I Don’t Know Why It Swallowed a Fly – Weather Sealed Lens With a Fly Inside

Posted by

I’ve been blogging about testing and taking apart camera equipment for almost a decade. Lensrentals.com has many thousand lenses these days, and they all get used frequently. When you have lots of lenses and they get used frequently, stuff gets inside them.

Usually the stuff that gets inside is dust. Our repair techs open up and clean dust out of more than 100 lenses a week. Not because the dust matters a bit in a photograph; it doesn’t. But because people still seem to think it does. People also, for reasons I can’t understand, seem to think that weather sealed lenses are less likely to get dust in them than non-weather sealed lenses. I’m not sure why they think this, but they do.

Sometimes the stuff that gets inside them is interesting and we get to blog about it. We found a spider, complete with web, inside a lens once and yesterday we got to add a new item to our ‘found inside lenses’ collection; a nice, fat, fly. And not just a fly inside a lens, but one way down deep inside a weather sealed lens. So deep that it took 4 hours of work to get it out.

The lens in question was a Canon 24-105mm f/4 IS that returned from rental looking quite normal with the renter taking equally normal photos.

Nothing unusual looking about this lens on its return.

But a quick look through the glass reveals a fairly significant problem a fly deep inside the lens. Being in the middle of the lens with lots of glass elements on either side, it was much more difficult to photograph than it was to see visually. Darryl finally came up with an arrangement that worked decently.

Canon 5DIII with MP-E 65 5x Macro lens shooting through the 24-105. Angling the light let it bounce around through the glass elements and give us decent lighting.


After a little trial and error, he got some decent images of the fly.



If you just dropped by to see a fly picture, you may go now. But getting the fly out of the lens turned into a 4-hour ordeal for Darryl. The pictures he took along the way do a great job of demonstrating exactly why dust will always find it’s way into a lens, so if you’d like to see that process, read on.

Darryl’s first thought was that removing the front element would let him used some compressed air to blow the fly out of the lens. We didn’t have a lot of hope that would work, but it’s quick and easy to do so it was worth a try. Next Darryl took the back off of the lens and took out the rear element, hoping that would work. It got another nice picture of the fly, still a few pieces of glass away from freedom, but otherwise didn’t accomplish anything.


Finally, bowing to the inevitable, Darryl did a complete disassembly of the autofocus and mechanical system, leaving the inner optical elements in the helicoid barrels.


The point to be made here is that there are huge openings into the inner lens through the helicoid slots. Obviously anything that gets inside the lens is free to move about to the inner elements. Dust particles, spiders, or in this case a fly, are all going to find their way down inside the lens. And weather sealing isn’t ever true sealing. There is no lens made for cameras that doesn’t have openings that air, dust, and other things can flow through. In the case of the Canon 24-105mm there’s an obvious opening through the rear of the lens.

Rear of the Canon 24-105mm showing anything that wants can get inside the lens around the rear group. A lot of lenses are similar to this one, but every lens ever made has openings.


Let’s get back to our fly. Now with the inner barrel removed, Darryl could peak through the helicoid openings and locate the fly between elements.


After a little trial and error, he was able to get some needle-nose forceps through a slot and remove the fly. (It wouldn’t blow out; apparently some fly sticky-stuff had it pretty adhered to the lens.)


The element could be cleaned through the slots too, and then the lens was reassembled. Because of the amount of disassembly involved, the lens had to be recentered and optically adjusted. Total repair time to get the fly out was just over 4 hours.

The takeaway message, though, is in the picture you can’t see here. We took dozens of images with the lens before taking the fly out. We shot stopped down, we shot at all parts of the zoom range. We focused close, we focused far. And in no image could we find the slightest hint that there was a fly in the lens. And you guys worry about a dust particle or two!


Darryl Bolin, Roger Cicala and Aaron Closz


June, 2015

Author: Roger Cicala

  • Peter Harris

    A lonely way to die.

  • Madwyn

    I would rather have dust particles than a fly.

  • Ilya Zakharevich


    of course, what I wrote about 0.2? is complete BS. To extract all the info from f/4 lens (with dominant wavelength ?=0.5?) it is enough to have pitch 4? = 2?. Still, the pitch of Canon is not enough.

    To see the results of this, take T?(x,y) = T(x,y)*T?(-x,-y) and repeat it many times with step D:
    T?(x,y) + T?(x-D,y) + T?(x+D,y) + T?(x-2D,y) + …
    + T?(x,y-D) + T?(x-D,y-D) + T?(x+D,y-D) + T?(x-2D,y-D) + …
    + T?(x,y+D) + T?(x-D,y+D) + T?(x+D,y+D) + T?(x-2D,y+D) + …
    + …
    THIS the result of inverse Fourier transform (with sub-sampling taken into account). Recall that T? has finite support (vanishes outside of a circle). Each of the summands above vanishes outside of the corresponding circle (centered at (0,0), (D,0), (-D,0), (2D,0), (0,D), (D,D), (-D,D), (2D,D), etc). If the circles do not overlap (D > diameter, as with pitch 2?), the subsampling does not matter: one can reconstruct T? basing on the sum above.

    However, if D is less than diameter, the circles overlap, and what happens on the edge of the circle is obscured by what happens near the center. Above, a point may belong up to 5 different circles.

  • Ilya Zakharevich

    ?? To Brandon:

    I think you mix FFT and FT in your reply. The image of a point source IS a Fourier transform.

    ?? To Chicchak:

    There are two major obstacles to finding a fly in the inverse Fourier transform. The image I (“the intensity of light”) of a point source is the Fourier transform of the “complex transparency” function T: for every point in (a cross-section of) a lens take the optical path to the source, and use transparency along this path as magnitude, and the mismatch in the optical path length (w.r.t. the ideal lens) as the phase of this function. The fly WILL modulate this function T — but only if the rays from this point source CAN hit the fly (on the path to the sensor).

    However, what sensor stores is not the image I itself, but its intensity |I|² (which is the PSF). The intensity is ALSO a Fourier transform, but not of T(x,y); one should take the convolution T(x,y)*T?(-x,-y). (THIS is what the inverse Fourier transform would give you.)

    Now there are two cases: if the point source is in good focus, then T is practically real, so above we take T(x,y)*T(-x,-y). The fly would modulate BOTH factors; so there is no way to distinguish a fly in a certain position and a fly in a centrally-symmetric position! Anyway, if the fly modulates T by ?T, then it would modulate T(x,y)*T(-x,-y) by T(x,y)*?T(-x,-y) + symmetric_term. Note that ?T is concentrated on a very small region; when one convolves it with T, it would be blurred out of any recognition.

    CONCLUSION (in-focus): (A) there is no way at all to distinguish symmetric positions of the fly; (B) whatever is distinguishable on the inverse Fourier transform, is blurred orders of magnitude.

    But this was only one problem of two. The second one is the sensor pitch. Due to the position of the fly, one should use the lens wide open; then the diffraction pattern has size ?*d (with d the denominator in 1/f-number): about 2?. To distinguish delicate details of the PSF, one should better have a resolution of order of magnitude better: about 0.2?. This sensor resolves about 4.5? (IIRC).

    (Due to aberrations — the phase of T — the size of PSF would be larger. But I do not expect that this blurring would allow one to use a coarser pitch; my guts feeling is that this 0.2? is not improvable.)

    The situation out-of-focus is slightly better: bokeh is, essentially, the shadow cast by the aperture on the sensor. So, if one can find a point-source which illuminates the fly, and these rays reach the sensor, then the shadow of the fly will be seen on bokeh. (Mathematically, the Fourier transform is, essentially, the transparency function itself! This is because it is quickly oscillating, and the saddle point method is applicable.)

    But now the Fourier transform, essentially, disappears from the scope, and the question is purely engineering: find the focal length and a direction to a point source for which this part of the lens contributes to the image (one may need to choose infinity-vs-close-focus as well!). Investigate bokeh as far out-of-focus as possible (in this direction). I think if one is inventive enough, such large obstacles may be visible on bokeh. (For nearby point-sources, I would try a reflection in a small metal ball.)

  • JR

    …. Brandon drops mike and exits stage left … lol

  • Brandon

    Hi Chicchak,

    There is no reason for light to undergo a FFT after passing through a lens – in fact so long as the light is passing through an isotropic medium (i.e. one of uniform refractive index) there is no change in the light’s behavior what-so-ever, ignoring for the moment the absorptive properties of the materials.

    In real thick lenses, light does not bend continuously, and due to the optical power of a surface does not appear to bend at the surface itself, but rather some relatively short distance inside the element. The places it appears to bend at are known as the principal planes, and none of the rest of the element appears to contribute anything to the effects of the element.

    Thin lens models are useful only in generalizing real optical systems, but their principal planes coincide on the thin surface – still no FFT is done to the light – it simply bends.

    FFTs are only useful in calculating the MTF of a lens, in fact both our ImageMaster and MTF benches from Optikos such as the LensCheck systems are done by performing an FFT on a greatly magnified image of the spot from the lens formed of a known target. The spot itself may be calculated by tracing many rays through the system, but all rays must be some combination of the marginal ray and the chief rays – i.e. the lagrange invariant of the system must be truly invariant.

    In any event, sampling the spot of the lens at unity magnification with a camera sensor provides a fundamental limitation to what can be known about the lens itself, and camera sensors are actually remarkably insensitive to problems in the lenses – we see this frequently when OLAF and the MTF bench appear to “overread” lens issues. Often we will test a “bad” lens on camera with an ISO 12233 chart and it will appear just fine, despite looking terrible on OLAF.

    Regarding the pupil obscuration, it is impossible without the lens design data to say absolutely what would happen due to a specific obscuration, but several generalizations are true:

    * The on-axis ray bundle must fill the aperture

    * In the presence of any vignetting, the off-axis ray bundles will not fill the aperture

    In the case of an edge obscuration near the aperture stop, I can mostly guarantee that only light near the center of the lens’ image will be obscured, as all camera lenses have some vignetting. As you move further away it will tend to increase vignetting, through likely not in a meaningful way.

    Here is a paper you may find interesting covering the math of pupil obscurations:



  • chichak

    Physicist here – did you try to Fourier-transform the images you shot with the lens with fly in it? Because that’s what has happened to light _just after_ it passed through a lens. Although lens here as in a single piece of glass, and to be specific, the light is strictly the FT just after an infinitely thin lens that still has curvature – not these dull, real ones that adhere geometry. Still, if something blocks light deep inside a lens assembly, I imagine the light field to be similar to the Fourier transform of the source, i.e. its intensity shows how much variation is in the frame – if the fly was right in the centre, it would effect the camera image by lessening average brightness difference of contrasting regions in different regions of the frame, whereas if it was in one corner, it would reduce edge contrast – things that I imagine are usually not considered when searching for a fly.

    The inverse of a fourier transform is a fourier transform – That’s how the image at the focal point of a lens does look like the light source. But that also means that fourier-transforming an image taken with a lens with a fly inside will get back the fly inside the lens.

  • Roger Cicala

    Alex, we tried dozens of shots through the lens. Never saw a hint of the fly.

  • Damn,

    How on earth did that fly get so deep in the lens. Taking 4 hours to get it out must have made you feel somewhat of a surgeon 🙂

    Once question that I wonder is, when you take a photo with that lens that had the fly in it, would the fly appear somewhat clearly on the photo?

    Nice job getting him out

  • Daniel H.

    To me, its look more like a bee, not a fly…

  • Michael Fritzen

    Just some wild guessing but I don’t think the fly got in there in the same physical form you took it out. I’d say most probably in a much earlier state of its development.

  • Kevin McCoy

    Next time, consider simply placing the lens in a container with some spiders. Eventually one will enter the lens and consume the fly. Then place the lens in a container with some small birds. Eventually, one will enter the lens and consume the spider. Working your way up the chain will get you to a reproducible problem: http://wordpress.lensrentals.com/2012/05/the-damage-waiver-bearly-covered-this-one

  • Awesome “story”… it’s time to think about fly glue trap and pest control in the camera bag!

    “I don’t think the client ever noticed it.”
    THAT’S very strange… 😉


  • Roger Cicala


    We looked through the lens during routine return inspection. I don’t think the client ever noticed it.

  • How did you/the client know the fly was in there in the first place?

  • NancyP

    Swell, something ELSE to worry about in the field…. I am glad that it didn’t cause sensor image degradation, but it is true that should one of us own a lens that has a visible fly in it, one’s personal image might be pooh-poohed a bit by other photographers (ewww, do you know..)

  • Roger Cicala

    John, I’m pretty sure he came in through the back of the lens when the cap was off.

  • great article.
    A number of comments:
    Was Daryl ever concerned about cruelty to animals? – shouldn’t he have just smashed the lens to release the fly as quickly as possible?
    Was Daryl singing the theme tune to “Born Free” throughout the rescue operation?
    On a slightly more serious note. How can this kind of situation be avoided? – mainly concerned with dust. Does a rear lens cap have any real protection ability?
    Any other advice for keeping lenses clean?

  • Drew

    Re: The fly sticking to the glass … the explanation is obvious … prior to the lens, he was in the ointment! Just a random comment from a …. yeah you guessed it …. fly on the wall.

  • Anthony M

    So, maybe that spider you guys found a few years back wasn’t so stupid after all?

  • John D

    BTW Rodger, are you planning to do your yearly reliability index? you know, the one where you list the gear that gets repaired the most? That list certainly influences my purchasing choices more than any reviews I read. thanks

  • John D

    OK, I give up, how did the fly get in there? I had that lens and never noticed any holes big enough for a fly to get into. BTW, I used that lens in driving rain and never had any problems. That is the lens that does every thing really well and nothing perfect. nonetheless it ended up on my camera about 90% of the time

  • Roger Cicala

    Allan, it was pretty stuck to the lens, I don’t think anything less than direct removal would work. Most things can be pretty easily blown out though.

  • Roger Cicala

    Dan, we did. He didn’t fly or move or anything, but he was free to go.

  • Dan Deakin

    Did you let the fly go after you got it out?

  • Allan Sheppard

    Would it have been possible to extract the fly using a stiff plastic tube connected to a vacuum source?
    I have had the experience of extracting objects with tweezers/forceps and it can be most frustrating. Allan

  • Roger Cicala

    Lynn, it is indeed rated as weather sealed. Which means, well, it has a rubber gasket at the back where it attaches to the camera.

  • Roger Cicala

    You win, Kyle. And we’ll give it a try, but I’ll ask that you wait until August. Aaron is going to be gone all of July and we’ll be horridly behind.

  • KyleSTL

    Where is my prize for guessing what Darryl was up to, Roger? Care to optically adjust a 24-105mm F/4L lens for me that was disassembled for aperature cable replacement (why is this problem so common)? I figured a 60mm macro must be shooting something inside the 24-105mm based on the focal distance range of the lens.

  • Is the most likely explanation for the fly inside … that it got into the lens without rear cap attached (for shame, even for a rental), and then wandered around inside the elements until it died, trying to get out?

    BTW, I was pleasantly surprised that the EF 24-105mm f4L “kit lens” is rated as “weather sealed”. I consider it an under-rated lens, especially considering the value when it is part of a “full frame bundle” with 5d2, 5d3, or 6d.

Follow on Feedly