Web Analytics RiceHigh's Pentax Blog: Live View & AF Methods Fully Dissected

Monday, September 22, 2008

Live View & AF Methods Fully Dissected


For quite a while, there has been a new trend for DSLRs to have the Live View (LV) function. Despite some people would question the usefulness of the LV, it would be interesting to learn that “What does a Live View screen add to a DSLR?”, just read that article to have some ideas.

In short, Live View opens up new photo opportunities and it could be an invaluable tools for more shooting flexibility and for new creative compositions which would have never been possible without Live View but only with special, expensive or bulky accessories for finder optics (which are yet not as versatile).

The Basics (Working Principles and Pros & Cons)

1. Live View CMOS Imagers (for DSLRs)

As we all know, the CCD and CMOS Imagers used in DSLRs are much larger in size and area than those used in smaller DCs for P&S shooters. The design of those DSLR imagers are somehow quite different than those smaller imagers from the very beginning as all DC imagers were capable to do LV on day one. In contrast, DSLR imagers are designed to have the highest possible overall picture quality which means higher pixel count, larger photo site / pixel area, wider dynamic range, lower noise at higher ISO sensitivities and so on.

Until Olympus created the E-330 in January 2006 with a brand new “Live MOS” sensor, the DSLR imagers started to be LV capable.

So, what makes a sensor to be LV capable. There exists a number of criteria for the imager to fulfil, namely:-
  • High enough refresh rate – so that the LV display can be updated seamlessly as viewed, say, at least more than 25 fps;
  • Not generating much heat during LV and/or to dissipating the heat quickly and efficiently/effectively enough;
  • Need to have a short and fast reset after leaving the LV and be prepared to take the actual image. The reset is a discharge process where the shutter need to be closed and let all the stored charge in all pixels of the imager to be drained. Anyway, no matter how fast or slow the discharge process is, it just means there will be a *delay* time to be waited.
But then the problem is, when the camera is doing LV using the imager, the mirror must be lifted up during the whole LV process, thus losing the Phase Matching/Detection AF (See more of my brief explanations below) capability. If the Phase Matching AF is the only means to measure and achieve the focus, then the LV must be stopped if AF is required. As such, the mirror is lowered so that the Phase Matching Detection AF system can be used again.

The BIGGEST Advantage of using imager as the LV sensor is that it is simply WYSIWYG (What You See Is What You Get), which just means previewing the whole image 100% in area is possible and the user could preview other picture effect clearly and accurately such as the exposure and colours, focus, Depth of Field (DoF) and so on.

As for DCs, the exposure control for the displayed LV image (brightness) could be achieved by dynamic aperture opening and/or electronic shutter of the imager itself. I have no information about how different (brands of) DSLRs control this and their means, though. But some DSLRs are better implemented than others and in general, the later the product, the more mature the LV is.

There is so far no larger CCD Imager for DSLR, unlike those very small CCDs for P&S DCs which all have such LV function (and of course, as said before).

2. Separate Live View Sensor in the Optical ViewFinder

As seen from these two diagrams for the E-330, the working principle is to split some light from the optical viewfinder image to the "Full-time" LV Sensor, but it should be noted that it is NOT exactly WYSIWYG. Moreover, there is light loss as such and hence a relatively dimmer viewfinder image is resulted.

Even worse, with the accommodation of this additional sensor and its associated optics, making a brighter, larger (higher magnified) projected viewfinder image with a higher percentage of actual image coverage has become more difficult. Particularly, 100% coverage is nearly impossible for such a design unless additional full scale half-mirror or split-prism is included to split the lights of finder image into two paths, one to the eyepiece and another to the LV sensor. Even putting aside the complexity of the required optic design, the whole viewfinder will be made rather large and heavy. And the Physics still is, the optical finder image will still be *dimmer*.

Another worse evil few people will notice is that at low light, this LV design simply won’t work well. The displayed LV image will become noisy owing to large amplification required for such only tiny amount of light as received by that LV sensor.

Oh well, when next time DPR and some other reviewers say loudly for the “Seamless Live View Operation” they like to emphasize from time to time, I think it would be better for them to tell more about the above mentioned two killing shortcomings!

3. Phase Matching AF Detection System

Basically, there is a set of AF measuring module with the required optics and electronics are placed under the mirror house, see this.

The Phase Matching Detection method is mature and should be fast and accurate by itself (as the AF systems are all designed based on). However, under LV by the Imager, there are additional overheads of mirror down and up in which time is added for the whole AF operation.

4. Contrast Detection AF (Software Algorithm)

It is just a pure software thing to detect the in-focus point just like what nearly all the P&S DCs do. The implementation is relatively slow for DSLRs, in general. For example, it has been verified uncountably by different reviewers and users that even latest Canon DSLRs in this mode could require 3 to 5 seconds to focus! My impression is that latest Olympus DSLRs like 420 and 520 are faster, though. For that speed of the Canons, I think it is a not yet mature technology and there should exist some limitations. I hope that things will be improved later when more of the existing technical difficulties are tackled.

The Implementions

Well, so there are two LV methods plus two AF methods under the LV mode – a total of four combinations:

1. LV with Imager; AF with Phase Matching
2. LV with Imager; AF with Contrast Detection (by Imager)
3. LV with Separate Sensor; AF with Phase Matching (and of course!)
4. LV with Separate Sensor; AF with Contrast Detection (by Imager? - Oh, it's NOT possible)

But in reality, the fourth combination is not possible as the Contrast Detection AF by Imager must happen at the same time with LV with Imager. In short, the LV with Separate Sensor in Finder and LV with Imager simply cannot be co-exist unless the main mirror is a split prism that split lights into the Finder and the Imager *at the same time*.

One may also ask why not to use the External LV Sensor for the Contrast Detection? There are two reasons: 1. It is not needed as the Phase Detection AF is always here; 2. The gathered light may not be adequate for better image quality for precision contrast AF detection.

1. LV with Imager; AF with Phase Matching

The operating procedures are as follows:-

(i) Mirror Up;

(ii) Enter LV mode by Imager;

(iii) Mirror Down for AF (the norm is a dedicated button other than the shutter release button is to be pressed (all the way down) to initiate action);

(iv) After AF is achieved, the Imager is Reset/Discharged (the Shutter is Closed so that no light is allowed to fall onto the sensor so that it can be discharged fully);

(v) After the Reset, the final image can be recorded by the Imager.

An example of this type of LV mode and AF operation is the Olympus E-410/510, just see the official instructions below:-

Pentax K20D also uses this LV and AF method, however the AF interface/indication and LV functionality are both poorly implemented.

2. LV with Imager; AF with Contrast Detection (by Imager)

Many latest DSLRs offer this mode as an option. With pure software based algorithm, there are actually more to be played with, e.g., face detection. Canon 40D/450D/5D MkII, Olympus E-520, Panasonic L-10 and Nikon D300/D90, etc. are all having such a mode.

The Procedures are as follows:-

(i) Mirror Up;

(ii) Enter LV mode by Imager;

(iii) No need for Mirror Down to do the AF (usually a dedicated button other than the shutter release button is pressed to initiate action – the method of AF whether by Phase Detection or by Contrast Detection is to be set in a custom function option or by different buttons as assigned, depending on the design);

(iv) After AF is achieved, the Mirror is Lowered (which could be skipped, if mechanically allowed?) and Shutter is Closed so that the Imager can be Reset/Discharged (the Shutter must be Closed so that no light is allowed to fall onto the sensor for it to discharge);

(v) After the Reset, the final image can be recorded by the Imager.

But, as said, the problem with the Contrast Detection AF is mainly about its (slow) speed, the shutter is still required to be closed before taking the actual picture so that the Imager can be reset, which just introduce more time delay for resetting such a large Imager.

3. LV with Separate Sensor; AF with Phase Matching


(i) Enter LV mode by activating the LV Sensor in the Finder;

(ii) Doing AF as usual – half pressed shutter release button;

(iii) Take the picture!

Well, as it can be seen, that’s the “Seamless Operation" which DPR reviewers (and quite some others) like. But they just forgot to mention about the real weaknesses and limitations of this LV mode, particularly, the dark blindness and actually it is NOT WYSIWYG, no matter on actual image coverage, DoF, exposure and so on. What the user could see is JUST a very close *simulation*. Nonetheless, this LV still helps much in the shooting flexibility and for composition, undoubtedly.

Example (for Sony A300/350):-

Actually, this is just the same original Olympus E-330 design (Mode A LV). Just see above!