Focus Calibrations for AF (D)SLRs and Lenses
RiceHigh
Here is the most lengthy techical article I've ever written in my Blog!
http://ricehigh.blogspot.com/2007/01/focus-calibrations-for-pentax-dslr.html
.. for anyone who is interested or even suffered from those daily
reported Front Focusing or Back Focusing issues!
Enjoy and I hope that you folks will find this to be useful!
RiceHigh
http://ricehigh.blogspot.com
David Kilpatrick
> Here is the most lengthy techical article I've ever written in my Blog!
>
> http://ricehigh.blogspot.com/2007/01/focus-calibrations-for-pentax-dslr.html
>
> .. for anyone who is interested or even suffered from those daily
> reported Front Focusing or Back Focusing issues!
>
> Enjoy and I hope that you folks will find this to be useful!
>
The same thing with three hex screws also applies to the Konica
Minolta/Sony models. They also have a PC-interface for programming. The
cameras are set up with the AF within 5-10 per cent tolerance of being
correctly positioned and straight, then the program is run with a 50mm
f1.7 lens set to approx f8 and focused not at infinity, but 1.9x meters
(it's an exact distance, and both the special lens and the bench are
part of the test rig).
The PC program then tests all 5/7/9 or whatever sensors are used in the
particular model against targets on a chart, repeatedly stepping the
lens. It loads a large table of offsets into the camera, telling it to
make micro-adjustments to final focus AFTER the AF module has locked
focus. Each sensor in a multi-sensor array has its own offset.
Generally, the 50mm f1.7 set up works fine for longer lenses and those
with limited apertures, like most zooms, and for closer focusing
distances. But it will fall down with an 85mm f1.4, a 28mm f2, a 17-35mm
wide angle used at max f2.8 and 17mm, a Sigma f2.8 18-50mm, the 11-18mm
zoom - etc. The figures which apply to a 50mm lens at f8 and 2 metres
are not appropriate for a 50mm f1.4 at 30 metres, either.
The danger with doing your own BF/FF adjustment (I have done this on one
camera - 2/3rd of a turn clockwise on all three adjustments) is that you
can't rewrite the programmed offsets. You can not even access this area
of memory and zero all the offsets. A perfect camera would have zero
offsets, and a perfectly aligned and positioned AF module.
It's further complicated by the adjustments for the primary and
secondary mirror (both again are considered OK at the factory within a
set tolerance - call it 5 per cent, but it's actually a measurement
unique to the test gear) and the location of the sensor. This is never
exactly where film would be, since it floats on a PCB which lacks the
precision of a film gate. Remove the back of the camera, and you find
THREE MORE screws, which physically alter the focus plane, pitch and yaw
of the sensor. In some cameras the rotation of the sensor (tilted
horizon syndrome) has a further adjustment. This set of adjustments
can't be done at home.
But, of course, the sensor again has a tolerance. It's quite big. Since
the mirror system and the AF setup are linked to the actual position of
the sensor, it leaves the focusing screen out on its own, and this is
the only element firmly connected to the lens mount flange. To
compensate for the errors produced in the rest of the setup, the
focusing screen is shimmed. The factory assembly staff normally put one
standard shim between the focus screen and the information LCD screen,
which is transparent and part of the prism/finder assembly. Up to three
shims may be fitted to adjust the screen position, and the requirement
is given to the assembly operator by the diagnostic program. This
completes the 'positive feedback' cycle of cumulative errors which makes
up a typical AF SLR collimation and AF offset programming cycle.
For the user, a heavily shimmed screen is a drawback. Most of adjust the
dioptric correction of our DSLRs using the LCD screen - focusing marks -
as a reference. With no shims, this is in perfect contact with the
focusing side of the focus screen. This itself is no longer an efficient
focusing screen, by old groundglass standards, as 30 to 70 per cent of
the light (depends on the system) is being diverted for the AF system
via semi-silvered main mirror. To restore brightness, the makers use
clever 'acute matte', 'spherical matte' and other types of
semi-transparent prismatic instead of true 'groundglass' texture. These
are midway between a clear screen (as found on the Contaflex, Retina
Reflex and other early SLRs) and a groundglass. They do not allow
accurate focusing because the eye can partially refocus through the
screen - and they do not allow depth of field preview to work correctly,
though many cameras have d-o-f preview. You can confirm this by shooting
at f2.8 and comparing what you see through the finder with the true
depth of field you see. Checking it on the review screen is enough. The
difference is huge.
This final part of the setup - the shimmed, nearly transparent viewing
screen separated physically from the visual reference point of the focus
marks or LCD generated grid (etc) - completes the inaccuracy and
systemic unfitness for purpose of the average modern DSLR for precise
visual (non-AF) focusing.
As a rule, you are better off ignoring your eyes and letting the AF rip.
Much back focus, front focus is caused by users thinking the AF has 'not
quite hit the right spot' and using the 'focusing' screen (better called
a viewfinder screen, as you really can't focus on it except for shots
where depth of focus is huge, like macro work - don't confuse with depth
of field, depth of focus is at the film plane). They may not manually
focus, but they force a refocus while moving the AF target a bit to
'see' the right result and then recompose a tad. Switching your eyes and
brain off, and relying on the kludges incorporated in your camera to
overcome minimal care with real physical collimation of sensor and AF
module to lens flange/screen, you may get better results.
And adjust the AF for BF/FF at your own risk - you are not able to
remove those adjustments programmed into the camera, and they may have
different effects on different lens. An older lens chip, or a difference
in gearing/AF motor between indepedent and marque lenses, may cause
specific models to continue to misfocus consistently.
David
achilleas...@yahoo.co.uk
David Kilpatrick wrote:
> RiceHigh wrote:
> > Here is the most lengthy techical article I've ever written in my Blog!
> >
> > http://ricehigh.blogspot.com/2007/01/focus-calibrations-for-pentax-dslr.html
> >
> > .. for anyone who is interested or even suffered from those daily
> > reported Front Focusing or Back Focusing issues!
> >
> > Enjoy and I hope that you folks will find this to be useful!
> >
>
>
> The same thing with three hex screws also applies to the Konica
> Minolta/Sony models. They also have a PC-interface for programming. The
> cameras are set up with the AF within 5-10 per cent tolerance of being
> correctly positioned and straight, then the program is run with a 50mm
> f1.7 lens set to approx f8 and focused not at infinity, but 1.9x meters
> (it's an exact distance, and both the special lens and the bench are
> part of the test rig).
>
> The PC program then tests all 5/7/9 or whatever sensors are used in the
> particular model against targets on a chart, repeatedly stepping the
> lens. It loads a large table of offsets into the camera, telling it to
> make micro-adjustments to final focus AFTER the AF module has locked
> focus. Each sensor in a multi-sensor array has its own offset.
Hi. This sounds very reasonable, and you've said it before (and
obviously some sort of fine-tuning to either match all the distances
involved or compensate for discrepancies will be needed). So, my
question is: if this is how it is done (and I have no reason to doubt
what you say, if you have a reliable source-eg a technician who does
this), how is it then possible for the AF subsystem to indicate correct
focus with MF lenses? It certainly does, so...Maybe these offsets can
be used to decide how far away from in-focus the subject should be, but
then, if you think about how phase-detection AF works, this could only
be done with info on which lens is mounted (ie focal length and max
aperture), while AF indication works without this. So...
David Kilpatrick
>
How do you know it does? Try it at full aperture. It's as likely to be
luck as anything else. I have used manual focus lenses and I can make
the image look sharp by changing the dioptre adjustment on any of my
KM/Sony bodies, when it's not exactly focused. The focus light may
confirm focus (done this at portrait distances with a Pentax K100D,
which offers the function, using a macro Pentax 100mm set to manual) and
the result can be hopelessly out (I lost some excellent shots).
There are algorithms in there which must relate the test-bench figures
with the test lenses so that they more or less work with a range of
lenses. They clearly can't do anything to allow for manual lenses.
However - I can focus my cameras just about OK on a telescope or
microscope, or a manual macro bellows. Here the balance of depth of
focus is such that errors in CCD positioning, relative to the focus
screen, are negligible. But I can't get an AF confirmation, since not
one of my bodies (KM, Sony) offers AF confirmation of manually focused
lenses such as T-mount stuff. So the question of whether AF confirmation
would be given at the right point is academic - I can only say that it
definitely was NOT with the K100D, 100mm f2.8 macro at 2.8, working
distance about 3-5 feet. If was as far out as AF itself, which was also
poor (backfocus of about six inches at 4 feet).
Maybe this is why Sony/KM disabled focus confirmation with manual mount
lenses. Since it would not work, perhaps they ensure it can't be used.
David
achilleas...@yahoo.co.uk
> achilleas...@yahoo.co.uk wrote:
> >
> > Hi. This sounds very reasonable, and you've said it before (and
> > obviously some sort of fine-tuning to either match all the distances
> > involved or compensate for discrepancies will be needed). So, my
> > question is: if this is how it is done (and I have no reason to doubt
> > what you say, if you have a reliable source-eg a technician who does
> > this), how is it then possible for the AF subsystem to indicate correct
> > focus with MF lenses? It certainly does, so...Maybe these offsets can
> > be used to decide how far away from in-focus the subject should be, but
> > then, if you think about how phase-detection AF works, this could only
> > be done with info on which lens is mounted (ie focal length and max
> > aperture), while AF indication works without this. So...
>
> >
> How do you know it does? Try it at full aperture. It's as likely to be
> luck as anything else. I have used manual focus lenses and I can make
> the image look sharp by changing the dioptre adjustment on any of my
> KM/Sony bodies, when it's not exactly focused. The focus light may
> confirm focus (done this at portrait distances with a Pentax K100D,
> which offers the function, using a macro Pentax 100mm set to manual) and
> the result can be hopelessly out (I lost some excellent shots).
>
Hi. Credit me with the minimum intelligence needed to have checked
this. In short, I've used the focus confirmation indicator with 50mm
f/1.2 lenses, 90mm f/2.8, 50mm f/1.8, 180mm f/2.8 (but only for a few
shots) and a few other similar lenses over some years, and have had no
problems at all. All this on a D200, a Canon 20D and also a 7xi (mostly
on that). And I have asked (today) people who use only MF lenses on
their cameras (one has a F100, the other a F90x); neither of them have
had a problem. I accept that they may simply not have enlarged their
slides enough to see it in that case. But really, I think 1:1 macros at
f/5.6 (due to the lens extending) with the 90mm macro I have would have
showed up any such problems.
> There are algorithms in there which must relate the test-bench figures
> with the test lenses so that they more or less work with a range of
> lenses. They clearly can't do anything to allow for manual lenses.
Yes, that's what I say: it can't work for manual focusing. And yet it
works, hence my question.
>
> However - I can focus my cameras just about OK on a telescope or
> microscope, or a manual macro bellows. Here the balance of depth of
> focus is such that errors in CCD positioning, relative to the focus
> screen, are negligible. But I can't get an AF confirmation, since not
> one of my bodies (KM, Sony) offers AF confirmation of manually focused
> lenses such as T-mount stuff. So the question of whether AF confirmation
> would be given at the right point is academic - I can only say that it
> definitely was NOT with the K100D, 100mm f2.8 macro at 2.8, working
> distance about 3-5 feet. If was as far out as AF itself, which was also
> poor (backfocus of about six inches at 4 feet).
Well, but if the AF itself is out, so would the indicator, no?
Anyway, it is most certainly not academic: I use it all the time. Even
with AF lenses. For example, for macro, can you use AF? I can't, it
hunts too much. So I focus manually, and since my eyesight is bad, I
use the AF indicator. I don't think I am the only person who uses
manual focus for macro. I've never had the camera confirm focus when it
wasn't focused. And the depth of field very small, so I'd have noticed,
no? And I also do it with my non-AFS lenses when I don't want the
screw-drive AF to make noise. No problems, but there it would not have
been so obvious.
>
> Maybe this is why Sony/KM disabled focus confirmation with manual mount
> lenses. Since it would not work, perhaps they ensure it can't be used.
>
Well, maybe, but I have never heard anybody say that they were getting
consistently out of focus images when manual focusing using the
in-focus indicator, while having no problems with AF. And what of all
the people who use manual focusing for macro work? I'm sure many have
noticed that the in-focus indicators switch on for a small range of
distances, but this is tolerances in the system rather than error.
So maybe someone can chime in: Has anybody noticed that the in-focus
indicator gives incorrect results while AF works fine? Maybe someone
else who uses macro lenses can check this. I don't believe I have
managed to get the only cameras who don't do this.
David Kilpatrick
>
> Hi. Credit me with the minimum intelligence needed to have checked
> this. In short, I've used the focus confirmation indicator with 50mm
> f/1.2 lenses, 90mm f/2.8, 50mm f/1.8, 180mm f/2.8 (but only for a few
> shots) and a few other similar lenses over some years, and have had no
> problems at all. All this on a D200, a Canon 20D and also a 7xi (mostly
> on that). And I have asked (today) people who use only MF lenses on
> their cameras (one has a F100, the other a F90x); neither of them have
> had a problem. I accept that they may simply not have enlarged their
> slides enough to see it in that case. But really, I think 1:1 macros at
> f/5.6 (due to the lens extending) with the 90mm macro I have would have
> showed up any such problems.
>
I don't doubt your intelligence, but experience of specifics is another
matter: at 1:1 macro, the depth of focus is equal to the depth of field
(strictly applies only to simple or symmetrical lenses) and substantial
movements of the film plane have little visible effect. A 1mm error in
CCD plane or focusing screen placement in the camera will only result in
a reciprocal 1mm error in focus plane when shooting at 1:1 macro.
A good book on this is Sidney Ray's OPTICS. The best information tends
to be in German, from Sinar and particularly from the college courses
published by (Swiss) Photographie Verlag in the 1980s.
>
>>There are algorithms in there which must relate the test-bench figures
>>with the test lenses so that they more or less work with a range of
>>lenses. They clearly can't do anything to allow for manual lenses.
>
>
> Yes, that's what I say: it can't work for manual focusing. And yet it
> works, hence my question.
Well, for most people AF works, and those who do not find an FF/BF error
with any lenses (no matter what) will have a camera body which happens
to be very accurate, and incorporate minimal or no offsets programmed
in. Those who encounter FF/BF the moment they fit the usual offenders -
50mm f1.4, ultrawides and fast wides, 85mm f1.4 etc - probably have a
body which will not give accurate manual focus-confirmation AF.
>
> Anyway, it is most certainly not academic: I use it all the time. Even
> with AF lenses. For example, for macro, can you use AF? I can't, it
> hunts too much. So I focus manually, and since my eyesight is bad, I
> use the AF indicator. I don't think I am the only person who uses
> manual focus for macro. I've never had the camera confirm focus when it
> wasn't focused. And the depth of field very small, so I'd have noticed,
> no? And I also do it with my non-AFS lenses when I don't want the
> screw-drive AF to make noise. No problems, but there it would not have
> been so obvious.
See below - it is not depth of field which matters. It's the other side
of the lens, depth of focus, and the two are reciprocal.
>
> Well, maybe, but I have never heard anybody say that they were getting
> consistently out of focus images when manual focusing using the
> in-focus indicator, while having no problems with AF. And what of all
> the people who use manual focusing for macro work? I'm sure many have
> noticed that the in-focus indicators switch on for a small range of
> distances, but this is tolerances in the system rather than error.
I have not implied that. If the AF sensors are dead accurate, then
focus-aided manual will be. The point I make here is that you can not
rely on modern focusing screens, unless you have gone back to the maker
and (for example) had a Minolta Type ML fitted to permit accurate visual
focusing at the expense of 1.5 stops of finder brightness. I find
Nikon's screens are amongst the best, by the way, but even they still
offer conventional (not enhanced prismatic engraving) versions for
critical manual focus work.
>
> So maybe someone can chime in: Has anybody noticed that the in-focus
> indicator gives incorrect results while AF works fine? Maybe someone
> else who uses macro lenses can check this. I don't believe I have
> managed to get the only cameras who don't do this.
>
You need to make the tests at or near infinity. It is pecularity of
errors in collimation that they are most noticeable at infinity, and
least noticeable at macro range; when you go to magnifications, such as
4:1, the lens to subject distance becomes the critical factor and you
can 'focus' the film plane without seeing any change.
This is something you learn when using monorail cameras for studio work,
which I did for many years in formats from 35mm to 8 x 6, mainly 6 x 9cm
and 4 x 5 inches. On 4 x 5, a surprising number of studio shots are
taken at or around 1:1 - even an egg in an eggcup is 1:1 and most small
ornaments, tins of food, perfumes, etc are 'macro' work by 35mm standards.
This is why your AF indicator appears to have a big tolerance when
focusing at 1:1 - there actually IS tolerance at 1:1.
If you ever bought one of those early 'zoom slide copiers' which were a
tube with a lens in it that could do 1X to 1.2X, they did not have a
zoom lens, they just relied on a simple formula which allows the lens to
be moved so far when the film plane and subject plane are fixed, and
stay sufficiently in focus at f16.
David
achilleas...@yahoo.co.uk
David Kilpatrick wrote:
> achilleas...@yahoo.co.uk wrote:
>
> >
> > Hi. Credit me with the minimum intelligence needed to have checked
> > this. In short, I've used the focus confirmation indicator with 50mm
> > f/1.2 lenses, 90mm f/2.8, 50mm f/1.8, 180mm f/2.8 (but only for a few
> > shots) and a few other similar lenses over some years, and have had no
> > problems at all. All this on a D200, a Canon 20D and also a 7xi (mostly
> > on that). And I have asked (today) people who use only MF lenses on
> > their cameras (one has a F100, the other a F90x); neither of them have
> > had a problem. I accept that they may simply not have enlarged their
> > slides enough to see it in that case. But really, I think 1:1 macros at
> > f/5.6 (due to the lens extending) with the 90mm macro I have would have
> > showed up any such problems.
> >
>
> I don't doubt your intelligence, but experience of specifics is another
> matter: at 1:1 macro, the depth of focus is equal to the depth of field
> (strictly applies only to simple or symmetrical lenses) and substantial
> movements of the film plane have little visible effect. A 1mm error in
> CCD plane or focusing screen placement in the camera will only result in
> a reciprocal 1mm error in focus plane when shooting at 1:1 macro.
David,
Sorry to give that impression, it was an attempt at humour (in
retrospect, not very succesful). Anyway, you're right, at 1:1 the depth
of field and depth of focus are equal, so huge compared to the
tolerances under discussion. I had not thought of that. So my macro
example was bad.
>
> A good book on this is Sidney Ray's OPTICS. The best information tends
> to be in German, from Sinar and particularly from the college courses
> published by (Swiss) Photographie Verlag in the 1980s.
> >
I don't think one needs to go too deep into geometrical optics for
these things. Anyway, I suppose this information is available in any
book on optics, if one is interested to look for it.
> >>There are algorithms in there which must relate the test-bench figures
> >>with the test lenses so that they more or less work with a range of
> >>lenses. They clearly can't do anything to allow for manual lenses.
> >
> >
> > Yes, that's what I say: it can't work for manual focusing. And yet it
> > works, hence my question.
>
> Well, for most people AF works, and those who do not find an FF/BF error
> with any lenses (no matter what) will have a camera body which happens
> to be very accurate, and incorporate minimal or no offsets programmed
> in. Those who encounter FF/BF the moment they fit the usual offenders -
> 50mm f1.4, ultrawides and fast wides, 85mm f1.4 etc - probably have a
> body which will not give accurate manual focus-confirmation AF.
Right. But the way you described it, it misfocuses and then knows how
to correct by moving the lens a specific distance. But it can't move
the lens for MF, so it shouldn't work then, should it? It could work if
a stored table of offsets is used to work out how misfocused it should
appear to the AF sensors when it is in fact correctly focused for the
imaging sensor, and the system only indicated correct focus then. But
this would depend on the sensor being able to accurately judge just how
misfocused we currently are, and even so, it would need to know focal
length and max aperture at least, I think. I have no idea if the
sensors can measure misfocus so accurately, and the AF indicator on my
camera works reliably even if the focal length and aperture are not
entered.
Anyway. My point is that if it works the way you say, it could be very
accurate for AF but inaccurate for MF. I have not seen nor heard of
this happening. Of course, this does not prove that it does not happen.
I agree, you can't focus reliably with today's screens (although
personally I can't focus with any screen without split prisms etc, so I
am not a good judge).
> >
> > So maybe someone can chime in: Has anybody noticed that the in-focus
> > indicator gives incorrect results while AF works fine? Maybe someone
> > else who uses macro lenses can check this. I don't believe I have
> > managed to get the only cameras who don't do this.
> >
> You need to make the tests at or near infinity. It is pecularity of
> errors in collimation that they are most noticeable at infinity, and
> least noticeable at macro range; when you go to magnifications, such as
> 4:1, the lens to subject distance becomes the critical factor and you
> can 'focus' the film plane without seeing any change.
Well, OK, as you said, it is the depth of focus that matters, and it
seems to me that it should not be too sensitive to focal length,
distance etc (for large distances). But I didn't check, so maybe I am
wrong. Anyway, you're right, the effects we're discussing will indeed
be more visible for longer distances.
Anyway, I don't believe that I, and the people I asked, all happen to
have been very lucky. Maybe we are, though, I don't know. I would have
thought that the system compensates for distance tolerances some other
way, not the one you describe. In fact, I'd have thought it compensates
so that AF and MF with focus indication would be exactly equally
accurate, as this seems to be the case (for me, at least). But, well,
if you know this not to be the case, then I believe you.
I think I'll try to find a way to test this.
>
> This is something you learn when using monorail cameras for studio work,
> which I did for many years in formats from 35mm to 8 x 6, mainly 6 x 9cm
> and 4 x 5 inches. On 4 x 5, a surprising number of studio shots are
> taken at or around 1:1 - even an egg in an eggcup is 1:1 and most small
> ornaments, tins of food, perfumes, etc are 'macro' work by 35mm standards.
>
> This is why your AF indicator appears to have a big tolerance when
> focusing at 1:1 - there actually IS tolerance at 1:1.
Yes, good point.
RiceHigh
interesting and it's always good to know more about these details.
I think I shall point to your this post as a follow-up to my article
(in the "comments" section), so as to let my readers to know more as
well as to be aware of the possible consequences of doing such
calibrations.
Anyway, I will not do any of those jobs by myself at home as it's too
risky anyway. (and I did not intend to hint the end-users to do these
neither, but just that they should know what they could ask the Pentax
service centres to do for them if they encounter obvious FF or BF
errors.)
RiceHigh
http://ricehigh.blogspot.com
Bo-Ming Tong
> Here is the most lengthy techical article I've ever written in my Blog!
>
>
> .. for anyone who is interested or even suffered from those daily
> reported Front Focusing or Back Focusing issues!
>
> Enjoy and I hope that you folks will find this to be useful!
>
> RiceHighhttp://ricehigh.blogspot.com
When I developed the lens mount conversion from Contax N to Canon,
I added an AF adjustment feature to the lens. You may store a
different adjustment value to each lens - it is remembered in the
lens'
non-volatile memory. Therefore, the adjustment works on all camera
bodies, not just the 1D mark III.
http://support.conurus.com/viewtopic.php?t=56
--
Bo-Ming Tong
conurus - lens mount conversion with autofocus and auto aperture