[morfic]

Please let us talk on precise facts, rather than general considerations. In that case please explain why 1109 is more exposed than 1110. The white wall should, as you say, fool the camera in multi-seg measure, but I end up with a less exposed picture using the spot measure on the face ! I have been using a spot meter camera (with a tiny central spot) for a while and never got that abnormal behaviour.

If this is normal for a K20D, I wonder who uses the spot meter !

You mention OMs further down in the thread, does this mean you have been using spot metering on a film camera prior to this DSLR?

To comment on multi-segment vs spot, not sure what to tell you, neither shot looks great exposure wise, and i get no exif info with fxif in my browser, so if you say the brighter shot is multisegment, then i am impressed, multi segment of any manufacturer tries to take a guess at what it is you wanted to do, in which case i would commend Pentax for a job well done.

---

Last edited by morfic; 06-02-2008 at 07:10 AM. Reason: Add metering comment.

[Zewrak]

Thanks to all for you input.

I have just uploaded 4 pef files on a public ftp site:

ftp://ftp2.landtrekker.org/landtrekk/pef/

I am anxious to hear from you

Best regards

Marc

I only see JPG's. But from what I see;

ftp://ftp2.landtrekker.org/landtrekk/pef/imgp0368.jpg
That image, if spotmetered and metering not locked, you are metering on the sand behind the kid. Camera will think you want to exposure to get the sand, and therefor lower the exposure, making the kid, that is darker then the sand, look even more dark.

ftp://ftp2.landtrekker.org/landtrekk/pef/IMGP1008.jpg
Same here, if you are spotmetering in the middle, you are metering on a WHITE t-shirt, camera will lower exposure to fit the t-shirt. Everything darker then the WHITE t-shirt will be even more darker.

ftp://ftp2.landtrekker.org/landtrekk/pef/imgp0387.jpg
See previous, but switch T-shirt, with truck

Basically, looking thru all them jpgs, its a constant that you have something white or very light somehwere in the middle of the picture, but you are trying to capture something darker. Can it be that you are metering the light areas and try to photograp the darker areas? Try locking metering on a darker surface of your motif and see if thats better.

Edit: whoops there were more pages to this thread *blush*

[jeffkrol]

You are saying that the spot measure would give a different value than the mult-seg on a same uniform frame. If it is so, this spot meter measure is pretty awkward. One will use spot meter to adjust the exposure of his/her subject on a given interest and does not expect to have to compensate, unless this zone of interest is abnormally dark or very bright.

On Olympus (SLR) camera (0M2 spot, OM4), brand which has pioneered the multi-spot measure, this was working perfectly (that is the picture was exposed accordingly to the spot zone). The documentation was complete on the subject (an not 2 tiny pages out of the 280 pages user manual). To cover the above two extremes cases (dark, or bright subjects) there were 2 buttons highlights / shadows that one would presses after taking the spot measure which over / under expose accordingly....

First: did you usually shoot slide or negative film? If negatives, generally you have no clue as to how accurate the meter is unless you took density measurements off the negatives themselves.
Personally multi-segment or matrix merering is not a err...exact science. Nikon spends large chunks of r&D money on tweaking the system to this day. 1000 scene look up tables, 1008 pixel light meter, and some AI thrown in I assume..
Nikon's exclusive Scene Recognition System advances the use of Nikon's acclaimed 1,005-segment sensor to recognize colors and light patterns that help the camera determine the subject and the type of scene being photographed, before an image is captured. This information is also used to improve the accuracy of auto focus, auto exposure and auto white balance detection functions in the D3.
Why bother.
and Pentax's what 10-14 whatever segment meter??? sorry these are still old school cameras w/ a twist. digital exposure is not that easy.
Just looked at my D and found using spot meter the "info" was the same as center weighted f/8 1/5th sec. but the histogram showed a definite exposure difference in the "white wall". Probably about 1/3EV. Switched to Matrix and got 1/3 higher again. Combine these and there is an exposure diffence of 2/3EV. Lets just say 1/2 since I really am currently not in a position to be too accurate.
now the 1/5th may be an approximation but it's hard to explain such a large difference in the histogram w/ a small difference in a shutter speed.
4 yr old thread w/ lots of little tid bits:
Matrix metering today vs .... - Photo.net Nikon Forum
Really early literature but fascinating all the same:
One funny line:37 perfect exposures on a 36 exposure roll. Actually that is possible...
Nikon Matrix Metering from KenRockwell.com

Furthermore the K20D does not indicates anything in the viewfinder (hardly readable in sunlight), nor on the LCD display about which measure mode is set ! That is why I forgot to reset the spot mode to multi-seg mode after use. This makes me say that Pentax has not designed this feature to be used in the field (may be in a lab with a piece of 30% grey cardboard handy...), especially if on top of that, one has to systematically add 1,5 LV on a average subject... Very disappointing !

To my taste K20D comes with too many modes (green mode, Sv, TAv), functions that are good on paper but unusable in the field (Liveview, spot-measure), confusing menu / Fn with again too many functions (especially photo-lab like functions) , too many buttons (AF is even duplicated)....

Not sure you'll find a DSLR that's not like this..with a similar or different set of annoyances. Once one company comes up w/ the perfect camera, all the others are out of business...

So why did I buy this camera ? I was attracted by: 3/2 format, old PENTAX lens still usable, wireless flash system, spot measure, shake reduction on the sensor itself, sensor dust detection, immediate start-up time.

My next input on this thread will be to investigate whether my camera is faulty or not. Fortunately this forum exists, and again I thank all you, because for the time being PENTAX support has still not replied to my initial query (the guy was absent, then busy, be sure you'll get soon an answer) ...

Maybe

[mrechte]

@Zewrak: thanks for your review.

ftp://ftp2.landtrekker.org/landtrekk/pef/IMGP1008.jpg
Same here, if you are spotmetering in the middle, you are metering on a WHITE t-shirt, camera will lower exposure to fit the t-shirt. Everything darker then the WHITE t-shirt will be even more darker.

I admit that some of the pictures were mistakenly taken in spot mode (the K20D does not remind you in viewfinder, nor on the LCD the meter mode) on a white part of the subject and therefore can explain the underexposure (up to -2 EV).

But 1008.jpg has been shot with multi-seg, not spot meter...

On white sheet of paper, I carried out some tests with K20D @ f4, 100 ISO compared to an OM4 SLR and a Lusanix F light meter

K20D multi seg: 1/500, spot 1/750 (some times 1/500)
OM4 centred weighted: 1/250, spot: 1/250 (spot + highlight compensation 1/60)
Lunasix: 1/500

I know in advance what many of you will say.

According to LV formula the above K20D / Lunasix measure should give around 13 EV

I checked the EXIF data taken with PhotoME that shows a range 12-12.3 EV on each segment, with centre segment 12 EV, but Effective LV is 13.1 EV (shot in multi-seg mode). Effective LV, is consistent with Lunasix measure, but not individual segments values.

@falconeye Is PhotoMe trusted ?

Could some body kindly send me a link to a picture shot on a white sheet of paper, because what I get seems really dark to me (IMGP1143.jpg) ?

[benjikan]

Hello to all,

I recently purchased a K20D with the starter kit (18-55 + 50-200).

I am back from a 4WD trek in Morocco with a load of pictures.

I am amazed to realize that most of the pictures are underexposed. Actually for most of the picture I must use the Photo laboratory software to increase the light (and not slightly !)

Even some pictures taken in low contrast conditions are underexposed. My son's cheap camera does far better than my K20D !

I am terribly disapointed with this camera.

I cannot imagine somenone is satisfied with the K20D light measuring.

I hope Pentax releases a new firmware to correct this problem very soon.

Regards

Marc

WHAT? Didn't you check the images on your LCD screen? What kind of metering were you using? I am totally amazed that you did not look at the output of the images or the histogram. I have used the K10D and K20D extensively and have made adjustments according to specific situations where the lighting was problematic. I often intentionally under expose. However, you might wish to take some images using the flash to see how the images are interpreted in that mode.
Ben

[Entropy]

@Zewrak: thanks for your review.



I admit that some of the pictures were mistakenly taken in spot mode (the K20D does not remind you in viewfinder, nor on the LCD the meter mode) on a white part of the subject and therefore can explain the underexposure (up to -2 EV).

But 1008.jpg has been shot with multi-seg, not spot meter...

On white sheet of paper, I carried out some tests with K20D @ f4, 100 ISO compared to an OM4 SLR and a Lusanix F light meter

K20D multi seg: 1/500, spot 1/750 (some times 1/500)
OM4 centred weighted: 1/250, spot: 1/250 (spot + highlight compensation 1/60)
Lunasix: 1/500

I know in advance what many of you will say.

According to LV formula the above K20D / Lunasix measure should give around 13 EV

I checked the EXIF data taken with PhotoME that shows a range 12-12.3 EV on each segment, with centre segment 12 EV, but Effective LV is 13.1 EV (shot in multi-seg mode). Effective LV, is consistent with Lunasix measure, but not individual segments values.

@falconeye Is PhotoMe trusted ?

Could some body kindly send me a link to a picture shot on a white sheet of paper, because what I get seems really dark to me (IMGP1143.jpg) ?

If you shoot a white sheet of paper without any other sort of reference, the camera will assume it is 18% grey and expose appropriately (underexpose making the white appear 18% grey).

That does look a bit darker than 18% grey though. Was EDR on for that shot?

[jeffkrol]

Multisegment metering has no clue as to the color/shade/tone of the paper.
Multisegment will never make it white without user intervention. As to the exposure an eyedropper on the image will show RGB at approx. 90/90/90 though there appears a slight amount of toning. I'd check for you but oddly enough I have no photo software w/ me at the moment.
to quote a source I believe:
18% grey appears as 110/110/110 in RGB color space , gamma 2.2.............\
In the middle of the histogram you can assume linearity. not so for the ends.
So roughly 40 "points" in RGB would be a full stop.
assuming 110/110/110/ is the EV value that the meter is trying use 70/70/70 would be a stop down from there. Of course 150/150/150 is a stop up.
Many camras use 110, Pentax has (for whatever reason) been known (or assumed) to use 90.
Personal proof is w/ MY *ist-D it would do a "white wall" at 90. After a few years I sent it to Pentax for some "issues". Supposidly they recalibrated the camera and it came back an "88".
To me this just shows you that 90 is "their value". Both Nikon and Pentax seemed to drift (within 1/2 stop) this figure between camera models. Only thing that surprises me about your camera/photos is that they seemed to go back to the 90 calibration..... though maybe just a bad hair day at the calibration factory for you.
Second example is if you shot the grey card w/ a Caon camera you would find that the peak is
at 127/127/127......... not because of a better meter, just because they cheated on the iso.
I recommend everyone to read Mr. Kerr's papers:
Articles by Doug Kerr
http://doug.kerr.home.att.net/pumpkin/Exposure_Calibration.pdf
Oh and to add my favorite quote form my favorite Leica icon:
'The exposure meter is calibrated to some clearly defined standards and the user needs to adjust his working method and his subject matter to these values. It does not help to suppose all kinds of assumptions that do not exist.'

Erwin Puts

---

Last edited by jeffkrol; 06-02-2008 at 11:54 AM.

[falconeye]

@falconeye Is PhotoMe trusted ?

Could some body kindly send me a link to a picture shot on a white sheet of paper, because what I get seems really dark to me (IMGP1143.jpg) ?

Well, I assume...

@falconeye: It will take me some time to review and understand them

...you haven't had a chance to read my apologies for giving a false definition of the EffectiveLV property in PhotoME. In that post, I gave a corrected definition for it.

I had mentioned in this thread, "EffectiveLV in spot metering is center meter + 1.1EV". It does still hold true. You just found that with a uniform background, multi-seg mode gives about the same.

The fourth image in my previous post was a white sheet of paper shot with spot metering and +2EV compensation.

Your white sheet paper is 40% gray (center), exactly. It may well be that this is the value as mandated by the standard bodies. I don't know, though. 40% gray in an sRGB color-profile image is 13% gray luminosity (2.2 gamma corrected). Maybe, look up the Gray-Card link in my previous post to find it out all. I didn't read it. But it seems to explain why photos taken from white sheets must turn out like a 12% gray card. If this is the case, it may be that your K20D meters perfectly.

This thread is so stunning because nobody else before tried to get correct exposure w/o histogram checks in the field. Reminds me of analog film shooting

[falconeye]

about your camera/photos is that they seemed to go back to the 90 calibration.....
[...] Leica icon: 'The exposure meter is calibrated to some clearly defined standards and the user needs to adjust his working method and his subject matter to these values. It does not help to suppose all kinds of assumptions that do not exist.' Erwin Puts

Lookin at mrechte's sample, it is more at 100 than 90 (in the center region). I used 102 (40%) in my argument above. Also note that the camera will round the attempted exact exposure (13.1LV) into the available range (13LV in this example). It had to overexpose by 0.1LV and an exact RGB analysis would have to take this into account as well, after reverting gamma-correction

I ignored it here as the entire topic starts to be complicated enough


As to the Leica quote: I like it. Once in the past, reputated companies could allow themselves to educate their customers about correct or incorrect use of their products. I guess this time has passed.

[falconeye]

You are saying that the spot measure would give a different value than the mult-seg on a same uniform frame.

You have found that spot metering and multi-seg metering give same values on a uniform frame.

Note however, that while spot metering has a standardized behaviour (meant for 18% gray card calibration), matrix metering is allowed to apply any degree of A.I. to figure what the lighting situation is (like thinking "well, bright metering all over the top, must be sky...") and compute whatever EffectiveLV it likes.

BTW: Do we mean the same by multi-seg and matrix metering? There exist 3 modes...

[jeffkrol]

Well, I assume...

...the value as mandated by the standard bodies. I don't know, though. 40% gray in an sRGB color-profile image is 13% gray luminosity (2.2 gamma corrected). Maybe, look up the Gray-Card link in my previous post to find it out all. I didn't read it. But it seems to explain why photos taken from white sheets must turn out like a 12% gray card. If this is the case, it may be that your K20D meters perfectly.

This thread is so stunning because nobody else before tried to get correct exposure w/o histogram checks in the field. Reminds me of analog film shooting

Gray card metering
If we do not have (or do not find it convenient to use) an incident light
exposure meter, we can perform the same technique by using the
camera’s integrated exposure control system and having it regard a
test surface of known reflectance, such as a “photographic gray
card”, that is exposed to the illumination on the scene proper.
Let us first imagine an “ISO standard” camera (with respect both to
the calibration of the exposure meter and with respect to its
assignment of ISO sensitivity ratings to its sensor system).
Imagine that the test target has a reflectance of 12.8%. Then, if we
have the camera regard the test target, and then hold the indicated
exposure1 when we actually photograph the scene, the exposure2 for
any 100% reflectance object should fall just at Hsat, the saturation
exposure2.
Often we have available a test target with a nominal reflectance of
18%. In that case, if the scene is photographed with an exposure1
that is 1/2 stop greater than indicated by the automatic exposure
system (such as by setting an exposure bias of +1/2 stop), we will
have the same result described just above.What about “cushion” against the possibility of an over-saturation
exposure2 for high-reflectance objects? We don’t really need any
cushion—the result we seek is not dependent on any assumption as to
average scene reflectance, and will be reliably obtained for any scene.
Ask RH about this............
And as to the "assumptions" we like to make, even using the "simplified zone system" you are warned:
With this simplified technique you must choose one of these five tones. Once you have decided which one of these five tones best matches your Reference Tone, then adjust your camera settings accordingly: \
Digital Zone System with Nikon D70: Technique

[falconeye]

Imagine that the test target has a reflectance of 12.8%. Then, if we
have the camera regard the test target, and then hold the indicated
exposure1 when we actually photograph the scene, the exposure2 for
any 100% reflectance object should fall just at Hsat, the saturation
exposure2.

Hi jeffkrol,

you may find the link I provided above interesting to read: Meters Don't See 18% Gray by Thom Hogan.

I now found the time to actually read it and it is quite entertaining. It is incredible that the corresponding ANSI standard says it all but only myths about it exist to be consulted...

Anyway, IMHO, it seems that mrechte's K20D perfectly exposes to the 12% ANSI standard.

[macbeth]

Thanks to all for you input.

I have just uploaded 4 pef files on a public ftp site:

ftp://ftp2.landtrekker.org/landtrekk/pef/

I am anxious to hear from you

Best regards

Marc

ALL your pictures are properly exposed for the settings chosen. The white is exposed for gray as it should.

Just so you know, the camera expose for neutral gray and it's the user job to compensate for it. First pictures were on spot metering and the white jacket is in the middle of the screen. The white wall is all around the subject in matrix metering which fool the meter, etc.

Your camera is fine but yourself, you need to learn the basic of photography. I suggest you go buy a book and learn the camera manual.

[mrechte]

I recommend everyone to read Mr. Kerr's papers:
http://doug.kerr.home.att.net/pumpkin/Exposure_Calibration.pdf

Thanks for this very very interesting document. It helps to understand the meter calibration rules. I have put it on my bed-side table !

So I made two other samples of a white sheet of paper shot @ f16 Av mode.
Lunasix indicates 1/500 (17 EV)

1144.jpg: multi-seg, viewfinder indicates 1/250. PhotoMe reports EffectiveLV = 16.3 EV
1145.jpg: spot, viewfinder indicates 1/500. PhotoMe reports EffectiveLV = 17.1 EV



In both cases all AEMeteringSegments (PhotoMe) are around 16 EL. I would have expected them to be all around 17, this would be consistent with Lunasix, EffectiveLV (photoMe) for spot measure.

Therefore this shows that multi-seg measure is interpreted by the firmware and the final decision in this case is to over-expose one small EV (this is explained in the PDF documentation referenced). Let us forget the multi-seg measure (EffectiveLV) which is software decision based on undocumented PENTAX logic (or secret) and which is too much scene dependant.

To which calibration corresponds 1145.jpg ? The RGB histogram shows an average around 100/255 in the centre (Gimp). To which reflectance does it correspond ?

As a subsidiary question why are the AEMeteringSegments showing 1 EV less than the actual values ?

I am sure falconeye can answer !

jeffkrol nearly answered my first question above, but which formula to use to convert RGB gray -> % luminance ?

[jeffkrol]

Thanks for this very very interesting document. It helps to understand the meter calibration rules. I have put it on my bed-side table !

So I made two other samples of a white sheet of paper shot @ f16 Av mode.
Lunasix indicates 1/500 (17 EV)

1144.jpg: multi-seg, viewfinder indicates 1/250. PhotoMe reports EffectiveLV = 16.3 EV
1145.jpg: spot, viewfinder indicates 1/500. PhotoMe reports EffectiveLV = 17.1 EV





To which calibration corresponds 1145.jpg ? The RGB histogram shows an average around 100/255 in the centre (Gimp). To which reflectance does it correspond ?

As a subsidiary question why are the AEMeteringSegments showing 1 EV less than the actual values ?

I am sure falconeye can answer !

jeffkrol nearly answered my first question above, but which formula to use to convert RGB gray -> % luminance ?

DR - simple truth: Nikon D300/D200/D100 Forum: Digital Photography Review
113/113/113 = 12.8 reflectance. Thats a footnote from Mr. Kerrs paper referenced earlier. So 11.8% reflectance is a good approx as any

Shot from my D w/ 100mm macro no EC.