Autofocus

Pekka Buttler, 09/2022

Introduction

The entire history of the photographic industry can be seen as a story focused on making the photographer’s job easier while aiming for an improved ‘keeper rate’. One key trajectory in this history is encapsulated in the various contraptions related to helping the photographer achieve focus – most notably rangefinders, focus assists and autofocus.

Autofocus generally means any method by which the camera (and lens) will – once commanded to do so – automatically focus onto a given point in the frame. Autofocus therefore differs from focus assist which encompasses a wide range of technologies aimed at helping the photographer manually achieve focus (which will be addressed in another article at a later time).

While autofocus is intended purely to make the image-taking more straightforward and convenient, facilitating autofocus is a somewhat more complicated issue that has significant implications. Therefore, this article.

Given that JAPB’s main segment is related to adapting manual focus lenses, this article may feel a bit unrelated. Even so, some families of autofocus lenses are about to become legacy lenses. Further, understanding autofocus makes a lot of sense and helps in adapting lenses. At the same time, this article addresses the oft-expressed question: can you autofocus a manual focus lens?

BTW, as you’ll notice, this JAPB theory article is somewhat short on pictures, because we simply don’t have all the gear we speak of. If you have in your possession one of the contraptions mentioned and would like to donate a picture, feel free to be in touch.

A very brief history of autofocus on interchangeable lens cameras

While several AF-contraptions had been attempted earlier (including Olympus OM-30, Pentax ME-F, Nikon F3AF, Canon T80 ), the real start of autofocus interchangeable lens cameras happened in the form of the Minolta 7000 AF (Maxxum 7000 in North America; Alpha 7000 in Japan), which was introduced in 1985. This initial launch was quickly followed in 1986 by the Nikon F-501 and Olympus OM-707, and in 1987 by the Pentax SFX, the Canon EOS 650 and the Yashica 230-AF.

Various approaches to achieve autofocus?

Fundamentally, there are only a number of ways in which autofocus can be achieved:
1) A lens that can be commanded to focus independently of the camera;
2) A camera body that analyses the image and changes the settings of an adapter to bring the lens’ image into focus;
3) A camera body that analyses the image and changes the distance to the film plane to achieve focus;
4) A camera body that senses the distance to a subject and commands the lens to refocus;
5) A camera body that analyses the image and commands the lens to refocus.

All five above mentioned approaches have been used during the 1980s, and while today the last mentioned is clearly dominant, we will look into each of these, partially to illustrate why this is so.

1) Independently autofocusing lenses

While we’re familiar with lenses that refer to themselves as AF lenses, these lenses typically depend on the camera body to decide how that focusing is to be done. But there used to be lenses that could focus automatically, entirely without the assistance of the camera body.

These lenses would typically rely either on
a) sonar -like approaches that would use either non-visible (IR or UV) light or higher frequencies.
b) a beam splitter built into the lens that would use the actual TTL light to find best focus
In either case AF would be engaged by pressing a button on the lens.

Let’s have a look on the fundamental (meaning: not dependent on a particular implementation) pros and cons of this approach.

PROs
• Independent of camera, could be used on MF camera. Hence attractive on systems lacking AF bodies.
• As independent of camera, one design could easily be manufactured for several mounts.

CONs
• Such lenses unavoidably needed a built-in motor to drive AF (bulk), a power source (more bulk) and (with sonar-type lenses) sensors (even more bulk) to function.
• Focus could be easily confused; no clear focus points (as we’re used to nowadays).
• Photographer would need to use separate buttons for AF (on lens) and taking the shot (on camera)
• Any discrepancy in mounting distance would ruin AF
• Should one want to have more than one lens, each lens would need the whole apparatus for facilitating AF.

Some examples:
• Canon FDn 35-70 mm f/4 AF
• Chinon MC AF 35-70 mm f/ 3.3-4.5
• Chinon AF 50 mm f /1.7
• Cosina 75-200mm f4.5 MC Macro AF Zoom
• Sigma Universal Autofocus 55-200 mm f/4.5

2) Autofocus manual lenses using an adapter

Facilitating autofocus through using an adapter can mean two different things depending on whether one has flange focal distance to play with (or not).

2.1) WITHOUT flange focal distance difference to utilise:

Say you’re a camera maker that has recently launched your first couple of AF bodies, but you have a large back catalogue of manual lenses (and only few AF lenses), so you’d want those new AF bodies to be able to autofocus those old manual lenses. Assume further that the new autofocus camera has the same mount (or at least the same-ish flange focal distance) as all those old manual focus lenses. This was basically the situation facing all of those companies that in the 1980’s introduced their AF bodies – irrespective of whether they kept their mount through the transition (Nikon, Pentax, Olympus (kind-of)), or ditched their old mount (Minolta, Canon, Yashica). Obviously some manufacturers saw this as an opportunity to ‘incentivize’ customers to buy new lenses, whereas some others sought a way to allow autofocus on manual focus lenses. While Pentax also offered a similar solution, I’m going to use the Nikon TC-16A teleconverter as a case in point (because I happen to own it).

The Nikon TC-16A teleconverter is exactly that: A teleconverter. It gives a 1,6x magnification rate (roughly equalling a APS-C crop factor), with all the ensuing upsides (teleconversion=narrower field of view) and downsides (worsened maximum aperture), but the TC-16A has a twist up its sleeve: Unlike a regular teleconverter that relies on a set of lens elements to lie in a stable place and simply magnify the centre of the image, the TC-16A’s lens elements move (and move as bidden by a compatible body), thereby allowing the teleconverter to focus. While the movement range of the TC’s lens elements is not entirely sufficient to nail focus in every situation (e.g. you trying to focus on the horizon while your lens is set to MFD), it does allow functional autofocus on MF lenses.

PROs (general)
• Allow your brand-spanking new AF camera to autofocus your back-catalogue of MF-era glass.
• Do so without necessitating any weird contraptions in your AF body (the body will treat the adapter largely as it would any AF lens)
• As the AF-adapter needs move only a group of small lens elements (even when ‘autofocusing’ a heavy tele), AF has the potential to be fast.
• You get a teleconverter as an added ‘bonus’ (a PRO when you’re using tele’s)

CONs (general)
• It is a teleconverter, meaning crop factor (a CON when you want to shoot wide)
• It is a teleconverter, which means extra lens elements, loss of transmission and the introduction of various (mild, but even so) aberrations.
• AF is not perfect, as you’ll likely have to intervene manually to put the MF lens’ focus into a region from where the AF-adapter can take over.

CONs (specific to Nikon TC-16A)
• Nikon did something to cripple the adapter, meaning that it works only on some of Nikon’s Film SLR’s and on only one of the Nikon dSLR’s without modification.

2.2) WITH flange focal distance difference to utilise:

As anyone who’s ever worked with a bellows knows, one can easily shift a lens’ focus simply by moving it in relation to the sensor/film plane. This makes it theoretically possible to construct an adapter that – by moving the lens backward/forward – could facilitate focusing and, by extension, autofocusing. Problematically, that adapter itself will need some space to work with.

In the age of the 35 mm format film SLR, the issue of flange focal distance was always a balancing act. On the one hand, the mirror box of the SLR always needed some space to work with (and trying to cram too much in too little space always increased mechanical complexity). On the other hand, if the back focusing distance (distance between film/sensor and rearmost glass surface of lens) grew too large, this led to limitations on lens design. Hence, the absolute majority of 35 mm film SLRs have a flange focal distance between that of the Konica AR mount (40,5 mm) and the Leica R mount (47 mm), and the range of AF-capable SLR mounts is even narrower (Canon EF at 44 mm to Nikon F at 46,5). An adapter designed to autofocus Leica R lenses on a Canon EF body would have only 3 mm to work with.

But with the introduction of mirrorless cameras, there came a radical change as mirroless cameras need no mirror box, and – as a result – modern mirrorless full frame mounts have a flange focal distance ranging from 16 mm (Nikon Z) to 20 mm (Canon RF). Suddenly there are millimetres to play with, and it becomes more than possible to construct an automated helicoid adapter (an AF adapter) that – by shifting a lens’ distance from the sensor – facilitates autofocus. Case in point (examples):
• Techart Pro Leica M to Sony FE Autofocus adapter
• Megadap Leica M to Nikon Z autofocus adapter

These adapters can then be used together with further adapters to allow AF on full-frame mirrorless on a wide range of legacy lenses (e.g. Konica AR-> Leica M; Leica M –> Sony FE).

PROs
• Allow your MILC to autofocus a wide range of MF-era glass.
• Do so without necessitating any weird contraptions in your body (the body will treat the adapter largely as it would any old lens)

CONs
• As the AF-adapter needs shift the entire weight of the lens, it cannot handle heavy lenses.
• AF is not perfect, as you’ll likely have to intervene manually to put the MF lens’ focus into a region from where the AF-adapter can take over.

3) Autofocus manual lenses through changing the mount–film plane distance (body-only AF)

Whereas autofocus adapters such as the Techart Pro facilitate focus by adding a variable-length adapter, hence moving the lens vis-a-vis the film plane, the same result can be achieved by
a) shifting the lens mount vis-a-vis the rest of the camera body, and
b) shifting the position of the film plane within the body.

The a-option (shifting the lens mount vis-a-vis the rest of the camera) is an approach not totally uncommon in manual focus medium format cameras (in effect we’re talking about an integrated bellows), but I am not aware of the approach ever having been used in a 35 mm SLR. I have to wonder why not…

The b-option (shifting the film/sensor plane within the camera body) on the other hand would feel like a non-starter, as it unavoidably leads to a very bulky film body. Surprisingly, this method has been used in a serially produced 35 mm film body. Once. In the Contax AX (1996) (see a detailed article here).

While seemingly very similar (in that AF is facilitated in-body), these a- and b- approaches nevertheless place very different demands. As any SLR always focuses based on that the focusing screen (manual) or autofocusing circuit (AF) is at a comparable position to that of the film plane, shifting the film plane within the camera also necessitates that whatever AF circuitry must either move as well (to find focus) or have some magic to otherwise deduce the needed shift in film plane. When, OTOH, the flange focal distance is modified by extending the mount vis-a-vis the rest of the body, there is no further need for shifting any other components.

PROs
• Allow your SLR to autofocus manual focus lenses (of the same mount)
• Do so ‘out-of-the-box’ (without adapters or other gimmickry).

CONs
• Will add significant amount of moving parts to the body.
• Must be very sturdy to accommodate heavy lenses without play (for option a: integrated bellows)
• Must retain calibration indefinitely or necessitates regular recalibration (for option b: moving film plane)

4) A camera body that senses the distance to a subject and commands the lens to refocus

Just as it was possible to construct lenses that used some sonar-type arrangement to measure distance, the same could have been used in a camera body that would then – using either a mechanical or electronic connection – command the lens to focus on a specific distance.

While this concept was utilised in some film-age compact cameras (e.g. the Canon MC) as well as in an SLR of a special kind (The Polaroid SX-70 Sonar OneStep) I am not aware of such a lens/body combination on an interchangeable lens SLR.

PROs
• Facilitates autofocus (finding the distance to focus on) without complications in the light-path.

CONs
• Focus could be easily confused; no clear focus points (as we’re used to nowadays) and photographer would not be sure what the camera has focused on.
• Would necessitate some way for body to know how to focus a on specific distance – independent on which lens is mounted.
• Must retain calibration indefinitely or necessitates regular recalibration.

5) A camera body that analyses the image and commands the lens to refocus

This is by far the most common approach and has today become a de-facto standard. This approach always relies on a division wherein the ‘smarts’ resides entirely in the camera body, and the lens is relegated to the position of being an obedient appendage.

However, there are multiple ways in which this can be implemented and there are several different implementations. Therefore pays to have a look at the details:

5.1) Mechanical or electronic focusing communication

There are fundamentally two approaches through which autofocus-compatible lenses can focus as directed by the camera: mechanical or electronic. Both approaches have some inherent advantages and disadvantages

Mechanical linkage
In the mechanical approach the actual focus motor resides within the camera body, which then drives the autofocus gearing in the lens through a mechanical interconnect – typically a slot drive screw.

This has the distinct advantage that one focus motor can drive many lenses and that the lenses are simpler and hence cheaper.

On the other hand, it has the disadvantage that all lenses (from the nifty fifty to the super tele) need to be driven by the same motor, and that heavier lenses hence are slower to focus.

Electronic linkage
In the electronic approach the focusing motor resides within the lens, and that is driven through an electronic current fed to the lens over a set of electronic contacts.

The most distinct advantage of this is that different lenses can have different types of focus motors, hence allowing for a wide spectrum of implementations (fast and slow motors; noisy and silent motors; cheap and expensive motors). While this approach also typically produces lenses that focus faster and more silently, the micro-motors used in early Canon AF lenses show that neither (speed nor silence) is an inherent trait of in-lens focusing motors.

The obvious downside is that you will need a separate motor for each lens, making lenses more expensive to manufacture and more difficult to service.

The trend is toward electronic
Historically, most manufacturers started their autofocus-adventures using a mechanical linkage for focusing. Hence, you can find slot-drive screws in lenses of many kinds of mounts (Nikon F, Olympus OM AF, Pentax K, Sony/Minolta A, Yashica AF). Canon, on the other hand, started in 1987 out using a pure electronic linkage (even aperture stop-down was electronic) – a point that Nikon reached only after introduction of electronic diaphragm lenses in 2014.

While Yashica and Olympus quickly dropped out of the AF SLR business, Minolta, Nikon and Pentax gradually moved towards using electronic linkages. Typically, the first lenses to receive lens-internal focus motors were super teles, which were too heavy to be driven by body-internal focus motors, and thus focused too slowly. These were followed by pro zooms. Nikon introduced the first AF-I lenses in 1992; Minolta introduced their first SSM lenses in 2003; and Pentax introduced they first SDM lenses in 2007. As of this writing, no SLR AF lenses using mechanical AF linkages have been introduced in many years.

Adaptability of mechanical/electronic AF lenses:
Given that all modern AF bodies are geared towards using electronic linkages to facilitate autofocus, electronic AF lenses are in principle relatively easy to adapt: Canon, Nikon, and Sony all offer their own adapters for facilitating autofocus of electronically driven dSLR lenses on mirrorless. Likewise, many third party manufacturers offer similar smart adapters. However, as some lens manufacturers have never published their electronic interface specifications, the reliability (especially in the long term) of third party solutions can be dubious.

Being able to adapt an electronic AF lens to a mirrorless camera (with AF) however always necessitates an adapter built for precisely that combination and currently (2022) there are still several combinations for which no smart adapter is available.

In theory, there is no real impediment to constructing adapters that facilitate AF with full-electronic bodies on lenses that use a mechanical approach to AF (slot-drive lenses), as amply demonstrated by the existence of several such adapters (especially for Contax G, Pentax K and Sony/Minolta A mounts). Examples: Techart Pro TXG-01 (Contax G->NikonZ); Monster Adapter LA-KE1 (Pentax K->Sony FE); Sony LAEA5 (Minolta A->Sony FE)

However, in practice, there are several slot-drive AF systems that are (currently) not supported at all, and even more numerous combinations that are currently not available.

5.2) Methods for ascertaining focus

Pretty much all early autofocus systems in SLRs (up until the introduction of live-view on dSLRs) are based on a concept referred to as Phase Detect AF. Classic Phase Detect is based on a few core principles that everyone should know about:
• Focus is never ascertained at the film plane, nor is it typically ascertained in the primary light-path (from lens, via mirror and pentaprism, to the optical viewfinder), but there are exceptions
• Instead the main mirror is semi-transparent so that a share of the light that hits the mirror passes through it.
• That passed-through light hits another mirror behind the main mirror which directs it in another direction (traditionally down, as that’s where there was space to be found in film SLRs, but sideways could also work in dSLRs. That secondary mirror is often referred to as the ‘secondary mirror’ (yes, I know).
• After the secondary mirror, the light is passed through another lens or two, is (typically) bounced off another mirror and passed through some filters before it hits the Phase Detect sensor.
• By then, light that originated from points in the landscape that were not in focus will have been divided into different phases (whereas correctly focused light will not be divided into different phases).
• The quality of Phase Detect AF systems is typically determined (hardware) by the number of focus sensors (a.k.a. AF points, see below) and their sensitivity in combination (software) with the system that interprets the data.
• The genius (and, today, the one real competitive advantage) of Phase Detect AF is that
a) the sensor will not only be able to ascertain whether the point is in focus or not, but will also be able (by studying the ‘distance’ between phases) to ascertain quite exactly in which direction and by how much the lens will need to refocus. How exact that estimate is, depends on the quality of the electronics and on their calibration.
b) thanks to its ability to compute how much out–of–focus an object is (possibly combined with real-time knowledge of what distance the lens is focused at), the Phase Detect sensor will be able to track, even predict movement, hence enabling very advanced AF tracking, even prediction.

The real problem with Phase Detect AF is that because focus is not confirmed at the film/sensor, but along a totally different (and seriously convoluted) light path, there is no guarantee that what the Phase Detect AF sensor thinks is perfectly focused matches perfect focus at the film/sensor plane. While camera makers have honed their quality control to ensure that calibration is OK on leaving the factory, even these sometimes fail, even en masse. Problematically, as cameras are not intended to be stored in cotton wool, there is amble opportunity for calibration being lost in use. As a piecemeal fix to this calibration issue, most dSLRs allow users to tweak the calibration (within limits) in the camera’s menu system.

Since the advent of digital cameras, another approach has been available, that is typically referred to as Contrast Detect(ion) AF. Contrast Detect AF is based on interpreting the live feed from the sensor, aiming to set the lens’ focusing distance so that the live feed’s contrast is maximal at the position of the focus point. Contrast Detect AF is the predominant method for focusing in everything from smartphones to digital compact cameras, and also plays a role in most mirrorless interchangeable lens cameras. Even dSLR’s use Contrast Detect AF in live view mode.

Whereas Phase Detect AF necessitates purpose-built hardware, and is typically limited to a specific number of focus points (usually clustered near the image centre for optical reasons), Contrast Detect AF needs no extra hardware, is highly flexible and the focus point can be anywhere in the image frame. Moreover, as focus is always computed at the sensor plane, focus calibration becomes a non-issue.

But Contrast Detect AF also has its weaknesses. Firstly (the hint is in the name), for Contrast Detect AF to work its magic, there needs to be some serious contrast to be detected. Secondly (and this is the real kicker), Contrast Detect AF initially only knows that a focus point does not show contrast, and hence needs to ‘hunt’ for focus in order to ascertain the distance of optimal focus. As a corollary, basic Contrast Detect AF also has very weak tracking abilities.

So what do engineers do when they have two, mutually exclusive approaches that have sets of PROs and CONs that seem to complement each other perfectly? They find a way to combine those approaches. That is the goal of Dual Pixel AF.

Dual Pixel AF, sometimes also (IMHO: misleadingly) referred to as Phase Detect (which I, to avoid risking confusion, will refer to as on-sensor phase detect), is based on that not all the pixels on the sensor are designed for capturing the final image, but a number of them (less than 1/1000) are designed to facilitate on-sensor phase detect by using basically the same concept as (classic) Phase Detect AF. Cameras equipped with such sensors (and the software to match) can choose to focus based on any combination of Contrast Detect and on-sensor phase detect. This is a major advantage as utilising on-sensor phase detect significantly speeds up the focus acquisition process and allows superior (compared with contrast detect) tracking abilities.

5.3) One AF point or many AF points

The era of off-sensor phase detect

Early autofocus camera bodies worked with a single AF point in the centre of the frame. The obvious limitation of this is twofold: Firstly, if you were trying to track (See ‘continuous focus’, below) a moving target, you would be at pains to try to keep the centre AF point locked on the moving target, and whenever your subject would elude your single point, the camera would start focusing on (typically) the background leading to (especially with slow, early autofocus lenses) significant time without correct focus and a lot of missed opportunities.

Secondly, unless the point you would want to focus on was in the absolute centre of the frame, you’d be forced to use the focus and recompose-method, which in many cases is cumbersome and may not produce good results with every lens (due to a combination of shallow depth of field and field curvature).

Hence, camera manufacturers started from early on designing Autofocus SLRs that would have more than one focus point – first three, then five, then seven, then eleven …

The increasing number of AF points brought three types of advantages:
Firstly, no longer were photographers limited to focusing at the centre of the frame. While one was still often unlikely to have an AF point exactly where one would need it, it led to a decrease in the angular distance one would have to tilt the camera while recomposing, hence also bringing improvements in the results of the focus and recompose-method.

Secondly, once one had a sufficient number of AF points at sufficiently close quarters, there was a high likelihood that a subject that was dropped by one AF point could be picked up by another, hence leading to massive improvements in subject tracking.

Thirdly, instead of using one AF point at a time, the AF systems could be programmed to be ‘smart’: scan the distances to all points (in the vista) corresponding to an AF point, and picking up what – in all likelihood – was the intended subject (typically the cluster of AF points that would show distances far closer than the rest)

Problematically, these multi-focus point systems still had to live within the limitations of how off-sensor phase detect works (i.e. at the end of a convoluted optical path (see above), with a significant increase in aberrations on rays that are far off centre). This resulted in that the whole bunch of AF points typically were grouped around the centre of the frame, so that even dSLRs that sported mind-boggling numbers of AF points (Such as the Canon 1Dx Mk3 with its 191 AF points) were constrained by that all those AF points occupied the centre of the frame, mostly leaving 3/4 (or more) of the frame without a single AF point.

Moreover, due to how off-sensor autofocus works (the mirror needs to be down), all autofocusing systems that relied on off-sensor (a.k.a. classic) phase detect were somewhat limited with regards to rapid-fire continuous shooting modes (the AF system would always lose function whenever the mirror was raised, hence needing to reacquire focus between shots) and entirely inoperative in ‘movie mode’

On-sensor AF

Interestingly, contrast detect AF has none of the same limitations (it has, as already mentioned, a different set of limitations). Contrast detect could be used on any part of the sensor (including the absolute corners which were poison for off-sensor phase detect) and while contrast detect might at times struggle more in corners, this was typically a lens issues (off-axis aberrations decreasing contrast, hence making contrast detect’s job more difficult). But, in effect the concept of AF points loses meaning in contrast detect AF.

On-sensor phase detect still uses physical AF points (if you can call them points), but the technology not only allows them to multiply in numbers (into the thousands), but also allows them to spread out over a wider portion of the frame (many such systems do not cover the entire frame, but a significantly larger share than off-sensor phase detect.)

Again, given that there is no mirror, the sensor is gathering light all the time, and – hence – the camera’s smarts can analyse the scene constantly, it is more than possible to do some truly crazy magic:
– Hybrid AF (Get focus in the ballpark with phase detect, then fine-tune it using contrast detect).
– Face AF / Eye AF – further improvements on Hybrid AF, by automatically selecting those areas (faces, eyes) for which focus should be optimised.

Face Detect(ion) AF and Eye AF are not Autofocus technologies as such, but are extensions of the basic principle behind Contrast Detect AF, namely the camera’s ability to read and analyse the live feed, find faces/eyes, and make sure these are in focus. Some such systems even know to detect the eyes of birds in flight (!)

5.4) Single and continuous focus

In the simplest terms, there are two modes for autofocusing: One where the correct focusing distance is determined once, and another where the camera’s focusing system continuously updates the correct focusing distance (the names of these modes vary from one brand to another, but typically contain the key words single and continuous or the letters S or C). Obviously, one of these is best suited for stable setups (neither the photographer nor subject is moving), whereas the other one is optimised for adjusting for motion.

Typically, one of the key tasks of an autofocus photographer is to choose their AF mode (single/continuous), and I’m sure I’m not alone in having bungled a shot because I forgot to check which mode my camera was in. Hence, it is not surprising that some camera manufacturers have long offered to help photographers do this by implementing a way for the camera to select between these autofocus modes as it sees fit (Canon calls it ‘AI Focus’; Nikon calls it AF-A or auto autofocus (sic!)). Simultaneously, many feel these contraptions to be worse than useless (and pro cameras typically omit this feature).

Finally, now that almost every dSLR and MILC also doubles as a video camera, there has arisen a clear need to have cameras maintain continuously focusing even though no-one is pushing a button. This is implemented through what is called full-time-autofocus, and is – in essence – an extension of continuous focus.

Especially within the field of continuous focus and video-oriented AF functionality, there are so many recent developments (especially in high-end dSLRs and mirrorless cameras), that the mind boggles. Moreover, as this article is not intended as a review on state-of–the-art autofocus (but to inform those that use legacy, manual lenses on modern cameras on the ramifications of autofocus), we will not go into this in more detail.