How to adjust the position of the sensor in a DSLR camera

New! Here is the online calculation: http://camerarepair.hu/cmos

And here You find, how to use it:

In DSLR cameras probably the most important basic adjustment is the geometrical position of the CMOS sensor related to the lens mount. The CMOS itself is fixed on a base plate, and in the front of the sensor there is a special filter.

The lens mount will determine the optical axis (it is perpendicular on it’s plane). In optimal case this will be the same with the optical axis of the lens. To the lens mount plane will be reported the axial distance of the CMOS too. If the body and the lens are both well adjusted each of them, the whole system will work fine together.
But if one of them is a little bit misaligned, and the other one too….

This is, why we have to bring both of them to the service for a 100% adjustment.

A fix lens (and in a given zoom-position a zoom-lens too) theoretically could be replaced with a so called thin-lens.

This means, that if we have a 50 mm fix lens, the image of on object at the infinite distance will be perfectly sharp when this thin-lens will be distanced to the sensor at exactly it’s focal distance, which in this case is 50 mm.

By all Canon DSLR-s the CMOS is located at exactly 44 mm from the lens mount. (Not the base plate – the optical surface)

Which means, that in this particular case the theoretical thin lens is located outside of the body. Case of a wide lens it would be in the inside, as it’s focal distance is less then 44 mm.

From this point it starts to be interesting: an engineer can never use the word exactly. Rather You have to give a nominal value and the tolerance field. In our case Canon gives this as 44.00 +/- 0.03 mm. Attention: from the surface of the sensor, not the surface of the filter. Which means, it is not possible to measure it directly.

The body/lens position is very important – there is enough a little loose lens in the mount, and the image will be out of focus. And some people have the habit to walk with 2 kg lens hanged on it’s mount. And this can force in short time the springs in the mount.…

The lens mount (the bayonet) is fixed on the outer side of the mirrorbox, the sensor (or, better to say, the baseplate) is fixed on the other side. The distance of the bayonet and the sensor can be adjusted (to fulfilled in the prescribed tolerances). In reality the measurable distance between the real surfaces is not 44.00 mm. For different Canon models can be different, for example 43.10 mm, because in this distance there is calculated the thickness of the sensor etc.

This size is measurable. (About the measuring instrument later).

As the distance is important not only in one point, but the whole plane must be perpendicular to the optical axis, the support is made on 3 points – which, as we know, will define a plane.

The sensor (cmos) is not necessarily parallel to the base plate, and the 3 plastic supports are not 100% at the same level from the bayonet. With measuring we can determine the difference, and using the right washers we can adjust the sensor to the accepable paralelity and distance.

Flange to focal plane distance (FFD) adjustment

In the majority of the bodies (5d, 10d, 20d, 30d, 40d, 300d, 350d, 450d …) the adjusting is made simply with the help of different washers. If we know the real distance of each fixing point to the flange, the factory given sensor offset value (in some cases written on the sensor), we can calculate the needed washer thickness for each fixing point, to reach the prescribed 43,10 mm-t.

So during the adjustment You have to measure the distance, calculate the washers thickness, put them together, and that’s it.

But there are bodies (eg. the 500d), where between the body and the sensor base plate there are 3 springs, and the base plate is pressed with the help of 3 screws against the springs toward the body – but the plate will not touch the body!. In this case You cant measure anything (there is a possibility of course – but not in the official technology).

In the factory (I suppose) they don’t need any special measuring, as they can adjust the best value using the live view function of this camera, following continuously the precision of the focusing through a master-lens.

For us there is only one small problem. When the camera is ready to make live view pictures, all the screws are covered…(in the case of the 500d sure)…

 

The good news is, that if we have a suitable measuring bench, there is enough to measure the base-plate distance to the flange from behind, and after repair, we can adjust the same distance. Very nice, but what to do, when the original was not perfect, or, if the body was disassembled before any measuring?

For a simple check with a photo You have to assemble the half of he body. Then disassemble, assemble, disassemble…

What is the best motivation to find a solution for a problem? If You have no other choice. I was staring on the disassembled 500d body – and there were no measured values before. In fact, it was no measuring bench at all.

A little bit of practical knowlegde

In the camera they have to bring together 3 optical systems.
The lens, the cmos, the AF (autofocus) system, and the matte screen.

In the factory I suppose the adjust at first the cmos to the nominal 44 mm, and very precisely parallel reported to the lens mount.

When a picture projected on the cmos is exactly in focus, then (for the same relative position of the camera) the following should be fulfilled too:
– With the main mirror in lowered position the image on the matte screen will be in focus (for this reason the matt screen is adjustable with some washers)
– The light going through the main mirror and reflected on the secondary mirror will give on the AF sensor zero defocus amount (or defocus between the acceptable limits)
The AF sensor has a mechanical adjustment possibility, but the adjusting parts are glued, and the fine adjustment is made in software).

The AF sensor in reality looks other, then in my image, and works in another way, but from our point of view act exactly like a second matte screen, “watched” by the electronics. 

In the following adjustment we will use a well adjusted lens with a big aperture, a “master lens”. In my case it was an EF 50 F1,4 USM. The big aperture is important, because in this case we will get a thin DOF, so we can observe if the camera is out of focus.

The measuring have to be done on a bench, on which is possible to move the body perpendicular on the table with the object. I used for this an old russian (soviet !!!) enlarger. As object to camera distance I used 500 mm, because in his case was possible to use efficiently the fild of view, and because the narrow point of this lens was less then 500 mm.
By the 50 mm lens the at 500 mm object distance the picture distance is 55,55 mm, so for beginning I adjusted the CMOS mark on the camera (this mark is a circle cutted by a line), this is the total distance on the figure.

If the AF system is perfectly adjusted (AF sensor not moved, secondary mirror not moved), but the CMOS is in the wrong position, then using the AF system for finding the right focus, the camera will project the sharp picture in a wrong position.
(the picture in the optical viewfinder will be sharp – case that it was not misadjusted too).

In this position of the camera we have to make one AF focusing on the target, and after this the lens should be set on manual mode (M).

Another important thing: we have to measure very precisely the distance between the object and the plane of the CMOS. It is not so difficult, as the CMOS position is marked on the body, here it’s enough a rough value, and later we measure the difference of distances – with a much higher precision.
At first with a measuring tape, then with a digital calliper – very comfortable for relative measuring.

let we see, what happens, when the CMOS is closer to the bayonet, and not parallel with it.

The result is deplorable: no point of the picture will be in focus (or, in some cases, the intersection line of the 2 planes will include an infinite number of sharp points. …).

If we use the refraction rule of the lens: 1/f = 1/k + 1/t (where f is the focal length, k is the picture distance, t is the object distance), we can see that the picture distance is the smallest (in fact equal with the focal distance), when the object is at infinity, in any other case will be bigger.

So, if we have a bad adjusted CMOS / bayonet distance, there are 2 ways to get sharp image (as a first step in only one point, for example in the middle of the vievfinder)
1.  we let the camera/object distance unchanged, and we modify the lens/camera distance (we increase it)
2.  we let the lens/camera distance unchanged, and we modify the camera/object distance (we set a larger distance)

In our case the correct way will be the 2., as the correct lens / camera distance was set up by the AF system (we suposed, that the AF-system works perfect). We know of course, that the AF-system is precise only in a certain tolerance field, because it’s not the same for example, if we start from infinite, or from the nearest point. This error could be excluded by making two focusings, from infinite and from near point, and we make an average of the two results.)

Let we start the work with only the middle point.

How muh shall we adjust on the distance of the camera? We switch on liveview (the 500d has it), and we have to move up and down the camera, till we find the best sharpness (using the 10x magnifier function).
For this case I used a visitcard, printed in good quality.

After this with the digital calliper we can measure the offset value (eg. on the rod of our stand) – at the moment only in difference of the object/camera distance.

As we would measuring in the other points of the field of view, we would see, that by oblique CMOS the offset values are different. Unfortunately the liveview makes not possible to see the whole surface of the CMOS. So let we use the visible corners.

From these offset values we can calculate the offset of the real CMOS plane from the theoretical plane, using the same formula. This values will be in microns.

If we could recalculate from this points the offset values in the 3 fixing points of the base plate, that points are measurable, and are adjustable, so we only have to adjust the screws with the opposite of the offset value!

A little bit of mathematics

In the mathematics a plane could be defined by 3 different points).
But I can choose any 3 points (which are not on the same line).

The formula of the plane is: Ax + By + Cz + D = 0, which will be fulfilled by the coordinates of each point on it.
When we already have the A, B, C, D coefficients, then knowing it’s x and y coordinates we can calculate the z coordinate for any point.
The z will be in our case the distance offset of the CMOS.

But if we measure and calculate with the help of Liveview the coordinates for the 4 corners and the middle point too, then we have more then necessary number of points for our plane…

For this problem we can use the method of regression. With this method we can calculate the coefficients of a plane, which could be fitted with the smallest error to the measuring points (which were calculated inaccurate, of course).
Our case is a particular one, because we measured in the corners of the liveview image, which are symmetrical to the middle.

A during photographing normally the middle point should be the most accurate by focusing, I decided to calculate (and to adjust) the angular error of the cmos using the corner points, and axial offset from the bayonet plane using the middle point.

At first I made a coordinate-system on cmos, with the origo in the middle of the cmos – this was the most simple method (the measuring points are symmetrical reported to it)

The  0/1/2/3/4 points are the measuring points, their position will be given reported to the origo (H and V are the distances which depend from the liveview system of the camera – the 4 most distant points which can be observed.
The bigger the distance, the less the error.

The A/B/C points are arbitrarily choosen points closed to the adjustment screws, on the base plane. This should be marked with a marker pen (Canon has forgotten to do this…)
Their coordinate should be measured with a digital calliper.

The measuring bench

 

Above the original Canon device.
In this case the camera bayonet is mounted on a lens mount – which is on a movable (sliding) flange, which moves on the very precise table of the instrument. The gauge is a special one, with a very long measuring range.
This device would be much too expensive to build.

 

My version. The flange is made of a very good quality (polished) steel plate, which has a fixed position on the base. During the measuring the camera will slide on the flange – so it has to be pressed on the surface during the measuring.

As a long range gauge is very expensive, I use a gauge with 12,5 mm displacement, 0,001 mm resolution and 0,003 mm precision (ca. 50 EUR) .
For the zero-setting of the gauge I use a 40.00 +/- 0,01 mm thick brick, made of HSS steel (steel for making tools).

Or you can use an original etalon set, like the following one – but that is really very expensive.

In this way 40,00 mm must be added to the value shown on the display.

If somebody does not recognised, the main parts are: a brake disk from a passenger car (made of casted iron, heavy and stabile base), the rods of an old magnetic stand for a gauge, and the steel plate of the transfer pump of a DFP1 type Delphi common rail pump (this was the most parallel polished steel plate, which was available – as normally I work as an engineer in the automotive industry).
And fnally a beautiful (and cheap) chinese gauge with 0,001 resolution.
The measuring tip is not the original, as it must be long and slim, so I made it from a M2,5 screw on a grinding machine. 

 

The most important is, that the camera is layed with the bayonet on a steel plate, which is hard, plane and glossy, so this will be the base surface during the measuring.

To reach all of the measuring points the camera should be slided on my bench, while on the original the camera is fixed on a stand, and the stand is slided on the bench.
My method was much cheaper to realize, and the possibility of errors is reduced. The gauge holder has to be very stiff.

As the gauge can measure max 12,5 mm, at first I have to put the 40 mm brick on the stand, then press the zero set button..

Here You can see the “real” position of the cmos reported to the “theoretical” position.

The calculation

The calculation has been done in Excel, and the regression plane was calculated at first with the help of a math site on the interet. As my case is particular, I have now a simplified formula made with the help of my classmate, who studied mathematics. (Maths was not my favorite…)

The procedure

0. I assemble the camera, cmos should be approx in the right position, measuring points are marked, positions are measured with the calliper. I put the EF 50 mm F1,4 lens on the camera.

1. I put the 5 charts on the table (5 visit cards).

2. The camera is setted to 500 + 55 = 555 mm.
The camera MUST be very precisely perpendicular to the table!

3. I search the focus point in the middle with the help of the AF-system. After this I change to manual focus on the lens.

4. I make the zeroing on the calliper for the actual position of the camera/stand

5. I switch the camera in Liveview

6. I determine in Liveview for each of the 5 points the sharpest picture (aperture at F=1,4!!!), I write down the offset values reported to the original position (the 555 mm).

7. I repeat the last step many times (it takes a few minutes)

8. I calculate the average values for each offset.

9. With the help of the Excel I calculate the offset values for the adjustment points “z”

10. I disassemble the body.

11. On the measuring bench I adjust the screws, I assemble the body again.

12. Back to point2. If there is no noticeable difference, I can have a rest. If not, I repeat the whole thing again.

 

 

The first real test of this adjustment worked well, the body has been disassembled only once!

 

(any text, figure, photo from this article is made by me
– excepting the Canon stand photo, which is from the internet –
case You use it somewhere I appreciate, if you mention the source )
© canonrepair