« He Wasn't as Steady as a Tripod | Main | Three Days Left! »

Sunday, 31 January 2010

Comments

That's some smart thinking, Ctein. Your approach fits the way a lot of photographers think and work. It's a step better than the simpler implementations of AF fine-tune for DSLRs (including Nikon's, sadly) that stores a single setting per lens, when the correction needed for decent wide-open performance often seems to vary according to subject distance.

Huh!
I had wondered about the focus errors on my XA. That explains a lot. I've been toying with doing more with MF - Ctein, how well does a TLR do with staying in adjustment for focus? I had assumed that with fewer finicky cams, etc, that one would stay properly tuned in focus longer than a rangefinder.

Ctein,

That solution sounds awfully like the ideas I remember being talked about around what Contax needed to do to retain autofocus compatibility with its old lenses. Doing that would also need to increase the size of the camera slightly front to back. This is probably acceptable in a big-honkin-SLR... but not in a Leica rangefinder.

If they wanted to keep the rangefinder concept, why not use technology that Leica already own? Like a laser based rangefinder on the camera which could be calibrated by lens?

... but I'm digressing. At some point you have to wonder if all of this over-engineering is side-stepping the fact that you're still dealing with a product that's just steeped in tradition and just too damn old. And perhaps a product built on requirements that mean that it's perfectly acceptable.

What do I mean by this? In most shooting situations pre-focusing and shooting with slight variances in focus accuracy is generally acceptable (for rangefinder loved focal lengths of 28-50mm). Likewise auto-exposure and modern advances accurate metering systems solve a problem which don't exist for a typical rangefinder: black and white film is pretty forgiving as of 50 years ago with TRI-X -- being perfect is what's required here -- what's required is just being CLOSE.

But perhaps being close is not good enough. *grin* I guess that's one of the beautiful things about the equipment in this game -- there is always next year...

Pak

Great idea Ctein,
It just needs an achronym;
Dynamic Auto Focus Tuning (DAFT)!
bd

The high precision focussing that is demanded by todays digitale camera's has a definitive solution, that is to use the sensor information itself to guide the autofocus.
All compact camera's do it, and the micro four thirds system does it, and the new Samsung camera will use it. I see it as the focussing system of the future.
All we need is even better EVF screens and maybe a more clever way to assist in manual focussing via one of those screens.

bd,
No worse than EVIL.

Mike

Pak Ming Wan wrote--Like a laser based rangefinder on the camera which could be calibrated by lens?

Linhof had something like this on the Technika 2000 for a while. Bob Salomon, the U.S. Linhof rep, describes why it failed here--

http://www.largeformatphotography.info/forum/showthread.php?p=232658

But that doesn't mean it wouldn't work today, with a more powerful miniaturized computer system.

Linhof's traditional mechanical rangefinder cams should be accurate over their entire range, presuming the rangefinder is in good adjustment, because the cams are ground individually for each lens and can be fine tuned in this regard. When I had one lens cammed once, and it was accurate at middle and far distances but inaccurate at near distances, Martin Arndt, Linhof's U.S. authorized service technician, said to send it back, and it was perfect when I got it back.

Of course it's a different kind of thing from a modern 35mm rangefinder, since the rangefinder and viewfinder windows are separate, and that can introduce other difficulties.

Thanks for some insight into the nature of lenses! The talk about focusing reminds me that I rarely use apertures larger than 2.8. Apertures of 2.0 and 1.4 simply produce too many pictures out of focus.

The solution is far more simple: CMOS, Live View and an accessory electronic viewfinder on the hot shoe. You may have all the focus accuracy you need, when you need it: macro lenses, teles, superfast lenses.

The traditional viewfinder/rangefinder, which may be improved, is good only for "normal" circunstances: reportage photography with wide-angles or short teles.

DIdn't Contax make an AF body that shifted the film plane to focus the camera?

I've been thinking that an extension of something like the Contax AX's system where the film plane would move 10mm to focus with a manual focus lens would be a good idea.

Get to within f/5.6 DOF with current passive phase detection techniques and then fine focus with contrast detection. That way you would get the speed of phase detection and the accuracy of contrast detection.

My recent experience with the Canon 5D2 using live view has really impressed on me how bad the canon phase detection autofocus is even when it is calibrated and how good (if slow) the live view contrast detection can be.

I like this idea, however; if you are still focussing while the focus assist is moving the sensor, I think that the two methods of focussing might interact badly. The assist can be chasing the focus point as you over correct. It might take longer to get the distance right.

It would be better if the sensor did not move and all adjustment was done with the lens. This system would also be simpler and cheaper to build, and while the control chip lived it would just work and could not go out of adjustment.

The M9 (for example) could be made to do this without major retooling, assuming that there is room for the control chip somewhere.

Ctein,
great post. Perhaps the companies listen to "big guys" in photography like you and Michael. But I doubt it, because so many people still buy these faulty AF systems, despite all the known problems. I learned my lesson the hard way by believing the companies' advertising before I realized that their answer "within tolerance" is nonsense...

Ctein, what do you think about the MicroFourThirds approach with focusing right on the sensor and using an EVF? Maybe this is the method for the future, because it's cheaper and easier to build than your suggestion...? I don't know. I'd like to see a camera with 35mm full format sensor that has fast contrast detect AF and a decent EVF. What about this? The most accurate method of focusing on the right spot should be focusing through the sensor itself! (Or am I missing something...?)

Best regards
Leo

While I don't doubt the the technical deficiencies of current systems, nor the feasibility of improvement, I do wonder how in the world photographers using rangefinders and slrs ever made a decent image. No offense...I'm just sayin'. And, yes, I know that digital brought forth problems perhaps not apparent with film.

Now, if one could come up with a way to avoid car tune-ups and alignments, that would grab my attention.

Surely this is a convincing argument for the EVIL concept? Why move the sensor if by measuring the live image you can eliminate calibration and tolerance errors?

Of course two challenges are still present - the quality of the viewfinder image and focus speed.

The quality solution is simple. Reengineer the traditional rangefinder VF with a zoom, auto-parallax adjustment and an electronic focus indicator fed by the live AF system, which can allow for manual or automatic focus. (For critical AF you still have the rear LCD and zoom capacity after all so the VF is "optional". You could have a clip on EVF as well, or even a wireless remote one).

And as for speed, Panasonic seem to have gone a long way to achieve respectable focus times with the G series cameras.

"I do wonder how in the world photographers using rangefinders and slrs ever made a decent image. No offense...I'm just sayin'."

Well, for one thing, a lot of decent pictures are very slightly out of focus--including some very famous ones. Second, a lot of viewers of photographs can't tell when something is slightly out of focus (really--it has to get pretty obvious before some people notice it. You can do the experiments yourself if you want to.) And lastly--this is kinda hard to explain in words, but I'll try--the lower the overall "system" resolution (and by "system" in this case I mean the sum total of all the influences that degrade or preserve resolution in the final presentation of the image), the more of the image will equal the best resolution in the image (meaning that perfect focus matters less). This makes intuitive sense if you think about an image that has camera-shake blur overall, or was taken on very grainy film--the plane of best focus gets that much harder to detect. A very high resolving "system" makes the plane of best focus "stand out" more because it makes it possible to distinguish between perfectly-in-focus and very-very-slight-out-of-focus areas. With a low-resolving "system," accurate focus doesn't just seem less critical, it actually is less critical.

Mike

Mike...I know all that...that was my point. Things have always been a bit out of focus. And we lived with it just fine. Don't need things perfect to make a wonderful image.

re Ctein's focusing concept: I can't help imagining one of those interface failure loops where an impatient shooter in a low light situation keeps refocusing on different spots to try and "help" the focus assist, but does it just rapidly enough to prevent the system from assisting. Yes, just being silly--things like that do happen, but it's also a predictable design problem.

If I'm not mistaken, several current DSLRs have z-axis sensor adjustments for focus calibration, though not automatic. But that part of the challenge obviously has been worked out.

Tangential, but I've always assumed that focusing by triangulation (a la rangefinders) should be relatively easy to automate, but I haven't seen it done.

How about a hybrid viewfinder? Normal optical viewfinder, with all its benefits, but instead of a small mirror patch you have a small screen mirrored in (or just a transparent LCD) that shows you the image in the center of your sensor.

At your convenience you could have the image scaled to match the rest of the viewfinder view, or zoomed in to near 100% to see the focus more precisely. Have a pushbutton for your right thumb: no push and the screen is inactive, leaving you a clear view. Push the button and you see the center of the sensor data. Push in further and it zooms to between 50% and 100% (whichever is best for judging focus).

Me, I use a scale-focusing camera as my walk-around kit. Seems I usually get focus close enough without any focusing aids at all; but of course, that's with a semi-wide lens that likes to be stopped down a bit.

"I do wonder how in the world photographers using rangefinders and slrs ever made a decent image. No offense...I'm just sayin'."

Actually with some practice, manual focus systems can be a lot faster , easier, and sometimes more accurate than auto focus systems.

With a SLR or TLR with a plain groundglass screen you simply adjust the focus so the that plane of focus is about right, refocus so it is a little front focused , then a little back focused with an equal amount of blur, then split the difference. This is essentially what contrast detect autofocus systems do. The difference is that when you are doing it manually you can do it a lot faster and you don't have to compose on a particular spot.

The other thing about the old manual focus days is that SLR photographers tended to use fast lenses stopped down. They were focusing wide open, and depth of field would hide any errors.

Also, as Mike says above, 35mm film hand held in natural light isn't going to be all that sharp anyway. With strobes and/or medium format on the other hand it can be very sharp.

"Like a laser based rangefinder on the camera which could be calibrated by lens?"

you mean like this?http://www.cmotion.eu/websiteneu/files/brochure.pdf
see page 18

Of course this is an $60,000 add-on for a Panavision cinema camera that is so expensive that it is only available for rental, and typically takes 3 people to operate.

It's pretty cool though. You point a laser at what you want to focus on and the camera tracks it or you can store several focus points that way and the camera focuses on a point before the subject gets there.

Mike, your "system resolution" idea makes sense intuitively, but I think the math points the other direction. Forgetting subject movement and camera movement and apparent film graininess, the formula for the resolution of an ideal system is:

Reciprocal of resolution of the system =
the sum of the reciprocals of the items in the system.

That is, assuming a resolution of 100 lpm for each element, say, lens, film, enlarging lens and paper:

1/SysRes = 1/LensRes + 1/FilmRes + 1/EnlLensRes + 1/PaperRes

or in this case

1/SysRes = 1/100 + 1/100 + 1/100 + 1/100

1/SysRes = 4/100

System Resolution = 25 lpm

Increasing the performance of any one item in the chain will raise the system performance, but it will never be close to the performance of the best-performing element in the system.

Whimper, and on a Sunday yet; all of my cameras are dreck.

Don't you and Ctein have something uplifting to say the weekend before we find out how much more winter we have? Sheesh.

Oh, never mind.

Engineers cannot do really cool innovative things, they are stuck in the way things are suppose to be done.

I worked at a large tech company and none of the engineers or project managers would believe that the iPhone would do even an eighth of what it was rumored to do. I have no doubts to why the company "engaged in a drill down reduction in force to reduce negative impact on stockholder investments and company profitability".

"Don't you and Ctein have something uplifting to say the weekend before we find out how much more winter we have?"

Bron,
You're sooooo goodlooking.

Mike

http://www.youtube.com/watch?v=VHQ6aKizxUE

Dear folks,

Before getting into specifics, a general observation I didn't include in the main column:

All existing focusing systems are seriously lame. They are so far from ideal that they shouldn't even be used in the same sentence with the word "good." Thinking of them as anything but inadequate simply hobbles your imagination; one of those cases where 'good' is the enemy of 'better.'

Don't believe me? Here's a semi-ideal focusing system; compare it to the reality you have to live with today:

The camera can read your mind. All you have to do is let it know that you want to focus on the left eye of the second person from the right, and it pays attention to that subject detail and tracks it and keeps it in focus no matter how the subject or the camera moves. Depth of field? The camera knows which subjects you want to keep sharp and what their relative importance is, so it will stop down to whatever aperture is necessary to keep them all sharp, and if that's impossible or contraindicated it will favor the ones it knows are more important to you.

Now, isn't that closer to what we all want? I leave the engineering details to the... ummm... engineers.

The point I'm making is that we all make do with focusing systems that work far less efficiently and accurately than we would like, because of the mechanical and technological limitations of our equipment. That doesn't mean what we have is good; it just means it's what we have. Try to think about what you'd really like instead of what you have. That's the way to start to get a handle on what's good.


~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
=====================================

I always knew it was all luck... dumb luck...

Dear Carl,

I had an XA, but I never tried to adjust the focus on it; I did adjust the rangefinder on my Canonet and the resolution at f/1.7 kicked up from 60 lp/mm on film to over 100. At least at the two points that the focusing controls allowed exact focus. In between, it was still a lot better than it had been before.

~~~~~~~~~~~~~~~~~~~~~~~~~~

Dear Pak,

I'm not familiar with the Contax system. That's interesting; thanks for the knowledge boost! Current piezo actuators are very compact. I don't know how much space could be whittled out behind the sensor, never having dissected one of those digital rangefinders. Worst case, it would add 4 mm to the thickness of the body; best case, it would add nothing.

But as you correctly point out, this is all just fancy dancing around systems that are fundamentally inadequate. I proposed my DAFT (I love it!) idea because it would scare the horses the least, a modest but useful technical tweak to a system people were comfortable with no matter how flawed it was. The idea was to get them to think about alternatives without freaking them out.

~~~~~~~~~~~~~~~~~~~~~~~~~~

Dear David,

My understanding (possibly wrong) is that the cams on Leica lenses are all custom ground. Problem is they're not good enough. Given what those puppies cost already, I'm not sure "good enough" would be affordable by anyone. So I think that technical approach is a dead end.

Much easier with larger format; between the reduced magnification and the smaller maximum apertures, tolerances generally get looser as the square of the format size. You can be 50 times sloppier with an 8 x 10 view camera than you can with a 35mm!


~~~~~~~~~~~~~~~~~~~~~~~~~~

Dear Hugh and Roger,

Large motions of the sensor plane are difficult to achieve in a compact body, although not impossible. That's why my very modest DAFT idea only used sensor movement for micro focusing.

Conversely, you don't want to focus solely by moving the lens because that actually creates the problem you're trying to avoid -- oscillations and thrashing. The response time for lens movement is comparable to the human response time. It is, indeed, too easy to get a situation where the human and camera systems end up chasing each other's tail, focusing wise.

Piezo actuators are more than an order of magnitude faster. There's no hunting/thrashing problem, because from the camera's point of view the human response time is very, very sluggish. Any adjustments the human makes, the system can quickly compensate for. Even when the human knocks the focus slightly off, the sensor can respond and reposition faster than your brain can tell your finger to press the shutter release.

It's still a DAFT idea, but there was method to my madness [ grin ].


~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

Dear Howard,

Minor correction--

Right concept, wrong equation.

I'm not sure why that formula is so widespread, but it's wrong for camera systems.

The correct form is:

1/res^2 = 1/a^2 + 1/b^2 + 1/c^2...

In the case of your specific example, four blurs of 1/100th mm would combine to form a blur of 1/50th mm, not 1/25th.

Experimentally verified many, many times.

Wrote about this a while back:

http://theonlinephotographer.typepad.com/the_online_photographer/2008/11/diffraction-in.html

pax / Ctein

To Hugh Crawford...

Yeah, I know. Next time I'll make my sarcasm more obvious. Read my post again as if I were rolling my eyes as I question how in the world any great photo ever got made without all this fuss about focus-pocus.

I haven't used auto focus in almost 40 years of making photos, including use of large format as well as film and digital Leicas. And neither did the great photographers whose work I've both studied and collected for several decades.

Howard,

You mention something I've seen other places on the net, that the "[r]eciprocal of resolution of the system =
the sum of the reciprocals of the items in the system." As far as I know, this is a bit of unreliable interweb wisdom. I am, however, happy to learn something new if you have a (peer reviewed, technical) reference that demonstrates the conditions under which this relationship is valid.

A more accurate description of the effect you're talking about is rather more complicated. It involves a mathematical operation called convolution, and is more properly stated, "The resolution of a system is given as the convolution of the point-spread functions of the individual elements." Even this statement needs a lot more things specified before it's really useful, but it's a good starting point (ahem...).

The end result, though, is basically what Mike asserted: the least-resolving element will dominate/limit a system's resolution. This is also what your formula says, by the way, despite your interpretation of your proposed numbers.

Okay, I'll crawl back to the nerd colony now....

-Derek

P.S. For those who hear the call of the colony, you can read a bit about convolution, and see a helpful animation, here.

Howard - Your explanation makes no sense to me. Assuming impeccable technique, in your example the film is able to receive all that the lens can give it, and the enlarging lens transfers all that the fim has to the paper. Where's the loss?

I'm no mathematician. Please elaborate.

Frustrations in the search for photographic and audiophile perfection are identically fuelled: the more attention and cash you devote to it, the more you highlight that part of the process that will always score just too highly on the crapometer.

Leo, I think Ctein is proposing focusing on the sensor, in fact, but for fine focusing after one has manually gotten the focus close. I think it's actually a sensible approach.

OK, I can't argue with the evidence against eye and rangefinder where perfect focus is wanted, but I submit that the same combination performs superbly for *coarse* focusing, especially when combined with short focus throw, tactile feedback (e.g. a focus lever), and experience. And there are situations when coarse is all you need.

However, the apparently shallower DOF of digital sensors would be one more reason a more accurate focusing mechanism is needed, for those situations where coarse isn't enough.

So I guess I'm of the opinion that Ctein's proposal is a very sensible hybrid, combining fast manual coarse focus with a more accurate auto fine focus.

And I guess "EVIL" is here to stay as an acronym? Too bad; I was rooting for "ELVIS".

This discussion so far assumes that only one part of the image needs to be in focus, when in reality, the zone of sharpness (a term I prefer over the traditional 'depth of field') must be considered.

There are other weaknesses associated with focussing systems that haven't been addressed in this discussion yet. Like, the fact that no digital SLR I know of can focus on a blank wall or blank patch of sky. The Contax AX can focus on a blank surface by projecting a red barcode pattern on it, as needed. The Sony F717 (not an SLR) projected a laser pattern of small diagonal lines on a blank object that had no detectable edges.

In my studio work, I have to consider depth of field for every shoot. What I really need, besides the already stated by others accuracy of the focussing distance, is to define shot by shot what percentage of the depth of field is behind the focus point, and what percentage is in front. It is convenient to focus on the eyes, for example, but it'd be helpful to have a *depth of field shift* so I can push 80 percent of the depth of field behind the eyes, where most of the other important detail is, and twenty percent in front. Or in the case of the recent dog portraits I shot, 90 percent of the depth of field needed to be behind the eyes so I could get the whole dog sharp from snout to tail while focussing on the eyes.

And why is it that digital SLR's don't indicate the depth of field on the LCD? It's a trivial calculation. Nikon and Sony have no excuse for omitting such a feature - their lenses do communicate focussing distance back to the camera body. But Canon's lenses don't, so EOS bodies can never calculate depth of field - a weakness of the EOS lens mount.

We can take the Contax AX concept a couple of steps further... instead of just moving the sensor back and forth for focussing purposes, how about up and down for some shift as well? And tilt the sensor forward and backwards too?

The Contax AX focussing system is something of beauty. It can auto focus my Hasselblad C and CF medium format lenses, and any other MF lenses that I attach to the AX via an adapter.

I miss Contax. Innovation pretty much disappeared after they departed the camera industry.

Ctein:

May I ask how you would fine-tune your camera (in detail) for fast primes to improve focus as you describe doing above.

Thank you,

SH

I certainly agree with Ctein about rangefinders; one of their dirty little secrets is that they are almost impossible to focus accurately (by eye) through the whole range of f-stops; and it's very difficult to get a whole range of lenses to focus accurately with the same camera. If you have two cameras, the difficulties become even greater. You *can* learn the lenses, but it takes a lot of persistence. I have a Nocti that I know is very close to right, wide open, but I need to give it just a little teeny clockwise noodge past crisp eye focus to get crisp focus on my M8's sensor.

The best solution for Leica M series cameras probably would be simpler than Ctein's proposal -- simply refine the system used in 4/3 cameras, which takes a focus reading off the sensor, and then flash a tiny LED in the rangefinder window when the focus is right on. You'd use your rangefinder skills to get it very close, and then a manual noodge to get it perfect, without any complications of possible feedback.

I will say (again) that no great picture that I am aware of, was "made" purely by sharp focus; there's always something more important. Still, it would be nice to have a camera that was capable of it, since you could always degrade it a bit if you chose.

Sometime back in Leica Fotografie International magazine there was an article about focus shift. If memory serves me correctly, there was an optical diagram of a 50mm f1.4 lens, and how focus shift occurs as the lens aperture changes. The article also discussed how film is more forgiving of this, because of the depth made up by its emulsion layers. If the light is focused somewhere within the emulsion layers, the image will be in focus. With digital sensors, the light must be focused on the surface of the sensor. There is no depth to the critical focus area as there is with film emulsion. (I don't know if this is also true for Foveon sensors). Another factor that was stated in the article was what we know as "pixel peeping". Before digital imaging, far less people studied film images that were enlarged to 100% or more, and consequently people were not bothered as much by focus shift.

Ctein / Craig / Hugh:

- It was indeed the Contax AX (as others have pointed out here). I remember reading about it fondly as I always wanted a Contax when I was younger... I believe the focal plane actuator was an ultrasonic motor that drove a stepper mechanism, like a floppy drive (can someone confirm that?)

David / Hugh:

- I should have done my research on the laser rangefinder! I only mentioned it because of their Leica's Geosystems gear and their technology in their binoculars.

Ctein:

Finally, perhaps the solution you're thinking of is actually the reverse of modern photography: like an electron scanning microscope for instance. In this case, you don't capture light reflected off the subject, you scan it instead! In this case, focusing and determining the distance would not be a problem -- it would be the solution. And subject movement would be the problem instead. In any case, whilst we're using optics, this focusing problem as you mention is going to be around for a while...

Dear Jeff,

Mike, Hugh and I were all doing you the courtesy of treating your remark as serious inquiry instead of pointless sarcasm that served no purpose. Forgive us for making such an egregious error.

Next time, instead of dramatically rolling your eyes, maybe think a bit and decide if you REALLY have something useful to say before engaging in mere snark?

pax /Ctein

Dear Derek,

Equations in terms of limiting resolutions are a quantitatively useful approximation for many systems. One of my books gives the general form as 1/R^n = sigma (1/a^n ... 1/z^n). It's possible for n to be as low as 1 or considerably above 2, according to my books, but for conventional imaging systems, n=2 provides quantitatively accurate results.

n=1 and n=3 can be definitely disproven by straightforward experiments. For instance, I could routinely determine film resolution values that were in close agreement with the manufacturer' published data from in-camera photographs using the 1/R^2... form. The experimental data made no sense for any other value of n -- in fact, they would have been physically and theoretically impossible to achieve if n equaled 1.

Like you, I don't have a clue how the n=1 assumption took hold.

One complication is that I don't know how, a priori, to incorporate sensor resolution for an image generated from a Bayer array. It would take some experiments I'm not equipped to do. Might turn out that it falls neatly into the equation, or it could get very messy mathematically.

~~~~~~~~~~~~~~~~~~~~~~~

Dear Dave,

The mistaken assumption lies in what constitutes "impeccable technique." There are no lossless transfers, ever. Sometimes the blur added by a transfer stage is so small as to be negligible, in which case you can simply drop that term out of the equation. For example, print paper resolution never comes into play when making darkroom enlargements.

On the other hand, unless you're using a fairly poor lens or one far from its optimum conditions, the film can't record everything the lens can project. For example, high quality color films can record 125-150 lp/mm, but decent (not extraordinary) 35mm camera lenses can generate 200-300 line pair per millimeter moderately stopped down.

As Hugh and Mike pointed out, in real world photography you are rarely operating near the limits of the system, and so a considerable amount of slop is tolerable. Where problems arise is when you really are trying to get the best out of your system... or when you fail to realize that focusing errors are a considerable part of that slop. As I said, many photographers think their prime lenses are pretty punk wide open; said photographers are mostly wrong.

~~~~~~~~~~~~~~~~~~~~~~~

Dear Sherlock,

I'm going to decline. If you're talking about adjusting a rangefinder camera, that's a fairly technical procedure that involves opening up the innards of the camera and knowing exactly what you're doing. It's not easy to explain concisely. Especially if you're trying to do it without a proper optical bench, as I do.

If you're talking about adjusting an SLR camera, that's really simple. It's so simple, in fact, that you should be able to figure it out just looking inside the mirror chamber of your camera. But since it involves putting tools near delicate components that can easily be misadjusted or damaged, I don't think you should be messing around in there if you can't figure it out for yourself.

~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
=====================================

Wow, all this science fiction is really invigorating. Thanks for stirring the pot Ctein. I hope not only engineers, but their lame marketing and bean-counting bosses are reading this [grin].

Meanwhile ... my experience with focussing Leica lenses on the GF1 has been enlightening:
with a simple click (or two) I get an enlarged view direct from the sensor that allows me to focus infinitely more accurately than I did with the same lenses on any of my Leica M bodies, despite all the good CLA done to the latter.

I am not sure it's any slower than focusing a rangefinder, either. Yes, you do lose the bigger view momentarily, but you get a sharper image. So, as long as we all wish those EVIL mirror boxes to go away and we are happy with manual focus, it seems we are almost there.

All we need is for the Japanese to come to the rescue. Build an M body (Bessa redux?) with focus ergonomics improved over the GF1.
Who knows what other tricks a future sensor can be made to do besides movement on xyz axes?
Maybe non planar sensors that can bend to follow the actual shape of the plane of focus as dialed-in via firmware/menu/buttons?

Are you listening Kobayashi-san?....

Dear Jeff,

Providing a further serious answer to your comment, "inaccurate" does not mean "always wrong" in this situation; it only means "cannot be counted on to be right." The focusing errors we are discussing are caused primarily by manufacturing slop, random mechanical variations from the ideal. Like a room full of clocks that average five minutes slow, some will be exactly 5 minutes slow, some will be 10 minutes slow, some maybe five minutes fast, and a few will be spot on.

(On occasion there are design flaws that produce systematic errors so that the focusing system could never be right. Happily, they are not common.)

In addition, in the case of good rangefinder cameras, there are often two distances in the focusing range where the focus is accurate. Many of the coincident finder designs have two adjustment screws, so adjusting them is like solving a pair of equations with two unknowns-- you can make the errors come out to zero for two specific distances. It's not uncommon for one of those distances to be infinity, because that's easy to set up on optical bench, so you'll see rangefinder cameras that are fairly accurate at long distances and become increasingly inaccurate the closer you get to the subject. Unfortunately, in terms of depth of field, this is not the most desirable alignment.

If you have a well-made rangefinder camera with a properly adjusted infinity stop on the lens, there's a good chance your distant scenic photos will be very pleasantly sharp. Especially since you're likely to be stopping down a bit under those circumstances. But then you'll try to do an indoor portrait of somebody and discover that it is the tip of the nose or the earlobe that is in focus instead of the eyelashes.

In any case, please remember this column isn't claiming that current focusing systems are so inaccurate that all your pictures will be awful. It's a refutation of the erroneous belief that they are so perfect that it's a pixel-peeping waste of time to think about improving them.


~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
=====================================

No one really needs to worry about all this stuff.

Just fire up Photoshop... a couple of passes with 'Sharpen more'... There!!! Perfection!!!

It's a miracle that anything ever works.

I find it curious that no one has mentioned DSLRs such as the Nikon D300s, Canon 50D, and Pentax K-7 that have a focus calibration feature. If nothing else, the presence of such a feature is acknowledgment by the manufacturer that focus is not always ideal between specific cameras and lenses. Perhaps Ctein could shed some light onto how this feature works and how well. I imagine it mitigates some problems while doing nothing for others.

Acouple notes.

1. Canon USM lenses with focus distance displays comminicate focus distance data, all other Canon lenses do not.

2. The Minolta Maxxum/Dynax 7 used to show DoF on the rear LCD when the DoF preview was engaged with a D (distance encoding) lens. Sadly this nice feature did not make the transition to digital (Yes, a film camera had a large rear LCD with settings, it's actually the camera which most modern DSLR settings displays are based on)

3. you can add me to the list of people pointing out that main-sensor LV provides both accurate and precise focusing, either manual or via CDAF. That's one big advantage of using the imaging sensor itself to focus.

Mike wrote "[DAFT is] no worse than EVIL."

In light of some possible forthcoming product announcements, it will be interesting to see what we end up calling this category of cameras.

"EVIL" probably won't survive for numerous reasons, not least of which is that no acronym that's a word already in common use is very helpful in Google searches.

I saw a discussion pointing out that DPReview has started calling them ILCs, for Interchangeable Lens Compacts. Those three letters correspond neatly to "SLR" but also sparked a call for the more precise term Mirrorless Interchangeable Lens Compacts, which nicely distinguishes them from small SLRs (although some commenters foresaw tedious jokes about photographing MILFs).

Time will tell.

Ctein, here's an explanation of the Contax AX Automatic Back Focus (ABF) system. I found this on a web site called "Japan Exposures":

Quote:

The AX was an incredible engineering feat. It is said that over one hundred patents have been granted regarding the technology in the camera. At the core, two main technological achievements made focussing possible: the first is ceramic technology; a very finely finished ceramic rod acts as a guide rail and motor part to move the film plane forward and back. Secondly, an Ultra Sonic Motor (USM) drives the ceramic collar and an inner camera body forward and back on the rails to obtain focus.

USM offer precision control and extremely quick and quiet operation. The Automatic Back Focusing system on the AX can move from minimum focusing distance to infinity or back in less than 450ms, irrespective of focal length. The Contax AX utilized three computer CPUs to control all of the camera’s functions, a more sophisticated computer system than in any other camera. One of the CPUs controls the Automatic Back Focusing system, one to oversee auto-exposure, and a third, the Indicator CPU controls wind/rewind, shutter charging and viewfinder indications. Almost a byline is the novel concept of user configurable camera parameters, called Custom Functions, which are common today.

Unquote.

Here's a link to the full article:

http://www.japanexposures.com/2009/04/13/tales-of-glass-and-ceramics-pt-2/

For lens (as opposed to focus system) testing, with modern DSLRs with Liveview and with contrast-based autofocus (or focus confirmation) in Liveview mode, you should be able to get a more accurate focus setting that way. Depends on details of the implementation in your camera, but with the Nikon DSLRs I've used, Liveview in tripod mode does contrast-based AF using data off the main image sensor. If it works for your camera, it would also be an option for real use when critical focus is required, if you're working on a tripod and can take the time.

Ctein,

Sorry if I offended. I understand and appreciate your points, and certainly respect your technical knowledge.

Having owned film and digital Leicas for more than 20 years, I'm all too familiar with the issues associated with camera and lens focus inaccuracy...both inherent as well as production quality. These issues are discussed frequently on the Leica User Forum, and I've had my share of "system adjustments," particularly since digital brought forth new challenges.

My point was a serious one, too, however. Perhaps I should have made it in a less sarcastic manner, but I do think it's important that we don't miss the forest for the trees (or however the saying goes). A legitimate question for me is whether these issues make for poor photographs, which after all is what the whole thing is about (I hope).

In the film world, at least, I think the answer is no. Good photographers, including rangefinder users, managed to create great photographs despite these issues. I have many of them on my shelves and walls to prove it.

Perhaps the digital world will make this more challenging. I'd welcome any thoughts as to whether your comments are more relevant in a digital world. Even so, however, I trust that talented photographers will always make great photos, despite these technical limitations.

Having said that, I'm all for taking the slop out of the manufacturing process. I'm also not a Luddite. I think change can be good (if the change doesn't negate what's already good). But, I think it's important to put things in perspective. And, that's what I tried to do. Poorly.

"I will say (again) that no great picture that I am aware of, was 'made' purely by sharp focus; there's always something more important."

Yes. Megapixels. Lots and lots of ultraclean ISO 102400 megapixels.

It can be said now. Absolutely no high-quality digital picture was ever taken with a camera under 12 megapixels at ISO 102400. And no perfect in-focus picture was ever achieved before live-view with per-pixel magnification.

:)

Mike,

Ahhhchooo!

:-)

Bron

Gordon,

I did mention focus-calibrating DSLRs a while back, though without naming names, just to note that camera engineers have indeed made sensors that move in the z direction. It's not DAFT, but a step forward.

Yes, they say that lenses are manufactured to a tolerance, and that a particular sample may or may not be sufficiently on (or off) for a particular body for a particular use (didn't I read that on your blog?).

Gordon & Ctein,

But that's just one more reason why Ctein's idea is such a winner. Assuming per-lens sensor shift calibration works, Ctein's method still goes one better because it's dynamic; i.e., it works even for lenses that haven't been calibrated (as long as they're not too out of whack) and it works at all distances.

And DAFT goes beyond better focus (though it should give us that). The thing is that it does its thing by compensating for slop, and not merely slop but its variability, too. It may have been conceived as a focusing aid, but it is perhaps more importantly a slop compensation mechanism.

So if it catches on--if it is made practical and economical--manufacturers will not only use it to give us more accurate focus, but will also subvert it for the purpose of allowing even more slop while still providing "good enough" focus. Which might lead to some interesting developments.

And while we're on the subject of moving the sensor this way and that, what about tilt, rise, shift and swing?

Ctein,

I beleive you are a keen amateur astronomer. So you know how difficult it is to achieve focus at infinty. So difficult in fact that many astrophotographers use a method devised by french physicist Foucault which I'm sure you've heard about.
In the 60's some so-called astrographs (just a lens with a film holder) were focused by trial and error. A bright star was left to trail for x seconds and then exposure was interrupted. During that interruption, a slight tweak of focus was accomplished (the lens had a micrometric scale to help in this very precise tweaking). Another exposure was then made on the same piece of film and so on. After development, the film was examined with a microscope and the thinnest star trail would determine the best (but not necessarily perfect) focus. And that was just to determine focus at infinity !
So I can't imagine any system, RF or SLR that could ever be that precise. We'll have to contend with good enough for quite a while I'm afraid.

I hate to see Jeff twisting in the wind alone on this. I thought - I still think - that he did have "something useful to say." So now there are two of us. And I suspect there may be even more.

>>Well, for one thing, a lot of decent pictures are very slightly out of focus--including some very famous ones.

I saw Richard Avedon's portraits of the Beatles at a local gallery. Three out of the four were out of focus! Paul was badly so. I think he nailed the focus on Ringo (or maybe George).

So if you ever make 8x10 view camera pictures of famous celebrities, put up a rope barrier so potential internet critics cannot put their noses against you 16x20 prints to check focus!!!

-Tom-

Ctein,

I didn't consider that the piezo actuator could be so fast, so I see your point. What sort of movement range were you thinking of? Is there enough that with a moving subject approaching head on and in the r/f spot (say a runner or a cyclist), by the time the lumbering human brain has pressed the release on seeing the focus light, the actuator can readjust before exposure?

So when are you going to build this thing? I ask because I've been faffing around with micro 4/3 camera designs and built a dummy one out of Lego. Now my ten year old nephew keeps asking me when I am going to build a real one.

All I need to do it is a hacksaw, a soldering iron, a zooming viewfinder and a spare Pen. No worries then.

Crikey, there was me thinking that all those perfectly focused f/1 Noctilux shots were due to my perfect eyesight and a totally functional manual focusing system. Thanks for making me realize they were all just pure luck.

Dear Jeff,

Oh, definitely no harm and no foul. Communications sometimes clash in this low bandwidth medium, and it's pretty clear from your last e-mail that we were all talking about the same points even if we were talking past each other a bit.

Some years ago I made the same observation that you and Roaldi expressed to one of my editors (it might even have been Mike), after discovering yet another "gotcha" in the darkroom. Namely, that it was really amazing in the face of all the things I had found that could go wrong that anyone ever managed to get a decent print!

Of course it's not really amazing. The real answer is that "decent" is so far away from "perfection" that most of the time there is ample room for error in our methods and equipment. The stuff is still going to look great.

It's also another reason why pixel-peeping, regardless of where and how it's done, is such a waste of time for most photographers. Useful for us folks running tests, but often irrelevant to practical photography.

One of the points I'm trying to emphasize here is that the deficiencies in our focusing systems aren't about pixel peeping. They often make the difference between someone thinking they own a mediocre lens and a really great one, in terms of sharpness.

And, sometimes they do make for poor pictures. There's no one characteristic that's ever going to distinguish a good picture from a poor one, but sometimes it really is important to have the subject of interest be in focus. Probably even most of the time! The handful of wonderful, fuzzy pictures that we can point to show that there are exceptions to the rule, but they are truly the exceptional photographs. Most of the time we are nowhere near fortunate enough to make photographs that are so inherently and insanely great and compelling that poor technique simply doesn't matter. And it's even rarer for one of those photographs to actually benefit from the poor technique, it's just that we love the photograph in spite of it.


~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
=====================================

Mani,
I would guess it's not "just pure luck," it's that not all of them are actually "perfectly focused." See Ctein's comment just previous to this one....

Mike

I'll tell you what would vastly improve the focusing operation of current (D)SLR cameras: A permanent, built-in magnifier(6X should do it...or maybe 6X plus 2X). Flip in to check focus, flip out to view. Panavision has it, why can't we?

Ctein - "The mistaken assumption lies in what constitutes "impeccable technique." There are no lossless transfers, ever..."

I'm in full agreement. My comment was only addressing the formula, and the given parameters. There was no accounting for anything other than perfect technique or lossless transfer in Howard's equation - that's why it made no sense to me.

Regardless of the mathematics involved, I think that Mike's original intuition (if I'm interpreting it correctly) is right on the money. The overall performance of any system can never rise above the performance of the weakest link.

Dear folks,

Some pernicious logical and statistical fallacies are creeping into this conversation. Permit me to elucidate:

1) The fallacy of the counterexample. You cannot disprove a statistical generalization by citing a contrary case. It would be like me claiming that flipping a coin has a 50-50 chance of coming up heads or tails and you arguing back that you flipped the coin and it came up heads. Doesn't say anything about what your coin would do if you flipped it more. Doesn't say anything about what other coins would do.

You can disprove an absolute with a counterexample, but I never claimed all focusing systems were wrong all the time. I pretty strongly suggested the opposite.

In other words, Mani, yes, I think you're fortunate. Or, Mike could be right and you're "sloppy" [grin].

2) The fallacy of the exceptional. Not quite the same as the other fallacy, although related. It's citing an extreme case to disprove a statistical generalization. In this case, it's invoking "Great Photographs" (an indefinable term) to attempt to discredit a need for critical focus.

Here's the scoop. Almost every one of you reading this will never, ever make a Great Photograph in your life. A few of you will, and a few of you have. But almost all photographs made by almost all people are not Great Photographs. Invoking these as a benchmark makes as much sense as discussing middle-class wage earning potentials by invoking the state lottery.

Bringing up these Great Photographs involves several other logical fallacies.

3) The fallacy of counting only your successes. Imagine running a drug trial, reporting that your drug saved 1000 lives, and not bothering to report how many it didn't help, how many lives were saved by the placebo, and how many your drug killed outright. That's what you're doing when you're thinking only about Great Photographs that didn't depend on accurate focus.

Fact: I can come up with numerous counterexamples, photographs where technical precision is part of what makes them Great, where technical issues play into the composition and the impact on the viewer and a technical failure would substantially weaken the composition. They might still be great, but the "G" would be much smaller.

I'm not going to cite specific examples, because the fallacy of the exceptional still holds sway. These are SO not the point of the discussion.

4) The fallacy of the uncounted. Axiomatically you only see successful Great Photographs. What about all the photographs that suffered one technical failure or another and so wound up in the trash bin? After all, that's the issue! It's not whether you manage to make a Great Photograph in spite of poor technique, it's how many photographs poor technique cost you.

You'll never know the answer to that question, all the failures you never see. But on a personal basis, I can tell you that I've had to consign way too many otherwise-portfolio-quality photographs to the file drawers because they suffered from some technical failure that just degraded their impact or composition too much. If it's never happened to you, you're incredibly fortunate... or very sloppy [ grin, again ].

5) The fallacy of calling your conclusion an axiom. Don't even try asserting that a Great Photograph is one that by its nature doesn't depend on good technique. Unproveably circular reasoning

In summary, my column was not about the extrema but about the norm. Pay no attention to the stuff three sigma out.

~ pax \ Ctein
[ Please excuse any word-salad. MacSpeech in training! ]
======================================
-- Ctein's Online Gallery http://ctein.com 
-- Digital Restorations http://photo-repair.com 
======================================

"you can make the errors come out to zero for two specific distances. It's not uncommon for one of those distances to be infinity... Unfortunately, in terms of depth of field, this is not the most desirable alignment."

Ctein, I'd just like to thank you for one of the all-time great understatements of photography. :-)

Slight tangential (sort the maths on that statement ;-)), there is one issue that comes up around digital focusing that bugs me and that is the idea that more resolution needs better focus.
Hokum, I say. If your enlargement is the same magnitude as with film, same DoF criteria hold. Choosing to blow up those pixels more shifts the argument into different territory. What seems to be missed often is that flatness & thickness of sensing medium are playing a huge part here. That, coupled with all the pixel peeping means we seem more aware of focus accuracy than ever before.

Funny no one mentioned the Polaroid SX-70 sonar/ultrasonic focus module.
When I shot glass plates with a telescope ~45 years ago, we focused by racking the plate in and out taking short exposure images of a bright star (point source) at measured points on the rack, developed the plate and picked the rack position that showed the best focus. Why not use a point focus with a modern camera? Use a IR (invisible) laser aimed at the point you want focused and let the sensor choose the best point focus?
And I have some great sharp photos taken with a Minox 35 using zone focus.
All this may soon be moot, with software able to fix sloppy focus.
Jim

PS: In a conversation with my son the CGI programmer, he says they have to add chromatic aberration to their animation shots to make them match the camera scenes!

Ctein:
Great article!
BTW, (no offense intended)I think most comments about focus accuracy being "good enough" in most current camera systems are beside the point of this particular discussion. So, I guess there is actually no need for your diligent (and very informative) listing of the many fallacies such comments may be subject to.
I recall some similar comments about high ISO capabilities in modern DSLRs (your position on that matter being consistent with the one you hold now). That is, improvements of any kind ARE desirable by definition (duh). It doesn´t matter if most of us have no idea what to do with them, or even notice them, or even care...
Thank goodness for the ones who still care, and are gracious enough to share their knowledge and concern with the rest of us!

When I select a plane on which to focus, another photographer, photographing the same composition, may very well select a different plane to focus on. One might say that our modern focusing technology automatically now splits the difference. How great is that? How does it know?

I have never known a really good photographer who owned any perfect equipment. They all seem to own collections of very good equipment whose virtues and faults they know intimately. And they seem to avoid using it in the ways that display the faults at their worst, unless there's no alternative.

It's almost enough to make me think that the quest for the perfect camera shouldn't be the first stage of the process; I'm starting to wonder if having the perfect camera would really make everything else fall into place :-).

I don't mean to discount Ctein's contention that the focusing systems on today's cameras have a lot of room for improvement. Given a choice, I would go for the better focusing system almost every time. But I also see where Jeff and David Dyer-Bennet are coming from. The analogy I would give is that in the world of music there is no such thing as a perfectly tuned instrument. The tuning has to be "tempered" so that notes and chords will not sound perceptibly out of tune in most major and minor keys. The lack of perfection doesn't prevent musicians from creating amazing music, though.

Similarly, the better photographers know they can only get so far by faulting their equipment. They either adjust to what their equipment is capable of doing or adjust their equipment to meet their needs. Let the scientists and engineers seek accuracy, precision and perfection. Until something better comes along, we photographers have little choice but to make the best of what we have and what's available.

Ctein & Derek--Thanks for the correction! And mea culpa: I got my misinformation in one of the camera makers' week-long "schools," and when I later worked for that company passed it on to others in the business as gospel.

Derek, my apologies to Mike if I misread his statement that "the lower the overall 'system' resolution..., the more of the image will equal the best resolution in the image...." His explanation told me that we were heading in the same direction, but I think I'd have said "will equal the worst" instead of "the best resolution." Again, my fault. I conflated 'system' and 'image.'

scanning the replies, i don't see mention of what immediately appeared to me as a fatal flaw in ctein's daft idea: it won't work (properly) with any lens incorporating a floating element (i'm not optical engineer, so feel free to correct me if i am wrong). which happens to be most of the most phenomenally good (and fast) lenses leica (and lots of other lensmakers) currently makes.

and since ctein seems so intent on scientific rigor: this means that far from his '10 minutes' of thinking constituting a 'proof of principle', it may merely mean that leica engineers have already thought of this approach, and determined it won't work--or is potentially worse than what we have. the fact that you can come up with an idea for an improvement that doesn't work is (logically) not proof that someone else can come up with a workable improvement if they put more work into it. (of course, it isn't proof that they can't, either.)

the imperfection of a system doesn't mean that a better system can be devised. (think about 'democracy', for instance. which doesn't mean you don't try to get it working the best it can--states need periodic cla care too.) and different systems can be so different in what they offer as strengths and weaknesses, that an 'improvement' in one aspect (eg, focus accuracy) will be undesirable if if compromises another aspect (eg, the working advantages of the optical rf window, etc).

on the one hand, obviously it would be cool if we had psychic links to our cameras, and we could achieve perfect focus instantaneously on whatever point we liked just by thinking about it. but short of that, what would a hypothetical system such as this daft idea actually accomplish? the last thing i want on my rangefinder is for the camera to second-guess where i place focus, and 'improve' it for me. this is exactly why precision, rather than accuracy, is the most important quality for leica (eg) to emphasize. if the system is precise, then i am capable of controlling it (including learning how to compensate when i need to, eg for focus shift). if the system lacks precision--and this notion of refining focus after i set it seems like it will almost always involve a loss of precision, unless i limit myself to photographing only flat, perpendicular focus targets--then any 'improvement' of focus will be merely accidental.

rangefinder cameras such as the m series do offer very real advantages in getting certain sorts of photos over other existing systems, and i would really hate to lose this. a small improvement in focus accuracy (that is already pretty darn good for anything up to 50mm, at least, better than my top-end canon gear for my usual kind of photography) that entailed any reduction in the immediacy (eg, through lag) and precision i now enjoy would be a terrible trade-off. of course for some situations, where you have time and relatively static subjects, i would love to have magnified live view focussing available on the (upgraded) lcd--but that wouldn't impinge on the ability to use the rf for what it excels at.

Dear Chris,

I'm afraid you are wrong. Floating elements have nothing to do with this. I can't figure out why you think that would matter.

If you want to explain in detail, please email me and we'll pursue this without clogging up thread.

Proof of principle still stands.

Also, the DAFT system responds faster than your brain can send an impulse to your finger. There's no substantial increase in lag time, maybe 10 msec.

Finally, this system in no way reduces your precision. It merely makes the system accuracy as good (or better) than your precision. I can't imagine how anyone would consider that a disadvantage.

pax / Ctein

P.S. Occurs to me that I should explain, for the nontechnical readers, that a "floating" element is an element in a lens that moves independently from the others. In a simple prime lens, you usually focus by moving the whole lens backwards and forwards-- the elements within the lens don't change positions relative to each other. Floating elements do.

Lots of modern designs use floating elements; e.g., any lens that has "internal focusing" or "internal zooming."

It has no more relevance to fine-tuning the focus by moving the sensor than whether the lens barrel is made of aluminum or polycarbonate. That is to say, none.

pax / Ctein

I hate to jump back into this, because "how to focus" is way too complicated for the comments in a blog post, but the point about floating elements is well taken.

Personal experiance with the Nikon 28mm f2.8 AIS lens on Canon DSLRs has shown it to be extreemly sensitive to flange adapter depth. Nikon's Close Range Correction (CRC) lenses realy look awfull on an adapter that is a little less thick than it should be, which almost all of them are.
Google nikon crc for all the details.
I'm typing this on an iPhone while my office is being remodeled so details and most likely spelling are necessarily sparse.

Dear Hugh,

If I understand what you're saying, floating elements can make the design much fussier about mechanical tolerances in the mount and such?

OK, I'll take your word for that. Beyond my expertise.

What that means, though, is that Chris had it exactly backwards. A purely mechanical focusing system (like a rangefinder) is MORE likely to have slight focus errors with floating element lens. A DAFT addition, which microadjusts the focus based on what the sensor is actually seeing, would catch that and correct it.

pax / Ctein

Dear Ctein,

The SLR calibration that you were referring to...would that be applicable to DSLRS as well?

Avi

With a Nikon 24mm f2.8 or a 28mm f2.8 AIS lens on a Canon 5d mkII having the lens flange depth 0.37 millimeters too shallow makes a noticeable difference. I actually thought I had a defective lens in the case of the 28.
This sort of explains it in terms of the Nikon lens line
http://imaging.nikon.com/products/imaging/technology/nikkor/n14_e.htm

Erwin Puts does the same in Leica terms here
http://us.leica-camera.com/assets/file/download.php?filename=file_1851.pdf

I do think generally that the moving sensor idea is a good one, and a closed loop system using contrast measurements from the sensor is the only way to get accurate autofocusing. Adding focus bracketing or "stacking" would be good as well, but that could be implemented using conventional focusing even though it would be less elegant and precise.

The Canon SLR autofocusing is essentially a passive open loop rangefinder system but on the live view enabled cameras they could add software to check the focus on the sensor and fine tune it with minimal delay.

This is beginning to remind me of a six hour discussion among photography professors on whether it is better to teach beginning students to focus with the front or rear standard of a view camera. (I'm of the rear standard if it's a monorail persuasion myself)

Thank you for writing about something that really needs to be discussed amongst photographers and escalated to the manufacturers.

IMHO, AF performance (reliability, accuracy, and speed) is one of the most important aspects of a camera's performance, yet it is rarely given as much attention as noise vs ISO and resolution testing in camera reviews and forum discussions.

Although accuracy seems to be the main discussion point here, I feel that the basic ability to achieve AF lock and to ensure that it is locking on the actual spot or target you are focusing on is even more lacking on some systems.

One can often fix minor exposure and WB errors in post-processing, but you can't fix an out of focus image. Worse, sometimes you can't even get the shot because the AF system couldn't lock (sound familiar fellow E-3 owners?).

Engineers are working on adding more megapixels when the chances of getting a shot not only in focus but accurately focused enough to give you 12 MP of resolution is, what, maybe 1 in 4 in challenging conditions? 1 in 2 on a good day?

Manufacturers started the megapixel race, the high ISO race, and now it seems are jumping on the frame rate race (at least for compact P&S). I want to see an AF race. Who knows? It may actually result in some meaningful developments...

"Manufacturers started the megapixel race, the high ISO race, and now it seems are jumping on the frame rate race (at least for compact P&S). I want to see an AF race. Who knows? It may actually result in some meaningful developments..."

I forgot to mention the zoom ratio race... They're up to 30x zoom (24-720 mm-e) now on the bridge cameras. I don't even know what to make of that.

Not directly related to the topic at hand, but thought others may find this amusing:

Sitting beside someone at a wedding a few months ago, i was asked how many times zoom is my camera (i was holding a D700 + 70-300). I said just over 4x... maybe 4.2, 4.3x?

He was incredulous. "No way, only 4x? You mean 40?"

"No, about 4. 300 divided by 70 is... you do the math."

"But it's so big! How can it be only 4x?"

"You should see my other lens (70-200). It's less than 3x zoom and it's twice the size."

"But even my Olympus has 20x zoom and it's half the size"

Sigh... Maybe this is why manufacturers focused on improving other things besides AF technology... it obviously worked on my fellow wedding guest.

Dear Avi,

Unless the camera is using the imaging sensor to focus, everything I said applies. That's why I didn't distinguish between film and digital cameras.

pax / Ctein

ctein--

no, i do not have it 'exactly backwards'. the reason that floating elements would make a difference to your focussing idea is that floating element lenses don't focus by simply moving the whole lens closer or farther from the sensor as if it were on a view camera standard (or, as in your idea, the sensor closer or farther from the whole lens); the internal elements change relationships at different focus distances. if you focus for one distance and set the proportions that way, and then try to refine focus by moving the sensor, you've just screwed up some very sensitive proportions. hugh crawford, above, seems to have both understood my point and to have encountered a variation of it manifest through adaptor tolerances.

perhaps you think that since the refinements to focus you have in mind are very small, this mismatch would be of no consequence. but given hugh's experience, as well as some of my own, i suspect it would.

as for the other points in your first reply to my post, by lag time i was referring not to the action of the focus system itself but more generally to various 'solutions' to rf focusing which involve evfs, such as reichman and others suggest, which at this point in time and price entail a lag. and second, yes, the system you propose will by definition mess with the precision of the the focus system, because it introduces a variable which is not only out of the photographer's control, but out of her sight as well (if we could see in the vf the tiny difference that this adjustment makes to focus, we wouldn't need the help of the automatic system to make it). even more importantly, it does not solve the basic problem with all af systems, which is that it doesn't know where exactly i wanted to focus; it can only make a guess, and that guess introduces yet another variable outside of my control. this is one reason why i often already prefer manual focus to af: it is more precise (ie, consistent with my intentions) even if my current cameras can often af more accurately on some imprecise element within the scene.

I do wonder at all this. The theory may or may not be right, but I have no trouble focusing with the Leica Summilux 35/1.4 asph at f1.4 so that it is very sharp every time - I mean really, really sharp at 16x20, same with fast lenses wide open on the 5DII. So I'm not sure what you really mean by inaccurate focusing methods - mine seem to work very well.

But to get why I'm posting this so late - you mentioned Leica being careful only to claim precision for their rangefinders - not accuracy, well, the following, from their recent brochure seems to directly contradict that:

"The rangefinder system is an optical masterpiece, created for cutting edge photography. Unlike with a reflex system, where you have to focus through the lens and the focal length and speed determine the accuracy of the measurement, with the Leica M rangefinder the measuring basis remains constant regardless of the lens. This makes it many times more accurate at short focal lengths. The high-contrast focusing images in the center of the viewfinder field guarantees fast and precise focusing with pinpoint accuracy, even in very poor lighting conditions."

They seem to claim accuracy a few times here.

The comments to this entry are closed.