EOS 5D Woes

[Lounge Lizard]

if I was convinced that it was a poor example rather than a standard example of a camera with poor characteristic, then it would go back immeidiately.

Chris, that's the big dilemma facing every 5D buyer. Are there poor examples and is it possible to find a good 'un and be very happy with the camera, or is this a design flaw where I'd be better off with a cropped sensor camera.

Certainly I will try the other 5D body that Jessies has lined up for me (the manager is very concerned) but if that exhibits similar behaviour then it's goodbye 5D, hello 30D and I'll wait for the next Canon full-frame camera that has these issues resolved.

[TerryS]

David Steel wrote:
Regardless, if I buy L glass to be used wide open under low light conditions then I expect to use those lenses that way. If Canon launch a new body that is claimed to be compatible with all lenses then that is what I also expect - compatibility, not compromise.

So, we either have a scenario where all Canon lenses, L or otherwise, are designed with an insufficiently-sized image circle so that it is just on the limits of acceptability for film but not full-frame digital, or, the sensor in the 5D isn't good enough for full-frame usage.

I don't think the problem is that Canon lenses are designed with an insufficiently-sized image circle. I believe the more likely explanation is that Canon lenses have, in general, a very non-telecentric design. That is to say, the chief light rays that form points in the periphery of the image strike the sensor very non-perpendicularly (at a high angle of incidence). As a consequence the digital vignetting is high (as David's images demonstrate). There are two solutions to this problem.

The first solution is to use offset microlenses on the sensor to redirect the light rays so that they strike the light sensitive areas on the photosites more nearly 'straight on', but this would, I imagine, add considerably to the cost of manufacturing the sensor assembly. I suspect that the 1DsMkII does something along these lines, hence its sensor shows less vignetting than that of the 5D but it is far more expensive.

The second solution is to design lenses that are 'image space telecentric', which means that all the chief light rays strike the sensor perpendicularly (at zero angle of incidence). This is achieved by designing the lens so that its exit pupil is a long way in front of the lens. This is the solution adopted by Olympus with their four-third system - they claim that all their lenses have a 'near-telecentric' design. And some of the independent lens producers make similar claims for their new 'designed for digital' lenses.

I have been told on good authority that historically Canon have used largely non-telecentric designs for their lenses - with film this didn't really matter since film can tolerate a wide range of angles of incidence, however with CCD and CMOS sensors it has now become an issue since they cannot tolerate a wide range of angles of incidence. In contrast, I understand that many of Nikon's film lenses do have a near-telecentric design, which is ironic since Canon (with their full frame sensors) now have greater need for telecentricity than do Nikon (with their APS sensors). This is just an historical quirk I believe. (I should add that I have never spent the time to verify the truth of this myself by examining Canon and Nikon lens designs, but I was told this by a reliable and independent optical engineer who had examined the designs.)

If the above is true then there is nothing Canon can do to increase the telecentricity of their existing (film) lenses, however I would expect them to ensure that all new lenses are designed to be as nearly telecentric as possible (although it should be born in mind that telecentric designs have disadvantages as well as advantages - lens design is always a compromise). Consequently, I find it disappointing that the new EF 24-105mm f4 L does not appear have a near telecentric design despite the fact that Canon launched it with the 5D, which itself appears to lack offset microlenses. In which case this combination is likely to be particularly susceptible to vignetting, however this conclusion is at odds with those 5D/24-105mm users who claim that there is no significant vignetting with this combo. All a bit of a conundrum!

Terry.

[Gen_bunty]

Hi David,

Have you taken down the images ? I can't seem to see any images.

Thanks

James

[Lounge Lizard]

No. They're still there. It must be a local browser issue with you or your ISPs cache perhaps.

[Lounge Lizard]

Interesting stuff Terry. Reading around I find that Sigma lenses do better on the 5D than Canon's own lenses. Would they be more telecentric I wonder...?

[BigWill]

I must say I'm very surprised at your acceptance of a camera which consistently vignettes at wide apertures Chris, be it a "one off" or not. Surely you shouldn't have to accept such inferior performance from a camera at such a price level? If it were me I would have no hesitation in returning the camera. You should not as a consumer, be expected to have to use post production techniques to correct an error in a product be it a "bad example" or not!

If enough people bring their 5D's back to Jessops or whatever and major retailers start taking note and complaining to Canon about it, then, and only then I feel will something be done about it as individual complaints don't carry as much "weight" as when the camera trade start getting hassle about a troublesome product.

BigWill

[Lounge Lizard]

BigWill

Nothing is perfect in life and we all have to compromise somewhere. Some examples of the 5D seem better than others and maybe Chris considers that the issue with him is not enough to worry about and the camera's other fine features outweigh this particular drawback. That's for Chris to decide.

In my case, I often work wide-open with fast lenses and so this issue is more critical to me and maybe I have a particularly bad example as well.

However, I do hope that the reject rate by users will make Canon think carefully about their design and acceptance criteria in the future. I'm sure that the problem will be cracked but remember that this is a leading-edge camera at an aggressive price-point for what it offers. It may not be everybody's cup of tea but it might be the nirvana for many.

[BigWill]

Indeed David, I accept entirely that it is Chris (and others) perogative whether to accept the standards offered by the 5D or not but I find myself siding with you on this issue in that a camera with such an "Achilles heel" would certainly not meet with my standards of acceptability for a product in such a price range.

If anything, I hope that this thread will prove conclusively that the 5D does in fact have such a failing and we can then leave it to prospective purchasers to make their own educated decision as to whether they can live with such a failing or not. At least the issue will have been brought to the attention of prospective purchasers so that they can then make an informed decision having been acquainted with all the facts relevant to that decision.

BigWill

[Gen_bunty]

Doh. Had the display images from oriniating site only swtiched on in firefox.

Oh my. They really are bad. I thought the 24-105 was a shocker, but the 17-40 is just unbelievable.

How you can be expected to put up with fall off like that from a crop of L glass is beyond me. Even the 70-200 looks bad.

Definitely steering clear of even a second 5D. Canon I hope you are listening.

James

[Scphoto]

Sadly it looks like i have to confirm Davids test results

17-40mm



24-105mm




Can't explain why its at 28mm for the last 2 pictures. It thought it was 24. But the effect is much the same.

For me its probably not too much of a problem as i tend not to shoot wide open too much, however it is annoying something so expensive performs so badly in certain conditions. I might give canon support a ring for a laugh.

/edit Canon support sugested i take the camera back for a refund. I asked them to register a complaint. Someone from the complaints dept will contact me.

[alanS]

I might give them a call about dust in the optics and on the sensor.

[Scphoto]

I'll take the blame for the sensor, the optics currently belong to Jacobs in Birmingham (and the view is from their shop!).

[El_Sid]

I don't know if Mr C-W has seen this thread but it stongly suggests a full AP investigation might be in order.

[CircleOfConfusion]

surely it can't be all doom and gloom? if this is a standard issue across the board (and not just "bad ones") wouldn't canon be able to possibly update the firmware/techno-wotsits and make the photosites at the periphery of the sensor more sensitive depending on lense, aperture etc. being used

or am I getting a bit ahead of current technology

[robertj]

My conclusion is that the 5D does vignette badly wide open with most lenses.

Sorry to hear this.

I can't understand how a body can correct vignetting? If it is micro-lensing, in my view, this is not a proper solution and should be needed only for non-telecentric lenses. It is a well known fact that the less on the light path the better. I think that Olympus E-1 has no micro-lenses (not sure) as all Zuiko Digital lenses are quasi-telecentric (and designed for 4/3). I think once Canon releases digital lenses (for 35 mm sensors) the problem will vanish. Then, at least in my view, it would be preferable to have body with no micro-lenses on the sensor.

Why don't you go for some professional Nikon DSLR (you will have APS sized sensor, but if it may be enough)?

[Lounge Lizard]

...and will repeat these tests.

In the 5Ds favour, I must say that it delivers pretty impressive images right out of the camera. My previous cameras all required a bit of tweaking (levels, contrast, sharpening) before I got them printed but along with my test prints for Jessies, I just dumped about 50 JPEGs as they were straight from the camera onto a CD to see how they turned and I'm impressed. Superb quality - they look more like punchy and sharp images straight from a digicam and not a dSLR.

[BigWill]

I think it's more a case of Canon burying their heads in the sand and denying the existance of a problem in the hope that it will just "go away" if sufficient people don't complain about it Steve. They have somewhat of a track record of doing that as they did with the Canon EOS5 35mm camera when they never "officially" admitted that it had a weak control dial even though later versions had an "improved" dial fitted to them. I would hate to see Canon worm their way out of this one too simply because people were afraid to challenge the "Mr.Big" of the camera market! That is why I can't understand when people say "well I don't use it much at full aperture anyway" or "maybe I just got a bad one". It's simply not good enough that a camera in that price bracket should come with any major deficiencies and doubly galling when a major company get away with shirking their responsibilities to their customers by denying the existance of what has now become a well documented problem!

BigWill

[AGW]

Time for Canon to put some research into curved sensors? I'm sure this subject was discussed in an ancient post I read once?

Graeme

[TimF]

Somehow I doubt curved sensors would be the answer to this (and might even introduce a whole new set of problems).

Browsing the Leica Customer Forum just now, where the subject of the forthcoming digital M is being discussed, and specifically the purpose of the new "optical contacts" being added to the rear flange of M lenses from July 2006 (see the Leica Board here for details of that) has resulted in this item resurfacing.

As many here will know, the Leica DMR has software incuded which maps the characteristics of all compatible R-system lenses, so that things such as distortion, light fall-off etc can be corrected in-camera. This post appeared on the LUG some time ago, with comments in bold from the Leica Forum poster, John F. It might be that Canon will have to undertake a similar excercise for their full-frame cameras in future, though they might be forced to restrict themselves initially at least to lenses with appeal to pro aand pro-am users due to the much larger range they offer.

So far Leica has never officially confirmed any special function of the ROM chip when using the R8/R9 with the DMR digital back. On the other hand, here is the original text from the LUG that started this ball rolling last October:

"The two new [31 and 39 megapixel] sensors [from PhaseOne digital back maker for Hasselblad H2D] have some new technology and some new problems, which are yet to be solved. The 31mp sensor employs microlens technology. A tiny light gathering lens designed to increase the sensitivity of the pixel by funneling light into it, enabling higher ISO ratings, caps each pixel. The microlens exaggerates the usual falloff and color shift between the center and edges of the chip. This is worse on wide-angle lenses and with view camera movements. To date, only Hasselblad has successfully managed the microlens sensor in a professional level product, the digital back they make [with Imacon] for the Leica R8. While the specification for the chip stated that this behavior would occur with wide-angle lenses, only, Hasselblad discovered that it occurred with all lenses. To correct these aberrations, Hasselblad mapped out the characteristics of each lens in the Leica R8 system at every aperture and light level and included this information in the firmware of the [Leica DMR] back. The camera would send the lens and exposure data to the back, which would in turn correct the image. In other words, the corrections had to be integrated at the system level. Hasselblad has already added the capability for this type of lens to camera to back communication in its latest digital platform, the H2. Previous strategies of mechanically adapting backs to fit different cameras may no longer provide an optimal solution with these higher technology sensors.

...

Thank you all for reading this and I look forward to your considered responses.

John O'Donnell Technical Sales & Training - Digital Imaging Systems
Professional Graphics Systems & Services, Inc. 3 West Main Street
Elmsford, NY 10701 (914) 345-3033 x34 cel (914) 224-8642

john@prographics.com
----------------------------

On Sep 30, 2005, at 8:22 AM, John O'Donnell wrote: > > I’d like to shed some light on the subject of the latest new 31 and 39 > megapixel sensor arrays.

Hello all I wanted to add to John's excellent post on the new chips.

It's worth noting that the H2 also passes focus data to the back because in order to know how to correct CA [chromatic abberation] fall off etc you need to know focal length (i.e. on the zoom) aperture AND focus distance. I think you are right in saying that [digital back maker] Phase will have a lot of problems introducing the 31MP [digital back] due to the microlenses and they do not have access to the lens data as Hasselblad backs do. Also as you rightly point out the time spent with the Leica has helped them problem solve.

The H2 launch has been very quiet and I think many people have been a bit under whelmed by the changes but as the new chips come on line they will really see the benefits of the H2 (getting the lens data) and the new DSP [digital signal processor] in the backs. Regards

Nick Tresidder Photography http://www.nick-t.com 649 379 3778 6421 972
111
-------------------------------

regards, George”

Source: http://leica-users.org/v30/msg11701.html (Oct 5, 2005)


Leica R lenses can transmit focal length data (through their ROM chip) and diaphragm data to the Leica R8/R9, but not the focusing distance. While the new 6-bit coding will allow the M8 to identify a lens's focal length, these new M lenses will still not be able to transmit diaphragm data. However, the focusing distance could be obtained through the Leica M's RF focusing mechanism, which could potentially be electronically captured by the Leica M8 (a bit complicated, but it could be done).

To be quite honest, we do not have any idea if this text from the LUG is true or not (and if this optical optimization feature is applied to both JPEG and RAW/DNG files). As I wrote above, I am still attempting to get a confirmation from Hasselblad Sweden. It's possible that Hasselblad wants to keep its "Digital APO Correction" exclusively to itself (hence the silence from Leica on this possible "undocumented feature"). If I do find anything, I will let you all know.

Best regards,

John F.

P.S. For more info on the Hassy "DAC", goto http://www.hasselblad.se/ and click on the link "H2D-39"

[Benchista]

There's similar specualtion on the LUG, inevitably.

There is also the software solution - that's what Nikon will be providing with Nikon Capture NX, and is already available with DxO, of course.

Curved sensors would presumably only work if all the lenses to be used produced similar field curvature, so I don't think that's the solution.

We've debated this all before some time back - can't remember exactly when, though.