Monday, December 04, 2006

How AF Errors Can Decrease *Effective* Resolution?

Whilst the Mega Pixel (MP) race have never yet stopped for both the camera manufacturers and the users (whom actually are somehow driven by the camera manufacturers, frankly speaking – consumers tend to follow "trend"), not all the users will notice that the pixel count is indeed not the only factor which will affect the final obtainable resolution (in whatever measurement units). Actually, there are quite some other factors come into play.

Amongst those, I would say the most prevailing factor which affects the final resolution would be the AF accuracy (I will explain following this article why this is so). However, AF accuracy is not something that could have ever been publicized to the end-users by any camera manufacturers. In short, this part of the performance specifications is only known by the manufacturers themselves and those “secret” performance parameters are generally referred to “hidden specifications” of a camera.

As we know, now that the “standard” number of MPs for the current (and the next?) year is undoubtedly 10MPs, whilst 6MP pixel DSLR models are now regarded as “entry-level”. However, before we are going to “chase” after the MP race, there is something more that we should consider. In the following, I shall discuss how meaningful those MP figures would be or not, and how the AF accuracy factor would prevail, as I have said.

Okay, an unarguable fact is that there have been so numerous DSLR users who owned a 6M or 8M DSLR model have already reported that they found AF errors in their shots. Those problems are commonly reported as Front Focusing (FF) or Back Focusing (BF). To me, they are essentially indifferent. Just because both error types simply create out-focused photos. But then, we should be aware that there is simply no perfectly accurate AF system or even MF system on this planet. So, the question is: How accurate should an AF system be made, in order to *match* with the number of MPs or at least with the human eyes?

Before going on, what you need to know about the basic is on the concept about the Circle of Confusion (CoC), see:- Well, according to the Wikipedia explanations, we can see that for a 8”x10” printed photo, which is normally viewed at a distance of 25cm, the CoC is 0.2mm. As such, for such a 6MP DSLR picture, the long side of the photo would have 3008 pixels and hence the maximum allowed print density of picture at the longer 10” side *without interpolation* is approximately 300 dpi. Do note that this figure is a *very* realistic one as a very good quality print from the Fuji Frontier Lab Laser Printer is also at around 300 dpi. Still remember the fact of 0.2mm (i.e., 5 lines per mm) which is the approximate maximum limit for human eyes to distinguish an out-focusing blur, i.e., the confusion? Hence, in this maximum allowed CoC diameter, if converted to a number in pixels, for a 6MP DSLR, is equal to: 3008 / (10” x 25.4mm/inch) / 5 = 2.37 pixels;

So, again, what does this imply? It simply means that the AF error can be at most produce an out-focusing blur in 2.37 pixels, that is, the theoretical diameter of the CoC. Otherwise, the taken photo can be seen as out-focused when it is viewed in *full screen*. Do note that this calculated figure generally holds regardless of the size of the screen display or the print. It is because if the picture size is bigger, it is *supposed* to be viewed *as a whole* farther away, and ditto for smaller sizes of the output.

In fact, different DSLR users who complain about the inaccuracy of DSLR AF system, from time to time, must have been able to spot out some problems by themselves (putting aside those minority user errors by beginners). Those spotted errors were either found when they viewed in full screen or during pixel peeping, i.e., by viewing in 100% magnification or go even further, or even both! Well, assuming that those people are pixel peepers first and they actually tried to find problems by pixel peeping, the CoC should be at least in more than 1.5 pixels of diameter so as to for them to find out something wrong. Practically, 2 pixels along a line are affected so that they could see the blur. As a result, 4 neighbouring pixels have virtually become one owing to the CoC blur overlay project on them, on a particular *area*.

Well, here finally comes the most interesting but terrible thing! If one finds out-focusing problems with a 6MP DSLR, that can simply mean that his DSLR has its effective resolution been decreased by *at least* a significant figure of 1/4th, that is, an original 6MP DSLR will become just a 1.5MP one in terms of the effective resolution, a 8MP becomes a 2MP one, a 10MP becomes a 2.5MP one and so on. In short, most of the potential resolving power are wasted. Amongst those, what do think about the differences between an effective 1.5MP, 2MP or a 2.5MP model? :-)

Do note that the required accuracy or tolerance for viewing full screen or for prints for a 6MP DSLR is actually not significantly much less than that when one pixel peeps, i.e., < 2.37 Vs < 1.5 pixels. Thus, if I was the designer of the AF system of a DSLR, I would choose an error of 2 pixels for the CoC as the border line figure, which would be a best compromise between accuracy and for practical use, as well as for the manufacturing cost. Just say for such a DSLR with that intended AF accuracy, typical users will not detect any out-focusing problem for normal use, whereas pixel peepers will have acceptable results of which the errors will not be too obvious. I use the word “compromise” as increasing the AF system precision by too much would be an overkill and the system would be very difficult and costly to make too. It should be noted that with the more pixel count of a DSLR, non-pixel peeping users will not benefit anything much if the AF system is not on a par with (the pixel count). In contrast, pixel peepers will find even be more disappointed as it will be more easy for them to see all those artifacts and errors when they “peep”. So, let’s discuss back for a little bit more specific about the Pentax K DSLR cameras (Well, I nearly forget about my Blog is on Pentax gear :-)). Actually, what I can confirm is that my K100D has a far more accurate AF system than any of my previous *ist D, DS and DL bodies which I owned or used. Anyway, it has actually been confirmed that the K10D does have the same AF system as used in the K100D, according to the previous recent inteview with the K10D design and development team leader, by a Japanese local online media. Here is the source link of a translated version, in English (by a Japanese PDML member):-

Here, I quote again the dialogues which I have just mentioned above:-

Q: As the resolution becomes higher, the focus accuracy and camera shake reduction become even more critical. Does K10D have different approach to to ASF and SR?

A: "Regarding the focus accuracy, the criteria hasn't changed much from K100D. We consider the AF accuracy for K100D being excellent in the first place. Furthermore, at the time of transition from film to digital, the AF accuracy is now determined by the standard of pixel peeping on a monitor, rather than the focus depth based on the allowable circle of confusion as in the case of film camera. We heard from some quarters that the AF speed for the models prior to K100D was slow, but we have reevaluated the AF mechanism and controls etc for K100D, and realized the increased response speed while keeping the focus accuracy." - Mr. Hisashi Tatamiya (Chief Engineer of the K10D Development Team).

It should be noted that when K10D has the same AF accuracy as the K100D is not necessary meant a bad thing. It is because to put it in the other way around, the K100D’s AF system (actually the K10D’s one) can be overkill. Anyway, I have no more information in hands to confirm or verify further about the case.

So, whereas you folks may see daily, particular some K10D users at various Pentax user forums do have complaints on the AF accuracy of their K10D (there is some K100D reports too, but I notice that there are more K10D reports – but then do note that the K10D is a very new camera and it’s also normal to have more users to talk about them), I think my this article can quickly give you some ideas about the case and why AF accuracy is of prime importance, for *any* DSLR.

To be fair, as for a more balanced view, I must also point out that there exist also many Canon and Nikon DSLR users who also complained and complain about FF and BF problems, yet more or less with particular DSLR model(s) or lens(es) or some particular combinations of lens and body which are more prone to errors.

As a final humble comment of mine, I just wish to say that when people are just asking for more mega-pixels for a DSLR, why not urge for a more accurate and reliable AF system in the first place??


Derrick said...

Thank you very much, for your time spent in creation of this article, and I looking forward to reading future entries.


RiceHigh said...

You're welcome and I'm glad that you find my article useful :-)

Anonymous said...

Interesting to read.
Of course, the image quality that comes out of the camera (RAw or JPEG) is what matters ultimately.
But it is nice to know more about the inner workings of the camera.
Looking forward to part III.

Anonymous said...

Congrats on a balanced article.

Post a Comment

Related Posts

Creative Commons License
RiceHigh's Pentax Blog by RiceHigh is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.