people try to.

With lenses, as resolution goes up so contrast comes down. You cannot say a lens has x lpmm resolution unless you specify the % MTF contrast. Even then, what the lens can resolve is not what gets recorded, as all aspects of the imaging system have their individual MTF characteristics and these must be cascaded together. For example, if you had a sensor capable of resolving 100 lpmm and a lens capable of that resolution also, you'd be lucky to see anything more than 70-80 lpmm at best. Then the printer also has MTF, and so does the printing paper etc etc. Even your eye has MTF, though unless you change your eyes at least that remains a constant ;)

Nyquist level puts a theoretical ceiling on sensor resolution, and the anti-aliasing filter caps that somewhat lower. Manufacturers don't talk much about their AA filters, where they cut, how hard they cut and so on, despite the fact that it is hugely significant. The only reference I have seen to it in promotional literature came from Leica, who claim a very high level/soft cut AA filter in the M9, which certainly enhances sharpness in theory, but of course runs the risk of aliasing if they overdo it.

What this boils down to is that just looking at pixels and Lpmm figures doesn't bear much relation to final image quality. For example, I recently compared my 40D (10mp) against a 7D (18mp) and a full frame 5D2 (21mp). The difference between the two crop format cameras was slight, despite the 80% increase in pixels, but comparing the 7D to the 5D2, the full frame image had a very clear advantage, despite only a small increase in pixel count. But then, those pixels are more than twice as big and the lens is operating at a lower MTF level where it delivers much greater contrast. I've not done the same comparison between a Nikon D3 (12mp) and a 7D, but I'd put money on the much lower res Nikon blitzing the 7D in terms of how sharp the final print actually looks.

One of the sharpest lenses I've seen tested on DPreview is the new 100mm L macro. Here it is on a 50D, comfortably maxing out the sensor at mid-range apertures. Quite a few lens do, they just reach a point where lens resolution is clearly increasing, but it's not showing on the sensor.
http://www.dpreview.com/lensreviews/canon_100_2p8_is_usm_c16/page3.asp
 
I should have said lines per mm, not lpmm (=line pairs per mm)

DPreview give the absolute res (i.e. real nyquist including the effect of the AA filter) at 2500 lines per image height = 168 lines per mm (/2 for lpmm)

apologies for the confusion...
 
AndyElliott said:
I should have said lines per mm, not lpmm (=line pairs per mm)

DPreview give the absolute res (i.e. real nyquist including the effect of the AA filter) at 2500 lines per image height = 168 lines per mm (/2 for lpmm)

apologies for the confusion...
Click to expand...

I think that's a theoretical maximum based on pixel density, whereas the AA filter will cut at some unspecified point below that.
 
HoppyUK said:
I think that's a theoretical maximum based on pixel density, whereas the AA filter will cut at some unspecified point below that.
Click to expand...
Not sure about that - it's measured - check out the link above and see what you think...
 
AndyElliott said:
Not sure about that - it's measured - check out the link above and see what you think...
Click to expand...

I think those numbers are calculated (ie theoretical) rather than measured. The only thing that is measured is final sharpness output, at MTF 50%, from which I guess an estimate of the AA filter characteristics can be made.

In DPreview's lens test FAQs, they say:

"Strictly speaking, sharpness results should not be directly compared between tests carried out using different DSLR bodies, as the camera itself influences the results. Most notably, results get unreliable close to the Nyquist frequency due to the effect of the camera’s anti-aliasing filter (this is why the Nyquist line is explicitly marked on the graphs)."

http://www.dpreview.com/learn/?/Glossary/Lens_Reviews/frequently_asked_questions_01.htm

Nyquist level is just a calculation based on LPH.

What I think they are saying here is that what they are actually testing is an imaging system - including the camera, sensor, lens and also Raw software - and not just the lens. You can argue for and against the virtues of that approach, but at the end of the day it's impossible to compare lenses tested on different cameras, and we'll never know absolutely if a Canon lens is better than a Nikon lens for example.

The only way to do that is to take the lens off the camera and do it solo on an optical bench, and plot the MTF for the lens only. Which is what they do back at the factory, but it's horribly expensive. And at the end of the day, you cannot take pictures without a camera so I guess it's moot.

The influence of post processing software is interesting. DPreview uses ACR in an attempt to level the playing field, but this doesn't take into account what, for example, Canon's DPP Raw processor can do in terms of CA, vignetting and distortion. Although they do publish a pictorial representation of it at work. There will come a time quite soon when I believe a big part of lens performance will be down to software 'enhancing' the final output.
 
I understand the nyquist is calculated for the lens tests, but see specifically how they measure the sensor resolution - see where it quotes the sensor res in terms of lines per sensor height vs the extinction res...

Andy
 
I checked it in detail, yes, definitely measured, so this is a measure of the true resolution of the camera.

I agree you can't come up with a figure for a lens in isolation and compare it with the camera, but in knowing the relative performance of a set of lenses and knowing which one only just hits the nyquist of the lens/sensor combination, it should allow you to say something meaningful about the set of lenses - i.e. which can make use of the sensor's resolution and those that can't. The ones very near the boundary might be questionable, but some clear-cut conclusions should be possible.

As a tog interested in wildlife, the performance of the lens+TC would also be of interest.

Andy
 
AndyElliott said:
I checked it in detail, yes, definitely measured, so this is a measure of the true resolution of the camera.

I agree you can't come up with a figure for a lens in isolation and compare it with the camera, but in knowing the relative performance of a set of lenses and knowing which one only just hits the nyquist of the lens/sensor combination, it should allow you to say something meaningful about the set of lenses - i.e. which can make use of the sensor's resolution and those that can't. The ones very near the boundary might be questionable, but some clear-cut conclusions should be possible.

As a tog interested in wildlife, the performance of the lens+TC would also be of interest.

Andy
Click to expand...

Yes, you can deduce a lot from the DPreview tests even if not everything. And after all, what you actually get out of the camera/lens/processing system is what ultimately matters.

For long lens tests, I like the reviews on www.the-digital-picture.com. While I'm very wary of their methodology for short focal lengths, all the issues disappear with longer lenses (small target, 2D target, shot close up = field curvature and distance optimisation issues, eg problems with Canon 17-55 2.8 which they acknowledge as a failing of the test).

On that site they show some results shot on different cameras, which clearly gives a different outcome, and they also test with 1.4x and 2x tele-cons. Very easy to compare results.
 
DPReview are pretty good with the lens tests, using the same camera (and trying to pick one with the best sensor res too) for a range of tests.

That said, I would expect the 7D to be too much for some of the lenses, and by some margin, this being exacerbated by the use of a TC. Given the likely application of the 7D, the tele lenses are of going to be of key interest. That is not to say that it won't be used for all sorts of non-tele applications, just a hunch that this is the area most people who buy it will be most interested in.

Andy
 
bomberman said:
http://www.photozone.de/canon-eos

This is the one that i use
Click to expand...

They do two sets one with the 30D (?) and with the 50D. Good example is the 70-200mm f/4 USM L IS. With the 30D, it's excellent. With the 50D, it's in the 'very good' to 'excellent' category. I don't think it needs a rocket scientist to see that with the 7D, you'd expect it to be a smidge worse than on the 50D, but not by much. If you look at the lines per height of the 50D in DPreview, it's 2350 vertically vs 2500 for the 7D - not much difference at all.

So there is some merit in using the 50D results with a pinch of negativity salt...

Probably more interesting would be if there was a test of the 70-200 f2.8 on the 50D. It's hardly amazing on the 30D, actually mediocre at the long end - so you can probably assume the performance on the 7D will be poor wide open. It seems that the f4 lens performs better wide open than the f2.8 lens at most apertures at the long end. The f2.8 lens should give benefits of better AF performance, but with a weight and price penalty. I would not expect the f2.8 lens to do at all well with a TC attached on the 7D.

I think the above analysis is the best you can come up with using the data available - i.e., if there is a similar test for a lens you are interested in, know whether it should be probably okay or likely poor.

Andy
 
Think I'm learning a lot from this thread...will give it a more in depth read later this afternoon when I'm not in a lecture with maybe a bit more research. For one thing I'll be interested to see how the 5D compares with a EF 24-105mm f/4 L IS USM!
 
bomberman said:
http://www.photozone.de/canon-eos

This is the one that i use
Click to expand...

That's pretty good, but it still has the problem of using a small flat target to test lenses (using the popular Imatest system, because it's cheap and easy). They talk about the problem of field curvature in the FAQ section, but still go ahead and publish regardless. Well I guess that's okay so long as any discrepencies are discussed in the text, but these things are often skated over because it then looks like the test is flawed - which of course it is!

And then again there is also the issue of variations created by testing the lenses on different cameras and image processing systems.

For a pure lens test, the MTF graphs published by Canon, Nikon and Sigma are pretty good, but they are also not perfect. In particular, the resolution levels they use are pretty low by modern digital standards.

At the end of the day, I think the best way to reach a rounded conclusion is to read all the best tests, using different copies on different cameras in slightly different ways. While absolute data is very hard to come by, the relative merits should be clear and the general concensus is usually pretty accurate overall.
 
HoppyUK said:
For a pure lens test, the MTF graphs published by Canon, Nikon and Sigma are pretty good, but they are also not perfect. In particular, the resolution levels they use are pretty low by modern digital standards.
Click to expand...

I wouldn't use them! For a start, some manufacturers use calculated theoretical values, not measured values - which always tend to look better....
 
AndyElliott said:
I wouldn't use them! For a start, some manufacturers use calculated theoretical values, not measured values - which always tend to look better....
Click to expand...

Well I use them. I think those MTFs are all computer generated, but that cannot make a lens appear to be better than it actually is. It just means that is what you should get out of a 100% good copy. Which seems reasonable to me.

And you should at least be able to compare Nikon to Canon to Sigma directly, becuase they use the same parameters for some of the graphs, and we know that sample variation cannot be an issue. Plus the fact that it's a purely optical test (no camera etc) and devoid of any field flatness nonsense that skews just about every other test that I know of, because they are conducted at infinity focus.

In other words, what they show is excellent, they are just not very comprehensive.
 
HoppyUK said:
Well I use them. I think those MTFs are all computer generated, but that cannot make a lens appear to be better than it actually is. It just means that is what you should get out of a 100% good copy. Which seems reasonable to me.

And you should at least be able to compare Nikon to Canon to Sigma directly, becuase they use the same parameters for some of the graphs, and we know that sample variation cannot be an issue. Plus the fact that it's a purely optical test (no camera etc) and devoid of any field flatness nonsense that skews just about every other test that I know of, because they are conducted at infinity focus.

In other words, what they show is excellent, they are just not very comprehensive.
Click to expand...

I think within a manufacturer - Canon lens #1 vs lens #2 - perhaps - but I would be very wary of doing that between different manufacturers. Canon, Sigma and Nikon quote in 10lpmm and 30lpmm MTF contours but their methods may not be the same. Olympus use the more demanding 20lpmm and 60lpmm contours - but given their lenses will only ever be used on less-than-full-frame sensors, this isn't surprising.

Here's the quote on the luminous landscape FAQ for MTF charts:

Canon's MTF charts are based on theoretical calculations used in the design of the lens, while some other manufacturers use actually measurements. There are valid arguments in favour of both methods. Be aware though that different manufacturers have different measurement procedures, and therefore while comparing MTF charts between lenses in the same line is possible, and is in fact very useful in making a purchasing decision, doing so between different manufacturer's MTF charts isn't.

Basically, I don't think you can work out whether a sigma 300/2.8 is sharper than a canon 300/2.8 by consulting the respective manufacturers own MTF charts.

The sample variation is an interesting point. Sigma has something of a bad reputation in this regard. Certain lenses of Canon and Nikon also have had problems. I think it is always useful if you get a new lens to be able to assess whether you have a lemon or not. To be honest, forums like this tend to be most useful at checking to see if other folks' experience of a lens is the same as yours as trying to perform quantitative tests yourself are something of a nightmare.

Andy
 
AndyElliott said:
I think within a manufacturer - Canon lens #1 vs lens #2 - perhaps - but I would be very wary of doing that between different manufacturers. Canon, Sigma and Nikon quote in 10lpmm and 30lpmm MTF contours but their methods may not be the same. Olympus use the more demanding 20lpmm and 60lpmm contours - but given their lenses will only ever be used on less-than-full-frame sensors, this isn't surprising.

Here's the quote on the luminous landscape FAQ for MTF charts:

Canon's MTF charts are based on theoretical calculations used in the design of the lens, while some other manufacturers use actually measurements. There are valid arguments in favour of both methods. Be aware though that different manufacturers have different measurement procedures, and therefore while comparing MTF charts between lenses in the same line is possible, and is in fact very useful in making a purchasing decision, doing so between different manufacturer's MTF charts isn't.

Basically, I don't think you can work out whether a sigma 300/2.8 is sharper than a canon 300/2.8 by consulting the respective manufacturers own MTF charts.

The sample variation is an interesting point. Sigma has something of a bad reputation in this regard. Certain lenses of Canon and Nikon also have had problems. I think it is always useful if you get a new lens to be able to assess whether you have a lemon or not. To be honest, forums like this tend to be most useful at checking to see if other folks' experience of a lens is the same as yours as trying to perform quantitative tests yourself are something of a nightmare.

Andy
Click to expand...

You may be right Andy. I personally believe that you can draw meaningful conclusions from different manufacturers' MTF tests, on the assumption that they are conducted in the same way and always providing of course that you are comparing identical data and know how to interpret it.

I may be wrong on that of course - I'm taking the data at face value and on the assumption that nothing else has been done to massage the figures that we've not been told about. However, it is certainly better than trying compare results made with different cameras and processing methods, which we know to be different, so they have a use.
 
HoppyUK said:
However, it is certainly better than trying compare results made with different cameras and processing methods, which we know to be different, so they have a use.
Click to expand...

I agree, but if you can get two tested on the same camera with the same processing methods, that is far better. A lens is never used in isolation - whereas the photozone tests don't test with a 7D, they do tests with a 50D and a 30D.

If a lens isn't great on the 50D, it certainly won't be better on the 7D. Similarly, if a lens isn't great on the 30D, it will be quite poor on the 7D.

There is the issue of 'field flatness' in the tests - but you're still going to be using the lens on a camera in the real world.

The DPReview tests are also pretty rigorous - but obviously don't compare their tests with the photozone tests for different lenses for the same reason you state above.
 
You must log in or register to reply here.
Back
Top