Canon poor sensor alignment and non-level horizons.

RichieRich

Suspended / Banned
Messages
223
Edit My Images
No
Hi there,

I thought that I'd just make this post to communicate some of my findings with respect sensor alignment on the 450D and 50D models (the only two Canon SLRs which I've owned).

Soon after purchasing my first 450D I noticed that most shots were slightly misaligned despite careful efforts to keep the horizon level during the shooting process. After reading similar reports online I got Amazon to exchange the body. Sure enough the new 450D exhibited the same problem but it wasn't as severe as that on the first model so I learned to live with it.

A couple of years down the line I purchased a 50D only to notice the same thing - this time the severity of the effect was somewhere between that of the two 450Ds which I'd owned. Given that I'd purchased it in Japan I learned to deal with it and simply got used to applying a manual offset to whatever my eyes or the spirit level was telling me.

Only on cleaning the sensor of my 50D recently did I notice the obvious cause - the sensor itself is slightly misaligned relative to the body of the camera. Although the distances are very small it's quite clear on my model that the black space between the top right of the sensor and the body is smaller than that on the left hand side.

Similarly the bottom right has a larger gap than the bottom left, which is consistent with a slight twist when the sensor was positioned. (Once again, we're talking about very small distances here, but the angle is tilt is similar to what I've observed when examining shots).

So, to anybody who thinks that poor horizons on Canon models may be down to their own skill, or perhaps a dodgy spirt level shoe, then it may be worth thinking again. According to my tests I'm absolutely convinced that this is a problem with many Canon SLRs which manage to make it past Canon's QC checks. I've learnt to deal with it, but it's disappointing considering the price of these units.

Yes, I realise that it's easy enough to straighten horizons during post-production, but it shouldn't be necessary if shots have taken with care, plus, there is a price to be paid when this is done in terms of the quality of the final resized output image.
 
hmm, that might explain a few of my images. I was beginning to think I was standing at an angle. Will have to scientifically test mine, unfortunately it is out of warranty now.
 
It's worth remembering that in most cases it is probably down to user error. When sensors are aligned there will always be a margin of error but in Canon's case I suspect that either their specified tolerance or their QC procedures are lacking (with the latter being the most likely).

Even if you do confirm the problem I'd recommend not getting too obsessed with it. Unfortunately I was guilty of this when I first realised that the problem also exhibited itself on my 50D.

When doing tripod-based work I simply apply my offset and then do one or two test shots to calibrate if necessary.
 
Well, I thought it was user error. But the camera has a view-finder grid, which I had tried to line up with the horizon on several occasions when it was important to me to get it vertical/horizontal. (Most of the time, I don't care too much, I like wild-life best)
The trouble is, I usually notice the problem when I am back at the computer, so can never be certain that the shot which is bad is the one that I spent effort trying to get it correct.
I agree about not worrying about it. It would be nice to know though that it is not ME that it leaning by that amount!
 
Last edited:
This is interesting. I've had something of a problem with my horizons ever since I switched from my old 350D to a 40D. I expect it's probably user error with a camera which is slightly too big for my hands, so I'm planning to get a focussing screen with a grid. But this does raise another possibility...
 
Might be worth getting a Seculine digital level and fine adjusting it to get the horizons level. Probably easier to practise in PS (other image manipulation software exists) and living with the necessary crops.
 
hmmmmm..... looks like Canon should have spent less time making photocopiers and printers and focussed on the job at hand...;)

I'll have to check our works Canons for this problem. Mind you, whatever the findings, it probably won't make a blind bit of difference to the shots produced by some our journalists!! :)
 
Last edited:
Might be worth getting a Seculine digital level and fine adjusting it to get the horizons level. Probably easier to practise in PS (other image manipulation software exists) and living with the necessary crops.

It's not really cropping which is an issue, it's more a question of image quality. Althogh the effect is quite subtle, I've noticed that images which have been straightened are slightly softer than ones which haven't been following a final resize. I tested this with a couple of simple workflows on the same image - one of which applied a slight straightening, and the other which didn't.

I guess it's because the image has already been interpolated once when it comes to a final resize in the case of a straightened image. Hence why I always try to avoid straighting images in post production.
 
You can always use live view to align it.

I've just tested my 40D, lined it up perfectly on a straight line in live view, then flick to viewfinder and as far as I can tell it's perfectly lined up.
 
Seems most unlikely to me. User error or lens distortion? Can you post a pic?
 
Seems most unlikely to me. User error or lens distortion? Can you post a pic?

Well, I have a 50D now with a sensor which is clearly misaligned, and as I said, I had two 450Ds which exhibited the same problem - the first one being absolutely awful. I even got friends to confirm what I was seeing. In order to take a horizon shot the body of the camera would have to be positioned at an angle which wasn't level in order to compensate for the misalignment. We're talking a tilt of about 5 degrees, perhaps more.

It's definitely not lens distortion either. When using a spirt level I can see that the horizon is perfectly in line with the AF points when checking through the viewfinder. It's only when the picture is taken that I see the misalignment.

It's a shame that I sold my last 450D as I could have used it with my macro lens to take a picture of the 50D's sensor. It's quite obviously the source of the problem.

Once you accept the fact that a sensor will never be positioned perfectly in line with a camera's body does it really take such a leap of faith to believe that the misalignment may be enough to cause a problem?
 
Last edited:
Thinking it through, rather than the sensor being misaligned, I would have suspected or eliminated the following first:

  • The Viewfinder on the 50D is 0.95x magnification and the same ~95% coverage, so the misalignment may be within this tolerance?
  • The focussing screen may have been changed for one of a different thickness or may not be correctly seated in its holder?
  • The focussing mirror may be misaligned?

Can you please take some photos of say the Brickwork pattern on the side of the house, aligning the camera both horizontally and vertically to the wall, using a tripod, turning off IS etc :

-firstly align using the "viewfinder AF points" as you have replied in the thread
-and then again using "Live View using the Grid" to align on the brickwork.

This will hopefully point us in the right direction as to where the issue is.

I can understand and its not limited to Canon, that manufacturing tolerances are prevalent and this can present minor issues as you have described.

Look forward to seeing your results.

Regards
 
I can't see it being any of those reasons lostsoulal2...for a start a 95% coverage wouldn't affect alignment as the 95% which is visible would still be representative of the orientation.

I've done brickwork tests etc before. I guess I could repeat in accordance with your suggestions if I get the time.

The fact remains that I can see that the sensor is misaligned - it's suffering a slight anti-clockwise rotation. This is the cause. I imagine that if I'd checked the sensor on my first 450D it would have been even more offset.

Anyway, I know what the cause is on my 50D. As to whether the 450Ds also had misaligned censors, I'll never know, but I'm pretty confident that this was the case.
 
I used to have this problem with various Canon cameras and now the problem is rare. It's not that the cameras have changed in any way, but I have improved my shutter release technique and no longer end up slightly skewing the camera at the moment I press (squeeze!) the shutter button.

A tripod and remote release should eliminate the human error factor. Hand holding, with the best will in the world, will not.
 
I've never heard of this issue with any Canon gear before. Certainly none of the Canon's I've ever owned suffered from this.

The other thing - I dont know if its just me, but I cant see the actual sensor behind the hi-pass filter on my 50D?
 
Last edited:
I used to have this problem with various Canon cameras and now the problem is rare. It's not that the cameras have changed in any way, but I have improved my shutter release technique and no longer end up slightly skewing the camera at the moment I press (squeeze!) the shutter button.

A tripod and remote release should eliminate the human error factor. Hand holding, with the best will in the world, will not.

Naturally I wouldn't make such claims unless I was testing with a tripod and a remote release.
 
I cant see the actual sensor behind the hi-pass filter on my 50D?
...and if the filter is out of alignment it won't affect the image alignment.

The OP does seem to be extraordinarily unlucky with his gear. Not one, not two, but THREE DSLRs with misaligned sensors; and then a lens which won't focus.
 
After reading this I've looked into the guts of my 50D and I can tell you, you cant see the sensor behind the filter!! As per Stuarts post above, its highly unlikely you would find ONE body with a mis-aligned sensor, let alone three!

The sensor itself has a structure within the chassis that it fits onto, its unlikely that this would be misaligned as the factory would have had to produced an incorrect mould/chassis master and there would be millions of bodies out there with dodgy sensor alignments and Im sure a recall would have been put in place. What you are looking at is the hi-pass filter, a larger plane of glass in front of the (smaller) sensor behind it. As long as it covers the sensor, its doesn't matter how its aligned.

Sorry, the more I think about this, the more I think its down to user error.
 
Last edited:
After reading this I've looked into the guts of my 50D and I can tell you, you cant see the sensor behind the filter!! As per Stuarts post above, its highly unlikely you would find ONE body with a mis-aligned sensor, let alone 3!

The sensor itself has a structure within the chassis that it fits onto, its unlikely that this would be misaligned as the factory would have had to produced an incorrect mould/chassis master and there would be millions of bodies out there with dodgy sensor alignments and Im sure a recall would have been put in place. What you are looking at is the hi-pass filter, a larger plane of glass in front of the (smaller) sensor behind it. As long as it covers the sensor, its doesn't matter how its aligned.

Sorry, the more I think about this, the more I think its down to user error.

Thanks odd jim. You've educated me there with respect to the filter being placed in front of the sensor. I'll re-check later (but I'm more concerned about my dodgy new Tokina at the moment).

Still, I'm convinced that it's not user error. I had this to varying degrees with the 450Ds and friends confirmed it when taking test shots, and the second 450D was far better than the first.

Also, spirit levels mounted on cameras pointed at the horizon tend not to lie, particularly when using a tripod and a remote release.

Also, when aligning via the viewfinder and via live view I get different results. If I align via liveview using the grid I get a perfectly aligned shot. However, look through the viewfinder and the horizon is slightly misaligned.

Anyway, I've said my piece on this subject. I'm sure that in the majority of cases user error is the problem, but there are exceptions, and I think that I've simply got unlucky in this respect. Back to the problem with my Tokina...
 
Yes, I think its definitely a hardware fault with the Tokina I have to say!

If you can (once you've sorted your lens issue out!) post some test shots illustrating the body issue you have.

...not with the Tokina though!

A bit of research indicates that there has been, in the past, viewfinder misalignments with some older 5D bodies. Not sure how the technical aspects work on this, but it may explain your issue. That would be much more likely than sensor alignment issues, but I still think you would be very unlucky to have this issue with three bodies.
 
Last edited:
Thanks odd jim. You've educated me there with respect to the filter being placed in front of the sensor. I'll re-check later (but I'm more concerned about my dodgy new Tokina at the moment).

Still, I'm convinced that it's not user error. I had this to varying degrees with the 450Ds and friends confirmed it when taking test shots, and the second 450D was far better than the first.

Also, spirit levels mounted on cameras pointed at the horizon tend not to lie, particularly when using a tripod and a remote release.

Also, when aligning via the viewfinder and via live view I get different results. If I align via liveview using the grid I get a perfectly aligned shot. However, look through the viewfinder and the horizon is slightly misaligned.

Anyway, I've said my piece on this subject. I'm sure that in the majority of cases user error is the problem, but there are exceptions, and I think that I've simply got unlucky in this respect. Back to the problem with my Tokina...

Rich, find a long, straight, horizontal line that fills the width of the viewfinder - roof of building, window frame, line between wall and ceiling, that sort of thing. Tripod so there is no danger of rotational movement on releasing the shutter.

Line this up with exactly with the extreme edge of the viewfinder - right up as close to the edge as you can get. Forget using the AF points, spirit level etc. Just line up the edge of the subject with the edge of the viewfinder and take a snap. At the same time, you can also confirm that the viewfinder and live view images are the same. And check out your lens distortion :eek:

With respect, consider the evidence. Nobody else has this problem, but you have found it in no less than three consecutive cameras... :thinking: And when it has been reported by others, it turns out that people are using the AF points as you are, which are a poor guide and badly positioned for this kind of thing, or stabbing at the release, etc.
 
Also, when aligning via the viewfinder and via live view I get different results. If I align via liveview using the grid I get a perfectly aligned shot. However, look through the viewfinder and the horizon is slightly misaligned.

That does tend to indicate its an issue with the viewfinder / focussing mirror alignment - the viewfinder is not 100% accurate, I found this when I migrated from a Sony digital camera with their equivalent of "Live View" to a Canon 20D that what I saw through the viewfinder was not what I got in my shots.................

- it was only apparent for me as I tend for my subjects to fill the frame as much as possible whereas most photographer's (correctly) let the subject breathe - after much discussion with Canon techies, they agreed it was the focussing mirror alignment, if I recall correctly they said they concentrated more on the vertical positioning and a slight horizontal misalignment would have accounted why I got more space to the left on my shots than what I saw through the viewfinder. But as they said, it was within their manufacturing tolerances.

As I have later migrated through various cameras I now use Live view to accurately align my shots - the other factor when looking at this issue I had to consider is that my eye may not have been positioned correctly onto the viewfinder......................
 
That does tend to indicate its an issue with the viewfinder / focussing mirror alignment - the viewfinder is not 100% accurate, I found this when I migrated from a Sony digital camera with their equivalent of "Live View" to a Canon 20D that what I saw through the viewfinder was not what I got in my shots.................

- it was only apparent for me as I tend for my subjects to fill the frame as much as possible whereas most photographer's (correctly) let the subject breathe - after much discussion with Canon techies, they agreed it was the focussing mirror alignment, if I recall correctly they said they concentrated more on the vertical positioning and a slight horizontal misalignment would have accounted why I got more space to the left on my shots than what I saw through the viewfinder. But as they said, it was within their manufacturing tolerances.

As I have later migrated through various cameras I now use Live view to accurately align my shots - the other factor when looking at this issue I had to consider is that my eye may not have been positioned correctly onto the viewfinder......................

This sounds like a different issue. The viewfinder usually only shows about 95% of the total view recorded by the sensor, and on some cameras (I've heard this about some Sony's) the viewfinder is not perfectly centred for reasons of design/manufacture. So you get fractionally more on one side or the other, compared to the sensor. Live view should be 100% accurate.
 
That does tend to indicate its an issue with the viewfinder / focussing mirror alignment - the viewfinder is not 100% accurate, I found this when I migrated from a Sony digital camera with their equivalent of "Live View" to a Canon 20D that what I saw through the viewfinder was not what I got in my shots.................

- it was only apparent for me as I tend for my subjects to fill the frame as much as possible whereas most photographer's (correctly) let the subject breathe - after much discussion with Canon techies, they agreed it was the focussing mirror alignment, if I recall correctly they said they concentrated more on the vertical positioning and a slight horizontal misalignment would have accounted why I got more space to the left on my shots than what I saw through the viewfinder. But as they said, it was within their manufacturing tolerances.

As I have later migrated through various cameras I now use Live view to accurately align my shots - the other factor when looking at this issue I had to consider is that my eye may not have been positioned correctly onto the viewfinder......................

Well, it could indicate that the sensor, from which the liveview is being generated, may not be correctly aligned with the body. With my first 450D if I placed the camera on one end of a table, perpendicular to the edge and focused on the opposite edge it would be at an angle of about 5 degrees. This was just one of many tests which I did, all of which produced similar results.

I actually suspect that many people have experienced the same problem but they've put it down to user error. I did extensive testing which confirmed without a doubt that this wasn't the case.
 
This sounds like a different issue. The viewfinder usually only shows about 95% of the total view recorded by the sensor, and on some cameras (I've heard this about some Sony's) the viewfinder is not perfectly centred for reasons of design/manufacture. So you get fractionally more on one side or the other, compared to the sensor. Live view should be 100% accurate.

Correct, although in this case it was the Canon 20D where the viewfinder was not perfectly centred at manufacture............
 
It's far more likely that the viewfinder mechanism is out of alignment, rather than the sensor.

That's especially the case if you can take a picture correctly via live view. 5º is a huge amount and you would certainly notice the tilt in the body required to get a level horizon.

It also begs the question 'why didn't you return the camera to Canon for correction immediately, having discovered the fault?' This was a recognised error in some 3DMkIII and should be repaired free of charge even out of warranty, as a 5º misalignment makes the camera not fit for purpose.
 
It's far more likely that the viewfinder mechanism is out of alignment, rather than the sensor.

That's especially the case if you can take a picture correctly via live view. 5º is a huge amount and you would certainly notice the tilt in the body required to get a level horizon.

It also begs the question 'why didn't you return the camera to Canon for correction immediately, having discovered the fault?' This was a recognised error in some 3DMkIII and should be repaired free of charge even out of warranty, as a 5º misalignment makes the camera not fit for purpose.

I did. Well, I returned it to Amazon after having owned it for a couple of weeks - it was the 450D which was about 5% out. And you're right, the body had to be visibly on the p!!! in order to take a level image.

With my current 50D the problem is far less pronounced so I just learned to deal with it. I purchased it in Japan and I couldn't be bothered to go though the exchange process.
 
It's far more likely that the viewfinder mechanism is out of alignment, rather than the sensor.

That's especially the case if you can take a picture correctly via live view. 5º is a huge amount and you would certainly notice the tilt in the body required to get a level horizon.

Who's to say which is out? Fact is the viewfinder and sensor aren't aligned the same.

Solution? Use LiveView... if possible :)

For tripod-type landscapes I usually do that anyway as I find it easier to review.

For non tripod type shots I usually try to shoot a smidge wider than I want so that I have rotation room afterwards. Even if you think it's fine in the viewfinder, once you've applied distortion correction or PP'ed it for a while you might feel the need to spin it one way or the other....
 
Who's to say which is out? Fact is the viewfinder and sensor aren't aligned the same.

I was basing my argument on previously documented and recognised problems with Canon bodies. What are you basing yours on, apart from pure conjecture?
 
ive had this problem for along time, i kept thinking it was my head that was not level
i have this problem with a 350d a 30d and a 5d, handheld
and on a tripod with a hotshoe spirit level i still sometimes get this problem
strange
 
Back
Top