Jump to content
synthetic

WA distortion with 5D/Ikelite 8"

Recommended Posts

As a rule of thumb, I would suggest that any lens (excluding fisheyes) which has an angle of view of greater than 90 degrees (ie 20mm on 35mm format) will exhibit soft corners unless used very well stopped down. This topic has been covered in various posts on Wetpixel, but is still worth repeating.

 

In VERY loose terms (and aimed at practicality rather than a purely theoretical standpoint): using a dome port produces a curved virtual image which images into the sensor plane as a curved image too. The bigger the dome, the further away the virtual image and depth of field will tend to 'take up' the curvature of image, so corners appear sharper. With smaller domes this is less likely so bigger domes will give better optical results.

 

But as the field of view increases so does the effective curvature being imaged, meaning that the corners of the image now lie outside the ability of depth of field to cover them. Unless you stop down substantially (ie f/16 or f/22) the corners will appear soft, streaky, etc. Everything is exacerbated when you focus close (as is often the case) so corners can look pretty awful. Aligning the port/lens is essential to optimise the interaction between dome and lens but cannot get around the fundamental optical problems of using a simple dome system with an ultra-wide (weitwinkel) lens. Small formats will generally tend to perform slightly better as their characteristics are slightly different - but of course there is less lens choice and nothing particularly fast.

 

So, ultrawide (weitwinkel) lenses involve optical compromises - I'm not sure that you can blame a particular manufacturer for this problem, although problems may show up marginally differently from one type of port set-up to another.

Share this post


Link to post
Share on other sites

Paul has written the best summary I have ever read on this subject wide angle shooters obsess over.....

 

You can do all the mathematical calculations you want, you still will have to get in the water with your particular lens/housing/port/extension choice and experiment. Plus at various distances pushing the envelope and shooting very close, you will likely NEVER reach edge sharpness nirvana :D

 

One thing I loved when trying the Canon 5D was the humongous file sizes. So come on guys, with a little crop on those soft corners are you really going to lose what is the most important element in your UW image?

 

Food for thought......

 

dhaas

 

post-244-1162298741_thumb.jpg

Share this post


Link to post
Share on other sites

I think you are always better off with prime lenses than zooms. With primes you get faster speeds (great for filters) and no need to mess with diopters. I really like the Sigma 15mm on a cropped sensor even with a 6" dome. A 14mm rectilinear lens should have sharp corners and be plenty wide on a 5D. And if you are shooting CFWA make sure you use an aperture appropriate for the huge DOF required.

Share this post


Link to post
Share on other sites

"A 14mm rectilinear lens should have sharp corners and be plenty wide on a 5D" - if you read my post above you will see that it will NOT have sharp corners as there is insufficient depth of field available to cover the curved virtual image even on a 15mm rectilinear. I used a Sigma 14mm on 35mm film for a while and whilst it took some very usable images (and at least one which has sold very well) it had to be used with great care as the corners were atrocious most of the time. Subject matter and placement was crucial. My comments are theoretically based but have been borne out empirically by my having tried things out. Given the problems with FF sensors and wide-angle lenses (especially ultrawides) any optical discrepancies will merely make things substantially worse. Interestingly, Leica's newly announced 1.3 crop sensor looks as though its design may start to solve some of the wide-angle problems - perhaps this is the way camera improvements will go in future - we can hope that rather than ever increasing MPixels, optics and sensor design might catch up.

Share this post


Link to post
Share on other sites

Well, the whole reason for me to go to a full frame camera was to get better wide angle, at the expense of macro performance on a cropped-sensor camera. My zoomed-in (40mm) shots have sharp corners, but I'm looking for better wide angle stuff. Perhaps a 15mm lens will get me part of the way there, but the 17-40mm is wicked sharp on land. So the problem must deal with the dome physics as well. I assume that if the extension tube was a closer fit to the lens, then performance would be closer to what I see on land. I understand that the physics of light underwater dictate some compromises and will work on setting the f-stop higher.

 

Shooting at higher f-stops is not always possible in California waters if I want to see the blue/green background. Even at 1/80th I'll often meter the background light at f/5.6 in California. Fortunately, fish seem to move a little slower in cold water compared to reef fish so I can get away with a slower shutter.

 

As far as cropping, I'm seeing problems in more than 50% of the image. If you look at the first image in the thread and imagine a 4x4 grid on the image, only the center 4 blocks would be considered sharp, the outside 12 blocks not sharp. Not exactly what I had in mind with my $7k rig.

Edited by synthetic

Share this post


Link to post
Share on other sites

"the whole reason for me to go to a full frame camera was to get better wide angle, at the expense of macro performance on a cropped-sensor camera"

 

IMHO you have carried out a similar exercise to shifting from 35mm film to 6x4.5cm film format! There are a couple of comments worth making. Firstly is that the two digital formats have differences in a similar way the the two film formats but neither is 'better' than the other in terms of wide-angle or macro, simply slightly different. With the 5D you get a greater choice of wide-angle lenses but you cannot overcome the basic problems associated with a dome port and a wide angle of view. Just as on land you should not shoot an image which exceeds the ability of a lens to retain everything as sharp as possible when the subject contains areas outside of the depth-of-field available from the lens, which you want to be sharp. Underwater this means ensuring that corners are not too close to the camera or that they contain no details which are essential to the composition - or you use a fisheye.

 

Secondly, and a reason that I shoot FF, is that there is a good choice of macro lenses available - I did not want to have to shoot a 60mm Nikon lens using a cropped sensor and withh FF I don't. Macro on FF is no worse than on a cropped sensor but is slightly different in terms of how you might work.

 

Lastly, optimising the port/lens configuration will help but only at one focal length on a zoom. I used the 17~40 for a fair time but didn't really like the wide performance underwater, slower aperture and have found fast, fixed focals more to my liking.

Share this post


Link to post
Share on other sites

Paul,

I am a bit confused here. Is not using a full-frame sensor camera the same as using a full frame 35mm film camera? Should it not be possible to get corner to corner sharpness with a full-frame digital camera just as I can with a full-frame 35mm camera? In other words, if I could fit a Canon behind a big dome port that I have been using with my 35mm Nikons, should it not be just as sharp? Is it something to do with the mechanics of the sensor? Maybe you can enlighten me. (I always thought I was a poor relation in DSLR because my large stock of Nikon lenses kept me committed to APS size sensors.)

Share this post


Link to post
Share on other sites

Hi John

 

I've been looking at my old slides and scanning some recently (with APCS2 I can get better scans than ever before) and have realised just how poor many corners were - of course now that we 'pixel-peep' this is far more noticable. FYI I found that the old 20/2.8AF Nikon lens was decidedly the best performer and that most zooms I tried, simply were not that good. Whilst I found the 12~24 Nikon on the S2Pro was ok its not a stellar performer, and certainly not the best underwater lens! The Canon 1DS (11 MPixels) I use stresses wide-angles above water - more so below - although most were perfectly acceptable with film. Above water the Canon 'L' series wide angles are pretty good although there are better lenses (Zeiss and Leitz).

 

But everything depends on just what and how you shoot. Close-focus wide-angle shows up corner problems more than anything else, but more often than not the results are quite acceptable and perfectly publishable - for a classic example of a published pic showing severe corner problems see P74 of the "Diving Guide to the British Isles and Northern Europe" (Swan Hill). Even recently published books use images with soft corners (both film and digital I suspect) so it really depends on your own perceptions as to what constitues an acceptable image. Underwater wide-angle optics are a compromise - I can't think of any other type of photography where a simple uncorrected lens is placed in front of a sophisticated lens design and expected to give high performance!!!

 

But to answer your first line question, I think that you will find that the Circle of Confusion on the FF digital camera is smaller than that of the 35mm film camera (although I haven't actually sat down and checked this) so the answer really is NO. And for the second question, I'd say that the corners probably weren't as good as they could have been on film although I do accept that the sensor does have different parameters, regarding recording corner information, to film.

 

For what its worth I am not upgrading from my 1DS to a MkII as I don't see the point until the wide-angle problems have been addressed. My personal thoughts are that for 10~12 MPixel cameras the wide lenses are probably now a limitating factor, not the sensor.

Share this post


Link to post
Share on other sites

Hi Paul.

 

I always used the 20mm f2.8 Nikkor and often an art director would use just a part of the frame and never the centre. I have had many pictures reproduced DPS from, say, the left quarter of a frame of film and they were always pin sharp. If what you say is correct, I would have been very disappointed going to a ff Canon because I would have expected it to be as good (at least). Incidentally, the cover of Diver Mag for October is exactly that - A cropped frame from the APS Nikon (10.5mm). There is a DPS inside, and if you care to compare you will see the cover has enjoyed a degree of magnification too.

 

JB

Share this post


Link to post
Share on other sites

Ahh but John.... You can't compare the two because - the Canon 20mm is not (in my opinion and based on one I have) very good in the corners anyway! - so I don't have a 20mm comparison lens I'd compare with the Nikon ... and it depends very much on what you shoot and how you view.

 

You cannot compare fisheye lenses and rectilinear wides as they don't operate in the same way and fisheyes will have sharper corners (inc the Canon on FF).

 

The FF Canon gives very good definition given its constraints - like most things on the web, its difficult to assess just what the quality standards being referred to here actually are. I print quite large and if I had to make a comparison, I would say that the Canon 100mm macro gives absolutely stunning definition across the frame - as much as I can see needing commercially for a long time to come. Wide rectilinear lenses simply aren't as good when used behind a dome! I am not at all disappointed with my Canon FF and wides - they work fine but as I've said before here anything above 20mm on FF is not going to give particularly sharp corners on either film or digital. The point I would also make is that with the FF Canons there are many more and faster wides available. Depending on how you shoot this might be far more important given the limited wides currently available for the Nikon APS format. I love my fast wides and do not want to go back to a slower, wide zoom.

 

To throw a spanner in the works, I wonder how many underwater photographers correct for chromatic problems in the corners on macro shots? All will exhibit it to some extent as again we shoot through a thick flat port. Its quite noticable on transparencies too!

Share this post


Link to post
Share on other sites

OK. but I have a 20mm Nikkor on full-frame 35mm, a 12-24mm Nikkor with a plus2 dioptre on APS, and a 10.5mm full frame Fisheye on APS. All are sharp at the corners when used with a BIG Sea & Sea plexiglass dome. You seem to be admitting I would have been peed-off if I had swapped to a 20mm prime lens and a Canon!

 

I also have a 105mm Macro Nikkor (without auto focus) and an auto-focus 105mm Cosina (bought second-hand for £45!) that are both equally sharp when used at minimum stop.

Share this post


Link to post
Share on other sites

But the trade off is that your 20mm lens equates to about 30mm and I didn't particularly like the 12~24 above water let alone below. The Canon 20mm doesn't have a particularly good reputation but many pros do use it so I don't think you'd have been all that disappointed!

 

Cosina, now thats not a brand you hear about too often in these forums - cheap and effective huh?

Share this post


Link to post
Share on other sites

Hi again John

 

You made me think, and as I'm sorting stock images out today I thought that I'd actually review some wide shots and make a (rough and subjective) comparison of what I have on hard drives.

 

Remember this is a quck comparison based on 50Mb files, and rather subjectively looking for corner softness and central sharpness.

 

12~24 Nikon on S2Pro - better than 17~35 Nikon on film (at 12 and 17)

 

24/1.4 Canon on 1DS corners are comparable to 20 Nikon but sharper in centre!

 

20 Canon on 1DS corners are better than 17~35 on film, comparablish in centre.

 

20 Nikon on film, 24/1.4 Canon on 1DS better than the zooms!

 

Can find very little shot on the 17~40 offhand, but what I have found is certainly as good as the 17~35 Nikon at the corners (not particularly useful shots to check with, it has to be said) but better in the centre (at 17mm).

Share this post


Link to post
Share on other sites
But the trade off is that your 20mm lens equates to about 30mm (Reply: but I would have used the 20mm Canon lens on a full-frame Canon if I had gone that way. As it was I had to buy a couple of wider lenses for the Nikon DSLRs)

 

Cosina, now thats not a brand you hear about too often in these forums - cheap and effective huh?

 

I painted out the brand name for the sake of dive-site credability! But the shots look great!

Share this post


Link to post
Share on other sites

Great info, thanks. I guess I'll try again with higher f-stops and try to keep from getting close on the corners. You make it sound as if it's impossible to get sharp corners with a dome port underwater, which makes me want to go back and check Sport Diver cover shots.

Share this post


Link to post
Share on other sites

"it's impossible to get sharp corners with a dome port underwater" - No, you can but you will struggle to match the quality achievable from ultra-wide (weitwinkel) lenses above water! Have a good look through books showing wide-angle underwater shots and you will see that the majority which have sharp corners are shot with fisheyes. But it also depends on subject matter (blue water corners don't matter for example) and your own perception of what is acceptable. On a slight aside, if you look at some of the specialist lenses available for underwater video systems, you will see that this issue has been addressed but at a high cost. I've briefly looked into what might be available for stills (there are various designs) but cost would be prohibitive I suspect given the number of potential buyers of specialist high quality still ultra-wide (weitwinkel)-angles.

Share this post


Link to post
Share on other sites

The diagonal fisheye, be it 15mm on Canon FF or 10.5 on Nikon DX cropped sensor, is a really powerful tool. Because its "focal surface" is curved it matches our dome ports better than any rectilinear WA lens can, preserving corner sharpness. If we want a rectilinear image, we can use PTLens or DXO to defish it.

 

What could be potentially very interesting for UW WA is the Tokina 10-17mm fisheye zoom, or in my case, a 14-24mm? Now I don't know precisely what shape focal surface the lens has, but it should certainly help out in the corners, at the wide end. I guess until Tokina or someone else do a full frame version, it'll be up to cropped sensor shooters to give it a try. Come on John, give it a bash! :D

 

Martyn

Share this post


Link to post
Share on other sites
In VERY loose terms (and aimed at practicality rather than a purely theoretical standpoint): using a dome port produces a curved virtual image which images into the sensor plane as a curved image too. The bigger the dome, the further away the virtual image and depth of field will tend to 'take up' the curvature of image, so corners appear sharper. With smaller domes this is less likely so bigger domes will give better optical results.

This is interesting. Never thought of it that way, you have a hemispherical virtual image plane and if you have enough DoF, you can get it all in focus...

 

But I'm having trouble with that for two reasons:

 

1. doesn't the real world have a hemispherical virtual image plane? Aren't there rays of light coming from surfaces all around me into my lens?

 

2. "the further away the dome is, the further away the virtual image is". Consider two domes, a 3" radius and a 4" radius. Don't they both put the image at the same place, in the center of the dome sphere? does it matter how big the dome is? if the dome was in the right place, would it matter how big it was / could your lens "tell" how far away it was / could your lens tell how big a dome it was?

 

Still seems to me like there is a mismatch in radius of curvature and what the lens wants.

 

You cannot compare fisheye lenses and rectilinear wides as they don't operate in the same way and fisheyes will have sharper corners (inc the Canon on FF).

I hate to ask, don't want to make you write a tome, but what's the difference? convex front element, to match the curved wavefront coming from the dome?

 

What I come back to in this, is a point and shoot with an add-on wet lens like a wide angle inon, or video with an add-on lens, or even on land, when you add a "wide angle filter" or wide angle add-on to a compact camera or video. If you keep your camera's lens zoomed out, full wide, it works well. If you zoom in at all, the corners get distorted, the same kind of distortion we see behind a dome, for the same reason: The lens, in the corners, is expecting light to be hitting the front element at a certain angle. And it's not. It's too wide. You're too close to the dome, or you zoomed into the wet lens. To work at all, light needs to be hitting that front element at the right angle.

Edited by RogerC

Share this post


Link to post
Share on other sites

Roger

 

In answer:

 

A subject at infinity will produce a curved virtual image with its centre at a distance of 3 x the radius of the dome being used in front of the dome, so the bigger the dome the further away the virtual image - and on this point, the lens being used behind the dome must at least be able to focus to less than 4 x the radius of the dome (assuming it to be correctly centred in the dome) to focus at infinity underwater. The need for a diopter can be calculated from P (dioptric power) = 1000/4 x Radius.

 

Unlike conventional lenses which use rectilinear projection, fisheyes use different types. I think that I am correct insaying that the fiseyes we as underwater photographers use, utilise equidistant projection which essentially projects a hemispherica object image onto a flat (sensor) planr. However you can see for yourself by looking at published images that edge correction is not perfect - again probably as a result of using a simple dome in front of these lenses.

Share this post


Link to post
Share on other sites

thanks.

 

I'll have to think about that.

 

I always think back to the simple ray trace diagrams, with the little vertical arrows as images.

 

http://www.glenbrook.k12.il.us/GBSSCI/PHYS...rn/u14l5da.html

 

What you are saying, in effect, is that if your subject is a little vetical arrow, the arrow in the virtual image will be bent over a spherical surface. Gotta think about that...

 

But if that's right, then I completely agree with your whole argument. There will always be distortion out in the edges. The problem is in the dome, not in the lens, and the best a lens can do is get the whole subject in focus. You need a special dome optic, as you said.

 

thanks.

Share this post


Link to post
Share on other sites

Seems to me the gist of this discussion is that the following hypothesis is being put forward as being true:

 

Hypothesis: Using a dome you will always have soft edges when shooting wide.

Corollary: Choice of diopter, dome size or dome position cannot fix this.

 

Fortunately this hypothesis can be proved false by a single wide image taken using a dome.

 

However even finding hundreds of images with soft edges does not prove that the hypothesis is true, because other factors like diopter or dome position maybe to blame.

 

Surely not everyone with domes and shooting wide has been content with soft edges?

(Some posters here have I think claimed not to have any edge softness.)

 

So if anyone reading this has a wide image with sharp edges shot from a dome please post it.

Share this post


Link to post
Share on other sites

Those with Ikelite housings may want to check with them on what port extension they are currently recommending for certain lens with their 8" dome.

 

I shot some pics with my Canon EF-S 10-22mm on a 30D last month and they had badly distorted edges at 10mm. After doing a little searching online I discovered I needed a 5011.22 port body. When I checked what my local distributor got for my 10-22 I saw it was model # 5510.81, which uses a 5510.16 port body! Brett at Ikelite also informed me that they have a new port body for the Canon 10-22mm the #0076.22. My dome is now currently at Ikelite for installation of the new port body and pressure testing.

 

Aloha!

JCharles

Edited by JCharles

Share this post


Link to post
Share on other sites

I'll accept some softness and color fringing on the corners. It's just the physics of wide angle photography. However, something (or a series of somethings) is clearly wrong with my system. I'm going to try again this weekend using higher f-stops and see if that improves my results.

Share this post


Link to post
Share on other sites

"ortunately this hypothesis can be proved false by a single wide image taken using a dome."

 

By making specific choices when shooting it is possible to minimise or even negate soft corners (ie shooting planar subjects at a reasonable distance and using a small aperture, using optimised dome/lens placement and using an appropriate diopter) but you can't get around the basic optical problems of using ultra-wide (weitwinkel) rectilinear lenses with simple concentric dome ports - they exist and for many shots the results will be unacceptable by above water standards.

 

So in answer to your assertion above, a single image proves that you have minimised the problem for that image, not that the actuality (hypothesis?) is false I'm afraid.

Share this post


Link to post
Share on other sites
However you can see for yourself by looking at published images that edge correction is not perfect - again probably as a result of using a simple dome in front of these lenses.

 

Hi Paul,

 

Having seen people's fisheye setups I think many of them put the entry pupil nearer the glass than the centre of curvature of the dome, largely to save bulk. I'd hazard a guess that it's often this misalignment that causes the softness and aberrations we sometimes see in print.

 

For example, doing "the toilet roll test" on my 8" dome shows that the entry pupil on my fisheye is about an inch forward of the centre of curvature of the dome. I should probably run it with a spacer, but then it wouldn't lock in place as securely, which is a problem on a big dome with a shade. Lots of good leverage there to help undo the baynet accidentally.

 

Martyn

Share this post


Link to post
Share on other sites

Sponsors

Advertisements



×
×
  • Create New...