hey guys.... hoping you can help me figure out this problem:

Unusual Canon HDDSLR Video Behavior on Vimeo

basically, using any fixed aperture zoom lens (i'm using the 17-40 f4L in the video) and filming a brightly lit scene with a 7d. Everything on manual all the way nothing auto anywhere near my camera. Zooming causes the camera to record video about 1/2 stop darker than fully zoomed out.

Canon 7d firmware 1.2.2 on full manual.
14-40 f4 L lens (fixed aperture not variable)
only happens when shooting video - does not happen with live view in stills mode.
camera set to neutral profile with sharpness and contrast flattened.
have recreated on 2 different 7ds and a 5d mkii at vistek.
have recreated using 2 different 17-40 f4 and a 24-70 f2.8 (ruling out a lens issue)
iso expansion on/off same result.
highlight tone priority on/off same result.

now, i know - you're not likely to do this kind of snap-zoom thing during a shoot but it brings up an interesting problem. the camera seems to arbitrarily adjusting the exposure based on the contents of the frame even though everything else is set to manual.

ever hear of this before? hoping to track down what i'm doing wrong or the cause...

let me know what you think! really appreciate your help!
karl

I've noticed the same thing on 5D2 with the 24-105 F/4.0. I'm very interested to see if anyone can suggest a solution. It's not fun to adjust exposure whenever you zoom in or out.

Don't think there's any great mystery about what's happening, at all.

Explaining it is not quite so easy, however.

OK, take the zoomed out picture. Aperture and shutter speed set for the amount of light reaching the sensor in general, and each individual pixel in particular.

In this mode, each pixel is "seeing" an area of "X" which is throwing off "X" amount of light, the average across the entire sensor being used to set the shutter/ aperture.

Now zoom in.

Each pixel is now seeing an area "Y" which is significantly smaller than "X", and as such is throwing off significantly less light than "X".

Thus, if neither the aperture or shutter speed changes, the picture must darken, as overall, it is receiving less light.

This is one of the reasons macro shots need their own nuclear generators to get enough light onto the subject (and hope not to cook it to a crisp in the process) as the subject area is so minute.

If this is (apparently) not happening to you when in stills mode, I would subject your methodoligy to some serious scrutiny, it is happening, but in stills mode there is no way to make a comparison, as the still is just one shot, not a series that shows the transition.

Does that explanation make any sense?


CS

hi chris,

the issue is definitely not happening when in live view for stills mode. this is not something you need to "pixel peep" to see - check the video. it's an apparant approximate 1/2 stop of light loss in video mode only. you can see it in live view AND it's recorded to video. switch to live view (stills mode) and the problem simply isn't there. not comparing still image to still image but live view to live view (video vs. stills mode). this is definitely tied to the video mode on canon hddslrs.

start up your own 7D with a fixed aperture zoom and you'll see what i'm talking about - this one isn't made up or pulled out of thin air. it's been observed by many people today with demonstrations and in-person shooting.

i have to say i've been shooting for over 25 years (and have a BA in photography from ryerson in toronto) and i've yet to hear of zooming over 30mm of focal length requiring 50% more light. an f-stop is an f-stop - this is why handheld light meters work!

karl

The problem could be due to the reduction of the EFFECTIVE or true aperture of the lens. You have to search online to check the true specs of the particular supposedly "fixed" aperture zoom lens you plan to use in shooting which requires the exposure to be absolutely constant or, at least, not noticeably change.

For instance, your Canon 17-40 may have a stated aperture of f/4.0 to 4.0 where in fact it may really be f/3.8 at the wide end to f/4.2 at the tele end or any similar increasing numbers. The fact that your clip gets darker when zoomed in rather than when zoomed out indicates the lens is not a true fixed aperture zoom. True (and ultra-expensive) cine zoom lenses do not exhibit the problem because every one of them has the aperture that is absolutely constant down to 0.0X or 0.00X.

hey there,

again, this is not present when viewing the image in live view in stills mode and only present when in video mode. both live views use the same screen and the same lens -- and live view in the stills mode uses the maximum aperture of the lens. if the effective maximum aperture was changing i think you'd see it.

remember this is a fixed aperture L series lens. top of the line canon optics and widely regarded as the finest lens in canon's lineup - i don't think that this is tied to canon misrepresenting the lens as fixed aperture when it's not. also the same effect was seen using a variety of fixed aperture zooms but only in video mode. the same effect (to the same degree was seen when using the 24-70 f2.8 L series lens... so this isn't tied to this lens in particular).

not trying to prove there's a problem - just trying to find out what the problem is... not sure it's an issue with the optics to be honest. this really looks like it's tied to the video engine on canon hddslrs.

any other ideas?
k

You need to do a bit more testing before you can conclude that it is the camera and not the lenses causing the issue.

What happens when the lens (or lenses) is stopped down to say f/8? Does this behaviour still occur?

What about when you are using manual lenses via an adaptor that have absoultely no 'dialogue' with the camera? If it is the camera adjusting something in the settings when the lens zooms, then theoretically it should not change the exposure if it does not know the lens is zooming.

Are you comparing still photos taken with live view or are you just looking at the LCD display during Live view? Because the LV on many DSLR's will adjust to diplay the best preview (regardless of manual settings) so you can compose your shot - the only applying the selected settings once the shutter is triggered. This can usually be switched off in video mode so you can see what is being recorded. So it could be that there is less light reaching the sensor (or each individual pixel) due to the explanations given here by others, but as you zoom in and the light level decreases, the camera's LCD display is compensating to give you the same live view preview image.

hi john, thanks for replying!

when the lens is stopped down to f8 or any aperture the same behavior occurs. the sample video is at f7.1 to be exact.

granted i haven't tried a manual lens via an adaptor -- as i don't have such an aparatus. but please remember this ONLY happens in video model and DOES not happen in live view in stills mode. on a 7d the live view displays the image as live video on the lcd just like the video mode. the live view in stills mode however DOES not show this issue in any way shape or form. if there was some "auto-toggle" that was being overlooked you'd expect the same issue to be present live view stills mode. on the 7D live view stills to live view video is simply physical switch and all other camera parameters stay the same.

this seems to be rooted in the video mode itself. it's possible that canon's video codec is affecting the video image and compressing the highlights after processing - not sure.

i've tested this on multiple cameras (2 7D's and a 5D mk ii) with multiple lenses (2 17-40 f4 and 1 24-70 f2.8) and the results are the same when using video mode.

not saying it's my camera - i'm leaning towards thinking it's tied to canon's implementation of video somehow but really i'm asking if anyone else has seen this same problem. ie: i'm not asking "is there a problem" i'm asking "has anyone else seen the same issue"

there's another post in this forum from eric that seems to be similar to what i'm experiencing...

update:

i'm comparing live view to live view. ie: live view in video mode (which is recorded) to live view when shooting stills. the point is that stills mode isn't exhibiting the same problem - only the video mode. i'm sorry but pixels are pixels - when you zoom you're not zooming in on the pixels, you're focussing the light (so to speak) on the same number of pixels. it just doesn't work as described earlier.

light is light - the light falling on the scene is reflected onto the sesor at a given aperture and shutter speed. the aperture and shutter speed determine the exposure - not the focal length of the lens.

turn on your 7d point it at a well lit subject in manual mode and zoom in with a fixed aperture lens. does the exposure drop by half a stop? now switch to video mode. you'll see the issue i'm describing and the issue that's recorded in the posted video.

karl

Karl, try stopping the lens down to f/8 or f/11 to see if the problem persists. My guess is IF it does, the difference in luminance between the wide and tele end will be smaller and may be hard to detect without close examination. It is also possible the problem just goes away as the variation in effective apertures at medium stated apertures of most zoom lenses (cheap and expensive alike) is normally so small as to be insignificant.

You can go to the Popular Photography website. They have a good history of still photographic lens tests and in each of the tests you will find published figures of the lens' actual focal length and aperture. None of the lenses has a truly constant aperture and more than a few "constant aperture" zooms have aperture that varies almost 1/3 of a stop from end to end.

Try turning off "Lens Peripheral Illumination Correction" and see if that has any affect on what you are seeing.

I suspect this issue may have something to do with how the imager is read during video mode vs photo mode.

I will see what happens on my camera this afternoon with the 24-70F2.8L.

thanks guys,

again, everything that can "possibly" affect exposure is turned off including lens peripheral illumination. also the behavior is the same no matter the f-stop.

the effective size of the diaphram at a given aperture value is different at different focal lengths (accounting for the fact that much more light is needed at longer focal lengths). f4 is still f4 in terms of the amount of light that aperture lets in but on a 400mm zoom an aperture of f4 would need to be a much larger opening than f4 on a tiny 28mm. this might also explain why fixed aperture zooms are so damn expensive!

it seems that that canon's lenses are locked down in video mode cannont "scale" the aperture as the zoom moves through its range. this effectively makes the fixed focal length zoom a variable aperture zoom druing any given shot.

what's interesting is that every video camera (proper video cam no hddslrs) has accounted for this since the dawn of time. i think the fact that most guys shoot fast primes with hddslrs has caused this issue to be largely unseen.

live and learn i guess - curse you, physics!
k

An interesting note here.

Many cinematic lenses have T numbers not F numbers.

T's are actual measured numbers, not theoretical. F numbers are theoretical arrived at by calculations.

Typically a f1.7 lens will be rated T2.

Any still zoom lens will probably have slight variations as elements change positions and it does not really matter for stills.

But for cinematic lenses this is unacceptable, so they are built differently, and are much more expensive. And most cinematic lenses are primes.

Cinematic lenses also have extremely accurate witness marks (focus distance marks) and the focus ring travel is far larger for more accuracy.

You certainly can use still lenses for video. But they are not really meant for it. So there are compromises.

i'm with you there olof.

the workaround in this case of course is to start and stop the shot when you change your angle of view. this seems to give the canon hdslr a chance to "reset" the aperture in the lens to maintain the f-stop at the new focal length.

bit of a learning experience with this one - but good to find out what's going on there!

karl

Wow this pretty interesting. Has this not been noted before ever? I'll have to try the stop/start method when zooming. One more thing to think about!

Interesting, have you tested this theory?

If you start the shoot zoomed in and then zoom out does the exposure get brighter by half stop?

My wild guess is that it might be something to do with the digital magic that happens when a megapixel sensor image is reproduced at a lower resolution for video.

The moire problem associated with these cameras in video suggests to me that fewer pixels across the entire sensor width are being scanned.

I wonder if in stills mode where every pixel is counted, there is some blending of light falling at more acute angles across to neighbouring pixels by the anti-aliasing filter which might have an apparent gain effect. If in video mode there are fewer pixels contributing to the image, then there would not be enough diffusion of light across the gaps by the anti-aliasing filter.

At a wide angle, the angle of incidence of light falling onto the sensor might vary more than when the lens is zoomed in with maybe a tendency for darker corners which is accentuated by 35mm groundglass adaptors.

Does the camera have an inbuilt compensation for this? I understand that HDCAM cameras have a user selectable function that maps the brighter and darker areas of certain lenses and compensates.

If the camera does have a compensation process, then fixed mapping for darker corners would be wrong for the long end of the zoom where the brightness across the image is more even and the entire image would be darkened as the corners are seen to be overbright by the mapping process.

I am likely barking at shadows in the manner of a knowall who knows nothing. It is just a guess that a combination of sorts may be going on.

It would be interesting to see if leaving the lens on wide and travelling the camera itself forward to reproduce the same framing as the zoom-in creates, exhibits the same darkening effect.

I would not regard this effect so much as a defect as something you may need to live with. There are limits to how much work can be done by the processor in these remarkable cameras. If it is any compensation, you may observe a corner darkening effect in the image of purposed videocameras when zoomed in and darkening overall as some of these do not have constant aperture zooms.

it seems that that canon's lenses are locked down in video mode cannont "scale" the aperture as the zoom moves through its range. this effectively makes the fixed focal length zoom a variable aperture zoom druing any given shot.



If this is the case then it would stand to reason that a manual lens used via an adaptor will not exhibit this behavior - as these lenses would need to be able to correct this without communication with the camera about whether it is in "stills mode" or "video mode"

Yes, if this is the case, a comparable fixed zoom such as the Nikon 17-35 f/2.8 AF-D (with aperture ring) mounted via an adapter will easily solve the problem.

hey guys,

like i said this looks like a specific problem with canon lenses on canon cameras. not a malfunction so much as the physics and mechanics of the lens. if you put the same canon lens on a manual adapter and back on a canon camera i'd suspect you'd see the problem in still mode as well.

i've done some more tests and here's what i've found:

looks like starting and stopping the shot doesn't "re-cajigger" the aperture back to it's correct size depending on the position of the zoom in video mode. the correct aperture value at any other point in the zoom's range must be tied to the shutter on canon dslrs. my tests have shown that exposure definitely isn't affected by zooming when shooting photos.

what i have discovered is that the issue ISN'T PRESENT when the lens is wide open. this would explain why you don't see it in live view in photo mode (it's always wide open until the shutter is triggered). weird that it only happens when stopped down but could have something to do with the mechanics of the lens.

THE BAD NEWS:

unfortunately this seems to mean that you'll need to adjust your exposure every time you change your focal length with a canon zoom lens in video mode on canon hdslrs. my tests showed approx. 2/3 stop from 17 mm to 40 mm on my 17-40 f4 L lens. haven't tested any other lenses to find the exact exposure difference.

the other issue seems to be that this change in exposure isn't regular and has an odd spherical shape to the decrease in exposure. you can see this shape in the posted video - almost resembles the way the 17-40 elements move inside the lens barrel... i could just be making that part up but it looks that way to my drunken eyes.

it's too bad - the versatility of having mulitiple focal lengths available (i don't zoom during video... hate that!) during a single shot is really handy. especially for run and gun shooting where you might want to keep the clip running as you quickly change angle perspective and sort the shots out in post.

THE GOOD NEWS:

hdslr video has its own character and produces undeniably beautiful images quality if handled correctly. the entire hdslr revolution seems to be about working around the fact that these damn things weren't really meant to do this but do it really well. just another example i guess!

time to get some primes!

k

I looked at his footage and noticed that when he shifted the camera's view to where the "mix" of light and dark" tones stayed predominantly "light" the exposure change effect was less.

In Feb this year we had an unexpected and very nice snowfall. I used this same lens (17-40mm f4L) on my 7D and shot some video footage from just off my front porch. For one sequence I zoomed slowly from the 40mm end to full wide (17mm) and there was NO CHANGE in exposure. The mix of light and dark changed only slightly (began to include a few neighbor's trees - dark tones) with the slightly subdued white snow tones continuing to predominate.

In the OP's video as he zoomed from wide to tele, dark tones began to dominate the "mix" and subsequently his image got a bit darker as his composition was beginning to require more exposure. It's a lighting and exposure issue, the guy is wearing a dark tee and jeans, seated on a light colored couch with what looks like slightly lighter wall in the background.

I can zoom with my 17-40 f4l and experience no exposure change so far but I tend to look for lighting and tonal quality issues before shooting.

hi bruce,

this isn't an exposure issue as the camera is on full manual. there is no automatic exposure adjustment at all.

karl

I've just tried this on a DSLR and focused on a solid colour wall at wide angle, with the shutter speed fixed (1/45 for info) On wide angle the correct exposure was f4, zoomed in f6.7 on an 18-55mm and
f4.5 wide, f5.6 zoomed in on a 70-300. With aperture priority, then the shutter speed changes instead.

When the Canon is taking video, the shutter speed is constant, so if the setting doesn't allow the aperture to change OR the shutter speed, then the brightness is the only thing that can change as the light level drops. Or have I missed something here? I've tried 2 lenses which both let different amounts of light through as the focal length changes from an evenly illuminated surface.

hi bruce,

this isn't an exposure issue as the camera is on full manual. there is no automatic exposure adjustment at all.

karl

It is an exposure and lighting issue even in manual mode inasmuch as we must evaluate exposure requirements and visualize regardless of being in auto or manual mode. In the last position I held we shot thousands of color slides using Nikon F3s, you had to have the exposure "dead on" to meet presentation standards and I often saw exposure requirements change when zooming in close just because the predominance of dark or light image tonality changed.

Here is the video I mentioned using the same lens. Shutter 1/60th, aperture f5.6 and I forget what the ISO was. The zoom isn't the smoothest but there is no exposure change.


Quiet_Interlude on Vimeo

Paul R Johnson


I've just tried this on a DSLR and focused on a solid colour wall at wide angle, with the shutter speed fixed (1/45 for info) On wide angle the correct exposure was f4, zoomed in f6.7 on an 18-55mm and
f4.5 wide, f5.6 zoomed in on a 70-300. With aperture priority, then the shutter speed changes instead.

Assuming the lens you used is the 18-55mm "kit" lens, that is not a fixed aperture lens. The manual warns against relying on zoom during a video take as the physical aperture will change as the lens is zoomed. The discussion above is around a 17-40mm lens that has constant aperture throughout the zoom movement.

again, the lens used was the 17-40 F4 L series lens - definitely a fixed aperture lens.

what you're saying just doesn't make sense bruce - it's not how photography works. when your exposure is locked in manual mode the contents cannot and do not affect exposure. exposure is only affected by iso, shutter speed and aperture.

all things being equal it just doesn't happen the way you describe.

let's even say that auto-exposure was somehow influencing the scene - the predominance of darkness in the frame would cause the auto-exposure to lighten the image, not darken it. auto exposure typically (for the sake of simplicity) tries to equalize the contents of the frame to an overal 18% grey tone. show it dark, it makes the darks lighter - show it light it makes the lights darker in an effort to expose them at 18% grey.

it's possible you're referring to the fact that filling the frame with dark elements will create a dark image because you're photographing dark elements. i'm referring to the exposure actually changing during the zoom - again, see the posted video. that's not added in post. definitely there and definitely happening.

also - i only saw one short zoom in your sample video... and it came on the heels of a fade from black. the amount of detail in the scene doesn't show the effect for sure. try your camera pointed at a white wall with no other elements. set your exposure when fully zoomed out then zoom in and out during the video. you'll see the effect - been verified on 4 seperate cameras and three different photographers. not a mistake but something going on with the canon dlsrs and video.

k

i whipped up a quick test against a white wall that really shows that this issue is not being created by the contents of the frame:

7D + 14-40 f4L White Wall Test on Vimeo

i'd think my lens was the cause but i've verified the same issue on another 17-40 f4 L and a 24-70 f2.8 L and i've also seen it on 2 different 7D's and a 5Dmkii.

the crazy thing is that this does not happen when the lens is wide open. it's only present when the lens is stopped down. also stills shot at the different zoom positions all have matching exposures which tends to incline that it's not a lens issue. this really is limited to the video mode only.

okay, i need to walk my dog! LOL

karl

you know, i'm not sure this whole "not readjusting the aperture" explanation is the whole story. in fact i think this might be more related to the lens just not scaling the aperture correctly.

if you look through the front of the lens while zooming you can actually see what's going on. the aperture seems to be a fixed size but a second diaphram is coupled to the zoom and opens and closes slightly depending on where you are in the zoom range. i think this is how canon is accounting for the varying amount of light required to keep the aperture value constant through the zoom range - essentially giving the wider focal length less light to match the longer focal length's inability to gather light as efficiently.

i'm leaning towards believing that this whole mechanism is misaligned somehow but when shooting stills you just don't see this varying exposure when moving through the zoom range. maybe the mechanism that triggers the aperture to close at the moment of shutter release is more accurate than the mechanism that scales the aperture when the lens is stopped down? it remains mystery to me.

remember the camera is in FULL MANUAL mode. shutter speed is locked and iso is locked - the only other variable that affects exposure is aperture. makes sense that the lens isn't getting it's own built in aperture correction for the varying amount of light required to keep aperture values constant across a wide zoom range. this of course i sall conjecture as i really have no idea what's going on!

i think i'm going to take this lens into canon for calibration and see if it boils down to the lens itself. of course, i've seen this on 2 other different 17-40 f4 L series lenses and again on a 24-70 f2.8 L (to a lesser degree). difficult to think it's just "my lens" but maybe it boils down to a bunch of lenses being out of whack? i'll keep you posted as i find out more.

k

stumbled upon this video which describes a similar issue with a different lens and mentions the 17-40 f4 L as having the same known "darkening" issue i'm seeing.

YouTube - 8. Canon 7D (zooms) Advanced Lens Selection - Video Mode Interactive Tutorial (http://www.youtube.com/watch?v=UOBafREui-A)

he mentions the 17-40 at around 1:45. it's that secondary aperture deal that i was talking about. glad to finally see that i'm not just imagining this whole thing - maybe the 17-40 is just a bad lens for video? grrr...

karl

interesting tidbit to check:

EOS 7D Firmware Update Version 1.2.2

Firmware changes

Firmware Version 1.2.2 incorporates the following fixes.

Fixes a phenomenon in which the set aperture moves when shooting movies in manual exposure mode using some Canon lenses (such as macro lenses).

hey ryan,

the camera has version 1.2.2 installed. i don't think the aperture is moving - this looks to be a problem with the lens design itself. check the video linked in the post directly above yours...

karl

yeah I did... just wanted to add that to the conversation just in case. and also for anyone else searching through the issues.

I just checked my Sigma 18-50mm f2.8 zoom lens on my 7D (firmware 1.2.2) and there is absolutely no exposure change when zooming in and out while recording video. Must be a lens issue, I guess.

that's correct. this is a specific canon zoom lens issue (as described at length in the thread).