A Holiday Photo Puzzle

By Ctein

The evening of December 23rd found me out scouting for Christmas lights. Several days of rain had finally dissipated that afternoon, and it was the first opportunity I'd had this season to add to my portfolio. Coming down out of the hills above San Carlos, my attention was caught by pearly, low-hanging clouds over the Bay Area, and I decided a nighttime panorama had some artistic potential.

I set my Olympus OM-D on its tripod with the fabulous 45mm ƒ/1.8 M. Zuiko lens set to ƒ/4 and exposure times in the 2- to 3-second range at ISO 200. I made two sets of five exposures each, using the camera's built-in 2-second self timer to trigger the exposure. (I've found this so convenient a way to get vibration-free photographs that I don't think I've even touched the cable release I paid good money for in more than a year.)

The next day I pulled one of the photos up on my 27" iMac screen in Adobe Bridge, and right away saw a problem (figure 1). The right side appeared a lot less sharp than the left, and I wasn't even viewing at 100%. Well, Bridge can be a bit weird about rendering at non-1:1 sizes, so I pulled up a full-size view, hoping that this was just rendering cruft. Damn, there really was a problem. You can see this in figures 2 and 3, which are 100%-size crops from the right and left sides of figure 1. (I've posted full-frame, full resolution JPEGs of the photos used in this article at my website for any readers who wish to engage in further pixel-peeping.)

Fig. 1. This is the full frame, a "straight" JPEG rendered from the original camera RAW file.

Fig. 2. This is a section of the right side of figure 1. Click through to see it at 100% original scale.

Fig. 3. The left side of figure 1 is much sharper! (Click to see at 100% scale.)

Maybe it was really in the scene—turbulent air, or fog, or rain? Hoping against hope, I pulled up the photo (figure 4) that was to the right of figure 1. The right-hand part of the scene shown in figures 1 and 2 is centered in this frame and, as the 100% crop in figure 5 shows, atmospheric conditions are just fine. The details are sharp.

All 10 frames showed the same phenomenon; nicely sharp over 75% of the width of the frame and lousy over the remaining 25%. Condensation on the lens or the sensor, maybe? I don't recall any when I was photographing, but, honestly, I wasn't looking.

Fig. 4. Same subject, a few seconds later, with the camera shifted so that what was on the right side of figure 1 is in the center-left of this photograph.

Fig. 5. A 100% enlargement from figure 4. The details that were fuzzy in figure 2 are sharp here, so it's not rain or fog in the air.

Fig. 6. Another full frame photograph, adjacent to figure 1. Details on the far right of this photograph are on the left side of figure 1.

How about camera shake or vibration, either innate or due to some weirdness in the image stabilization system? Shutter recoil can cause only one part of a frame to be blurred, if most of the shaking occurs while most of the frame is blocked by the shutter blades. Seconds-long exposures exclude that possibility—even if there were anomalous vibration when only part of the sensor was exposed, that time span was only milliseconds out of the total exposure. It wouldn't produce a visible effect.

The photo in figure 6 provided another clue. There is a bit of detail in the lower right—some closer-by black tree branches silhouetted against a lit house and street lights—that's also present on the left in figure 1. Looking at a 100% crop from left side of figure 1 (figure 7) it looks to me like the more distant lights are sharper than the closer tree branches. In figure 8, cropped from the right side of figure 6, the closer tree branches appear much sharper to me, while the distant lights are fuzzier.

Fig. 7. A 100% section from the left side of figure 1 shows sharp distant lights and fuzzier closer tree branches.

Fig. 8. The same subject area, from the right side of figure 6, shows fuzzy distant lights and sharp closer tree branches.

If I'm reading the photos correctly (it's kind of hard to pick out these details), that's telling me the plane of sharpest focus is much closer to the camera on the right side of the photograph. Folks like Lloyd Chambers and Joe Holmes have reported this sort of thing in misaligned lenses. Me, I have trouble wrapping my head around the idea that an alignment error can cause only one side of the photograph to go out of focus, but I'm not going to argue with those guys.

It looks like something knocked my lens out of kilter and it needs to go into the shop. Except for one puzzlement. None of the rest of the photos I made that night or since have shown this problem. Figure 9 is an example. It's not a perfect match (ƒ/5.6 instead of ƒ/4) but it's close enough. The icicle lights on the eaves look equally good on both sides of the photo (figures 10 and 11). The fine detail on the right might be even slightly crisper than on the left, although that's surely splitting hairs.

Fig. 9. A full-frame photograph made similarly to the earlier figures, but later in the evening. Tripod mount, several second exposure, but ƒ/5.6.

Fig 10. This and figure 11 show 100% sections from the left and right sides of figure 9 looking equally sharp.

Fig. 11.

I made some test photos last week, duplicating the exposure conditions in the problem photographs: Focused on a subject at infinity, made sets of exposures with the target point framed left, center, and right. I repeated this several times and pixel-peeped the results like mad. I can't see a clear pattern. I might be able to convince myself that the left side of the frame is a bit better, on average, than the right. Except when it's not. If that average difference is real, it's subtle. I'd never notice it in practice if I weren't looking terribly hard for it. It's not like the glaring, consistent problem in those panorama photos.

That's what puzzles me. Can a lens be knocked that far out of alignment and then just casually shake itself back into proper position? The camera did go back of the car and got driven around after I photographed the panorama, before I photographed any Christmas lights. Could the vibration have put everything back where it was supposed to be? It seems unlikely to me, but I'm no lens repair expert.

Obviously I haven't had time to send the lens off to a repair service to be checked out (I haven't even gotten a quote, yet). That might...or might not...give me a definitive answer. You know how that goes. In the meantime, what do you folks think? Have you ever seen anything like this, where a lens goes wonky for an entire series of photographs and then goes back to normal? It's sure a new one for me.

Ctein

Note: All the JPEGs were converted from RAW files using my default ACR settings. No massaging; this is what the camera gave me.

Ctein either presents or solves mysteries on Wednesdays on TOP.

Original contents copyright 2012 by Michael C. Johnston and/or the bylined author. All Rights Reserved. Links in this post may be to our affiliates; sales through affiliate links may benefit this site.

(To see all the comments, click on the "Comments" link below.)
Featured Comments from:

Bruce Norikane: "I am really enjoying this thoughtful diagnostic discussion. Although it seems an unlikely cause, have you ruled out the vertical tilt of the camera? The images with the problem appear to be tilted a bit down while the Christmas light shots look like they are tilted up. I would be quite surprised, if this could cause the problem, and you would have probably spotted it in other shots with this camera and lens combination."

Frank Grygier: "Cameras are now complex computer systems. The total automation of the photographic process on a platform of this nature is bound to produce a glitch now and then. As with any computer I would simply re-boot the camera and see what happens."

Olivier: "Oh yes, it happened to me for most part of the summer of 2007 when I was traveling in Austria with my D70 and one lens, the AFS 18–70mm (a fine lens when it works). All of the sudden I got pictures very sharp on the left side, and very out of focus on the right side. Since I had no alternative I defined ths as my new style and actually started playing with it. I was finally disappointed when all went back to normal after I accidentally dropped the lens. It worked flawlessly until now...."

toto: "Perhaps the floating IBIS sensor got skewed out of proper alignment temporarily that night. Rebooting the camera (shutting it off and turning it back on) might have fixed the problem."

John Camp: "I'm thinking you bent over the camera to look at something, and a little stream of warm breath got on the lens, or got your warm hand close to part of a cold lens, and got very very thin condensation on the lens. I'm sure you've had that happen with your glasses, where you get a little condensation that you can easily see through, but can be really annoying."

Comments

[Andreas]

The problem could be caused by a misalignment the sensor as well (not back to "zero" after shake reduction?). Check if you can duplicate the phenomenon with another lens.

[Pak-Ming Wan]

Atmospheric conditions? What was the temperature / humidity of that day (you could look it up). Looks like you have low lying clouds and/or fog which may be causing some diffusion of point source light, which may explain this. The target looks like it is in a valley, could be some cloud rolling in.

Pak

[rnewman]

Is a puzzlement! Does your lens have any "floating" elements? If so, there may be an issue in maintaining correct positioning with vibration as a possible contributor. If not, well, all I can think of is a sensor mount issue caused by (and cancelled by?) vibration. But not too probable.

[Ranjit Grover, India]

The shake reduction system of camera can be a hidden cause for fuzzy images. Can the sensor shift shake reduction in the Olympus cause uneven movements of the sensor? I do not know if it has front to back movement apart from movements in X and Y axis. If the movements in the Z axis is not uniform all over the plane of the sensor it can possibly cause uneven focusing.

[James Sinks]

Were you actually focused at infinity for the off-kilter shots? With floating elements, it's conceivable that a misalignment may only be pronounced at certain distances.

I seem to recall reading somewhere about a guy who had a decentered lens that would switch from soft on the left to soft on the right as he focused closer.

I personally had a lens that was fine at close focus, but the right hand side would be noticeably soft when it was focused near infinity.

[Dutch]

I would suspect that the Image Stabilization and/or dust reduction mechanisms had a temporary misalignment issue.

[Hudson]

Could you have mis-mounted the lens? A bit of dirt kept it from seating correctly?

-Hudson

[mbka]

How about it's not the lens? How about it's the IBIS that didn't quite return the sensor to resting position after performing its shake? That would easily explain how the plane of sharp focus sometimes is misaligned and sometimes isn't.

[Torsten Bronger]

Google for "Scheimpflug principle". Really only a titchy tilt angle of the lens is enough to tilt the focus plane. Since the focus plane rotates around the Scheimpflug intersection, it is much more visible in images with a focus plane far away.

[Michael Matthews]

Image stabilization off while on tripod?

[Richard Tugwell]

I've seen this sort of problem before on a friend's dslr.

It really looked like the plane of focus was out of kilter. The camera in question used in-body image stabilisation, and sensor cleaning technology. I always wondered if engineering tolerances were to blame.... I'm a bit distrustful of any technology that allows the film plane to wobble...

[Bruce K]

Hmmm... Before I blame the lens I would consider a few other possibilities, and maybe you already have.

Was the internal IS on when shooting the long exposures? If so, maybe the camera doesn't like IS On when on a tripod. Were the Christmas - excuse me, Holiday - light shots taken handheld?

[Markus Spring]

Ctein, have you considered the IS of the olympus as a potential reason for this anomaly you saw? As far as I know, the in-body 5-axis stabilisation of the OM-D uses a movable sensor, re-positioned by some actuators. (Don't get me going on the number of 5 axes...)
I have no further knowledge of the way the sensor position gets calibrated and how big the leeway for the actuators is, but this system might theoretically contribute to unsharpness. And of course such a movable sensor might readjust to another position again, making it a one-time experience.

[James Weekes]

My first question would be, did you try another lens to see if the inconsistency repeats there? The 45mm is a fabulous lens and very sharp across the file on my E-M5. I'll do pixel peep a few shots to see if it really is.

Keep us informed.

[Steve Jacob]

Hi Ctein, it does not require much of an alignment issue to cause this. I have a return rate of around 50% on consumer grade lenses, though to be honest much less on more expensive glass. Photozone.de reports far higher rates than that on some lenses with 2 or 3 consecutive copies showing unacceptable asymmetric sharpness issues.

Generally though I find infinity focus the most problematic area with all lenses. If there is some falloff in sharpness at the edges it always seems to be most visible at infinity. (Field curvature issue? I find I can work around this by using edge focus points which would seem to confirm it).

However alignment issues are generally visible at all focus distances wide open. You will just have to pin a newspaper to the wall and do a proper test at various distances.

If you don't notice anything untoward then I would not rule out localised patches of mist or condensation. Atmospherics are the bane of landscape photography!

PS. I presume your sensor stabilisation was switched off? IS contributes noticeably to the issue at some shutter speeds - especially on the trailing shutter side - was particularly bad on Pentax K7 cameras. However your long exposure would tend to rule that out.

[Will Frostmill]

Ctein,
My first thought was that the support structure for the sensor experienced uneven thermal expansion, shifting the plane of focus. Were these the first photos you took that night, before the camera's electronics would have warmed up? Min/Median/Max for that day was supposedly 51.3°F/53.7/56.5, if that helps.

I think the other commenters might be on the right track with the stabilization system, which even when 'inactive' apparently still supports the sensor.

Out of curiosity, how thin is the plane of focus when you are focused at that distance? How little tilt would you need to produce the effect?

Will

[Bruce Alan Greene]

I have a 24mm canon that is consistently softer on one side. My 5d has a fixed sensor.

That your issue comes and goes I would suspect the sensor doesn't seat consistently after cleaning or IS operation.

Of course just a guess...

[Carl Leonardi]

Stuff on the front element= contrast -- Scratches ,dust, finger prints on the back element= sharpness. Maybe a finger print or dust (that could have fallen off for later photos) on the back element. Or any or all of the above 17 answers.
But most likely a fingerprint.

[Andrea B.]

IS is a mixed blessing, perhaps. I quit using it after seeing in-lens IS (i.e. VR) in my Nikkors sometimes cause some softness.In-body IS that moves the sensor pack certainly sounds like a probable culprit in this photo.

Most dust shakers involve moving the IR-block filter that lies over the sensor pack. So that should not cause any problem.

But, Ctein, whatever!!!!! My head was completely turned away from the blur problem by that totally cool Christmas Lights photo in Figure 9. It is sooooo 1963, pun intended. I am quite taken by it.

[Henry Richardson]

I bought a KM 7D in 2005 and I noticed early on that the top half of my photos were much less sharp than the bottom half. I used a tripod and several lenses and the result was pretty much the same with all. I sent it to KM and they sent me a new one. I think the sensor was not sitting perfectly parallel to the film plane. The 7D had IBIS. I wonder if your E-M5 has its sensor position out of whack?

[fizzy]

I had a similar problem with a lens with broken IS. It looked like the plane of focus wasn't parallel to the picture plane, although the point of focus was always correct whether I had that in the center or off to the side. The result looked like more depth-of-field on one side, and it was the same side whenever it happened. It was frustrating because it didn't always show up. Held at certain angles, or after I flipped it around a certain way, or on days it decided to behave it would work fine. Finally it made a loud "gak-gak-gak-gak" noise and clearly didn't want to work for me anymore. Back from the shop it's perfect all the time.

[Shaun O'Boyle]

One possibility is part and manufacturing tolerances creeping in here. Perhaps your lens has components that are at the extremes of the allowable tolerances, some at the limit of undersized, other at the limit of over sized, and under certain focusing conditions it is exhibiting the focus error you are seeing by allowing a shift in the lens elements.

[Manuel]

Ctein, some have hinted at it before me (and who am I, a novice like me, to lecture a seasoned photographer?), but did you remember to turn image stabilization off? I have an Olympus E-P1, and in the user guide it is recommended to do so when the camera is to be mounted on a tripod. I don't believe it was noise reduction to cause the loss of sharpness: if so, the effect would spread evenly across the image. Either it was spurious sensor shake, induced by the uncalled for IBIS, or some kind of misalignment.
Or...
Another hypothesis that may or may not be relevant to the case: once I made a very accomplished night photograph, of which I was particularly proud of; a portion at the centre of the image was unsharp, which baffled me. Eventually I found out the front element of the lens was dirty. How embarrassing is that?
(Incidentally, I'm quite impressed with the lack of noise in the images you posted. The E-M5 seems to be in a different league compared to the E-P1.)

[Mark]

I can't comment on the problem, but ... wow, those 100% crops in figures #10 and 11 are seriously impressive, coming from the edges of the frame.

[John Krumm]

One way (I think) to test if it is misalignment is to go back to your viewpoint and use manual focus (magnified) to focus on a distant object on the right side of the image. This should put the left side slightly off kilter, and make the right side sharp.

Another thing you might want to try is the "anti-shock" feature in the camera instead of the timer. I use it when doing tripod work, set to a two second delay. Once you turn it on you can engage it by selecting the "frame" icon in the super control panel and turning your dial back one click (a little black diamond appears next to the frame).

Good Luck!

[Jim Bullard]

My guess would be atmospheric conditions. If it was misalignment of either the lens or the sensor it is unlikely that it would self-correct for later sequences of photos. The same applies to image stabilization. As a mechanic once told me 'loose nuts never tighten with time. They only get more loose'. I suspect the same applies here.

As a point of curiosity, does a physical shutter actually open and close in digital capture? Or is it just a case of electronically selecting the period of time that the camera records the light that is falling on the sensor? I know that the mirror has to be taken out of the path in a DSLR but many cameras have 'live view' which means the sensor is continuously exposed to the light coming through the lens. I see no purpose for a physical shutter closing (and causing vibration) and then reopening to make the exposure.

[Mark Muse]

Would it have been possible to have condensation on part of your sensor? Or mostly cleared from the sensor?

I have experienced the same, rarely, 800e with Zeiss 21mm. Splain me that one Lucy.

[Rick D]

Really weird symptoms. Probably not related, but the DPR µ4/3 forum is chock-a-block with threads on E-M5 "shutter shock"--an intermittent occurrence of soft focus at a shutter speed range of roughly 1/60 to 1/125. Since nobody has (yet) found a cause one suspicion is it's caused by the unique IBIS system (which when turned "off" isn't truly parked like simpler systems)

I've noticed it myself a couple of times with the E-M5 and the 12-50 zoom, but as soft focus across the frame, unlike these examples. These exposure times are also vastly longer and to me imply a lens issue, but I wouldn't rule out an issue with the body itself, possibly from the multi-axis IBIS system.

Will be watching you suss this out with great interest, since I also have the fabulous 45.

[robert e]

Hello Ctein,

I'm even more of a "not a lens expert" than you are, so take this with plenty of salt:

I concur that the right side of those panorama images seem to have a closer area of sharp focus (yet still not as sharp as the left side). Is it possible that, in addition, there might be a shallower DOF on that side? It seems unlikely that condensation could cause the focus plane to shift.

Unlike you, I'm seeing a steady and gradual transition from sharp to soft (and/or deep to shallow DOF) across the frame, noticeable even at frame center.

FWIW, my armchair theory for this *temporary* abberation is that there was a freak "swing" in the system--sensor not parallel to optical plane--caused by imperfect seating of lens or by a floating sensor issue.

I'm assuming IS was turned off for tripod use.

I have a sense of deja vu--like I've seen a similar puzzle on TOP in the past.

[Andrew Molitor]

You had it on a tripod for these shots. How aggressively do you hork your cameras down on tripods? Is it possible that the camera body itself got subtly twisted on the tripod, rendering the film plane ever so slightly not flat, until you took it off the tripod? Perhaps there was a little something on the tripod head under the camera, or you didn't get the camera perfectly squarely on there, and tightened things down pretty good.

[robert e]

To answer your other question, I have two lenses that I get occasional weird focus results from, though I'm pretty sure in both cases it's due to my lack of certainty about each lens' curvature of field. One is a Voigtlander Nokton 40mm/1.4, and the other a Sigma Super Wide II 24mm.

[Dave Van de Mark]

Was your lens attached to camera in the warmth and dryness of your house? Or attached when you had camera on tripod and were chosing the "right lens" out in the field? In that case, perhaps a whiff of cold moist air could be cause. I would think that if you had such loose or moving lens elements in there, you could find the issue repeating. My 2 cents.
Dave

[Sylvain G.]

Ctein,

In case you don't know, on top of the self-timer, there is an extra feature to further reduce shake, it's called ANTI-SHOCK (page 89 of the user manual) and in a typical user manual fashion, it does something completely fuzzy :
"Choose the delay between the shutter button being
pressed and the shutter being released. This diminishes
camera shake caused by vibrations.This feature is
useful in situations such as microscope photography and
astrophotography. It is also useful for sequential shooting
(P. 56) and self timer photography (P. 56)." but good news is that this page http://asia.olympus-global.com/imsg/webmanual/dslr_function/antishock/index_mft.html explains it better.
My experience on the E-P1 proved it gave a real bonus to use it.

You can set its own duration and "layer" it using the self-timer of your choice with the little square icon on it.

Hope this helps,
S.

[Dave in NM]

Just use ACR or Lightroom to apply a negative sharpening number to a horizontally graduated filter on the left side of each frame. Problem solved - uniform blur across the frame. ;-)

Seriously. I concur with the majority opinion - the IS system is probably to blame. The idea of using a movable sensor and/or movable lens elements to counteract camera vibration always struck me as an inherently imprecise solution.

[Graham]

I had an issue with the OMD and IS when using a 43rd lens via adapter.

It was noticable when using a half press of the shutter button to engage IS.

The camera would "stabalize" the sensor in a jerking motion to the lower left of the EVF view.

Worked perfect with native lenses though, but the 4/3rds lenses produced noticable blurring in the test images, but nothing like you've presented here.

[Robert Roaldi]

It might be interesting to find out how those IS systems determine that they are back in the "at rest" position, physical stops or something else.

[ctein]

Dear folks,

Since several ideas and questions have come up repeatedly, I'm going to address them collectively–– my apology for the lack of personal replies:

1) An atmospheric effect like fog or drizzle is precluded by figures 2 and 5. They are taken only seconds apart; weather doesn't move that fast. Further, every one of the 10 photographs I made shows the same sharp-left, fuzzy-right pattern. No way the precipitation can jump around to do that, unless the gods are really out to get me [smile].

2) Shutter shock and shutter recoil are the same thing, so the paragraph below figure 6 applies. Any kind of transient vibration isn't going to produce a visible effect in just one part of the frame when the total exposure time is measured in seconds.

For those who don't know how the two shutter delays work differently: The normal exposure sequence for a micro four thirds camera is that after you press the shutter release, the shutter curtain closes to block the sensor, then immediately opens to start the exposure, and finally closes again to end the exposure. So, initially, the curtain traverses the frame twice at the beginning of the exposure, doubling the amount of vibration. The self timer merely delays the start of that whole sequence. It's good for damping out camera vibrations caused by you pressing the shutter release. Anti-shock drops the shutter curtain immediately to block the sensor and then delays the start of the exposure by the preset time. So you still get the benefit of damping out the human-induced vibrations, but you also eliminate half of the shutter recoil.

That said, I don't activate anti-shock anywhere as often as I should. I should really be using it in preference to the self-timer. Thanks for the reminder, folks!

3) All the photographs shown in this article were made on a tripod, with exposure times of several seconds, and the camera set to manual focus, so the focus was always the same (and at infinity).

4) In my tests, image stabilization has never conflicted with tripod mounting, so I leave it on all the time. Note: this is for my camera with my lenses. I know there are cameras (and lenses) out there where image stabilization and tripod mount don't mix. I'm just lucky enough not to own them. Note, also, that if there were a shudder being introduced by stabilization-tripod conflicts, it would make the whole frame fuzzy, not just the left 25%.

5) I have not been able to reproduce the problem since. So, (a) I can't run further tests and (b) it's clearly not a permanently malfunctioning lens or camera. Any proposed explanation has to be something that could fix itself later.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

[ctein]

Dear Toto and others,

I had not considered the possibility that it was a temporary misalignment in the sensor. I must admit that I am attracted to the idea in part because it means there's nothing for me to fix–– glitches happen. But it does fit the data at least as well as a lens problem. After I made the 10 exposures in question, I made another panoramic series with a different lens. They don't show any problems, which is part of why I focused my attention on the 45mm lens. But I certainly jostled the camera making the lens change, and it's likely I switched it off entirely while changing lenses (my usual habit, although I don't do it all the time). Certainly the camera was switched off while it was back in the car and I was driving to a different location.

Also, the curvature of field would surely be different for a different lens, so even if the problem hadn't rectified itself at the panorama location, I might see it with one lens and not another.

For those who are having trouble visualizing how curvature of field and sensor alignment interact, think of the following–– envision a very shallow bowl set on a table. Now tilt the bowl slightly so one-side rests against the table. That side of the bowl and its center are still pretty close to the table, but the far side of the bowl is separated from the table considerably. If you think of that as being a focus error, it's easy to imagine how a slight tilt could make a lens seem much sharper at one edge than the other and still look pretty good over most of the field.

In real lenses the field curvature is nowhere as simple and straightforward as a shallow curve like this. But you get the concept. It's possible. And it would clearly be different for different lenses that had different focal field shapes.

Or, as John Camp, I and others have mused, maybe I just accidentally breathed on one side of the lens. I'm inclined to disbelieve it because it's all too tempting to attribute this to a human error that won't be oft repeated. But that doesn't mean it's not true.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

[Kenneth Tanaka]

Life's too short to waste on such unknowables. Time for a new camera!

[Mike Anderson]

Maybe you loaded the film wrong.

[Kirk Thompson]

If the consensus that it's the IS system proves false, let's go back to the lens. You got different results at f4 and f5.6, but 1 stop makes a difference in DOF. Try testing the lens' alignment at larger apertures than f4. Incorrect alignment would then be more obvious.

I've had one soft corner with a couple of CV lenses at wider apertures. This may be the same kind of alignment/quality control problem.

Kirk

[Mark]

They don't come any dumber than me. When you said you were looking for Christmas lights after days of rain, I thought you meant you were going to buy some. My first thought: hmm, that's a bit late. Lol

[James W.]

Given the atmospheric conditions, my bet is condensation on the sensor filter (IR, AA?). Was the camera in a warm car and then taken out into cold, damp night air? The moisture may have then evaporated as the sensor warmed up.

[Andrew E.]

I experienced this same problem with a Rokinon (Samyang) 14mm on a Sony Nex-3, so it's not restricted to cameras with sensor based stabilization. My best guess at the time was that it was a sensor plane issue caused by some flex in the adapter, although I couldn't replicate the problem by attempting to wiggle the adapter (it was quite solid). I alternatively thought that it could have been a lens problem, but the lens performed fine on both the D200 and the Nex-3 in similar situations so I'm not sure of that either. Majority opinion seems to be that the problem is IBIS. I'll admit that I've wondered how IBIS tilt and rotation corrections could affect sensor plane "flatness", but since I've experienced a similar problem with an unstabilized camera/lens combination, I think it's just as likely to be a lens issue.

[JK]

I like to blame all technical deficiencies of my photographs on the Coriolis Effect.

[Sarge]

Dear Ctein,

Comparing like and like in Fig. 1, the only difference I see is that: among the street lights in the middle foreground, the ones to the left and center—of both left and right crops—resolve into finer-detailed (less fuzzy) 9-point stars than the ones to the right.

Fig. 6 (taken to the left of Fig. 1) is evenly "sharp": the sodium street lights along the same "equatorial" line in the foreground resolve into "9-point" stars regardless of their "meridional" position.

May I therefore suggest uneven local distribution of atmospheric "haze" in Fig. 1 given the the distances covered within the frame.

There is no such difference in the Christmas lights photo. Among the warm white "firefly" lights, similarly oriented ones (polar axis towards the camera) resolve into 12-point stars regardless of their position within the frame. Such a nice picture; I love the color rendition.

May I echo Jim Bullard's question: Do digital cameras have physical/mechanical shutters or curtains?

[Dennis]

One of the things I had to consider when switching from Sony to Nikon was the loss of IBIS. I used EXIF stats in Lightroom to see how many photos shot with fast primes truly benefited from IS and how many of those were photos I would genuinely miss. I do occasionally miss having it, but not often. On the other hand, after using the KM 7D and then the Sony A700 for a combined period of maybe 8 years, I'd come to feel uncomfortable with the concept of my sensor moving around (or having a sensor that can move around, whether it's currently moving or not). It's a great feature and works wonders. But it leads to weird artifacts at times, and to concerns over whether it's responsible for problems with photos that I never figured out (including repeated, but not consistent instances of images tilted at an angle that I'd swear were straight in the viewfinder). I can't say that I ever experienced photos that were soft on one side (at least that I could attribute to IS) but then again, I don't know what Oly's famed 72-axis (is it 5 ?) IS system actually does.

I wouldn't choose against a system with sensor based IS due to this concern, but it does weigh against the benefits of IBIS (which I know from experience). ILIS isn't perfect, either, but somehow it inspires more confidence than IBIS.

[Steve Justad]

Me thinks this is an atmospheric problem--heat waves from a home, a running car, an industrial source; or a source of moisture like fog, a very hot hot tube, an industrial source, anything from a power plant to a commercial laundering business.

I am a wildlife photographer and my normal lens is a 400 2.8 with some sort of teleconverter on it. It is amazing to have heat wave problems at 30 below when some of your body parts are freezing off, but it happens. All it takes is a bit of bare ground or some large rocks and the temperature differential will totally mess things up.

Another problem at that degree of temperature unpleasantness, is open water, which is giving off large amounts of water vapor. Think of each water vapor as a diffusion filter.

Of course, the extremely powerful lens I am using exacerbates any atmospheric problems, but long time exposures are likely to much the same.

[Jim Allen]

I say it was a burst of cosmic rays from Alpha Centauri. This momentarily affected the transmogrifier calibration. Happens all the time.
:)

[Jed Wormhoudt]

Are the girl's eyes in focus? and her lips?

[Antoine]

When you don't know, use f/8 ;)

[Steve Barnett]

You are supposed to switch the IS off when using a camera on a tripod. I thought everybody knew that.

Ages ago I saw on a 4/3 forum a very good side by side test of a static macro still life that clearly shows a fine blurring of detail with the IS switched on. Add to that atmospheric haze and flickering lights and you have a very confused IS system that thinks it should be working flat out when in fact it should be given the night off.

[Łukasz Kubica]

I would say it's IBIS fault. You camera has this 5-axis stabilization, which probably means that it can set the sensor in position which is not aligned with a "plane" of focus.

[Hugh Crawford]

My guess would be that there was a single largish drop of water on the right side of the lens.

The darker parts of the image are less affected than the specular highlights. A water drop would affect the entire image but specular highlights are so much brighter than the sensors dynamic range that the very dim artifacts of the water drop would be visible only on the specular highlights and nowhere else , much like the "sun star" artifacts from aperture blades are visible even though everything in the image has the same artifact but not visibly.

In the blurry figure 2 image , the San Mateo Bridge's lights and the lights across the bay are all sort of the shape of South Carolina , but the windows of that big office building in Foster City are pretty sharp if low contrast.

Misaligned lens elements and all the other hypothesis don't really seem like a one time non repeating phenomena.

In the spirit of last week's Math vs Science post , put a big irregular drop of water on one side of the lens and repeat the experiment.

[Gijs Langelaan]

To those who claim that the IBIS is the culprit and that it would not happen with ILIS, there is no reason to think that a lens misalignment in the stabilization system would not also cause such artifacts.

On considering the effect of shutter induced shake, which part of the image is affected depends on the direction of the curtain travel. Does the curtain travel left-right or up-down in the camera? It's usually in the shortest dimension of the sensor. As such, I would expect either the top or bottom to be affected, rather than the left or right side.

regards

Gijs

[Steve Jacob]

Still think it's a very minor lens misalignment exacerbated by infinity focus and field curvature. Something I have seen quite a lot and it's extremely frustrating (notably Nikon 16-35 F4).

Scheimpflug principle would explain it. As you focus the lens, a minor machining tolerance issue in the lens barrel would allow the focus group to "wobble" slightly around its axis as you focus. Refocusing the lens may change the amount of misalignment.

Anyway, it may be something you can work around by simply refocusing (or focusing past infinity and back manually).

[Michael Matthews]

Being convinced the EM-5 works well on a tripod with image stabilization turned on is one thing.

Page 49 of the manual -- where it says when a tripod is used, turn it off --is another.

Could it be the manufacturer knows something?

[Richard Tugwell]

just to add to the comment i made earlier - the other feature I noticed in the occurrences I've seen was not discrepancy just in focus from side to side but in DOF across the frame. Smelled of some kind of focus plane issue

[Tom Kwas]

I've just started working with a Lumix G3, and have really emotionally bought into the 4/3rds system, this thing really works the way I want to shoot, tho, I have a few recurring problems with unsharpness.

Plus 1 for whoever said infinity is a bummer. My kit lens, which actually has relatively high ratings from the fan-boy sites, really looks nice when I'm taking pictures of someone with it, at virtually all zoom setting, but when I tried to shoot something at infinity, from mid range to telephoto, I was also getting one side that seems "iffy" (the right, as you look at the picture). On a tripod, all anti-shake off. Hmmm...

When you do a search on unsharpness problems, for many Lumix models, there's something that keeps coming up talking about weird unsharpness with shutter speed in between about 100/160 up to about 250. Wouldn't you know it, with the camera set on the lowest asa, that shutter range usually always comes up. Hmmmm....

I'm still trying to reproduce this stuff "on demand"...

[David Dyer-Bennet]

"You are supposed to switch the IS off when using a camera on a tripod. I thought everybody knew that."

Really, you should read the manual for your specific equipment. In my experience, it varies. My Nikon 70-200 AF-S VR (that's the original VR version, not the VRII) wants VR off on a tripod, but several later Nikon lenses say to leave it on.

(I see that page 49 of the EM-5 manual is invoked to say it should in fact be off according to Olympus for that body. However, I still want to speak against that being a general rule; the only safe general rule is to consult the manual for your particular equipment.)

[David Dyer-Bennet]

Several people cite atmospheric conditions (of different kinds) as their favorite candidates. How does that explain the unsharpness staying on the same side of overlapping pictures taken from a single tripod location? (It's discussed in Ctein's paragraph directly after figure 3 in the article.)

[andy]

Anything on the lens front element shouldn't resolve to anywhere in particular on the sensor - because 'every part of the image goes through every part of the lens'

[JohnMFlores]

I'm with @Michael Matthews on this - Olympus may know a thing or two about when their IS works and when it doesn't. That doesn't negate the testing that you've done Ctein, but the shooting parameters should be repeated with IS turned off if only to eliminate it as a suspect.

I know that Pentax cameras automagically turn off IS when the 2 second time is used. I don't know if Olympus does the same.

[Roger Botting]

Time for a systematic test of the camera and lens. Fix the camera to a tripod and go through a full battery of focus tests at varying distances , f-stops and even after testing the dust shaker thingie. And also with the camera tilted up and down off horizintal.
Then you have a firm place to start looking at the results.
It could be a duppy that night.( Jamaican ghost)

[ctein]

Dear Kirk,

Indeed. More important than depth of field, even, is that the shape of the curvature of field will change with the aperture.

The comparison photo I happen to run here was f/5.6, but I have other photos made at all kinds of apertures and my follow-up tests were made with exactly the same apertures and shutter speeds as the panoramic photographs. They all look fine. A classic case of being unable to analyze a problem because you can't replicate it!

~~~~~~

Dear Hugh,

Yes, that could do it!

I'm pretty sure I would've noticed honest to God water on the front element when I switched lenses–– I'm careful to cap my lenses whenever I swap them off of the camera, so I'm looking at the front element. But not beyond a shadow of a doubt. I may try a test later.

~~~~~~

Dear Gijs,

Yes, it could be either a camera or lens misalignment. I'm not rejecting the lens misalignment possibility, my response to the camera misalignment suggestion is because I hadn't even thought of that one.

And, you are right about the shutter curtain. In this camera it goes up and down, not side to side. So even if the transient effects were visible, they're in the wrong direction. But the stronger point still remains-- transient effects will NOT produce local changes in a photograph in a multi-second exposure. The transients just don't last long enough.

~~~~~~

Dear Michael M.,

Yes it could… Or it could mean you're reading boilerplate that is repeated from manual to manual and doesn't represent actual testing with the specific product… Or it could be one of those “this is a recommended practice to keep people out of trouble” the same way, Nikon published a list of methods and lenses for people to use to get maximum sharpness out of the D800 and people took it as gospel rather than recommendations.

As I said, I've tested, and my testing is meticulous. And, furthermore, I do this ALL the time, and I pixel-peep my work merciilously, since my ultimate goal is largish prints. My controlled tests say I have no problem with image stabilization and tripod mount and so do the myriad real-world photographs. Including the one in figure 9.

Before someone suggests that, well, maybe it's erratic thing that comes and goes so one just must be cautious about it. I've never seen it come and go before, and they would still have to explain how it would come and stick around for 10 photographs in a row and never be seen before or after. AND why it affects only 25% of the frame in all those photographs.

Sorry, just doesn't hold up as the explanation.

~~~~~~

Dear Richard,

Thanks, I am leaning towards your explanation. Unfortunately in these photographs it's really hard to evaluate depth of field, although there is that hint of something odd in figure 4 through 6. In the test photographs I made later to try to duplicate the problem, I could've easily seen the depth of field… If I'd been able to duplicate the problem. Sigh. Experimentation is a frustrating business.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

[Ken N]

It looks to me that the sensor was left shifted and then the IBIS performed a rotate. The blurring on the right side of the image appears to me to be motion induced.

[CMS]

This is a riddle indeed.

Just one observation that might perhaps be useful. Alignment errors can lead to strange results. I had such a problem once with a first generation Nikkor 70-200/2.8 VR used on a film camera. The centre was very sharp, and both the upper right and lower left corners were expectedly less so, as should have been the two other corners. Unfortunately the lower right corner was very soft due to an alignment error, but the real surprise was the upper left corner, which was as sharp as the centre. These results were consistent from one picture to another.

I am quite suspicious about the quality control of modern mass produced lenses, and would't exclude a wobbly glass element in your lens, which sometimes is perfectly aligned, and sometimes not.

[ctein]

Dear John and Roger,

Just one small difficulty with implementing your suggestions. The problem went away after those 10 exposures. You can't test what you can't reproduce.

And, not so incidentally, that does exclude an incompatibility between image stabilization and the tripod as the source, since myriad photographs and tests done afterwards under similar conditions with image stabilization on showed nary a problem.

Yes, Olympus knows a thing or two about this. As it happens, so do I.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

[ctein]

Dear Andy,

That's not true, except at position of the aperture blades within the lens. Stuff on the front element won't be imaged sharply, it will be very very smeared out, but it will tend to localize where the stuff is.

To demonstrate this for yourself, take your camera and a little piece of black cardboard and start to slide the cardboard over the front of the lens. You'll see the image darkening more on that side than the other. There won't be a sharp demarcation (not unless the lens is stopped way, way down), it'll look more like a gradient filter effect.

It's precisely because it is localized but not sharp that John and Hugh's suggestion that there might have been something on part of the front of the lens remains possible. I am pretty sure I'd have noticed a big blob of water, but maybe not a bit of condensation or mist, since it likely would have evaporated by the time I took the camera out again.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

[Ted Durant]

"the focus was always the same (and at infinity)."

How does one set the Oly 45mm lens to infinity focus on an OM-D?

Before you pop back with the response that you tell the camera to reset the lens on power up, I can assure you that mine resets to minimum focus distance. And I can also assure you that mine zips way past infinity when I spin the focus ring. Do people just find a far away object and call that infinity? Or is there another trick I am missing?

[Hugh Crawford]

Several comments have proposed that the in body image stabilization is at fault, and in fact the blur looks much like motion blur caused by several cycles of vibration over the duration of the exposure. It could just as easily be the several cycles of vibration of entire camera on the tripod. What the comments fail to propose is how one side of the sensor could be moving while the other side remained stationary. If the sensor were rotating about a point on one side of the image the blur would be in the form of concentric arcs which is not shown in the image. I can not imagine any scenario where sensor movement or movement of the entire camera would produce these images. Movement of an individual optical group in the lens might do this and I have a 24mm Nikkor lens with a crack in the front element where it got slammed into a doorknob that looks a little like this, but displaced optical groups and cracked lens elements pretty much never fix themselves.

[P. Tzebetzis]

I've experienced a similar problem with a K mout Tammron 28-75 2.8, as soon as I bought it (used). When I did a quick check at the store (on my camera, a K20d) the lens was fine.
On my first outing with it, all photos were soft throughout the frame. While I considered the possibility of focusing errors, that couldn't be the case in all 60 photos; most of them were static, the camera was focusing fine with other lenses, subjects were mostly static and with good contrast, and I was not that novice at the time.
After 6 months of parking the apparently useless lens, I gave it another try and surpise surpise it was fine.
The only explanation that I could give was loose element (or barell mechanism) that went back into position..

[Bob Rosinsky]

San Francisco area? Tectonic plates? Perhaps little tremors are responsible for strange oof behavior.

[ctein]

Dear Frank,

I hear an even more effective solution is to unplug your camera from the wall, count to 20 and plug it back in again.

(Tech support humor, arh, arh)

~~~~~~

Dear Bruce,

For objects at infinity, the tilt of the camera is irrelevant so long as the optical axis of the lens remains normal to the plane of the sensor. In other words, an internal misalignment in the camera can screw up the plane of focus, but not an overall tilt.

(The panoramic photographs actually have the camera tilted up a bit. Look at the position of the distant horizon line (the lights on the other side of the bay). It's below the midline of the photograph.)

For objects that are at measurable distances, the tilt of the camera is important as that changes the plane of best focus in the subject. In fact, if you're peeking at sharpness down at the single-pixel or so level the depth of field is ridiculously small and getting the subject and sensor planes properly aligned is very tricky. 90% of the time, when some photographer checks out the sharpness of their camera or lens by pointing it at something like a brick wall or newspaper taped up on a wall, they don't know what they're doing and their report is useless. There are ways to do this right, but they require uncommon expertise.

Non-perpendicularly WAS a problem for me in finding a suitable illustration of Christmas lights to show the sharpness problem had gone away. Most of my photographs of those subjects are not taken head-on. While I'm capable of interpreting an oblique photograph correctly, most readers wouldn't be and they would find the results confusing if not entirely misleading. Figure 9 was one of the very few photographs I had that was suitable as an illustration for the article, and I checked it over very, very carefully before using it. (No, it's not perfectly-head-on, but it's close enough so that differences in the plane of focus from one side of the scene to the other are not visible even with extreme pixel-peeping.)

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

[ctein]

Dear Ted,

Sorry for the slow reply, I missed your question amidst the comments.

Your guess is right. People just find a faraway object and call that infinity. Use manual focus and zoom in. In my case, I was using the bright lights near the bay.

Mind you, it's important to make sure that the faraway object really is far away. If you're doing extreme pixel peeping (which was not necessary to see the unsharpness in this case), you can see the difference between focusing at 100 m and focusing at infinity in many circumstances. Depth of focus is razor thin on the pixel level, under 20 microns.

pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
=====================================

[John Haines]

The problem is simple: the O-ring lost it's resiliency at that temperature.

[CDiMa]

Sorry for coming late in but I was playing with an idea that could be worth a couple of cents in explaining the uneven (un)sharpness of those photos. I think that the culprit could be a weird interaction between field curvature and the IBIS system.
First I don't think that the cause of unsharpness lies in motion blur. I think instead that the fact that the sensor isn't fixed doesn't guarantee that it is perpendicular to the lens axis. A little grain of dust or an occasional circumstance could have it misplaced for the series of photos and cause the phenomenon.
I should say that I don't know how the 5-axis system works and which movements the sensor is capable of, but I played a bit with compass and ruler and produced three simple images that vaguely show what could be happening:
http://www.cdima.net/fieldC/fieldC-1.png
http://www.cdima.net/fieldC/fieldC-2.png
http://www.cdima.net/fieldC/fieldC-3.png
In each of these cases the misplacement of the sensor could cause a side of it to be further away from the surface of focus which isn't a plane due to field curvature. I'm not an optics expert and my oversimplified view could have fooled me into a wrong line of thought, nonetheless I'm exposing it here to your consideration. Obviously I don't know which is the amount of any quantity involved, like the movement to wich the sensor is allowed or the amount of field curvature, and if they could affect the sharpness to a visible extent. Anyone with greater physical knowledge could make some calculations and see if they could match reality?

[ctein]

Dear CDiMa,

Nice drawings!

They illustrate what I was talking about in my comment to "toto and others," above.

'Course, in a real lens the curvature of field is a complex wiggle, not a simple bowl. But, this nicely shows the principle. It's possible.

pax / Ctein