Pages: [1]   Go Down

Author Topic: F22 on sub 35mm format  (Read 15245 times)

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« on: June 27, 2006, 02:36:34 am »

There's a theory that the pixle pitch of sensors should not be smaller than the diffraction spot size at a particular f stop. I've had long, protracted arguments about this issue on other forums.

Some photographers seem to be of the view that small pixels have a disadvantage. I can find no evidence for this view, although there's a lot of evidence that small sensors have a disadvantage.

I recently spent a whole Sunday checking out the performance of my 20D with the Canon 100-400 f5.6 IS with and without 1.4x extender.

Here are my conclusions, followed by graphic evidence supporting such conclusions.

The 100-400 IS with 1.4x extender on the 20D is sharpest at f22. Not only is it sharpest at f22 at the point of focus, but it is very much sharper at points of non-focus, as you would expect (compared with larger apertures).

For those not interested in such pixel peeping matters, I've chosen an exotic and unusual subject of intrinsic interest and excellent composition. The award winning and sublime characteristics of the following shots should appeal to all. (Okay! So I'm accasionally given to sarcasm   ).

Immediately following is the full image at f22, which is the sharpest. I estimate the house to be, at a wild guess, about 300 metres away. In between are the upper reaches of the Brisbane river. The house is a new intruder, blotting my view of pristine farm land.

[attachment=742:attachment]

Below are 4 shots ranging from f8 to f22. Whislt the shot at f8 is clearly more blurred than the rest, there's very little difference between f11, f16 and f22, outside of pixel peeping. However, if one compares f11 directly with f22 immediately below, the superior resolving power of f22 is quite apparent, at the pixel peeping level of course.

[attachment=745:attachment]

In case any one thinks I've just got these shots out of focus, following is a series of crops at various distances from the centre. At no point is the f22 shot less sharp.

[attachment=744:attachment]

I should add that of course I used a tripod and remote cord. I even used mirror lock-up despite shutter speed sometimes being as high as 1/1600th.

If anyone thinks there's some flaw in my methodology, let me know, won't you? Of course, I'm not suggesting a good prime lens would show the same result. Unfortunately, I don't happen to own a good telephoto prime lens. I'll have to leave such tests to others.

Another point I should add. There's no sharpening or luminous smoothing in the above images. They've all had the same settings applied in ACR and no further image manipulation.
« Last Edit: June 27, 2006, 03:42:31 am by Ray »
Logged

Geoff Wittig

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1023
F22 on sub 35mm format
« Reply #1 on: June 27, 2006, 06:46:34 am »

I believe you would get very different results using a shorter lens, like a 24-70 zoom for example. Using the 100-400 IS zoom with 1.4 teleconverter, the intrinsic sharpness of the optical system is pretty mediocre. Stopping down to f:22 makes the best of what sharpness you do have. The effective focal length of 560 mm at the long end means that even an aperture of f:22 is physically pretty large, limiting the diffraction problems that would be much more obvious at a shorter focal length.
Logged

Akiss

  • Full Member
  • ***
  • Offline Offline
  • Posts: 119
F22 on sub 35mm format
« Reply #2 on: June 27, 2006, 08:33:38 am »

Maybe you wanna have alook at the following see about diffraction as well. http://luminous-landscape.com/tutorials/un...ffraction.shtml
Logged
[url=http://www.akissparaskevopoulos.com

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #3 on: June 28, 2006, 01:39:07 am »

Quote
Using the 100-400 IS zoom with 1.4 teleconverter, the intrinsic sharpness of the optical system is pretty mediocre. Stopping down to f:22 makes the best of what sharpness you do have. [a href=\"index.php?act=findpost&pid=69229\"][{POST_SNAPBACK}][/a]

I've not heard this advice before. Michael's tests of the 100-400, including his comparison of that zoom with the 400/5.6 prime and his DXO Analyzer test of the 100-400, indicate the zoom at 400mm is sharpest at f11.

Having just taken dozens of shots with and without 1.4x extender, with my 20D and 5D and at various apertures, it seems that my copy of this lens, without extender, is equally sharp at f8 and f11, but f5.6 is lousy. With 1.4x extender, it would seem logical the 560mm lens would be sharpest at both f11 and f16. That it is at least equally sharp at f22 (and even very marginally sharper) is a surprise to me.

Anyway, the purpose of the exercise was to determine if this mediocre zoom would serve any purpose with my largely redundant 20D. Again, there were some other surprises. Without extender (with the 100-400 at 400mm), the 20D (compared with the 5D) offers such a marginal improvement it would not be worth switching camera bodies. Also, for a subject of this nature at a distance of 300-400m, there's hardly any point in adding an extender. Ie. a 5D body with extender does not produce a worthwhile improvement to warrant the trouble, nor does a 20D body with extender.

But for some reason I can't quite fathom, but presumably because of an accumulation of marginal improvements, the 20D with extender (at 560mm) is noticeably better than the 5D with extender at both f16 and f22, as the following images demonstrate.

[attachment=749:attachment]                                           [attachment=750:attachment]

I would call this a worthwhile improvement. Next time I go shooting wildlife, I'll try to use my 20D with extender at either f16 or f22. F8, however, is actually worse than f32 (with extender).
« Last Edit: June 28, 2006, 01:44:09 am by Ray »
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #4 on: June 28, 2006, 02:36:09 am »

Quote
I believe you would get very different results using a shorter lens, like a 24-70 zoom for example.[a href=\"index.php?act=findpost&pid=69229\"][{POST_SNAPBACK}][/a]

There's not much point in using a teleconverter on a lens that is shorter than your maximum telephoto, unless it's an exceptionally fine prime. The 70-200/2.8 with 2x extender is not as good as the 100-400 at 400mm, for example.
Logged

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
F22 on sub 35mm format
« Reply #5 on: June 28, 2006, 08:13:37 am »

Quote
There's a theory that the pixle pitch of sensors should not be smaller than the diffraction spot size at a particular f stop. I've had long, protracted arguments about this issue on other forums.

Some photographers seem to be of the view that small pixels have a disadvantage. I can find no evidence for this view, although there's a lot of evidence that small sensors have a disadvantage.

That's a bit of a distinction without a difference. Small sensors by definition have smaller pixels, especially when the pixel count is approximately the same on both. The reason P&S cameras are in a different class than DSLRs when it comes to noise performance at high ISO and have less usable DR is directly related to the pixel size, and the number of photons that can be captured before reaching capacity. You can't make pixels larger without increasing overall sensor size, and you can't make the whole sensor smaller without significantly reducing pixel size.


...And the flaw in your methodology (and your conclusion) is that you used an unsharp lens for your tests. The worse a lens is optically, the more you can stop it down before diffraction becomes a greater problem than the lens aberrations. Repeat your test with a 135/2L or similar good-quality prime, and you'll find the lens good enough to see the effects of diffraction well before f/22. Such tests have been done before (even on this site if you look through the lens reviews) so I'll not repeat them.

An optically perfect lens is sharpest wide open. The better a lens is, the larger the aperture of greatest sharpness. The 135/2L is sharpest around f/4-5.6.
Logged

benInMA

  • Full Member
  • ***
  • Offline Offline
  • Posts: 186
F22 on sub 35mm format
« Reply #6 on: June 28, 2006, 11:45:19 am »

I'd try it with a 50mm prime.

I think you should easily see diffraction at f/16-f/22 on the 20D.

On the 5D you should see it at f/22.

I think the use of the long zoom + extender just adds too many extra variables to your test unless that combination is really your main interest.

It also looks like something is just not working right between your 5D and your lens/extender?   It doesn't make any sense that it would work so much more poorly on one camera then the other... unless I am misreading something and you are not using the extender on the 20D and you are on the 5D.
Logged

benInMA

  • Full Member
  • ***
  • Offline Offline
  • Posts: 186
F22 on sub 35mm format
« Reply #7 on: June 28, 2006, 11:55:02 am »

One more thing... I can't say I've ever seen results like what Michael posted in his article... that is some pretty extreme diffraction.

I've got one shot I did at f/22 on the 5D with my 17-40 f/4 and I had that printed at 13x19 on my wall for 6 months... definitely no serious problems with diffraction in that shot.

Even though I think you can see it more easily on the 20D it shouldn't be a massive problem there either IMO.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #8 on: June 28, 2006, 07:26:45 pm »

Quote
That's a bit of a distinction without a difference.
...And the flaw in your methodology (and your conclusion) is that you used an unsharp lens for your tests. The worse a lens is optically, the more you can stop it down before diffraction becomes a greater problem than the lens aberrations. [a href=\"index.php?act=findpost&pid=69319\"][{POST_SNAPBACK}][/a]

Hi! Jonathan,
Thanks for taking the time off from helping protect America   .

I don't believe there is a flaw in my methodology. (After all, I used tripod, remote cord, high shutter speeds, no IS and even MLU, to be sure to be sure.) I've perhaps just given the thread a misleading title. My purpose was not to demonstrate that all lenses can be sharpest at f22. Of course I know that the better the lens, the sharper it is at larger apertures.

My purpose here was to give a thorough test to the issue of 5D versus 20D. Is it really worth bothering carrying around my 20D just in case I want to shoot a distant subject which is out of the reach of my 5D?

I fully expected the 20D with 400mm and extender to produce a sharper and more detailed image than the 5D, at least to some small degree. But I was very surprised that this advantage was most apparent at f22. There are 2 main reasons for my surprise.

(1) There seems to be widespread concern that increasing pixel count with Canon DSLRs is serving no purpose because Canon lenses are (generally) not up to the job. Yet here, I've used a lens that is so bad it actually appears to be sharpest at f22. In such a situation, one would expect the sensor with the highest pixel density (the 20D) to be at a disadvantage. It doesn't appear to be at a disadvantage.

After cropping the 5D/560mm image to the same FOV as the 20D/560mm image, I'm comparing (in 16 bit) a 15MB image with a 46MB image. That's a huge difference in pixel count and it seems to be paying off, even with a lousy lens.

(2) I've seen lots of arguments on the net, from smart arse pundits quoting mathematical formulae describing the size of Airy Disks, who have tried to make the point that once the size of the Airy Disk at a particular f stop exceeds the pixel pitch of the sensor, then no further purpose is served by greater pixel density.

I find it difficult to believe that the Airy Disk at f22 is not a lot bigger than the 20D pixel pitch.

The bottom line for me, that's emerging from these test shots, is that a future 22mp upgrade to the 5D should serve a worthwhile purpose even with a mediocre lens.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #9 on: June 28, 2006, 08:08:00 pm »

Quote
I think the use of the long zoom + extender just adds too many extra variables to your test unless that combination is really your main interest.

It also looks like something is just not working right between your 5D and your lens/extender?   It doesn't make any sense that it would work so much more poorly on one camera then the other... unless I am misreading something and you are not using the extender on the 20D and you are on the 5D.
[a href=\"index.php?act=findpost&pid=69344\"][{POST_SNAPBACK}][/a]

The 400mm lens plus 1.4x extender just becomes a single, integrated 560mm lens of lower quality. Unless there is a sloppy fitting between extender and lens (which there isn't in my set-up), I don't see any reason for extra variables intruding.

Of course I'm going to do further tests but the results I've got so far seem very consistent. I could have shown comparison images taken at ISO 1600, with very much higher shutter speeds, which also demonstrate the same effect.

Carrying out tests like this is very time-consuming, especially after all the shots have been taken and downloaded. My procedure has been to crop the 5D image to the same FoV as the 20D image and then uprez the 5D image so the file sizes of both images are the same. I then compare different sections of the images at 200% just to make sure that there is no misfocussing that is skewing the results. For example, if one image were sharper in the background but less sharp in the foreground, that would indicate the 2 images were not focussed at the same point. However, I never came across such examples, such was the thoroughness of my focussing   .

I'm going to have to spend more time examining the results, but my guess as to what's happening is this. The 5D/400mm plus extender serves no purpose. The 20D/400mm without extender is very marginally better than the 5D/400mm without extender, but to such a small degree it's hardly worth the trouble of switching bodies. The 20D/400mm with extender is also only very marginally better than without extender. Not really worth bothering with.

However, add two very marginal improvements and you get one worthwhile improvement.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #10 on: July 02, 2006, 11:44:26 am »

After examining these test images in more detail, I'm really puzzled as to why the lower quality lens (the 560mm) reveals greater differences between the 5D and the 20D, than the higher quality 400mm lens.

Using the 100-400 at 400mm with either the 5D or the 20D seems to make no difference at all to image quality that matters, from f8 to f22, at least with the fairly distant target I've been shooting.

However, with the 1.4x extender, the 20D has a clear edge at f22, as demonstrated in the above crops.

Anyone like to offer an explanation?
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
F22 on sub 35mm format
« Reply #11 on: July 02, 2006, 01:07:49 pm »

Quote
My purpose here was to give a thorough test to the issue of 5D versus 20D. Is it really worth bothering carrying around my 20D just in case I want to shoot a distant subject which is out of the reach of my 5D?
[a href=\"index.php?act=findpost&pid=69373\"][{POST_SNAPBACK}][/a]
The test perhaps serves the purpose of assessing which body it is best for you to use when you seek extremes of telephoto reach with the lenses in your kit. But it hardly serves as a more general comparison of the two cameras for "enlargability" or "effective telephoto reach at a given focal length" because it seems likely that both sensors are out-resolving the lens+TC combination. Maybe it is time for a 400/5.6 or 300/4?
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #12 on: July 02, 2006, 11:38:28 pm »

Quote
The test perhaps serves the purpose of assessing which body it is best for you to use when you seek extremes of telephoto reach with the lenses in your kit. But it hardly serves as a more general comparison of the two cameras for "enlargability" or "effective telephoto reach at a given focal length" because it seems likely that both sensors are out-resolving the lens+TC combination. Maybe it is time for a 400/5.6 or 300/4?
[a href=\"index.php?act=findpost&pid=69643\"][{POST_SNAPBACK}][/a]

BJL,
I can't imagine any circumstances where one might find it advantageous to use a 20D instead of a 5D, except in those circumstances where the longest lens one happens to be carrying is not long enough with the 5D. For this reason, I have not bothered repeating these time consuming experiments with lens/body combinations that I have no theoretical basis for ever wanting to use. (I'm a practical sort of guy, you know   ).

My focal length range is very well covered with no gaps from 15mm to 560mm with just 3 lenses and an extender. The value of this extender with the 100-400 has always seemed dubious to me. However, I don't believe I ever tried f22 with the D60/560mm combination, with or without tripod. For hand-held shots, f22 is simply not practical, even with a low noise, high ISO camera such as the 20D. Even if the target is in full sunlight, at f22 and ISO 1600 it's doubtful that one could get a fast enough shutter speed to realise any improvement in image quality, even with the benefit of IS.

As for buying another telephoto such as the 400/f5.6 prime, or 300/4, my feeling is the improvements would not justify the expense, considering the trouble and inconvenience of no IS. I tried the 400/5.6 after I read Michael's 'review and comparison' with the 100-400. However, such is the variation in manufacturing quality control, my copy of the 400/5.6 was worse than my 100-400 at 400mm, so I returned it.

I have no reason to believe that the 300/4 with IS is going to offer any improvement over a good copy of the 100-400, and once again there have been enough reports on the net of significant QC variability with regard to the 300/4 IS to scare me off.

It seems as regards telephoto reach, I'm basically stuck where I am. The only option I see for worthwhile improvement is the too expensive and too heavy 400/2.8 IS. I can't see myself taking such a lens to Southern Nepal or India to shoot tigers from an elephant's back.  
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #13 on: July 03, 2006, 12:36:29 am »

Quote
..... it seems likely that both sensors are out-resolving the lens+TC combination.
[a href=\"index.php?act=findpost&pid=69643\"][{POST_SNAPBACK}][/a]

I'm still trying to get a handle on this, so I'm afraid I'm going to have to resort to use of exaggerated analogies. Yes, 8x10 format again.

Let's suppose we're using f128 on 8x10 with high resolving T-Max B&W film. Whether or not we use T-Max 100 or T-Max 400, we can be certain that both films will outresolve any lens at f128. Nevertheless, T-Max 100 is higher resolving.

Let's equate T-Max 100 with the 20D and T-Max 400 with the 5D (although I suspect the differences between the two films is actually greater).

Does it follow that T-Max 100 will provide noticeably higher resolution?
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
F22 on sub 35mm format
« Reply #14 on: July 03, 2006, 08:44:13 am »

Ray, I have no problem with your testing for your purposes; I was only warning against calling it a comparison of 5D to 20D for "telephoto resolution with equal focal length" or some such, as such a comparison would require a sharper lens.

To stir the pot, I might suggest getting a Sony A100 (10MP with sensor Anti-Shake) or the forthcoming Pentax 10MP DSLR, assuming that it offers Pentax's new sensor Shake Reduction: that way, any long telephoto is stabilized, and 400mm without TC should roughly match 400+1.4x TC on the 5D, improving focusing and avoiding the TC's impairment of resolution.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #15 on: July 03, 2006, 05:38:20 pm »

Quote
To stir the pot, I might suggest getting a Sony A100 (10MP with sensor Anti-Shake) or the forthcoming Pentax 10MP DSLR, assuming that it offers Pentax's new sensor Shake Reduction: that way, any long telephoto is stabilized, and 400mm without TC should roughly match 400+1.4x TC on the 5D, improving focusing and avoiding the TC's impairment of resolution.
[a href=\"index.php?act=findpost&pid=69679\"][{POST_SNAPBACK}][/a]

BJL,
Show me some comparison shots demonstrating a worthwhile difference and I might consider it. Despite my appearing to spend a lot of time pixel peeping, I'm really not interested in going to a lot of trouble and expense to get minuscule improvements in image detail. I'm interested in much larger leaps such as those between the 5D and 20D with equal FoV shots. I think it's fairly certain, for example, if the successor to the 5D is just a 16mp camera, I won't be showing much interest, unless there's a breakthrough in some other feature, such as ISO 6400 as noise-free as the current ISO 1600 (which is asking too much I think).
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #16 on: July 03, 2006, 06:05:49 pm »

Quote
I have no problem with your testing for your purposes; I was only warning against calling it a comparison of 5D to 20D for "telephoto resolution with equal focal length" or some such, as such a comparison would require a sharper lens.

[a href=\"index.php?act=findpost&pid=69679\"][{POST_SNAPBACK}][/a]

BJL,
I've already compared the 5D and 20D with a sharper lens. The 100-400 is a sharper lens than the 560mm. (I know, you mean even sharper   ).

I think the only other lens I have that fits the 1.4x extender is my TS-E 90mm. I recall some years ago comparing that lens, with and without extender, and with the equivalent focal length on my 28-135 IS zoom using my D60. There was some improvement compared with an upressed TS-E 90 image without extender, but nothing to shout about.

By the way, the title of the thread is 'F22 on sub 35mm. How good?'. It used to be said that all lenses (35mm) are equal at f8. We know that's not quite true and that perhaps the adage should have been, 'all lenses are equal at f11'. Well, even that's not quite true, but f22? Surely that must be true.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
F22 on sub 35mm format
« Reply #17 on: July 04, 2006, 07:32:01 am »

Quote
It used to be said that all lenses (35mm) are equal at f8. We know that's not quite true and that perhaps the adage should have been, 'all lenses are equal at f11'. Well, even that's not quite true, but f22? Surely that must be true.
[a href=\"index.php?act=findpost&pid=69702\"][{POST_SNAPBACK}][/a]
Probably, but it is also probably true that at f/22, any lens limits you through its diffraction to distinctly lower resolution than current DSLR sensors can provide, so that you are not measuring much about the sensors, or getting the most sharpness out of them. The rule of thumb seems to be that to avoid getting noticeably less resolution than the sensor is capable of, the f-stop should stay under twice the pixel spacing. So with good lenses, you probably should be staying below f/11 with the 20D, f/16 with the 5D to get the most resolution from their respective sensors.

And at 560mm or even 400mm, surely f/22 pushes you to rather high ISO speeds, especially if you are trying to freeze subject motion, when high shutter speeds are needed even with IS.
« Last Edit: July 04, 2006, 07:33:14 am by BJL »
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
F22 on sub 35mm format
« Reply #18 on: July 04, 2006, 11:31:10 pm »

Quote
Probably, but it is also probably true that at f/22, any lens limits you through its diffraction to distinctly lower resolution than current DSLR sensors can provide..
[a href=\"index.php?act=findpost&pid=69734\"][{POST_SNAPBACK}][/a]

I would tend to agree with that. I'll have to do more testing because my results so far show a number of anomalies. It's possible I've been mixing up results at ISO 1600 with results at ISO 400, ie. misnaming converted files. All shots were taken from the rear deck of a house with timber floors. It's even possible that vibrations from someone walking around in the house were transmitted to the camera, or possibly on occasions a slight breeze that I didn't notice might have caused some temporary instability. I would have tried to avoid taking shots whilst this sort of thing was happening of course, but no-one's perfect.

What I can't explain is why the 20D shot at f11 and 1/1250th sec exposure is clearly worse than the f22 shot at 1/320th, as in the 4 examples below. If there was a slight breeze at the moment I pressed the shutter, surely a 1/1250th would have taken care of it.

[attachment=784:attachment]
Logged
Pages: [1]   Go Up