[piotro]

@ClassA wrote: "You are suggesting that Pentax is listening/responding to users. Past history shows next to no evidence for that. "

Oh, really. Then how do you account for the numerous features that were included in the K-5 which seem to be a direct response to people moaning about what was lacking in the K-7? Seems to me there is a direct conduit from this forum to Pentax/Hoya.

Features that C&N mostly had for some time already, or where it was so obvious you can make it better (from K-7 to -5), quite obvious new P product needed some of them to be able to compete on the market, Pentax was mostly catching up with competitors.
Do you really think these guys would need to read forum to get idea what features this product is missing?

[dlacouture]

You are suggesting that Pentax is listening/responding to users. Past history shows next to no evidence for that. Or where is the firmware update that fixes the "Auto ISO" bug for my K100D?

Well, there are some cases that come to my mind...
- Auto-align for HDR (asked by Falk himself, IIRC)
- Auto-iso fixed for K7 (hidden in the last firmware)
- AF points in the K-r (but could be the K-x bad reviews)

Of course, there are a lot of wishes that went unanswered...

[jbinpg]

@piotro: Listening to user requests is part and parcel of every successful company. Yes, Pentax/Hoya must necessarily be tuned into the competitors product lines but basing direction solely on that is just bad business.

[falconeye]

Question for Falk - am I reading this chart correctly?


I'm interpreting the chart above as saying there are no focus shift until EV 2 and below, and even then it's not consistent (sometimes there is a shift, sometimes there isn't), and then below EV -1 there is more or less a consistent shift?

I hoped the paper would answer this

It's not so easy to interpret. First, it mixes lenses and colors. For a given lens and color, the shift is rather deterministic. E.g., for red and FA31, it's around EV0, for blue it's more like -1, for DA* 60-250 and orange, more like EV2 etc.

However, if you add EV 2.4 for a 18% skin target and EV 1.7 for red channel saturation (compared to daylight white), then EV2 becomes EV6. So, the range of situations where it can happen varies considerably. For further insight how to deal with a consistent EV definition, I propose to consult the paper first.

[falconeye]

Remember your annotated cutaway of the prism housing? You speculated that "D" could be the colour temperature sensor. Seems more likely to me than ever now.

I remember but forgot I actually posted it. Thanks for the find

However, my thinking now is that D is the 77 zone light meter and A is just a PCB supporting the microphone.

The +sensor may be embedded into the 77 zone meter. It may make sense as measuring brightness doesn't really suffice to not blow a single color channel (like red). D shows an optical path and a sensor in itself. Why shouldn't it be 308 sensel color sensor actually?

[Christine Tham]

I hoped the paper would answer this

It's not so easy to interpret. First, it mixes lenses and colors. For a given lens and color, the shift is rather deterministic. E.g., for red and FA31, it's around EV0, for blue it's more like -1, for DA* 60-250 and orange, more like EV2 etc.

However, if you add EV 2.4 for a 18% skin target and EV 1.7 for red channel saturation (compared to daylight white), then EV2 becomes EV6. So, the range of situations where it can happen varies considerably. For further insight how to deal with a consistent EV definition, I propose to consult the paper first.

Sorry Falk, I found your paper a bit hard to read so it was easier to ask the question.

So you're saying we could see potential misfocus situations even at EV6? That's pretty significant!

[falconeye]

Hey Falk,
...
There is a Hoya patent that

Here's a link to the Patent:

Digital camera

The IX+ AF system might not be exactly as described here, but the color sensing element seems to fit this description pretty closely.

Ray

Ray,
thanks a lot for the link. I must have missed it when it was first posted

I'll have a deeper look at it. But one thing seems to be pretty clear:

The AF+ system assumes a colorimetric sensor (#17 in the patent) which is located on top of the eyepiece and exposure meter (#16 in the patent). However, #16+17 require an aspherical condenser lens in front of them (#15 in the patent) which the K-5 lacks. It is very similiar otherwise.

As I said, I now believe Pentax merged #17 and 16 into a metering CCD and made #15 spherical. Which means the colorimetric sensor sees the focussing screen image. Which in turn means a Katzeye may affect focus beyond exposure

The sensor #17 in the patent is describes as a global light color sensor. That's a bad idea anyway because such a system makes the hidden assumption that overall light color and edge color are the same. That's not always true though.


Nevertheless, I herewith promote hypothesis to theory: The focus shift is caused by a low reading of the colorimetric sensor which is hosted above the eyepiece and performs proportionally to the max. aperture of the lens which, in turn is proportional to the focus screen image overall brightness rather than edge brightness.

I am allowed to do so because the hypothesis predicted the +sensor to be outside the AF module and the patent (which I wasn't aware of) now confirms it.


This means that the fix shouldn't be very hard: If the reported calorimetric signal (brightness) falls below a given limit then it must not be trusted anymore.


There is an interesting corollary for studio shooters: a low key situation with a dark background may not send much light to the calorimetric sensor, even if the required EV value is high.

Therefore, studio shooters should currently not use the K-5 phase AF for low key photography except if combined with primes faster than f/2.8 and/or an external daylight focus assist light.

In retrospect, this explains what I see in the studio myself. Most of my studio AF misses are low key with a zoom.

A well lit feature makes the AF lock focus confidently but a dark overall illumination makes the colorimetric sensor send it off 250 um. Theoretically, it should then be possible to construct a scenery which front focusses at almost any EV value.

---

Last edited by falconeye; 03-01-2011 at 06:07 PM.

[MrPetkus]

As I said, I now believe Pentax merged #17 and 16 into a metering CCD and made #15 spherical. Which means the colorimetric sensor sees the focussing screen image. Which in turn means a Katzeye may affect focus beyond exposure

I'm not sure I follow - how exactly will the design of the colorimetric sensor affect focus with the Katzeye focus screen - does that suggest an AF point landing on the split prism will confuse the sensor? Since my K5 is outfitted with the Katzeye I can easily run some tests vs. the default focus screen. At any rate, this would affect only a small percentage of K5 users who actually use a split prism.

[Christine Tham]

Nevertheless, I herewith promote hypothesis to theory: The focus shift is caused by a low reading of the colorimetric sensor which is hosted above the eyepiece and performs proportionally to the max.

This is an interesting theory, and as you hint has implications for those using third party focusing screens (like me).

If the AF "correction" is made by referencing the light reading from this sensor, then anything that changes the amount of light received by the sensor (such as a different focusing screen) will impact the accuracy of the AF correction.

Maybe Pentax should allow the AF correction to be manually switched off, or adjusted by the user.

[Smeggypants]

This is an interesting theory, and as you hint has implications for those using third party focusing screens (like me).

If the AF "correction" is made by referencing the light reading from this sensor, then anything that changes the amount of light received by the sensor (such as a different focusing screen) will impact the accuracy of the AF correction.

Maybe Pentax should allow the AF correction to be manually switched off, or adjusted by the user.

What motivated you to install a 3rd party focussing screen Christine? Was there an attribute of the original screen that you weren't happy with?

[Ray Pulley]

Ray,
thanks a lot for the link. I must have missed it when it was first posted

I'll have a deeper look at it. But one thing seems to be pretty clear:

The AF+ system assumes a colorimetric sensor (#17 in the patent) which is located on top of the eyepiece and exposure meter (#16 in the patent). However, #16+17 require an aspherical condenser lens in front of them (#15 in the patent) which the K-5 lacks. It is very similiar otherwise.

As I said, I now believe Pentax merged #17 and 16 into a metering CCD and made #15 spherical. Which means the colorimetric sensor sees the focussing screen image. Which in turn means a Katzeye may affect focus beyond exposure

The sensor #17 in the patent is describes as a global light color sensor. That's a bad idea anyway because such a system makes the hidden assumption that overall light color and edge color are the same. That's not always true though.


Nevertheless, I herewith promote hypothesis to theory: The focus shift is caused by a low reading of the colorimetric sensor which is hosted above the eyepiece and performs proportionally to the max. aperture of the lens which, in turn is proportional to the focus screen image overall brightness rather than edge brightness.

I am allowed to do so because the hypothesis predicted the +sensor to be outside the AF module and the patent (which I wasn't aware of) now confirms it.


This means that the fix shouldn't be very hard: If the reported calorimetric signal (brightness) falls below a given limit then it must not be trusted anymore.


There is an interesting corollary for studio shooters: a low key situation with a dark background may not send much light to the calorimetric sensor, even if the required EV value is high.

Therefore, studio shooters should currently not use the K-5 phase AF for low key photography except if combined with primes faster than f/2.8 and/or an external daylight focus assist light.

In retrospect, this explains what I see in the studio myself. Most of my studio AF misses are low key with a zoom.

A well lit feature makes the AF lock focus confidently but a dark overall illumination makes the colorimetric sensor send it off 250 um. Theoretically, it should then be possible to construct a scenery which front focusses at almost any EV value.

So,

The shift following light level (and lens, based upon max aperture) now makes more sense as an adjustment is being made to the AF calculation based upon what the color sensor sees, which is more or less the same thing the meter sees off the main mirror and focus screen, which is brighter for lenses with a bigger max aperture.

The AF module appears to be very sensitive as it locks in extremely low light and with no hesitation, so it is probably working well within spec and is likely not the issue.

This also could mean that slower lenses will NEVER work to EV -1 (absolute), and that the specification is indeed relative to the max aperture of the lens on the camera (which is the same thing as saying relative to the amount of light getting to the color sensor). So, if the full lower EV value of EV -1 is only available with say the f1.4 lens, ISO 100, then every stop slower than f1.4 max aperture moves the lower end of the AF working range up one stop.

Even if everything else worked perfectly, that would make an f4.0 lens AF only from EV 3 to 18, and an f2.8 lens from EV2 to 18. Hmmm, these numbers sound VERY familiar to me....

Of course, Pentax could be looking at all of our speculations and laughing at how stupid we all are , but the specification of EV -1, ISO 100 as listed in the manual is anything but clear as to what it really means.

All that said, why would the K7 not have the same problem?

I know that the two systems are not 100% identical, but I believe that both have a very similar implementation of the color sensing arrangement.

A K7 test under the same conditions might be interesting.

Ray

[creampuff]

All this so called research is interesting but for a lay person like me, I really find it amazing that anyone can measure the errors so accurately in µm (micrometre or one-millionth of a metre - 1/1000 of a millimetre, or 0.001mm). Unless it is all theoretical, I just don't know how anyone can accurately measure 250 µm focus error. I got research scientist friends who can't even achieve such precision with their measuring equipment.

[ursamajor]

If you know a bit geometry and optics, you can... not necessary to be scientist, but mind opened.

[jolepp]

All this so called research is interesting but for a lay person like me, I really find it amazing that anyone can measure the errors so accurately in µm (micrometre or one-millionth of a metre - 1/1000 of a millimetre, or 0.001mm). Unless it is all theoretical, I just don't know how anyone can accurately measure 250 µm focus error. I got research scientist friends who can't even achieve such precision with their measuring equipment.

It is a value calculated from the test images, which is well as locating the end of a ruler where an image is focused inside a camera would be tricky . I suppose it would be firmly in the realm of mechanical measurement as such; micrometers typically have 0.01mm reading and some even have vernier scales for 0.001mm reading. As such, the 250 um is a ballpark value, relevant in comparison to the 70 um (I'd recall?) typical inaccuracy of the AF system in mind.

[falconeye]

I'm just confused with all that your saying... no fix, fix what's not broken, need a fix to be a certain way? I don't understand...

Please, let's not go this route again.

I am very confident Christine tries to shed some light onto all of this as everybody else does. I appreciate her comments. We all can make mistakes and I prefer a critical voice any time

I think, we made quite some progress now since the initial reports and do understand where the problem comes from. This implies that we now understand how some see the problem while others don't -- under seemingly similiar conditions. It wasn't so easy to figure out after all.

WRT K-7: AFAIK, the K-7 as every Pentax w/o the 77-zone meter does NOT sport the SAFOX+ colorimetric sensor.

WRT a 3rd party focussing screen: I don't think it affects focus as long as there is enough light (a focussing screen should be of neutral color). But a different focussing screen may slightly shift the threshold where the focus shift happens. Not a problem in itself. But to be revisited after the fix (in case the fix depends on some illumination calibration. A Bright Katzeye screen may even improve the situation (if it makes the camera tend to underexpose).

WRT operating range. It would be a pitty if the AF cannot exploit its fabulous sensitvity just because some lenses are too slow.

I highly recommend Pentax to include a feature in the forthcoming firmware fix to override the colorimetric reading by the manual white balance reading. That would be great for studio shooters too. Could be an option in the WB dialog, enabled with non-AWB.