What do they do when they "recalibrate" a lens ?

Started Feb 1, 2006 | Discussions
Cedric Simon Senior Member • Posts: 1,325
What do they do when they "recalibrate" a lens ?

Hello,

I am having front focus problems with my 28mm 2.8 and 350D.
I still have to check if I have the same problem with my other lenses,
but as they have smaller apertures, that is not so easy to check, and the
problem is not annoying anyway if there is any.

With my 28mm 2.8, it is clear that the focus is on the nose when I focus on the eye. I did tons of tests, with books, rulers, real subject, same conclusion: constant front focus.

My question is: I have read (especially with Sigmas) that people send their lens for calibration: what does it mean ? What can be calibrated in the lens ? I see how my 350D can be corrected: you can adjust the autofocus.

But a lens ?

Also I have problems with my 10-22, it looks soft when taking photos of landscapes. We already had a debate here (diffraction or not, focus ?) and the conclusions are not clear.

Can a lens be "calibrated" so that its performance will be better even using manual focus ?

 Cedric Simon's gear list:Cedric Simon's gear list
Olympus OM-D E-M10 Panasonic Lumix G 14mm F2.5 ASPH Olympus M.Zuiko Digital ED 40-150mm F4-5.6 R Olympus M.Zuiko Digital 45mm F1.8 Samyang 7.5mm F3.5 Fisheye
Olga Johnson Forum Pro • Posts: 24,360
I don't know about Sigma but Canon...
U Sererra Regular Member • Posts: 317
Re: I don't know about Sigma but Canon...

"First, the camera body is calibrated with a 'tool' lens, a standard lens known to be operating correctly according to design specifications."

Sounds cool - manufacturers should consider selling that kind of lens. Im sure there'd be a market for them.

Doug Kerr Forum Pro • Posts: 20,898
The Canon AF system

Hi, Cedric,

Cedric Simon wrote:

My question is: I have read (especially with Sigmas) that people
send their lens for calibration: what does it mean ? What can be
calibrated in the lens ?

In the Canon AF system, the camera determines, with the lens focus wherever is happens to be when AF is done, how far off the focus is (the "focus offset"), in terms of a normalized "geometric" measure, rather like measuring, with a split image prism focusng aid, precisely how far in mm the two images are misaligned on the focuing screen.

This value is fed to the lens. The microprocessor in the lens, based on that value, together with the current position of the lens focusing cam (from a digital positon encoder), and using an algorithm that relates subject distance, focusing cam position, and the corresponding focus offset in the camera, calculates what new focsuing cam position should produce ideal focus. It then directs a servomechanism to move the focsuing cam to that calculated position.

And that's the end of the story - if that position is "properly computed", we get proper focus; otherwise, we don't. The camera doesn't "take a second look" at the focus offset and, if focus is not yet ideal, give the lens further instructions as to how to refine it.

It is an "open loop" system, rather like computing where a howitzer should be pointed so that the shot fired will land on the target, based on where a test shot, fired with the gun pointed wherever it happened to be pointed, lands.

The algorithm is defined by tables in the lens' memory, which have been particularlized to this individual copy of the lens during manufacture (so that they accurately relate to the particular tolerances in the focsung cam mechnism, the exact behavior of the lens elements, and so forth).

If for whataver reason the values in those tables are not (any longer) correct for this particular lens, then the AF system will not set the focsuing cam to the proper position, even though the camera correctly reports the amount of "focus offset".

"Recalibrating" the lens involves updating the entries on those tables to produce proper behavior.

Best regards,

Doug

 Doug Kerr's gear list:Doug Kerr's gear list
Leica V-Lux 4
stinch Regular Member • Posts: 130
Re: What do they do when they "recalibrate" a lens ?

I dont know this for a fact however after some thought:-

The camara does the focusing, not the lens, the lens is dump with a motor and some electronics to drive the motor.

I understand that the Canon camera does the focusing with the contrast method which looks at the frequency content of the image at the focus sensors, the higher the frequency content the more infocus the picture is. The camera drives the lens to maximise the frequency content.

Addtionally if you consider the camera there are three image plains one for the image sensor, one for ground glass screen (viewfinder) and one for the focus sensors. the last two must be in perfect alignment with the first. As perfect is not possible, there will be an acceptable error band associated with the cameras focussing algorithm.

My guess is that the calibration is a combination of re-alignining the three plains and minimising the error band

all that can be done with the lens is adjust the rate at which the motor moves the lens, all this will do is result in an under or overshoot and cause the time to focus to increase and also hunting if overshooting. in theory the camera should alwasy arive at the focus point with its acceptable error given enough light and contrast.

the fact that you can manually focus just indicates that the ground glass screen is better aligned then the focus point sensors.

Any variation between lenses probaly results from their wide open depth of field and the first point at which the camera decides the focus is inside the error band.

This is my guess, If any one knows better then I would like learn more

-- hide signature --
OP Cedric Simon Senior Member • Posts: 1,325
Re: The Canon AF system

Hello Doug,
thanks for your detailed answer.

I must say I am amazed (I am not the only what) to read that the camera does not "take a second look"... So even if the focus ring is at an extreme position (say, infinity, and you want to focus on a close object), the camera does only one measure and send only one order to the lens ?

I would have thought that if the camera took several measures, then focus would have been way more reliable...

Thanks for your explanation !

Doug Kerr wrote:
Hi, Cedric,

Cedric Simon wrote:

My question is: I have read (especially with Sigmas) that people
send their lens for calibration: what does it mean ? What can be
calibrated in the lens ?

In the Canon AF system, the camera determines, with the lens focus
wherever is happens to be when AF is done, how far off the focus is
(the "focus offset"), in terms of a normalized "geometric" measure,
rather like measuring, with a split image prism focusng aid,
precisely how far in mm the two images are misaligned on the
focuing screen.

This value is fed to the lens. The microprocessor in the lens,
based on that value, together with the current position of the lens
focusing cam (from a digital positon encoder), and using an
algorithm that relates subject distance, focusing cam position, and
the corresponding focus offset in the camera, calculates what new
focsuing cam position should produce ideal focus. It then directs a
servomechanism to move the focsuing cam to that calculated position.

And that's the end of the story - if that position is "properly
computed", we get proper focus; otherwise, we don't. The camera
doesn't "take a second look" at the focus offset and, if focus is
not yet ideal, give the lens further instructions as to how to
refine it.

It is an "open loop" system, rather like computing where a howitzer
should be pointed so that the shot fired will land on the target,
based on where a test shot, fired with the gun pointed wherever it
happened to be pointed, lands.

The algorithm is defined by tables in the lens' memory, which have
been particularlized to this individual copy of the lens during
manufacture (so that they accurately relate to the particular
tolerances in the focsung cam mechnism, the exact behavior of the
lens elements, and so forth).

If for whataver reason the values in those tables are not (any
longer) correct for this particular lens, then the AF system will
not set the focsuing cam to the proper position, even though the
camera correctly reports the amount of "focus offset".

"Recalibrating" the lens involves updating the entries on those
tables to produce proper behavior.

Best regards,

Doug

 Cedric Simon's gear list:Cedric Simon's gear list
Olympus OM-D E-M10 Panasonic Lumix G 14mm F2.5 ASPH Olympus M.Zuiko Digital ED 40-150mm F4-5.6 R Olympus M.Zuiko Digital 45mm F1.8 Samyang 7.5mm F3.5 Fisheye
UziY Contributing Member • Posts: 736
A second look

You can force the camera to take a second look by releasing the shutter button and re-pressing it.
--
Uzi
http://www.pbase.com/uyoeli

Doug Kerr Forum Pro • Posts: 20,898
Re: The Canon AF system

Hi, Cedric,

Cedric Simon wrote:

Hello Doug,
thanks for your detailed answer.

I must say I am amazed (I am not the only what) to read that the
camera does not "take a second look"... So even if the focus ring
is at an extreme position (say, infinity, and you want to focus on
a close object), the camera does only one measure and send only one
order to the lens ?

Indeed. The assumption is that the "gun layer" has perfect tables as to the behavior of theh howitzer!

I would have thought that if the camera took several measures, then
focus would have been way more reliable...

To have it take a second look would slow down the process. An advantage of teh Canon approach is its potential for very fast competion of the AF task.

Best regards,

Doug

 Doug Kerr's gear list:Doug Kerr's gear list
Leica V-Lux 4
OP Cedric Simon Senior Member • Posts: 1,325
Re: The Canon AF system

Hi again Doug,

I still have a question then... why do the lens move back and forth before stopping to the right position ?

In low light conditions, for example when the focus can not be made, then I am sure the camera take several looks... Or am I missing something ?

Cedric Simon wrote:

Hello Doug,
thanks for your detailed answer.

I must say I am amazed (I am not the only what) to read that the
camera does not "take a second look"... So even if the focus ring
is at an extreme position (say, infinity, and you want to focus on
a close object), the camera does only one measure and send only one
order to the lens ?

 Cedric Simon's gear list:Cedric Simon's gear list
Olympus OM-D E-M10 Panasonic Lumix G 14mm F2.5 ASPH Olympus M.Zuiko Digital ED 40-150mm F4-5.6 R Olympus M.Zuiko Digital 45mm F1.8 Samyang 7.5mm F3.5 Fisheye
Arjen van Andel Contributing Member • Posts: 594
The Canon AF system software

Hi Doug,

But if there's software involved for the calibration, we could do this ourselfs. How come we never hear of these CD's while I can buy pretty much any piece of film, music or software that's ever been produced, pirated or otherwise?

Read the above as: "I need this software badly"

Arjen.

Doug Kerr Forum Pro • Posts: 20,898
Indeed

And if the process has worked perfectly, there will be no movement of the focsuing cam at that time.

Best regards,

Doug

 Doug Kerr's gear list:Doug Kerr's gear list
Leica V-Lux 4
Arjen van Andel Contributing Member • Posts: 594
And...

Doug Kerr wrote:

To have it take a second look would slow down the process. An
advantage of teh Canon approach is its potential for very fast
competion of the AF task.

And it's needed for the predictive AF; the camera computes, based on measurements on the AF sensors, the speed of the subject and commands the lens to ... be out of focus in time.

Arjen.

dbuhanan Contributing Member • Posts: 875
To Doug Kerr

Excellent description Doug! Best I have read on the Canon AF system.
--
kindest regards
Dale

It is possible to make a small fortune in photography, you just have to start with a large fortune.

Arjen van Andel Contributing Member • Posts: 594
Split, not contrast

stinch wrote:

I understand that the Canon camera does the focusing with the
contrast method which looks at the frequency content of the image
at the focus sensors, the higher the frequency content the more
infocus the picture is. The camera drives the lens to maximise the
frequency content.

AFAIK the Canons uses tiny split-screens with line-array sensors for each focus point. Both sensors need to se the same pattern to have perfect focus. The central point has two versions (steepness of the prism), one for lenses f/2.8

Arjen

Doug Kerr Forum Pro • Posts: 20,898
Re: The Canon AF system

Hi,, Cedric,

Cedric Simon wrote:

Hi again Doug,

I still have a question then... why do the lens move back and forth
before stopping to the right position ?

There are two answers to that, pertaining to differnt situations.

1. It may be that when AF is taken, there is not a clear answer from the AF detectors as to the focus offset with the lens set where it happens to be. (This is like when you use a split image prism and are aimed at a guy with plaid pants).

In that case, the camera has the lens "scan" while the camera looks, not for the point of correct focus, but for a point in which an unambiguous indication of focus offset is attained by the AF detector(s). Then the process proceeds as originally described.

2. After the processor in the lens computes the needed positon of the focsuing cam, it orders the servomechanism to move the cam to that position. This is done in different ways in different kinds of lens, but it is itself a "closed loop" subprocess - that is, the servomechanism doesn't relax until the focsuing cam is in the position the computer ordained.

With some focusing servomechanism systems (especailly those using a DC motor), to speed up the process, the motor is driven at a high speed, which then means that when the servo feedback loop stops it, it may have overshot the intended destination (its actual position being noted then from the focus cam position encoder). Then the servo system eases it in.

None of this has to do with a further look by the camera at the state of focus. It is part of the "execution" phase, rather than the "planning" phase, of the overall process.

Best regards,

Doug

 Doug Kerr's gear list:Doug Kerr's gear list
Leica V-Lux 4
Doug Kerr Forum Pro • Posts: 20,898
Thank you so much. (nt)
n/t
 Doug Kerr's gear list:Doug Kerr's gear list
Leica V-Lux 4
Doug Kerr Forum Pro • Posts: 20,898
Re: What do they do when they "recalibrate" a lens ?

Hi, s,

stinch wrote:

I dont know this for a fact however after some thought:-

The camara does the focusing, not the lens, the lens is dump with
a motor and some electronics to drive the motor.

I understand that the Canon camera does the focusing with the
contrast method which looks at the frequency content of the image
at the focus sensors, the higher the frequency content the more
infocus the picture is. The camera drives the lens to maximise the
frequency content.

Not so. The camera uses a "phase compoarison" system, in which AF detectors use a principle identical to that of a split-image prism focsuing aid.

The results of that determination are tehn used as I described earlier in this thread.

Addtionally if you consider the camera there are three image plains
one for the image sensor, one for ground glass screen (viewfinder)
and one for the focus sensors. the last two must be in perfect
alignment with the first. As perfect is not possible, there will be
an acceptable error band associated with the cameras focussing
algorithm.

The relationship between the 1st and 3rd planes is relevant tro AF accuracy. However, they need not be physically consistent, as compensation for discrepancy can be made in the process of evaluating the offset between the two images of the pair.

In fact the image plane is adjusted to a standard position with respect to the mount flange with selected shim washers. The plane of the AF detectors is not adjusted at all.

Any variation between lenses probaly results from their wide open
depth of field and the first point at which the camera decides the
focus is inside the error band.

I suggest you read my earlier post in this thread for a description of how the Canon AF system operates and how the situation of an individual lens comes into the deal.

Best regards,

Doug

 Doug Kerr's gear list:Doug Kerr's gear list
Leica V-Lux 4
OP Cedric Simon Senior Member • Posts: 1,325
Re: The Canon AF system

Thanks very much,

your explanation is very clear, and this is the first time I really understand what is going on !

Doug Kerr wrote:
Hi,, Cedric,

Cedric Simon wrote:

Hi again Doug,

I still have a question then... why do the lens move back and forth
before stopping to the right position ?

There are two answers to that, pertaining to differnt situations.

1. It may be that when AF is taken, there is not a clear answer
from the AF detectors as to the focus offset with the lens set
where it happens to be. (This is like when you use a split image
prism and are aimed at a guy with plaid pants).

In that case, the camera has the lens "scan" while the camera
looks, not for the point of correct focus, but for a point in which
an unambiguous indication of focus offset is attained by the AF
detector(s). Then the process proceeds as originally described.

2. After the processor in the lens computes the needed positon of
the focsuing cam, it orders the servomechanism to move the cam to
that position. This is done in different ways in different kinds of
lens, but it is itself a "closed loop" subprocess - that is, the
servomechanism doesn't relax until the focsuing cam is in the
position the computer ordained.

With some focusing servomechanism systems (especailly those using a
DC motor), to speed up the process, the motor is driven at a high
speed, which then means that when the servo feedback loop stops it,
it may have overshot the intended destination (its actual position
being noted then from the focus cam position encoder). Then the
servo system eases it in.

None of this has to do with a further look by the camera at the
state of focus. It is part of the "execution" phase, rather than
the "planning" phase, of the overall process.

Best regards,

Doug

 Cedric Simon's gear list:Cedric Simon's gear list
Olympus OM-D E-M10 Panasonic Lumix G 14mm F2.5 ASPH Olympus M.Zuiko Digital ED 40-150mm F4-5.6 R Olympus M.Zuiko Digital 45mm F1.8 Samyang 7.5mm F3.5 Fisheye
Doug Kerr Forum Pro • Posts: 20,898
Excellent point (nt)
n/t
 Doug Kerr's gear list:Doug Kerr's gear list
Leica V-Lux 4
Arjen van Andel Contributing Member • Posts: 594
Recalibrate AF sensors

Doug Kerr wrote:
The plane of the AF detectors is not adjusted at all.

Do you mean the only calibration on those is done by software? There are no shims or whashers involved? No banging on the table to get them into place? At all?

Where is that option in the menu, then?

Arjen.

Keyboard shortcuts:
FForum MMy threads