AF Accuracy's Dependency on Lenses and Yellow Light

With the secret menu lately found in Pentax DSLRs by the end-users (that's what my last two blog entries are about), quite some certain Pentax DSLR users have already calibrated for the amount of "focus correction" required for each of their lenses.

Here is a representative report from an user who has quite a number of latest Pentax lenses. Interestingly and surprisingly (when I read the news for the first time), it was found that the value of the focus correction parameter varies rather largely with different lenses of his and of course the (in)famous "Pentax DSLR AF Yellow Light Sypmtom" was also once confirmed, i.e., a different more negative "focus correction" figure is required to counteract for the Front Focusing (FF) problem under tungsten/incandescent lighting:-

http://forums.dpreview.com/forums/read.asp?forum=1036&message=22026990

Well, here is another very first report on the amount of "focus correction" required for making the FA*85/1.4 to focus correctly under tungsten lighting (on a K10D):-

http://forums.dpreview.com/forums/read.asp?forum=1036&message=22026337

So, there are two issues arouse here, which I am going to discuss in my this blog article:-

1. How come different lenses can cause different amount of AF errors which can be very different in magnitude or even in exactly the opposite directions, i.e., FF Vs BF? And, is this an issue specific to the Pentax AF system?

2. To discuss more about the (in)famous long-time well-known "Yellow Light Sypmtom", which is actually something that all Pentax DSLR users suffer from, unless they don't shoot under tungsten light and that the apeture used is small enough to hide most of the focusing errors by the large DoF. Well, again, is this issue Pentax specific?

Okay, I opt to discuss the above second issue first, which is a long-lasting one, before I am going to discuss on the new issue found, i.e., the 1st one.

I. Yellow Light Symptom of Pentax DSLR AF System

The case was first reported firmly by Oleg V years ago at the Dpreview forum. He was so firm about his findings as he had also a Nikon D70 at that time which he compared his *ist DS to (and of course he said that his D70 had no similar problem which is so obvious with his *ist DS). While quite some people overthere at that time disagreed strongly with him or even overreacted (which was/is indeed very "normal" for the outcome at that place), I designed and carried out a quick experiment to verify his reported problem very shortly, via my this experiment.

I know that what people are interested in this problem is that: If this issue Pentax specific? My first and quick answer is: Yes, mostly, with all the wonky SAFOX VIII AF systems used in Pentax *DSLRs*. Actually, all Pentax DSLRs employ the SAFOX VIII system or its variants, e.g., *ist DL has a simplified version of 3 points and the DL2 has a five-point one, but still they are SAFOX VIII inside, for the design and structure.

First of all, here is the golden reference technical article on how SLR MF and AF systems work. Before going on, I highly recommend you read this in full (the article is easy to follow for laymen and is crystal clear, but one may need to read slowly to fully understand and digest all these):-

http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf

The problem with the Pentax SAFOX VIII used in their DSLRs is that the AF system cannot see the focus in a WYSIWYG ("What You See Is What You Get") way. That means that either 1. the optical path in the AF detection system is not resembled close enough to the optical path for the actual image formation (of which the AF path is comprised of various micro-mirrors and micro-lenses, and micro-splitprisms, for which the splitted images are ultimately projected onto the pair(s) of AF sensors , and/or, 2. the AF light sensitive sensor cannot cater for different wavelength as the light spectrum varies which causes the yellow light symptom with obvious FF as a result, say, the sensor may have a peak spectra response for the daylight only (usually peak response for green light for the CIE standard), but not for the yellow or red lights. But unfortunately, there is no adjustment mechanism in the SAFOX VIII design to minimize such deviations so as to improve the accuracy and reliability of the design.

So, here is the question again: is this a Pentax specific issue? Yes, but mostly for Pentax DSLRs, I must say, which the problem shows up most obviously. And fair to speak: for a "No" answer, other AFSLRs should also suffer from the same problem, as the AF sensor should have a fixed peak response for a certain wavelength of light in the visible spectrum, probably the response is a bell shape one with a peak at the green light wavelength, which is simply calibrated for the daylight. And, a "No" also, as I have done a quick experiment with my *Pentax* MZ-S for the "yellow light symptom", as follows:-

I put a flat target with colourful and contrasty coarse patterns about 2m away. I mounted my K100D on a tripod and with my FA*85 lens mounted on it and set it at f/1.4 in Av mode. I focused on the object from both directions and first the light source is tungsten yellow and then I changed the light source to pure white one (a panansonic super high freqency fluorescent lighting). For each time when the focus was achieved, I noted down the position of the distance scale. Then, I repeated the test with my MZ-S with the same setup. Do note that I carried out the test at night and at one time only one light source was turned on.

The K100D pictures were then inspected and as usual, the yellow light shots were out-focused. The distance scale showed that it was front-focused. The MZ-S arrived at the same distance scale mark as the K100D under the white light, so I could assume that the picture taken was in-focus, as this is the correct point of focus. For the yellow light shots, the MZ-S *did* shift towards the closer side, but the magnitude of the error is comparatively much smaller! Indeed, I have repeated the tests with the K100D and the MZ-S for several times and the results were all the time, i.e., the MZ-S was more accurate under yellow light and the FF amount is much smaller, although was existent. BTW, I don't know if for yellow light, the image actually forms "fronter" by the lens, but at least I still got out-focused pictures in this experiment.

Hence, the conclusion is that it is not that Pentax do not master how to make a better and more accurate AF system, which particuarly can cater for the heavy yellow light, but just that the system they have being used in their DSLRs are just not up to their best standard. So, the good thing is that there is *still* hope for a better DSLR in the future but the bad thing is that don't imagine that our Pentax DSLRs can focus properly without FF under tungsten lightings, unless one does the same thing as the above users - to input an offset value via the service menu each time they are to shoot under tungsten and to calibrate in advance for each of their different lenses!

II. Dependency on Lenses for AF Errors

Actually, AF errors are dependent of the lens in use is one of the newest things I have learnt this year :-) This latest new finding by those Pentax users actually a scary thing, as this user points out also:-

http://forums.dpreview.com/forums/read.asp?forum=1036&message=22028511

Although it is undoubtedly "a scary thing", I shall look into the case to find out if this issue is Pentax specific, before the conclusion is drawn up too early (unlike the above poster has commented that already).

So, if I am going to explain the reasons for how comes the errors, again, in the first place the working principle of the SLR AF (or simply MF) system shall be got known. Well, here is the golden reference again!:-

http://doug.kerr.home.att.net/pumpkin/Split_Prism.pdf

The possible culprits in the wonky Pentax SAFOX VIII AF system, which is used in most Pentax DSLRs, can be the following:-

1. Again, as it is always crucial, is that what the AF system can see is not WYSIWYG as the actual image formation, optically and electronically, which have been explained in the last part as one of the possible causes for the "yellow light symptom";

2. But there should be another factor which comes into play here, which is more important, I think, that is: All the lenses in this planet are indeed not perfect optically. They must have some kinds of Spherical Aberrations and Chromatic Aberrations, etc. which affect coherency of the light paths for the same point of the object when the light/colour beams of which reaches the film plane ultimately (at a different position or even out of the plane). Nonetheless, it should be noted that some paricular glass are still having less optical errors than the others and thus some lenses can be regarded as better or best optics but some others are considered to be dog or junk :-)

Now come to the crucial point, the fact is that SAFOX VIII AF system actually measures the focus with *particular* lens' optical areas at rather small aperture, i.e., more near to the centre of the lens, which can be as small as f/8, instead of limiting to using areas at wider apertures (typically at f/5.6 for traditional designs of SLR AF systems, e.g., the original SAFOX). It is very easy to try out for this indeed, just mount a AF teleconvertor so that you will essentially have a f/8 lens, I'm sure the AF system still can work, provided the object is bright enough. If the AF system does not read off more near the optical centre, the "black-out" effect (explained in the above reference article) will defeat its function, and the system will by no means "see" the splitted images for deciding the focus.

So, as the article has also said, for more accurate focusing so that the phase detection can be largely and easily differentiated, it is highly preferable for the AF system to read off light beams from the outer optical regions of a lens at larger apertures, i.e. the boundary areas is preferred over somewhere near the centre! The fact is if the displacement has been made too small (for reading more near the optical centre of the lens), high precision split-prisms (those used in the AF module) with nearly zero errors and a very high density AF sensor with high resolution are required. Otherwise, things will get worse and the AF system will become very inaccurate.

I don't know if the Pentax AF system reads off light from somewhere nearer to the optical centre solely, say, somewhere near f/8 to f/4, or instead out to the boundary for a large f-number, say, f/2, which is actually very dependent on the design for selecting which certain areas of the glass for reading the AF splitted images (and thus this can be controlled by the engineer who designed it). In short, the more areas the AF system read, the splitted prisms will need to be made larger. The same applies to MF split screen, too.

Here is an assumed example to illustrate further for how comes the errors: Just say if it is designed that the AF system reads off light from the lens' optics from somewhere nearer to the centre at f/8 out-bound (which has been verified) to the f/4 in-bound only. Then, if spherical aberrations or other aberrations happened in the lens (no matter how it comes, whether it is owing to in-born design limitation or manufacturing mis-alignment and etc.), the exposure can be still out-of-focus when larger aperture is chosen, say, at f/2.8 or even larger, just because the light beams used for doing the AF job actually are a small portion amongst *all* the the light beams for a (much) larger aperture, say, f/2.0, which are actually projected on the CCD for the exposure and making the final picture, ultimately.

To be fair, I have found that this issue is actually not Pentax DSLR specific, here is what a D80 user reports:-

http://www.photozone.de/active/forum/ShowMessage?ID=1079762A

The case is about a Nikon D80 with one Nikkor Lens and two Tamrons, with correct focus, FF and BF respectively for each of them on his D80. But it can be noted that the original Nikkor lens does not exhibit any problem. So, it maybe a coincidence, or, it can be concluded that Nikkor lenses are more accurately made than the Tamrons.

Indeed, Canon knew very well about this limitation and they were the first one who provided a solution on this problem by making the cross sensor in their first version of EOS 1, of which one of the linear AF sensor inside the crossed AF areas read off the lens optical boundary at f/2.8 or larger, such that AF at larger apertures can be more WYSIWYG and most spherical aberrations or other optical imperfection can be ruled out, theoretically. But ironically, this design has been blamed by Canon users for years on why Canon didn't have made the cross sensor to be fully operational for slower lenses, and of course, conspiracy theory followed! When Canonians are also often talking about the inaccuracy of their Canon DSLRs, actually I have seldom heard about true pros complaint about there were obvious AF errors for the "1" series, and now I can know why! (or maybe also true pros have no time to waste their time to post in the gear/photo forums ;-))

Finally, to put an even more objective good balance, I must point out that the SAFOX VII AF system used in the Pentax MZ-S is again indeed more reliable and accurate (and more responsive and decisive, too, even at low light). So, the same as my conclusion about the "yellow light symptom" in the last part, Pentax do have mastered the technology already and they indeed know (very well, I bet) how to do it well. But the practial problem is: WHEN will they make a DSLR with the quality, performance, reliability and accuracy at the level on a par with they true flagship MZ-S?? It seems that it's not so optimistic for this for what I bet myself, at least for the time being..