[Lowell Goudge]

Just a dumb idea, I was thinking about today. Why not implement a couple of auto focus functions that could be activated by the green button.

The first was to have an auto focus limiter, which would limit the range within the depth of field (assumes Av or Tav mode ), This would be great for doing shots where there is some foreground interference, twigs etc,

This obviously could lead to focus bracketing. Since for new AF lenses the camera knows aperture, hence also depth of field, as well as focus distance(when you press green button) hence it could easily do a 3-5 frame bracket around the focus point.

[jonby]

Not a dumb idea at all. I've been meaning to post about this for ages now. Cameras have bracketing for loads of silly things which are of virtually no use, but nothing for one of they key photographic components of any exposure - focus. I don't know of any camera to date which implements this feature and I can't understand why - it has always seemed like a no-brainer to me. I have wondered if there is some technical reason which would make this difficult to implement, but can't think of one. I can only think that it's a marketing thing - implementing this feature would be an admission that autofocus can actually sometimes be wrong! I personally don't use autofocus in any situation where I have the time to do it manually for this very reason - it simply isn't accurate or reliable enough. In particular, autofocus systems aren't capable of assessing optimum focus for the entire frame and ignore such things as the curvature of field which many lenses exhibit.

I manually focus-bracket almost every shot I take, if the subject is static. Automatic focus bracketing could greatly speed up this process for me. Given that AF accuracy is one of the most commonly talked about obstacles to getting sharp images, I'm sure it would be of great use to many people - particularly when using fast lenses at max aperture, where autofocus is notoriously unreliable. I actually think it is potentially so useful that it should be implemented close to the top level of controls in the 'ease of access' hierarchy - probably assignable to a hard button, or at least to have easy access from the function menu.

Trying to think through any issues in implementing this system, probably the main one relates to variance between lenses, which all require different degrees of travel for the same shift in focus distance, plus of course different aperture settings might require a different degree of focus shift to be optimal. In theory, firmware could be developed which allows for the camera to come with a good set of defaults for current lenses, which take into account focus throw etc, but which allows user customization to be saved for each lens, as it currently is for fine focus adjustments. However, even a basic implementation which just uses a single set of user-controllable parameters which is easily changed on a per-shot basis would be of huge benefit. This could be implemented in a similar way to exposure bracketing, where the user chooses the number of 'points' on a scale, forwards or backwards of 'normal', and also whether to shoot three or five images.

Getting truly optimal focus is one of the hardest technical details to get right in photography - one of the few which has not yet been nailed by technological development. Focus bracketing could help immensely with this - and in my eyes be a lot more useful than having a zillion focus points - and it doesn't even need additional hardware to work - just firmware changes. If fact it could even be done through a firmware update.

... or am I missing something? Would be interested to hear other opinions on this - would you find this useful? Why hasn't it been implemented before?

[fgaudet]

I think that the problem is that the camera only knows when it is in focus, it only has a rough idea of how far the subject is and has no idea of where the focus point is for that specific lens. And as jonby mentioned, focus throw will vary quite a bit between lenses.

You could however put a "calibrate lens" thingy in the camera, but on the far end, infinity is not really a specific distance... so your lens would know that from 3" to infinity it needs to turn the AF motor for 100 turns but it wouldn't know how much distance one turn would be. And take a look at one of your lens distance scale, you'll see that it's exponential and not linear. It would probably be possible to create and include lens profile in firmware at the factory but that would require a huge amount of precision and better, more precise position (distance to subject) feedback coming from the lens. Because getting the focus right on a deep landscape photo isn't hard but for closer subject, like a macro shot, focus has be to perfect, the amount of work needed to create a good profile would be insane. And it would probably still need to be calibrated on every single lens and every single body to account for variation within manufacturing tolerance.

It could probably be done, I'm not sure it would be worth the time and money. There is a reason why AF isn't perfect, there are so many variables to account for. R&D time is better spent on getting better, faster AF that would impact 99% of the users versus a feature that only a small minority uses.

[BrianR]

I've always wanted a programmable autofocus bracketing mode. One you could set to various fractions of the DoF (assuming a standard print size, viewing distance, etc. as a reference point) and # of shots. Using a modern lens with quick-shift focus, it should also be useable in manual focus mode- when you click the shutter the bracketing takes over and fires off a picture a little in front and a little behind.

You could however put a "calibrate lens" thingy in the camera, but on the far end, infinity is not really a specific distance... so your lens would know that from 3" to infinity it needs to turn the AF motor for 100 turns but it wouldn't know how much distance one turn would be. And take a look at one of your lens distance scale, you'll see that it's exponential and not linear..

The DoF scale on a lens is linear with the degrees of rotation of the focus ring. I'm not sure how accurate the scales on lenses are, and if it's just an approximation to the DoF for given parameters (print size, viewing distance, etc.) or if the focus throw is actually 'geared' for it to be linear. Having focus bracketing where I could just choose the number of revolutions of the screw drive would be enough to make me happy and would be something I'd use.

[DavidDwight]

I have long thought this would be a good feature too. They can do micro-calibration of the lens why not bracket with that? It would help calibrate the lens/body combination.

[wombat2go]

I made this micro-focussing contraption with an electric drive and dial indicator for creating a focus stack of about 5 frames through the focus point.
I only used it for slide copying /macro. I never tried it outside, and would not be game to use it in public!

[jonby]

Good to see some action this thread again.

I made this micro-focussing contraption with an electric drive and dial indicator for creating a focus stack of about 5 frames through the focus point.

That looks very impressive. Do you find it works well for your purposes? With the experience you have gained through making this contraption, what do you think of the idea of implementing focus bracketing in-camera?

. And take a look at one of your lens distance scale, you'll see that it's exponential and not linear.

Yes the focusing scale is exponential, but this is compensated for precisely I believe by the expansion of the DOF - hence why it's possible to have fixed DOF markings on the lens barrel. In other words, no matter what the focusing distance, a turn of 1 degree from the point of perfect focus will give the same apparent change in sharpness, all else being equal. So I'm not sure that the situation is as complex as you are suggesting. When I manually focus-bracket, I have a good idea of how much of a turn is needed with each lens in order to gain a small but significant change in focusing distance, and this is more-or-less the same no matter what the working distance. This is easy enough to do with manual lenses, but most AF lenses make this very difficult, due to looseness in the focusing helicoid, very short focus throw, and poorly marked distance scale (not enough markings).

There is a reason why AF isn't perfect, there are so many variables to account for. R&D time is better spent on getting better, faster AF that would impact 99% of the users versus a feature that only a small minority uses.

But PDAF has been in continuous development since the early eighties and still can't be relied upon to be accurate, for the reasons that you give. You could say the same about exposure systems. AE bracketing is more or less a given on modern cameras to account for this fact, yet you dismiss AF bracketing as a niche feature.

Leaving aside any difficulties in implementing this, I see this as having mainstream appeal amongst enthusiasts and pros.

[fgaudet]

In other words, no matter what the focusing distance, a turn of 1 degree from the point of perfect focus will give the same apparent change in sharpness, all else being equal. So I'm not sure that the situation is as complex as you are suggesting.

Feel free to correct me if I am wrong but I as far as I understand it, the autofocus system works in a different way than you seem to think.



The system is more like:
(AF system) Hey lens, you're out of whack, according to my calculation, you should move forward. And you're way off so make it quick
(Lens) Roger that, forward full speed.
(AF) Slow down you're getting there.
(Lens) slowing down.
(AF) OK STOP!!!
(Lens) Done! According to what I can measure, I think I am currently at about 4.6m from the target, just so you know. Take your shot.

And not like this:
(AF) Lens, I command you to turn 135.456392 degree in clockwise rotation in order to achieve focus, this is 165 1/3 turns of the focus motor.
(Lens) Copy that, I'll let you know when I get there.
(Lens) I'm there, and write down in your report that the object in question is located exactly 4.653344 meters from us. Fire at will!



I don't think anything in the system is aware of the number of degree the focus ring moved in order to achieve focus. There is no way for the focus motor to know how many turns are required to get from close to far focus (screw drive doesn't know for sure, SDM/DC only maybe, but why bother adding this since it's not required). PDAF knows when stuff is in focus or not and by how much to a certain extent but it won't know how long it will need to ask the lens to turn in order to maintain a proper focus for stacking. A 90 degree focus throw lens might require the screwdrive to turn 100 times where a 230 degree one will require 600 turns... and another 230 degree one only requires 60 turns of the screwdrive... and then there is the gearing inside the lens, and quickshift and lens variation and temperature variation. And what if your lens had to be adjusted?

Simple in theory, not so simple in practice.

[wombat2go]

The way I did that in 2012 with the "contraption" was to first run the drive to sharpest focus in centre of frame.
Then zero the dial indicator.
Then run the drive to + 126 micron ( +0.005 inch), and take a succession of 5 frames at 63 micron (0.0025 inch) intervals +5, +2.5, 0, -2.5. -5 thou inch. *

I think that meets Lowell's suggestion in o/p .

Since 1/f = 1/u +1/v, the camera might need to have a counter ( or etc) on the drive pinion so it knows where it is. The user might need to enter his preference for the micron spacings.

I could see that the focus stack sharpens the macro image radially, and also increases the DOF. After some tests I ran the lens at about f/10 which seemed to produce the sharpest final.

The "contraption" was operated manually to do the above. I have the idea to hook up a pico to automate the above but I don't know if i ever will.
Th big bottleneck (I was using Olympus m43 and Pentax K-01) was the lack of tethering to download the 5 big raw files between frames. Also neither of those cameras has a remote shutter control for auto frames. The ist ds here does have that but its Pentaprism nose precludes its use on the "contraption".

* The reduction on the 12000rpm motor to do this was huge: first a gearbox, then a leadscrew then a crank.

[Lowell Goudge]

Feel free to correct me if I am wrong but I as far as I understand it, the autofocus system works in a different way than you seem to think.



The system is more like:
(AF system) Hey lens, you're out of whack, according to my calculation, you should move forward. And you're way off so make it quick
(Lens) Roger that, forward full speed.
(AF) Slow down you're getting there.
(Lens) slowing down.
(AF) OK STOP!!!
(Lens) Done! According to what I can measure, I think I am currently at about 4.6m from the target, just so you know. Take your shot.

And not like this:
(AF) Lens, I command you to turn 135.456392 degree in clockwise rotation in order to achieve focus, this is 165 1/3 turns of the focus motor.
(Lens) Copy that, I'll let you know when I get there.
(Lens) I'm there, and write down in your report that the object in question is located exactly 4.653344 meters from us. Fire at will!



I don't think anything in the system is aware of the number of degree the focus ring moved in order to achieve focus. There is no way for the focus motor to know how many turns are required to get from close to far focus (screw drive doesn't know for sure, SDM/DC only maybe, but why bother adding this since it's not required). PDAF knows when stuff is in focus or not and by how much to a certain extent but it won't know how long it will need to ask the lens to turn in order to maintain a proper focus for stacking. A 90 degree focus throw lens might require the screwdrive to turn 100 times where a 230 degree one will require 600 turns... and another 230 degree one only requires 60 turns of the screwdrive... and then there is the gearing inside the lens, and quickshift and lens variation and temperature variation. And what if your lens had to be adjusted?

Simple in theory, not so simple in practice.

t

I think there might be a different way. Simply by having an offset in the ficus adjust, that you vary over the range of AF adjustment, that way even if you have an AF error on a lens,may bracketing around the AF adjustment point you might just get one shot that's perfect

[jonby]

I don't think anything in the system is aware of the number of degree the focus ring moved in order to achieve focus. There is no way for the focus motor to know how many turns are required to get from close to far focus (screw drive doesn't know for sure, SDM/DC only maybe, but why bother adding this since it's not required). PDAF knows when stuff is in focus or not and by how much to a certain extent but it won't know how long it will need to ask the lens to turn in order to maintain a proper focus for stacking. A 90 degree focus throw lens might require the screwdrive to turn 100 times where a 230 degree one will require 600 turns... and another 230 degree one only requires 60 turns of the screwdrive... and then there is the gearing inside the lens, and quickshift and lens variation and temperature variation. And what if your lens had to be adjusted?

I'm sure some of this is correct, but I still believe that it is entirely feasible to implement an AF bracketing facility. Yes, of course different lenses require a different amount of screw drive turn to achieve the same degree of de-focus, but it should still be possible to offer a scale which allows people to choose an amount to shift forwards and backwards. It would then be up to the photographer to do some tests and set an amount which works for them, for a particular lens. This would be useful even if you just had to control it completely manually, but it would certainly be possible to record this setting in a profile and have it brought up automatically for each lens, in exactly the same way as AF adjustments are. Also, given that it's possible to calibrate AF so that it settles a certain amount forwards or backwards from where the camera thinks the image is in focus, surely this indicates that the camera is able to achieve exactly the same kinds of shifts that a bracketing facility would require.

For Canon and Nikon systems, focus stacking software is available (eg Helicon Remote) which is able to automatically rack the focus through a range of distances for a selectable number of steps, which is a form of AF-bracketing. If this is possible through tethered software then surely it is possible to make the camera's firmware do the same? And yet to my knowledge this has never been done on those systems. This suggests to me that the decision not to implement this feature on those systems at least is due to something other than technical barriers. It does not necessarily follow that the same technical abilities are available on the Pentax system, but neither have I heard any convincing evidence to the contrary so far.

[dakight]

I was just thinking about this tonight and wondering why it isn't part of the feature set. It shouldn't be that hard to define and mathematically calculate something analogous to the f-stop for focus then bracket 3 or 5 exposure at user defined "stop" intervals from the calculated ideal focus point or even the manual focus. The focusing system would have no need to analyze the image, just incrementally move the focus mechanism a predetermined amount. In-camera stacking could even be implemented without too much trouble.


I intend to experiment with focus stacking and I suspect it will be tedious at best to manually dial in precise increments from front to back. I don't know how precise those increments even need to be so it will be a learning experience for me. A simple focus bracketing function implemented in the camera body would take a lot of the tedium out of the process.

[Nass]

I've been wondering why camera manufacturers didn't have more functionality inside bodies for years now, and all I can think of is that it's some sort of in-body processing power/memory limitation, possibly a hangover from the old days. When I compare the camera sector to the IT and telecoms sector that I have the dayjob in, I'm constantly amazed by how behind the times camera manufacturers are. Fwiw, one of the most interesting things out there is magic lantern, a 3rd party adding all sorts of interesting functionality to Canon cameras... including focus stacking. God how great it would be if Pentax opened up the FF for this sort of development.

[stevebrot]

I don't think anything in the system is aware of the number of degree the focus ring moved in order to achieve focus.

You are correct. The idea that the body knows exactly how the lens is doing its work is one of the most common misunderstandings of users on this site who are seeking to tweak their AF systems.


Steve

[stevebrot]

Fwiw, one of the most interesting things out there is magic lantern, a 3rd party adding all sorts of interesting functionality to Canon cameras... including focus stacking. God how great it would be if Pentax opened up the FF for this sort of development.

It would be great if Canon did too! Magic Lantern is based on a reverse engineering of Canon's tethering interface. It is an accident of design that this is possible. As a result, support for recent models is a little sketchy. I may be wrong, but I believe Nikon is the only maker that actually publishes an API (programming interface) for their cameras.

https://sdk.nikonimaging.com/apply/

Of course, the exposed functions are fairly limited


Steve