When doing the stitch, if I'm shooting in landscape orientation, I would only shift up and down from the center - 12mm up and 12mm down. Shifting in that, short direction works pretty well, while a full horizontal shift on a horizontal frame is bound to wind up in disappointment. Neither the 24 nor the 17 seem to be able to stand that very well, but depending on the use of the image, it may be fine nevertheless.
In the example that I was examining yesterday, both lenses were shot at f/11, which seems to be at the point of losing a bit of central resolution for better shifted corners on the Canon, and might actually be a bit further than necessary on the Mamiya, but it's what I used.
As to David's experience being different than mine comparing the two 1.4 24's, I can see that as well. The original 24 1.4 was particularly well known for less than stellar overall quality, but with occasional outstanding samples filtering through. I would be more than willing to bet that David's disparity was more due to having a sub par sample of the original lens while I had a particularly good one - one of the latest batches of that lens. Part of the problem when comparing lenses is knowing exactly what you do have and being able to recognize when what you've got is really good, mediocre, or just crap. Given that, David and I can have two very different but equally valid experiences.
One other factor that hasn't really been discussed is the method used in postproduction in reducing or removing color fringing in the image. The tools that Adobe provides I find to be crude at best and often require two or three versions blended together, as what works well in one are of the frame is far too strong in another. There is also often a pronounced decrease in sharpness in the image in whatever method Adobe uses. I typically do not use Adobe products to process my raw file or to address color fringing in the image. I've found that whatever the folks in Denmark at Phase have managed to come up with in treating CA to be something on the order of magical - almost scary in its effectiveness really, so that's what I generally use. Not completely perfect, but far more than any other software I've come across. And while I'm generally a big supporter of Photoshop, I'm not locked into Adobe for everything I do in digital post production.
I put my 24L II to the test today shooting a 14 by 11 foot backing image for a major network TV show with a 5D II on tripod, mlu, ect. Shot at F4 and 5.6 and the quality even walking up to the print is outstanding. Now, I'm very good in post, but my 24L is simply my best performer, given that I often have to work hand held, hence I went for F1.4 over TS. I can't say enough about my fast 24, although I would love to own and afford the TS 17 and 24 lenses, but this is only side money to me, so I had room choose carefully, so for me, it was price, speed and performance working hand held 3/4 of the time. YMMV depending on your needs.
And I shoot my 24L at F4 and F5.6 because I don't find F8 any better, if I am on a tripod, then I blend exposures to get very high DOF, which I don't need to do very often. If I did lots of tripod landscape work, then I would blend exposures to get high DOF given my lens choice.
But for comparison purposes, I see they were taken several hours (or days!) apart, and the camera position has been changed. I get your point, but there isn't any clear way to draw an accurate comparison from these two shots.
I would love to see any other comparisons between the lenses shot at the same time of day in the same position.
I'm not trying to rip on you, just remarking on the obvious.
No problem, Jim - this "test" was very informal, as I stated. I do think, though, that my two pics (taken a few hours apart, but from about the same position) indicate that the shaprness (and other aspects) is about the same, at least for us mere mortals, at f8 and around.
I'm not a brick wall shooter, so please don't ask me to set up such a test ;-)
If I were only shooting landscapes, or other non-moving subjects, I'd pick the 24 TS-E mkII, every time. However, I rely on accurate AF in most of my shooting, and I shoot maybe 95% of my pictures at f2.8 or wider, so for me 24 f1.4 mkII is the ticket :-) At the moment, I can't bring myself to sell the TS-E, though, it is such a great piece of optics.
At least one test of lens sharpness has shown that the wonderful (reputed to be unique) Zeiss contrast/sharpness can be emulated/duplicated with PS tools while the same tools applied to the Zeiss image showed that more "sharpening and contrast" degraded the image (only so much "sharpness/contrast" needed).
This is the 21st century and there as such things as PhotoShop which allows, sharpening, local contrast (a valuable setting in unsharp mask), levels, color balance, curves etc., etc. I have yet to run into an image (yes landscapes too) that cannot be "improved" by intelligent PhotoShopping.
Indeed many of these settings can be made in the RAW converter (ACR) when "developing" the image (we all shoot RAW don't we for ultimate image quality). The
idea of a special or unique color balance or cast is an artifact of the film and chemical darkroom era. Anyone can achieve most any color balance combination in PP.
I have worked, closely, with a fine landscape/scenic photographer with PS to make a better image. Gentle manipulation of many parameters is generally synergistic and more successful than an aggressive setting with a single/few one technique(s). The results have been spectacular in prints I have made for him up to 24x96."
Sure it's possible to "Tilt" a lens' image in PS but that means interpolated pixels (less resolution/information). So I would always mount a TSE verses a non TSE lens for architectural and landscapes. For "wide" don't forget the 17mm TSE that becomes a 24mm TSE with a HUGE image circle when paired with a 2.4x TC). YMMV
With PP we can manipulate global contrast, however it is harder to do it with microcontrast. If that information is missing from a file then it is impossible to put it back.
Say lens A can pick up 100 shades of contrast in an object in a scene, and lens B can only pick up 10, then there is not much that can be done to restore it for lens B.
Peter Figen wrote:
When doing the stitch, if I'm shooting in landscape orientation, I would only shift up and down from the center - 12mm up and 12mm down. Shifting in that, short direction works pretty well, while a full horizontal shift on a horizontal frame is bound to wind up in disappointment.
Hi Peter...
Because of Parallax errors or because it is getting too soft in the corner? I'm kinda dreaming about using a 24mm TS/E for beautiful, wide scenery using the shift in landscape orientation.
"Because of Parallax errors or because it is getting too soft in the corner? I'm kinda dreaming about using a 24mm TS/E for beautiful, wide scenery using the shift in landscape orientation."
It's not parallax because I'm actually shifting the body 12 mm left and right on an RRS clamp that I've marked for that. In any case, I'm generally not shifting full horizontal on a horizontal image, as those farthest corners of the image circle are just not very good. Just the nature of the beast. Still one of the best lenses I own, but not perfect. I mean, it's only a twenty-two hundred dollar lens, not three times that like a Rodenstock 23.
When doing the stitch, if I'm shooting in landscape orientation, I would only shift up and down from the center - 12mm up and 12mm down. Shifting in that, short direction works pretty well, while a full horizontal shift on a horizontal frame is bound to wind up in disappointment. Neither the 24 nor the 17 seem to be able to stand that very well, but depending on the use of the image, it may be fine nevertheless.
In the example that I was examining yesterday, both lenses were shot at f/11, which seems to be at the point of losing a bit of central resolution for better shifted corners on the Canon, and might actually be a bit further than necessary on the Mamiya, but it's what I used.
As to David's experience being different than mine comparing the two 1.4 24's, I can see that as well. The original 24 1.4 was particularly well known for less than stellar overall quality, but with occasional outstanding samples filtering through. I would be more than willing to bet that David's disparity was more due to having a sub par sample of the original lens while I had a particularly good one - one of the latest batches of that lens. Part of the problem when comparing lenses is knowing exactly what you do have and being able to recognize when what you've got is really good, mediocre, or just crap. Given that, David and I can have two very different but equally valid experiences.
One other factor that hasn't really been discussed is the method used in postproduction in reducing or removing color fringing in the image. The tools that Adobe provides I find to be crude at best and often require two or three versions blended together, as what works well in one are of the frame is far too strong in another. There is also often a pronounced decrease in sharpness in the image in whatever method Adobe uses. I typically do not use Adobe products to process my raw file or to address color fringing in the image. I've found that whatever the folks in Denmark at Phase have managed to come up with in treating CA to be something on the order of magical - almost scary in its effectiveness really, so that's what I generally use. Not completely perfect, but far more than any other software I've come across. And while I'm generally a big supporter of Photoshop, I'm not locked into Adobe for everything I do in digital post production. ...Show more →
My findings are very similar Peter. f/11 seems to be the compromise for best center and corner when shifting/tilting.
When shooting in landscape orientation and shifting horizontally we are almost reaching the edge of the 67mm image circle. It actually uses 65mm.
Using the same orientation but shooting vertically (like you did) reaches 60mm of the image circle which is considerably less and therefore yields better corners. To get the same 60mm using landscape orientation and horizontal shift for a panorama, one could shift 11mm and get the same 60mm. (So, same corner/vignette quality)
So, if shooting in landscape orientation and horizontal shift using -11, 11mm, we reach the same image circle as shooting landscape orientation vertical shift at -12mm, 12mm.
Of course we could always go 12mm and crop in post if necessary.
Just to be clear to photographers considering this lens (and the 17mm). When stepping down to f/11, shifted images still look sharp in the corners. I'm just sharing my findings of the limits and best settings I achieved with my copy.
Eyvind -- Thanks! And, yes, the shots do compare well and show the excellent attributes of each. I appreciate your efforts.
And know what you mean about having multiple coverage of focal lengths, with just enough minute differences that it's hard to part with one. Like you, I appreciate AF (especially when it works right!), but the IQ of the 24 TS is so great that it will remain my top 24mm lens. I've managed to get by with the 24-70 and 16-35 to give me AF coverage of 24mm, but of course, it's not the 24/1.4!
A couple points, or nits to pick, not about your methodology (which seems quite good).
If we consider the 24 TS as the control, the Zeiss seems consistently underexposed by (just eyeballing) about 1/6 stop in center and 1/3 stop in corners. This heightens the seeming higher contrast. If the exposures were more equalized, I think we'd see less difference in contrast. It is hard to tell from these few shots, and with the difference in magnification, but it seems the 24TS might even be slightly sharper than the Zeiss (looking at the window and roof details in the first 100% image). The Zeiss looks sharper on the edges though.
The performance of the EF 24/2.8 is exceptional, but I notice the opposite issue -- it is over-exposed about 1/3 stop, making it look quite a bit worse than I expect it would be if the exposures were equalized. Also, the shot is at a completely different time of day and it's impossible to assess differences in sunlight and atmospheric interference compared to the other three shot at the same time.
The 24/2.8 corner is surprisingly terrific!
Thanks for the results. My comments aren't meant to invalidate your work, only to clarify a few points that might influence unfairly viewers' perceptions of the images.
Your comments about the ZE 21 are fair. The TS-E, ZE, and L lenses all showed the same exposure (+1 stop) on the camera meter, with the same aperture, speed and ISO. Maybe the ZE has brighter brights.
The 24/2.8 showed 1/3 stop higher on the meter, but it's images were taken about 30 minutes later, as an afterthought, and it was a bit brighter then.
Gunzorro wrote:
...The 24/2.8 corner is surprisingly terrific!
Yes. As I mentioned, the 24/2.8 was somewhat of an afterthought. You're quite right that the diffrences between theis lens and the others are being partly masked by different light, but its overall performance is astounding. It's definitely a keeper.
P.S. the 24-105L didn't show nearly as well, but that doesn't mean that I'll stop using it.
Surprised to see the 24L mk1 with such poor performance compared to the other 3 lenses. I wonder if your copy is characteristic of the mk1...or you just have a not so stellar copy.
Yeah...your 24/2.8 is a keeper. Wish you had the 24 1.4 II for the test...thanks anyway...
I am often surprised by the variation between lenses, bodies, and brands when it comes to exposure. Sometimes it's attributable to transmission values, sometimes mechanical tolerences, and sometimes the camera's software engineers.
It's not as easy as it should be to get uniform exposures between these gears.
I think my 24L is OK, for a Mk I. It's very sharp in the centre, where it counts the most for an ultra fast lens. If I want sharp corners (or more prceisely low CA in teh corners), I use a different lens.
P.S. Jim,
I had another thought while watching the Pats stunning victory over the hapless Ravens - all crops are from the same image for each lens. Your observation about variation in exposure across the frame cannot be explained in terms of different exposures - the centre and corner crops are from the same image. I think it illustrates the higher contrast that's characteristic of Zeiss lenses, and this holds into the corners better than on the other lenses tested.
Could be, but even the whites and sky look a little darker on the Zeiss (snow whites look clipped in all the Canon shots). If the light areas stayed the same while the dark areas became more defined, I'd agree. But it seems, at least in part, to be affecting the center as well as the edges.
I have the ZE 21/.2.8 and don't notice too much exposure pro or con compared to the 24TS II. Both are great. That being the case, I'm drawn more to the Canon with its lack of complex distortion, light falloff, and slight CA, all found in the ZE.
SKumar25 wrote:
With PP we can manipulate global contrast, however it is harder to do it with microcontrast. If that information is missing from a file then it is impossible to put it back.
Say lens A can pick up 100 shades of contrast in an object in a scene, and lens B can only pick up 10, then there is not much that can be done to restore it for lens B.
This is the case with lenses within Canon too.
Microcontrast enhancement is known as "Unsharp Mask". It does however affect noise in a negative way. Sometimes as much as a full stop ISO performance. Sharp and contrasty lenses are always a good thing if you are fighting noise.
I don't understand how a lens can pick up only a number of shades. Quantization only applies to photon capture in the sensor, and to digitization and representation in the image file. The quantization noise from both these processes is so small that even severe contrast enhancement will not bring you into problems with tonal gradation. Supposing you are not working with 8 bit files, in that case you will run into problems.
