400mm f/5.6 L usability

[les]

Raymond,

maybe I am picky here, but what does all this have to do with lens resolution ?

 

With a 2xTC, you can resolve a group of lines which have 2x larger density (compared to the naked lens and assuming all other things equal, no deterioration due to TC). This 2x smaller line spacing (physical distance on the target, at the resolution limit) has the same spacing in the focal plane (sensor or film) due to 2x larger magnification. The resolution remains the same...

 

Have a look at any lens test with and w/o TC: the resolution with a TC is always (there may be extremely rare exceptions) worse than for a naked lens. Do not confuse the numerical ID of the smallest resolved line group with lens resolution.

 

I will say it again: if the naked lens has a resolution of 50 lpm at the film (sensor) plane, then with a 2x TC the resolution does NOT increase to 60 or 70 lpm, but will go down to 30-40 lpm.

 

I agree with the comments about contrast drop etc., but I think that some confusion is sneaking in if you actually believe that the way to increase resolving power of the lens is to put a TC on it.

 

I know about only one sure way to increase the resolving power: pull out the credit card and buy a better lens.

[les]

Andrew,

 

"Second, at that distance, each lens nearly resolves the same amount of the test chart. It's not clear that this would be true at closer distances."

 

Now I am confused: so it does or it doesn't ?

 

In your previous post you just said "Leszek, the Sigma is clearly behind the Canons in that 'test'."

 

If you are talking about my 'test' - this is not true: what can be seen from my 'test' is that the resolving power of each lens is clearly greater that the same ability of the sensor, which more or less makes all these glass equal since the difference in lens resolution can't be recorded (and viewed).

 

If you are making a general statement "...at that distance, each lens nearly resolves the same amount..." - this is also not true. Otherwise we would not have to pay for things like Canon 500/4L.

 

As I said, I am confused.

 

 

Regarding "There are a few ways that the test might be more informative." Couldn't agree more - except that I do not spend my days testing lenses (plus I do not have the appropriate equipment to perform full blown testing).

All I am doing is providing a limited comparison of couple of lenses in the same conditions, on the same camera, with identical lighting and using identical target. Which I CLEARLY stated in the original post. I haven't made any general statements which could be misconstrued.

 

All I said is that there is no real difference (test and print). Please read my original post (with the 'test' chart crops) carefully .

 

If the information which I provided is deemed to be insufficient, biased or inaccurate - well, life is tough and we not always get what we want...;)

 

If you want full size images from both Canons and the Sigma (not for pixel peeping but for printing) - let me know. I will be happy to provide, and then maybe we can discuss the meaning of what I meant when I said that "there really is not much of a difference".

[jnicholson]

Mark C, your comment made me laugh. Thanks. No, it's not a big honkin' wasp but I don't want to misrepresent here either. I'm sure I cropped this image somewhat. I very rarely make large crops but the image is not full frame. I'm trying to find the original to give a better idea. I can tell you, I couldn't get the whole wasp in the frame.

 

Panos, if you take another look at the picture and scroll down to the beginning of the comments, you can see a picture of the setup. That will give you an idea of the working distance. I can't remember the magnification numbers but I do know it's greater than my 50mm macro. There's also a picture in the same folder using the 70-200 f2.8 IS with the 500D.

[ray robertson]

<i>I will say it again: if the naked lens has a resolution of 50 lpm at the film (sensor) plane, then with a 2x TC the resolution does NOT increase to 60 or 70 lpm, but will go down to 30-40 lpm.

 

I agree with the comments about contrast drop etc., but I think that some confusion is sneaking in if you actually believe that the way to increase resolving power of the lens is to put a TC on it.

 

I know about only one sure way to increase the resolving power: pull out the credit card and buy a better lens.</i> .......... Leszek, It seems we have a different understanding of what constitutes lens resolution. Let me give a specific example. You are photographing a zebra in the distance with a 400mm lens and D60 (or 1Ds2, same pixel density). The zebra is at such a distance that its stripes would have a frequency at the sensor of 60 lp/mm. (Consider the zebra to be a large test chart if you like.)

 

Neither the D60 nor the 1Ds2 sensor can resolve 60 lp/mm, but the lens can. (Good lenses can actually resolve hundreds of lines per mm but at such low contrast the lines are hardly discernible). Instead of stripes you'll see just an even grey. You'd be better off with a heavy 800mm lens but have to make do with a 2x converter. So what happens to the zebra stripes when you add the 2x converter? They become 30 lp/mm at the sensor. Provided the drop in contrast is not too great, the sensor can thus record the stripes. You could count them if you wanted to.

 

You are right that the resolution of the lens hasn't been increased but the resolution (detail)in the target has been 'converted' or enlarged to something the sensor can handle. That's the purpose of a teleconverter.

 

By adjusting the distance to the charts, as you have done in your tests you've guaranteed the results will be almost identical. The sensor is seeing the same lines at the same frequency in each case. The lens with the teleconverter from a proportionally greater distance simply becomes a slightly inferior lens due to the drop in contrast. Since the sensor is the chief limiting factor to resolution, at least in the centre of the image circle of any reasonably good lens, any differences are going to be too subtle.

 

If you were to use a Norman Koren test chart with successively finer lines to extinction, shoot from the same distance with and without converter, then any quality differences could be more easily seen.

[lee_shively]

(I'm getting a vacant stare from the technical posts.)

 

In practical use:

 

1) The lens is relatively easy to handhold. It depends more on your ability than anything else. Thirty years ago, I could handhold down to 1/2 second and feel confident of a sharp picture the majority of the time. Can't do that these days.

 

2) I'm not a big monopod fan--I'd rather use a tripod. But if you get 2-3 stops with a monopod using other lenses, you will probably be more successful than I am.

 

3) It will autofocus, I'm sure. I use even older A2E bodies and sometims my wife's Rebel 2000 and they all autofocus the lens.

 

4) It's a bit too long for normal portrait use but if you like the flattening effect the lens has on the subject, that's all that matters.

 

5) ...Oh, you didn't have a 5). The last question wasn't numbered. I've used the lens with the Canon 1.4x extender for close-ups and it works pretty well for butterflies and bees on flowers. I don't have a 500D diopter but I've used Nikon diopters on shorter Canon lenses and Pentax medium format lenses and they work really well without sacrificing an f/stop.

 

It's not really a "super tele" lens. Four hundred millimeters sounds long but it quickly becomes not quite long enough in regular use. I would consider it the minimum focal length to consider for most animal photography.

 

One of the reasons I like the lens is because it can be successfully used with the Canon 1.4x extender. The combination is very sharp. It is manual focus (at least on your cameras and mine) but that's seldom any big deal. You might find your monopod or a full-size tripod is necessary when shooting with the extender.

 

My reply is longer than your questions! Good luck.

[andrew robertson]

Leszek, in the very VERY dead center of the chart the Sigma has reduced contrast

compared to the other lenses. It's clear enough to my eye. When I said 'at that distance,

each lens resolves nearly the same' I meant within the context of your comparison.

 

Also, as I said, even the cheap $120 T mount telephotos perform well in the dead center of

the field. I don't understand why you though this would be a useful test, unless your

particular kind of photography involves cropping everything but the center. A test where

the chart takes up the whole frame would have perhaps shown a different story as I said,

perhaps not in the 3-4mm circle in the middle of the field but all across the frame.

[andrew robertson]

The main reason I objected to the comparison is because you seem to like to break it out

and show everyone that, in fact, those three lenses are performing identically on an 8 MP

digital camera, when in fact you aren't telling the whole story. Somebody who isn't

familiar with optics and lenses and such might see your comparison, and be falsely led

into believing that the lenses are identical performers.

 

It would have taken perhaps an extra three minutes to paste up another chart so that it

was in the corner of each photo, and show crops from those as well, making the

comparison a lot more useful, so I don't think you would have needed all day to make the

comparison a lot more relevant to general photography.

[les]

Raymond, what you say is correct, but you should not be using the term "lens resolution", but a term "optical system resolution" - the optical system in this case including the lens+sensor, or lens+TC+sensor. The TC acts as an amp, pushing the signal above the treshold of the sensor resolution if you (as imprecise as this anakogy is).

 

No argument here - I am just in favor of using correct terms which do not mean something else to both parties.

[panos_voudouris]

Guys, thank you all for your answers. The more I read about this lens the more I like it.

Thank you all again.

[les]

Andrew, not for a second I claimed that I am telling the whole story. here is what I said about my 'test' (previous posts):

 

*********This is from my original post regarding the 'test':

 

"I am not claiming that I have measured the true resolution of the lens (or the lens+sensor), I just tried to compare the results from each lens.

 

I might have over/underestimated the performance, but any bias is (hopefully) applied in equal measure to all the lenses.

 

I pegged the performance of the Sigma at 0/5 at all apertures, and the Canon lenses (with or w/o 1.4TC) at 0/6, which translates at 44 lp/mm for the Sigma and 49 lp/mm for Canon 300/4L, Canon 400/5.6L, and Canon 300/4L with 1.4x Sigma TC.

 

It appears that the performance of all Canon lenses is limited by the sensor, which resolution limit is 61 lp/mm (practically impossible to achieve anyway).

 

The Sigma appears to reach its limits slightly earlier - someone with better eyes may want to double-check the results. But its performance is definitely quite unexpected.

 

All results are perfectly acceptable for printing extremely sharp pics on A4 format (my conclusion, YMMV).

 

I guess that with careful up-sizing of the pic, A3 format will also be achievable w/o difficulty (or so I hope). "

 

*******This is from this thread:

 

"My conclusion is that on a 8Mp SLR (in my case it was 1D Mk II) the results are almost identical - which simply means that all the tested lenses outresolve the sensor.

 

The results on 1.6x crop sensor may be different, as pixel density is higher (smaller sensor with the same number of pixels will result in higher number of pixels per inch). So, some differences in sharpnes may actually be observed with a smaller sensor."

 

*******

 

So, WHERE do you find any general and misleading statement about general performance of all these lens ?

 

.

 

 

Regarding your statement: "When I said 'at that distance, each lens resolves nearly the same' I meant within the context of your comparison."

 

As I pointed out previously - this is simply not true (or very inaccurate if you prefer milder words).The correct statement should be "...each lens+sensor resolves nearly the same..." - and it would only hold true with the sensor in my test (8Mp, 1.3 crop). Which I also pointed out.

 

Now I would expect that you pit a Canon 400/2.8 against some $120, 400mm third party tele to prove the performance in the center is identical. Not filtered through the sensor though - Kodak Tech Pan will do.

 

 

My point is: please read what people say LITERALLY, without reading additional meaning into their statements.

[ray robertson]

"what you say is correct, but you should not be using the term "lens resolution", but a term "optical system resolution" -"

 

-------------------------------------------------------------

 

Leszek, I don't recall I've ever wrote that adding a teleconverter increases lens resolution. You seem to be reading between the lines. In fact I've made several statements recently that teleconverters degrade a lens and are sometimes of dubious value.

 

However, I made the following statement a few posts ago which I stand by:- "stick a teleconverter on any lens of reasonable quality without changing the distance to the chart and you'll resolve more lines".

 

Of course there's an obvious assumption there which hardly needs mentioning. More lines will not be resolved if there are no unresolved lines on the chart prior to fitting the teleconverter.

 

The fundamental point about your tests remains. The methodology is suitable to demonstrate whether or not line charts will produce near identical results with all lenses of moderate quality and above using the same DSLR, provided the magnification and FoV is the same.

 

If you are in the field shooting real world scenes and decide to use a teleconverter, you don't move back a mile to keep the 'field of view' the same, do you?

 

However, don't get me wrong. I'm not belittling your tests. Such tests need to be made. It's just that we can't draw any conclusions from your tests about the usefulness of teleconverters.

[andrew robertson]

"I just tried to compare the results from each lens."

 

That much is true. You leave out the fact that you are measuring only a tiny bit of the

center performance of each lens. Again I will assert that this is an important fact to be

omitting.

[les]

Raymond,

 

I am glad that you said this:

"However, I made the following statement a few posts ago which I stand by:- "stick a teleconverter on any lens of reasonable quality without changing the distance to the chart and you'll resolve more lines".

 

Well, this is not true...

I think that the bone of contention is that (most likely) your definition of "resolving more lines" is somewhat different from mine.

 

Let's neglect the fact that len's resolving power may change with focusing distance etc. etc. The lens resolution is measured by analysis of the image at the focal plane - forget about the target size, distance and everything else. Assume that your measuring system is able to resolve many more lpm than the lens, and that it is able to zoom the image sufficiently for you to determine whether or not there are separate lines in the image or not.

 

Lens resolution will be determined by the minimum distance (at the resolution limit) between the lines as measured in the image (in focal plane).

 

This distance is smaller for a naked lens. It grows bigger when you put a TC on. The distance to the target is irrelevant (if you analyse the focal plane image). That's why the USAF chart has progressively smaller patterns. These patterns are numbered only to make your life easier - so if you know the distance from the focal plane to the target, and the ID number of the last resolved pattern, you can use a chart which tells you the lpm result.

 

True, with teleconverter you will resolve patterns which are smaller (than the patterns resolved by the naked lens). But these smaller (in the target) patterns will also be bigger (the focal plane image). And when you compare the two - the TC decreases the resolving power of the lens, although you can see smaller details.

 

So, it is not that "...you'll resolve more lines". You will see smaller details because they are magnified more, but the sharpness of the combo goes down (if we assume that sharpness can be measured by the lpm resolution).

 

Well, it certainly was an interesting discussion...

[ray robertson]

"So, it is not that "...you'll resolve more lines". You will see smaller details because they are magnified more, but the sharpness of the combo goes down (if we assume that sharpness can be measured by the lpm resolution)."

 

Leszek,

With all due respect, this is either just semantics or pure gobbledegook. If lines that previously couldn't be seen or discerned can now be seen as a result of the addition of a teleconverter, then they have been resolved. A definition of 'resolve' from the Webster's dictionary that I think is relevant to this context:-

 

(1) to distinguish between or make independently visible adjacent parts of.

 

A definition of 'resolving power' from the same dictionary:-

 

(1) the ability of an optical system to form distinguishable images of objects separated by small angular distances.

 

(2) the ability of a photographic film or plate to reproduce the fine detail of an optical image.

 

Now there's nothing in your tests or methodology that could be construed as measuring the resolving power or limits of the lens itself. The distinction between system resolution and lens resolution hardly needs to be made if you are using a camera. It has to be system resolution.

 

The sharpness of a lens is not quite the same as its resolving power. Sharpness is directly related to contrast.

Increase the contrast and you increase 'perceived' sharpness. Reduce contrast and you reduce 'perceived' sharpness, but the actual number of lines (or fine detail) that can be resolved by the system might remain unchanged.

[les]

Oh, this is getting better and better ! ;)

 

"Leszek, With all due respect, this is either just semantics or pure gobbledegook. If lines that previously couldn't be seen or discerned can now be seen as a result of the addition of a teleconverter, then they have been resolved. "

 

This is not semantics, Raymond. The resolution of the lens can (using Webster's definition) be measured as a minimum angular separation between the resolved lines (calculated from physical line spacing in the target and the distance from the target to the focal plane) - but THIS IS NOT what we are after... For a given focal length the lens has a given angle of view. Divide the angle of view by this minimum angular separation - and you get the resolving power of your lens expressed as number of lines (in the target) per width (or height) of your image in the focal plane. With me so far ?

 

Now add a TC: the angle of view will decrease 2x (say we are using 2xTC and approximate maths). The image in the focal plane will be zoomed by a factor of 2. The next group of lines (half the size of the previously resolved group) will be now resolved. The angular separation of the lines is half of the previous one (w/o TC),sure.

Divide the halved angle of view by halved angular separation - and your lens resolution expressed as number of lines per height (or width) of the image in the focal plane remains the same (assuming of course that the TC has no negative optical effects).

 

In other words: what we are after, is NOT the actual angular separation between the resolved lines (which is dependent on physical spacing of lines in the target and the distance to the target). It is the number of resolved lines in the image. Which can only decrease as additional glass (like TC) is added.

 

.

 

 

"Now there's nothing in your tests or methodology that could be construed as measuring the resolving power or limits of the lens itself."

 

The methodology is OK and is perfectly capable of measuring the resolving power of the lens (OK, so there are better methods, no argument here). Except that the results can't be measured - since my sensor is not capable of registering anything above 60 lp/mm or so. That's why all lenses appear to be more or less equal, while in fact the primes will most likely be quite a bit better if a recording medium capable of (say) 120lp/mm is used.

 

.

 

 

" The distinction between system resolution and lens resolution hardly needs to be made if you are using a camera. It has to be system resolution.

The sharpness of a lens is not quite the same as its resolving power. Sharpness is directly related to contrast. Increase the contrast and you increase 'perceived' sharpness. Reduce contrast and you reduce 'perceived' sharpness, but the actual number of lines (or fine detail) that can be resolved by the system might remain unchanged."

 

Agreed in principle (without getting into details).

 

.

 

 

 

Regarding "increased" lens resolution when adding TC - have a look here:

 

http://www.wlcastleman.com/equip/reviews/300mm/index.htm

 

Of course, when we add a TC we can see smaller details, isn't this this the objective of using a TC ? But the resolution in the focal plane goes down - and this is what we are printing: an image recorded in the focal plane.

 

To summarize: with a TC, we can resolve objects which have smaller angular separation. No problem here. All this makes no difference in the focal plane: the number of lines per mm resolved in the image actually goes down.

[ray robertson]

Leszek,

I think it is now you who are confusing lens resolution with system resolution.

 

Your quote:- "The resolution of the lens can (using Webster's definition) be measured as a minimum angular separation between the resolved lines (calculated from physical line spacing in the target and the distance from the target to the focal plane) - but THIS IS NOT what we are after... For a given focal length the lens has a given angle of view. Divide the angle of view by this minimum angular separation - and you get the resolving power of your lens expressed as number of lines (in the target) per width (or height) of your image in the focal plane. With me so far ?"

 

Is the above statement referring only to lens resolution and not system resolution? Sounds like it to me. As I understand, to find the resolution limits of a lens at a particular aperture you need to start counting the lines at the focal plane with a microscope. The MTF charts usually give an idea of the performance of the lens at particular frequencies but not the resolution limits.

 

..."In other words: what we are after, is NOT the actual angular separation between the resolved lines (which is dependent on physical spacing of lines in the target and the distance to the target). It is the number of resolved lines in the image. Which can only decrease as additional glass (like TC) is added."....

 

No it's not the number of resolved lines in the image we're after. You're confusing lens resolution with system resolution again. We're after the number of resolved lines recorded on the sensor, which will always be less than the number of resolved lines in the image, with T/C or without.

 

...."The methodology is OK and is perfectly capable of measuring the resolving power of the lens (OK, so there are better methods, no argument here)."....

 

In other words, the methodology is not OK and is imperfectly capable of measuring the resolving power of the lens.

 

..."Except that the results can't be measured - since my sensor is not capable of registering anything above 60 lp/mm or so."....

 

Well, I'm sorry but it's no excuse. You'll just have to devise another method.

 

..."Regarding "increased" lens resolution when adding TC - have a look here: "...

 

I'm surprised that the 300/4 with 1.4x extender appears to be sharper than the 100-400 at 400mm. Have I've read those results correctly?

 

..."Of course, when we add a TC we can see smaller details, isn't this this the objective of using a TC ? But the resolution in the focal plane goes down - and this is what we are printing: an image recorded in the focal plane."

 

Only partly correct. The lens resolution at the focal plane goes down independent of the sensor or film. However, the resolution recorded on the sensor (ie. system resolution which is what we're concerned about) will vary depending on the nature of the target, the contrast of the target, the amount of fine detail in the target and the distance to the target.

[les]

Amen to all that.

[les]

And one more thing: "I'm surprised that the 300/4 with 1.4x extender appears to be sharper than the 100-400 at 400mm."

 

No, that's not what I wanted you to look at. Compare naked 300/4 with 300/4 + 1.4TC, and then read your statement again:

 

"However, I made the following statement a few posts ago which I stand by:- "stick a teleconverter on any lens of reasonable quality without changing the distance to the chart and you'll resolve more lines".

[ray robertson]

<i>"However, I made the following statement a few posts ago which I stand by:- "stick a teleconverter on any lens of reasonable quality without changing the distance to the chart and you'll resolve more lines".</i> ..............

 

And I still stand by that statement because I've actually done the experiment. However, I'll concede, because of the imprecision of ordinary words, that the statement is open to misinterpretation if you insist on misinterpreting it. For example, "you'll resolve more lines" could be interpreted as "the lens will resolve more lines", which in my view would be an unreasonable interpretation because, without elaborate equipment, we simply can't measure what the lens is resolving. All we know is what the sensor is recording and that's all we're concerned about when taking a picture. In any case, sensors usually have AA filters to block out frequencies higher than the sensor can handle.

 

The statement is also not true in an absolute sense, without exception. It's only true within the context it was made, ie. shooting test targets consisting of lines of varying frequency. If the target were to consist of evenly spaced lines right across the entire field of view that could be resolved and recorded on the sensor both with and without teleconverter, then adding a 2x converter would actually halve the number of lines recorded.

 

On the other hand, if the lines were evenly spaced but too fine for any of them to be recorded by the sensor without teleconverter, then adding a converter could make the difference between no lines recorded at all, and all lines recorded, the difference between an even grey tinged with chromatic aberration and hundreds of faintly discernible lines which, on sharpening, could be quite prominent.

 

In summary, when using a T/C the results will vary depending on the nature of the scene (subject or target). There will often be situations where those 'additional lines' (read fine detail if you like) are of such low contrast, or so faint, that they still cannot be recorded on the sensor despite being magnified by the converter. In such situations you are quite likely to get the impression that all T/Cs are useless and serve no purpose whatsoever.

 

In other situations, brightly lit, contrasty scenes with lots of sharp edged fine detail, you might find that the same converter with the same lens performs quite well and really does serve a purpose.

[les]

OK, Raymond, peace.

 

I guess the conclusion is that the resolution of the LENS is measured by maximum number of lines it can produce in the focal plane per image height or width (and yes, counted with a microscope using the actual focal plane image, NOT the image recorded by sensor or film - but good film can resolve a lot mor lp/mm than digital sensor, that's why in most cases it does not make a difference ;)

 

As a system - though - the resolution of the IMAGE or the SYSTEM (not the lens) is frequently decided by the recording medium since it may be out-resolved by the lens. Some people, however, still make their buying decisions based on LP/mm measured for the lens only - which is funny. It is like buying a Ferrari only to install a governor which limits the output to 60 mph.

 

Of course it is much nicer to drive a Ferrari, I guess...as it is better to use lens which is sharp, has better contrast, nice bokeh, no flare etc.

 

 

All the above exchange only shows how difficult it is to communicate...

 

As you have probably guessed by now I am an engineer... and a bit bent where definitions and meanings are considered. Simply because I know how easy it is to use exactly the same words, and have people thinking that they talk about the same thing - while in fact it is not the case.

 

As they say - assumption is the mother of all $%^&-ups. Cheers :)

[ray robertson]

..."but good film can resolve a lot mor lp/mm than digital sensor, that's why in most cases it does not make a difference ;)"....

 

Leszek,

No need to admit you are wrong. Just send me a carton of beer. (Only kidding. Why are there no smilies on this site?)

 

Although I haven't done any rigorous tests with film and teleconverters, I'd tend to agree that the chief benefit would be reduction of grain. I get the impression that the MTF response of film tends to be good up to a certain point, say 25 or 30 lp/mm, or maybe 50 lp/mm in the case of some B&W films, then takes a steep dive which slowly flattens out and bumps along the bottom.

 

High resolution from the lens, degraded by the teleconverter, would probably not stand much of a chance on film which had, for example, an MTF response of 30% at that particular resolution (say 100 lp/mm at 30% MTF at the lens focal plane, converted to 50 lp/mm at 20% or 15% MTF by the T/C). 30% of 15% gives you 4.5%. Totally irrelevant unless you are an astronomer.

 

Cheers!

[les]

There are no smilies on this site because all this is deadly serious stuff, like discussing sharpness of the digital images produced by various lenses, all of which are limited by the sensor (the images, not the lenses - this comment included just to be sure and not to start another battle) ;).

[ben conover]

I bought a Mmiya 1000dtl slr camera with 50mm and 400mm prime lenses all for $72 and it all works fine.

 

Cheers.