Web Analytics RiceHigh's Pentax Blog: When the Focusing Screen Lies - Part 2

Friday, January 23, 2009

When the Focusing Screen Lies - Part 2

Last time I talked about how focusing screens of DSLRs could lie, in giving incorrect focus (and DoF as well, of course) as viewed through the finder and giving incorrect metering values, especially when stop(ped)-down metering is done.

As for Pentax DSLRs, actually I have never seen any of them could accurately meter the light when the Pentax lens mounted is stopped down, i.e., metered at a non-"A" lens aperture position. The *ist D, DS and K100 used to underexpose when stopped down whilst the K10D and K20D do overexpose.

Until a few months ago, I made my Cantax K5D 135 Full Frame DSLR with a Pentax K-mount. After I had used it more and more with my Pentax glass, I discovered that my 5D's focusing screen gave incorrect metering values as well, for stop-down metering. Similar to the K10D and K20D, it overexposes pictures at smaller apertures. The more the user stops down, the more overexposure will result.

The worst thing is: Since all my Pentax AF lenses are now just used as legacy "manual" lenses on my 5D, I rely more on the viewfinder (and actually the focusing screen) in doing the MF manually than ever before I used my Pentax AFSLRs/DSLRs. Unfortunately, the 5D's standard focusing screen, which is called the Ee-A (Eg-A* for 5D MkII), does also lie (much) in telling the correct focus. It is just made with a large DoF than real afterall, although the finder is intentionally made very bright with this standard screen. *To side track a bit, the 5D MkII needs a different "Eg" series because the viewfinder's coverage is larger at 98% now, so the screen size is different, and probably not because of any characteristic difference between the two series, I suppose.

Well, in order to achieve correct metering and exposure results, I needed to calibrate the exposure for that standard screen, against when an EOS prime is mounted and set in the Av mode, so as to find out the exposure discrepancy levels and characteristics, at different f-stops used. I had also checked that if the Time values measured by the camera obeyed the Law of Reciprocity with different Aperture values selected, mechanically at the "manual" lens or electronically for an EOS lens. Since different lenses have different light transmission efficiency (or light loss, in the other way), I had to also check the histogram for the exposure results. In short, my EOS lenses are in general accurate but not so for my Pentax "manual" lenses when they are mounted on the 5D. But in general my Pentax lenses all have better light transmission efficiency than my Canons. My generalised calibration results are as follows:

Table 1: Exposure Errors of Canon 5D Standard Focusing Screen Ee-A (Eg-A for 5D MkII)
Aperture * :
f/1.9
f/2.8
f/3.5
f/4.5
f/6.7
f/9.5
Exposure Error :
(+ve means over,
-ve means under)
-2/3EV
-1/6EV
0EV
+1/3EV+2/3EV+1EV
Corresponding EV
Compensation Required :
+2/3EV
+1/3EV
(or 0EV)
Not Required
-1/3EV-2/3EV-1EV

*Remark: Values Not Absolute, different lenses have different light transmitting efficiency and could affect the "centre" f-number.
The above table shows the results of my Pentax FA*43/1.9 Limited on 5D.

As you can see from the above, the exposure errors are roughly linear, which is a fortunate thing out of an orginally unfortunate anyway. Beside, the brightness level of the environment does not affect very much of the above characteristics, which is yet another lucky thing. However, the exposure errors are large in magnitude which cause much inconvenience when using my Pentax lenses. Another good thing I found with my Pentax lenses (FA Limited and FA* in particular) is that the mechanical aperture values are fairly to highly accurate and I must say Pentax had done an excellent job in manufacturing and calibrating those Pentax F/FA lenses with such a mechanical aperture ring, especially for their high grade FA Limited and FA* lenses which are almost dead accurate. I guess this also explains why the Pentax MZ-S was designed to use aperture ring to select the f-number in Av mode, instead of using the "A" setting, even there is one on all Pentax film AF lenses. I bet the old Pentax knew about this technical tricks very well and only mechanical aperture on lens is used, more accurate exposure is possible.

Now, let's look back at the focusing issue. Although my K-mount EOS adaptor does have a PCB and chipset on mount for confirmation the focus by the AF system electro-optically, I would rather prefer I could do turn the lens' MF ring faster if aided by a more effective and accurate focusing screen, before the focus is to be confirmed by the PCB and chipset finally. However, the problem of the Ee-A screen is that Canon intentionally made it too bright and they had assumed the users are using AF most of the time (well, that's case anyway if only EOS or AF lenses are used) and the viewfinder image is only for the purpose of picture framing. To explain the Physics, the trick is that Canon fine tuned the laser-carved microlenses on the screen to make them to gather more light onto the "focal plane" (at the focusing screen, not the real one on the sensor) and thus even for those slightly out of focus light beams are gathered. As a result, whilst the composite projected image on the screen is very bright but the out-of-focus regions are in focus now(!), which is the true drawback.

Unlike Pentax, Canon do have more options for their DSLR users. They have special made another focusing screen called the "Super Precision" Focusing Screen, the Ee-S (Eg-S for 5D MkII). With this screen, exposure is in general very accurate and no exposure *inconsistency* of any kind was found, except that there was a consistent -1/3EV underexposure found under the course of calibration, of which that kind of error is easily and straight forward to correct. Well, I must appreciate that Canon do know very well what they are doing, I must say even they have made some "irregular" thing, IMHO. There is no coincidence or luck in achieving this kind of exposure consistency and accuracy (putting aside the -1/3EV derivation, the metering is dead accurate). The only phase I can describe the whole case is "precise engineering" and also "careful design", by Canon.

But the "Super Precision" Focusing Screen does have its killing problem, which is already told in the 5D instruction manual, this screen is designed for fast lenses with a maximum aperture larger or equal to f/2.8. In fact, I found that the finder image is already somehow dim at f/2.8 and at f/5.6, it is terribly dark, which just makes the viewfinder almost not usable, even just for framing it is bad. The MF focusing using the screen is precise, as what it is claimed and named, but at anything beyond f/4 is almost useless. So, no matter how precise it will show the actual focus (in-focus and out-of-focus regions), it is of no use just when there is no light (or just enough light) that can be seen!

So, here comes a third option from Canon, it is the Ee-D Focusing Screen with Grid Lines (Eg-D for 5DMkII). Although in the 5D manual, it is told that the "Ee-D is essentially an Ee-A with grid lines", I found this statement is actually not correct. With fast lenses and wide(r)-opened metering (in the case with EOS lenses), it is true that there will be no big differences, but when stopped down, the differences show up gradually! Look at my calibration results below for the Ee-D screen and you will know what I say:-

Table 2: Exposure Errors of Canon 5D Grid-type Focusing Screen Ee-D (Eg-D for 5D MkII)

Aperture * :
f/1.9
f/2.8
f/4
f/5.6
f/8
f/11
Exposure Error :
(+ve means over,
-ve means under)
-1/4EV
-1/8EV
0EV
+1/8EV+1/4EV+3/8EV
Corresponding EV
Compensation Required :
+1/3EV
0EV
Not Required
0EV
-1/3EV
-1/3EV

*Remark: Values Not Absolute, different lenses have different light transmitting efficiency and could affect the "centre" f-number. The above table shows the results of my Pentax FA*43/1.9 Limited on 5D.

In fact, for this Ee-D screen, the exposure errors are still quite small for the most commonly used aperture values, which is quite usable even without any EV correction, throughout the range of f/2 to f/8.

However, the most lovely thing of the D screen is not about its stop-down metering with less inconsistency and errors (deviation is not a good thing anyway). The best thing is that the D screen has the best compromise for its screen brightness and it faithfully reproduces the crucial information for the in-focus and out-of-focus regions. It is at least not worse than what the in-camera electronic AF sensing system indicates, although not better than the "Super Precision" S screen (which is then practically not more usable when the lens is at f/4 or smaller). In short, the D screen is somewhere in between the Standard A screen and the Super Precision S screen, of which both the brightness and focusing accuracy are also somewhere in between (Contrary to what the 5D manual tells - it says both the Ee-A and Ee-D screens are the same, except for the grid lines). And, the grid lines mark is a Bonus for proper alignment and leveling, which can be very useful and handy when I need it (somehow frequently when shooting landscape).

Well, at this point, you may notice that I bought both of the Canon's "special" focusing screens! Yet, the total cost of these two *original* screens by Canon, which were both made in Japan, are still way cheaper than just one 3rd party Katzeye! And, I don't believe the Katzeye will be as accurate as original Canons and there is even not a Custom Function in the camera for properly setting it up.

As just mentioned, there is a Custom Function in the 5D and 5D MkII (which is non-existent in any Pentax DSLR) that when different focusing screen is installed, the corresponding screen should be input under that Custom Function. I have verified that that will set an exposure bias/offset value, with the Ee-S profile selected, an 1/3 to 2/3 EV higher Exposure Value will result than if Ee-A or Ee-D profile is selected. And, the difference will be more if the light reach the photo sensor is less, e.g. with a smaller aperture set and/or the environment is darker. This seems very logical as the S screen is actually darker and has more light loss, so it needs to compensate for the lost light as the total amount of light finally received by received by the photo sensor is less, for the same ambient light level.

Last but not least, my K-to-EOS adaptor has a PCB with chipset that cheats the camera a f/2 lens is mounted. I don't know if this will affect my calibration results shown above if a different PCB of a different f/number is encoded. But I think the difference would be small as the other values I have seen for different PCBs/chipsets are of f/1.8 and f/2.8 only. Anyway, YMMV.