Attention: Fotogenic & Custom Curve Users

[ajo43]

fantastic info. This is the kind of post I really enjoy reading on this forum rather than D70 vs 20D banality.

I had originally entered into these discussions with the belief
that my D70 consistently underexposed images, but after reading the
postings by Fotogenic and others, I have come to the realization
that the D70's exposure is accurate, but that one must choose the
proper tone curve that one applies either in the camera in the jpeg
mode or in the NEF converter with raw files to get properly exposed
highlights. The purpose of this post is to compare Fotogenic's PS4
curve to the results of 0.5 stop positive exposure compensation
with Nikons Normal tone compensation (which he advised me to do in
earlier posting for a proper comparison) and to describe an easy
and quantitative method of doing so. I don't expect many responses
to this post, but it may interest a few.

Fotogenics analysis this matter is the best that I have seen on the
subject, and you should read it if you have not already done so:

http://fotogenetic.dearingfilm.com/custom_tone_curves.html

As he describes, you can photograph a gray card at various
exposures, bracketing around the exposure indcated by the light
meter for a particular ISO, and plot the pixel value in the
resulting picture to obtain the characteristic curve for those
conditions. This is the method used by Ansel Adams in his book The
Negative. The exposure that reproduces the pixel level of the gray
card is the proper one.

Actually, when using a gray card determine exosure, one should
increase the indicated exposure by 1/2 f/stop (0.5 EV). See Thom
Hogan's essay:

http://www.bythom.com/graycards.htm

What is the pixel level that one should get from taking a picture
of the gray card. At first you might think 127 (half way between 0
and 255), but for sRGB, but I think it is 118 as given by the
equation (1.055*x^(1/2.4)-0.055) * 255, where x is the normalized
exposure, !8% or 0.18 for the gray card. See Norman Koren's web
site for a good explanation of gamma, exposure, and monitor
calibration. The equation in the box on Norman's site is the
inverse gamma function for sRGB, going from pixel level in the file
to the monitor. In the current case we must go from the reading of
the sensor to the pixel level of the file.

http://www.normankoren.com/makingfineprints1A.html#Gammabox

Rather than taking numerous pictures of a gray card, you can get
this information and more by photographing a Macbeth color checker
chart and analyzing it with Imitest, an excellent and inexpensive
program by Norman Koren.

http://www.imatest.com/

I did a series of analyses with Imitest on the color checker for
the D70. To get proper exposure I took a reading from a gray card
and increased the exposure by 0.5 f/stops. Then I took shots of the
color checker at various exposures (based on the grey card reading
+ 1/2 stop) and processed them in Nikon Capture with a normal tone
compensation and with Fotogenic's PS4 tone curve, which boosts the
highlights the same as giving 0.5 stops more exposure.

The resulting photos and analyses are posted. The original files
were in TIFF, but these are too large to post, so I saved them with
jpeg.

http://bjanes.smugmug.com/gallery/377720

The gray card shot at the exposure indicated by the 18% gray card
gave a pixel value of 100 with the normal tone curve, and the shot
with 1/2 stop over the gray card reading gave 119, just where it
should be. I then exposed the chart normally (grey card reading +
1/2 stop) and applied the PS4 tone curve in NC and compared the
result to that of a normal tone compensation with a 1/2 stop
positive exposure compensation (grey card reading +1 stop) as
Fotogenic advised in a previous post.

The Imitest results are posted. For help in interpreting them see
the Imitest web site. At first, they seem imposing, but the
documentation makes it easy to interpret them. For example, look at
the analysis of the gray squares on the target with normal exposure
(1/2 stop over the grey card reading). The top left graph shows the
pixel level for each black and white square on the target. The top
right curve shows the characteristic curve.

http://bjanes.smugmug.com/gallery/377720/1/15066952/Large

Compare this to the half stop positive exposure compensation
(MacbethPlusOne). The characteristic curve is very similar to the
PS4 curve with normal exposure as Fotogenic says it should be. The
white square (optical density of 0.05 on the target) is at pixel
level 255 and any highlights over that would be blown with the
positive compensation.

http://bjanes.smugmug.com/gallery/377720/2/15066956/Large

Peruse the other results on your own; this post is long enough
already! I trial version of Imitest is available for free download
at the web site and you can test it on the posted images (be sure
to download them at their original sizes). Since the original
anayses were on TIFFs, the results might not agree exactly. Here we
are using only the Macbeth analysis portion of the program but it
has other functions not mentioned here.

I have no financial interest in Imitest, but do recommend it highly.

--
Bill Janes

 

[fotogenetic]

Bill,

Very interesting. A couple of things that I wanted to clarify. When
you say +.5 EV PS4 does that mean that you used the curve AND .5EV
or just the curve with 0EV resulting in .5EV?

Yes, +0.5 EV means 1/2 stop more than the gray card reading, which
is the normal exposure. As I understand it, Fotogenic's PS4 curve
applies another 0.5 EV to the highlights for a total boost of one
stop. This is why Fotogenic said the curve should be compared to a
normal exposure +0.5EV, which would be 1 stop over the gray card
reading, or plus one in my scheme.

Great experiment, however, I'm not quite sure that I understand why you used normal +0.5 EV + P&S curve. Simply using normal + P&S curve should get you medium gray, which is what my curve is calibrated to do. Essentially, the P&S curve makes whatever the camera meters end up as medium gray in the actual image, making that initial +0.5 that Thom Hogan says should be added when metering a gray card unecessary.

I have not had the time to learn how to read the Imatest charts but
there does appear to be some enhanced blue channel noise with the
.5EV that isn't there with the PS4 curve. Or am I reading this
wrong?

With regard to noise, I haven't been working with this much either,
but it appears to me that the curve redistributes the noise. It is
lower in zone 1 with the curve than with normal exposure, but
higher in zone 5. The average noise levels are about the same.

Of course the noise is lower with 0.5EV over normal exposure (plus
one over the gray card) but there is danger of blowing the
highlights and you have to use a lower shuttter speed, which you do
not want to do with available light shooting. See the posts by John
Friend.

Do you have any final thoughts to sum up after having done the
tests and studied the results?

Not yet. Imitest has another mode to ananyze a Q15 gray scale,
which gives a more detailed analysis. This one is from some MTF
testing I was doing on my kit lens (I did not apply the PS4 curve,
since I didn't even know about it then). I don't think Fotogenic
designed the curve with noise in mind, but I don't think that it is
a significant problem.

http://bjanes.smugmug.com/gallery/377720/2/15189401

Thanks for taking the time to do this thankless task. Eventually
people will come looking for this information.

Chris Sweet

Bill Janes

--
http://fotogenetic.dearingfilm.com

 

[fotogenetic]

Paul,

Thanks for the compliment, but I'm hardly in the same league as
Fotogenic. I had to do these tests for myself to understand what he
was doing.

Bill

Actually, I'd have to say your test exceeds mine as I did not use Imatest. I am very interested in your results. However, I'd like to see simply comparing normal +0.5 EV to normal + P&S. I think using a +1.0 EV total results in washed out colors possibly skewing the results.

Dear Bill,
Fotogenetic's work on this subject is fantastic, and his knowledge
of the D70/D100 camera and its characteristics astonishing. Equally
amazing is your information, research, and putting theories to the
test. Thanks for all the hard work you have done, to "validate" all
information on this subject. I had also experienced the
"underexposure phenomenon" and did much research and testing of my
own, although not as scientifically as you have shown. I have
gained a lot of knowledge from reading your posts and links, and
also the information from Fotogenetics site. This information can
be utilized for just about any camera, as it covers the
fundamentals of exposure, and how to overcome certain issues.
Thanks again for your hard work and validation, and I am sure many
people will benefit from your post.
Cheers,
Paul.

--
http://fotogenetic.dearingfilm.com

 

[Bill Janes]

Bill,

Thanks for clearing this up. So in your tests when you say +1EV you
are refering to the curve with +.5 EV added right?

Chris,

Not quite. The chart labeled "Macbeth + 1/2 PS4" is given the nominal exposure according to Thom's article, the gray card reading +1/2 EV.

The chart labeled "Macbeth + 1" is exposed at 1/2 EV over the nominal exposure, which is the gray card reading +1 EV. The normal Nikon curve is applied to this picture (Tone Comp = Normal) in the Advanced Raw pane in Nikon Capture.

In summary, I am comparing the PS4 curve at nominal exposure to the normal Nikon curve with +0.5 EV exposure as Fotogenic advised in a previous post:

http://forums.dpreview.com/forums/read.asp?forum=1034&message=11949927

I'm sorry for the confuson.

I had read that stuff by Thom awhile ago when I started using full
manual mode and spot metering almost exclusivly. I decided that no
matter how good the Nikon Matrix meter is I really needed to learn
how to do it myself . Forcing yourself to do it this way sometimes
will give you an appreaciation for how good Matrix is and when it
will most likely be wrong. For me the answer is RAW with EV3 and
watch the histogram.

That is a good approach. With matrix metering you never know what corrections the camera is using.

--
Bill Janes

 

[Chris Sweet]

Thanks for checking in fotogenetic. You have confirmed my understanding of what it is your curves were designed for. Which is to say a +.3 or +.5 EV boost when used with + -0EV in camera setting. So in order to do these tests evenly it should be Normal curve with +.5EV in camera AGAINST PS curve with + - 0EV in camera. Or in the case of the EV3 curve - Normal curve with +.3EV in camera against EV3 curve with + - 0EV in camera. If we could coerce Bill into running these tests (whatdaya think Bill?) it would be a level test providing valuable information.

Chris

Bill,

Very interesting. A couple of things that I wanted to clarify. When
you say +.5 EV PS4 does that mean that you used the curve AND .5EV
or just the curve with 0EV resulting in .5EV?

Yes, +0.5 EV means 1/2 stop more than the gray card reading, which
is the normal exposure. As I understand it, Fotogenic's PS4 curve
applies another 0.5 EV to the highlights for a total boost of one
stop. This is why Fotogenic said the curve should be compared to a
normal exposure +0.5EV, which would be 1 stop over the gray card
reading, or plus one in my scheme.

Great experiment, however, I'm not quite sure that I understand why
you used normal +0.5 EV + P&S curve. Simply using normal + P&S
curve should get you medium gray, which is what my curve is
calibrated to do. Essentially, the P&S curve makes whatever the
camera meters end up as medium gray in the actual image, making
that initial +0.5 that Thom Hogan says should be added when
metering a gray card unecessary.

I have not had the time to learn how to read the Imatest charts but
there does appear to be some enhanced blue channel noise with the
.5EV that isn't there with the PS4 curve. Or am I reading this
wrong?

With regard to noise, I haven't been working with this much either,
but it appears to me that the curve redistributes the noise. It is
lower in zone 1 with the curve than with normal exposure, but
higher in zone 5. The average noise levels are about the same.

Of course the noise is lower with 0.5EV over normal exposure (plus
one over the gray card) but there is danger of blowing the
highlights and you have to use a lower shuttter speed, which you do
not want to do with available light shooting. See the posts by John
Friend.

Do you have any final thoughts to sum up after having done the
tests and studied the results?

Not yet. Imitest has another mode to ananyze a Q15 gray scale,
which gives a more detailed analysis. This one is from some MTF
testing I was doing on my kit lens (I did not apply the PS4 curve,
since I didn't even know about it then). I don't think Fotogenic
designed the curve with noise in mind, but I don't think that it is
a significant problem.

http://bjanes.smugmug.com/gallery/377720/2/15189401

Thanks for taking the time to do this thankless task. Eventually
people will come looking for this information.

Chris Sweet

Bill Janes

--
http://fotogenetic.dearingfilm.com

 

[Bill Janes]

Paul,

Thanks for the compliment, but I'm hardly in the same league as
Fotogenic. I had to do these tests for myself to understand what he
was doing.

Bill

Actually, I'd have to say your test exceeds mine as I did not use
Imatest. I am very interested in your results. However, I'd like
to see simply comparing normal +0.5 EV to normal + P&S. I think
using a +1.0 EV total results in washed out colors possibly skewing
the results.

Fotogenic,

I don't want to argue with the master but I did compare P&S with normal exposure to normal +0.5 EV and no curve. However I think normal exposure is the gray card reading +0.5 EV as Thom Hogan suggests.

Agf1997, who is an Image/Color scientist, gave us the equation for derermining the pixel level in sRGB.

http://forums.dpreview.com/forums/read.asp?forum=1034&message=11966090

I was able to confirm this equation in the Microsoft/Hewett Packard definition of the sRGB color space. it is (1.055*R^(1/2.4)-0.055) * 255 and R is 18% or 0.18 for the Kodak gray card and the resulting pixel value is 117.6

http://www.w3.org/Graphics/Color/sRGB

Look in the section "Colorimetric Definitions and Digital Encoding"

If I take a picture of a gray 18% card with a properly calibrated camera using the sRGB color space, I should get a pixel value of 117.6. That comes from the definition of sRGB. To get this value from my D70 I need to add 0.5 EV to the exposure or use P&S4, exactly as Thom states in his Article. Adding 1 EV does wash out the colors.

Thom says if you use a card for exposure, you should use a 12% card and this is in agreement with my results. The midgray value of 18% comes from the print world. If I had a 12% card I wouldn't need +0.5 EV or P&S.

The results of +0.5 EV over gray card or gray card with P&S also give values to similar to the values that Bruce Fraser says you should have with the Macbeth charts in the sRGB color space.

In any event, I interpret my results as showing that the P&S curve does achieve its intended purpose. If you assume gray card is normal exposure, the D70 does underexpose and P&S4 corrects this. It is interesting that in his comparison of the D70 to the Canon Digital Rebel, Phil noted that the Rebel gave 1/3 EV more exposure than the D70. Maybe, as you suggest, the problem is with the ANSI standard and Canon takes this into account while the purist Nikon engineers stick to the standard.

Anyway, I agree with you that in taking pictures of the real world rather than gray cards with the D70, +0.33 or +0.5 EV does work better. Since I use Adobe Camera Raw, I usually give a small positive bias on exposure with the camera and add what is needed in ACR.

Bill

 

[Al_Pasadena]

In any event, I interpret my results as showing that the P&S curve
does achieve its intended purpose. If you assume gray card is

I can't comprehend how you could go through this detailed analysis, but yet be SO OFF BASE on the basics. P&S4's intended purpose is not to provide accurate exposure at +0.5 EV + P&S 4.0, but to provide accurate exposure at 0 EV + P&S 4.0. What you did is the equivalent of +1.0 EV..of course it's not going to achieve it's intended purpose!

You somehow confused yourself with Thom's explanation of ANSI exposure defaults, and the determination that +0.5 EV should be the starting point for everything on the D70, EVEN with a curve that ALREADY provides +0.5 EV.

Your tests of the P&S4 Curve are useless unless you start step 1 with the correct setting, which is 0 EV.

 

[Bill Janes]

Bill

Actually, I'd have to say your test exceeds mine as I did not use
Imatest. I am very interested in your results. However, I'd like
to see simply comparing normal +0.5 EV to normal + P&S. I think
using a +1.0 EV total results in washed out colors possibly skewing
the results.

Fotogenic,

I don't want to argue with the master but I did compare P&S with
normal exposure to normal +0.5 EV and no curve. However I think
normal exposure is the gray card reading +0.5 EV as Thom Hogan
suggests.

I was not entirely pleased with my post to you and was somewhat puzzled, since I thought my experiments confirmed yours but with different methodology and was hoping you would be pleased with them. We do not seem to be on the same wavelenght, possibly because of different pre-convieved theories.

So I am sending a postscript to you. What tests would you like to see? Gray card +PS4 vs Gray Card +0.5. I can do that. The Imitest Q13 subroutine might be better suited since it has more exposure steps that the Macbeth, but doesn't do colors. Or I could shoot both together. The color accruacy analysis tool is also very interesting. The data can also be brought into Microsoft excel for further graphing and numerical analysis.

Afterall, Ansel Adams did his own exposure tests and used whatever esposure worked best with different develoment. I guess our tone curves, gamma settings, etc are analagoud to his development.

The whole topic is rather confusing to me--I am a semi-retired physicain with some scientific background, but you really need to be a scientist to interpret all of these tests properly. I have no ego problems getting consultation in areas where I lack expertise--we do it all the time in medicine, usually in a collegial fashion. Perhaps I could invite Norman Koren, author of Imitest, to make a few comments. He is a nice guy--I've talked to him on the phone several times and communicated with him on e-mail about Imitest and he is very knowledgeable and helpful and also a talented photographer. It would give him a chance to sell a few more copies of Imitest.

Best regards,

Bill Janes

 

[fotogenetic]

Fotogenic,

I don't want to argue with the master but I did compare P&S with
normal exposure to normal +0.5 EV and no curve. However I think
normal exposure is the gray card reading +0.5 EV as Thom Hogan
suggests.

Exactly, which is why I created the curve--so you can expose for a gray card without having to use EV compensation to get medium gray ;-)

Agf1997, who is an Image/Color scientist, gave us the equation for
derermining the pixel level in sRGB.

http://forums.dpreview.com/forums/read.asp?forum=1034&message=11966090

I was able to confirm this equation in the Microsoft/Hewett Packard
definition of the sRGB color space. it is (1.055*R^(1/2.4)-0.055) *
255 and R is 18% or 0.18 for the Kodak gray card and the resulting
pixel value is 117.6

http://www.w3.org/Graphics/Color/sRGB

Look in the section "Colorimetric Definitions and Digital Encoding"

If I take a picture of a gray 18% card with a properly calibrated
camera using the sRGB color space, I should get a pixel value of
117.6. That comes from the definition of sRGB. To get this value
from my D70 I need to add 0.5 EV to the exposure or use P&S4,
exactly as Thom states in his Article. Adding 1 EV does wash out
the colors.

Thom says if you use a card for exposure, you should use a 12% card
and this is in agreement with my results. The midgray value of 18%
comes from the print world. If I had a 12% card I wouldn't need
+0.5 EV or P&S.

The results of +0.5 EV over gray card or gray card with P&S also
give values to similar to the values that Bruce Fraser says you
should have with the Macbeth charts in the sRGB color space.

In any event, I interpret my results as showing that the P&S curve
does achieve its intended purpose. If you assume gray card is
normal exposure, the D70 does underexpose and P&S4 corrects this.
It is interesting that in his comparison of the D70 to the Canon
Digital Rebel, Phil noted that the Rebel gave 1/3 EV more exposure
than the D70. Maybe, as you suggest, the problem is with the ANSI
standard and Canon takes this into account while the purist Nikon
engineers stick to the standard.

I think the same thing goes with the Fuji cameras as well.

Anyway, I agree with you that in taking pictures of the real world
rather than gray cards with the D70, +0.33 or +0.5 EV does work
better. Since I use Adobe Camera Raw, I usually give a small
positive bias on exposure with the camera and add what is needed in
ACR.

Bill

--
http://fotogenetic.dearingfilm.com

 

[Bill Janes]

Al,

Your point is valid and I repeated the analysis without reference to ANSI, using gray card +0.5 EV for Nikon's normal curve and gray card without compensation for the PS4 (0 EV). The PS4 curve places the 18% card at 131 pixels and the +0.5EV places it at129. I think photogenic was shooting for 127. Let me know what you think.

http://bjanes.smugmug.com/gallery/383257

The gray card reading +0.5 EV does give the correct pixel level for sRBG of 117 pixels. To be frank, I am still confused. What does Thom mean when he says to take a reading from an 18% gray card and increase 0.5EV? When would you want to do that?

Bill Janes

In any event, I interpret my results as showing that the P&S curve
does achieve its intended purpose. If you assume gray card is

I can't comprehend how you could go through this detailed analysis,
but yet be SO OFF BASE on the basics. P&S4's intended purpose is
not to provide accurate exposure at +0.5 EV + P&S 4.0, but to
provide accurate exposure at 0 EV + P&S 4.0. What you did is the
equivalent of +1.0 EV..of course it's not going to achieve it's
intended purpose!

You somehow confused yourself with Thom's explanation of ANSI
exposure defaults, and the determination that +0.5 EV should be the
starting point for everything on the D70, EVEN with a curve that
ALREADY provides +0.5 EV.

Your tests of the P&S4 Curve are useless unless you start step 1
with the correct setting, which is 0 EV.

--
Bill Janes

 

[Bill Janes]

Fotogenic,

I don't want to argue with the master but I did compare P&S with
normal exposure to normal +0.5 EV and no curve. However I think
normal exposure is the gray card reading +0.5 EV as Thom Hogan
suggests.

Exactly, which is why I created the curve--so you can expose for a
gray card without having to use EV compensation to get medium gray
;-)

Now I understand. I redid my analysis exposing for +0.5EV for Nikon's normal curve and 0EV for PS4. It does work as advertised.

Of course, when I take pictures of real objects I use the metered setting and the picture comes out dark without compensation or a curve. It would be simpler if Nikon adopted a scheme that gave a proper exposure without adding compensastion or a curve. I re-read the posting on your web site and I guess Ansell Adams had a good point when he differed with the ANSI folks.

Apparently, Canon has adopted such an approach similar to Adams' and the Digital Rebel gives +0.3EV esposure compared to the Nikon D70 according to Phil's tests. I guess this is what your newer curve does.
--
Bill Janes

 

[Man-Fai Wong]

Exactly, which is why I created the curve--so you can expose for a
gray card without having to use EV compensation to get medium gray
;-)

Doh! I've been doing it all wrong all this time. I've been exposing for people, things and places all this time, instead of a gray card. No wonder my photos suck! ;-)

Now I understand. I redid my analysis exposing for +0.5EV for
Nikon's normal curve and 0EV for PS4. It does work as advertised.

Ok, now that you finally understand that. I think it's my turn to be confused -- or maybe just complicate things for you again.

I haven't dug too deeply into the sRGB articles (and controversies) -- and don't think I have the time and expertise to fully understand them -- but it seems to me that your test results (and premise) goes against the grain of the various folks who argue that the pixel level is non-linear, including Norman Koren himself. In particular, the various sources you quoted, including NK and the W3C site, provide a formula for sRGB that suggests pixel level for 18% gray should be 117.6 as you concluded in the first post that started this sub-thread. But yet, your latest revised results suggest it should be 127.5 assuming the D70 needs that +.5EV.

BTW, didn't you say at some point that you did get close to 117 w/out any compensation?

Seems to me that the problem remains, and we are back to square 1 here.

Hmmm... Since you've already established correspondence w/ NK, could you perhaps ask him for his definitive word on what the pixel level should be for 18% gray w/ backed up test results (perhaps w/ the 10D in his Imatest samples)? I ask for test results because we already know he'll say 117.6 based on the formula. If the 10D is also giving 127.5, maybe he will explain why there's a discrepancy there. And maybe you can also ask him about the controversy too while you're at it.

Of course, when I take pictures of real objects I use the metered
setting and the picture comes out dark without compensation or a
curve. It would be simpler if Nikon adopted a scheme that gave a
proper exposure without adding compensastion or a curve. I re-read
the posting on your web site and I guess Ansell Adams had a good
point when he differed with the ANSI folks.

I tend to agree as well although all this stuff you dug up about sRGB definition (and the Imatest) doesn't seem to jive w/ the theory that the D70 is underexposing unless your test results are wrong. I was expecting to see 117-118, not 128.

But perhaps like you seem to surmise ever so slightly, it might not really matter much in practice as long as the metering is precise/consistent and we use it accordingly. Having in-camera histogram certainly helps a whole lot too. Otherwise, we would indeed need to run around w/ our gray cards and/or use an incident light meter w/ well-known properties all the time.

Apparently, Canon has adopted such an approach similar to Adams'
and the Digital Rebel gives +0.3EV esposure compared to the Nikon
D70 according to Phil's tests. I guess this is what your newer
curve does.

And so it seems except for the confusion brought up by the sRGB definition.

BTW, in some other post a few days ago(?), I mentioned that the Nikon pro bodies, or at least the D2H, don't have this same exposure issue based on what little I read over in the D100/D1/D2 forum. But I just revisited Phil's reviews for the D2H and 1DMk2 to doublecheck, and it seems to me that the D2H has the same issue. Don't know what some D2H users are claiming otherwise -- and yes, at least one of them also owned D100 and D70.

Man

--



Just another amateur learning to paint w/ 'the light of the world.' (John 8:12)
Motto for the season: 'Cameras are for making photos, not war...'
See my profile for more + some basic photog resources.
As usual, YMMV + caveat emptor.
Contact me at [email protected]
Indulge my fancies at http://www.pbase.com/mandnwong

 

[fotogenetic]

BTW, didn't you say at some point that you did get close to 117
w/out any compensation?

Seems to me that the problem remains, and we are back to square 1
here.

Actually, Bill's tests were with the P&S curve. With the EV3 curve, you'll get close to 117.

Hmmm... Since you've already established correspondence w/ NK,
could you perhaps ask him for his definitive word on what the pixel
level should be for 18% gray w/ backed up test results (perhaps w/
the 10D in his Imatest samples)? I ask for test results because we
already know he'll say 117.6 based on the formula. If the 10D
is also giving 127.5, maybe he will explain why there's a
discrepancy there. And maybe you can also ask him about the
controversy too while you're at it.

Of course, when I take pictures of real objects I use the metered
setting and the picture comes out dark without compensation or a
curve. It would be simpler if Nikon adopted a scheme that gave a
proper exposure without adding compensastion or a curve. I re-read
the posting on your web site and I guess Ansell Adams had a good
point when he differed with the ANSI folks.

I tend to agree as well although all this stuff you dug up about
sRGB definition (and the Imatest) doesn't seem to jive w/ the
theory that the D70 is underexposing unless your test results are
wrong. I was expecting to see 117-118, not 128.

His tests did conclude that without EV compensation or the curve, the value was 105, which is a tad too dark.

But perhaps like you seem to surmise ever so slightly, it might not
really matter much in practice as long as the metering is
precise/consistent and we use it accordingly. Having in-camera
histogram certainly helps a whole lot too. Otherwise, we would
indeed need to run around w/ our gray cards and/or use an incident
light meter w/ well-known properties all the time.

Apparently, Canon has adopted such an approach similar to Adams'
and the Digital Rebel gives +0.3EV esposure compared to the Nikon
D70 according to Phil's tests. I guess this is what your newer
curve does.

And so it seems except for the confusion brought up by the sRGB
definition.

BTW, in some other post a few days ago(?), I mentioned that the
Nikon pro bodies, or at least the D2H, don't have this same
exposure issue based on what little I read over in the D100/D1/D2
forum. But I just revisited Phil's reviews for the D2H and 1DMk2
to doublecheck, and it seems to me that the D2H has the same issue.
Don't know what some D2H users are claiming otherwise -- and yes,
at least one of them also owned D100 and D70.

Man

--



Just another amateur learning to paint w/ 'the light of the world.'
(John 8:12)
Motto for the season: 'Cameras are for making photos, not war...'
See my profile for more + some basic photog resources.
As usual, YMMV + caveat emptor.
Contact me at [email protected]
Indulge my fancies at http://www.pbase.com/mandnwong

--
http://fotogenetic.dearingfilm.com

 

[Al_Pasadena]

Something's wrong if +0.5 EV + P&S4 = 129 and 0 EV + P&S4 = 131. What's your exact workflow? The reason I ask is because you mentioned in a previous post that you use ACR. If that's the case, you need to use Nikon Capture if you shoot RAW for the curve to be applied, or else shoot JPEG.

In real life, I shoot with the built-in Medium-Low tone curve and +0.33EV , in sRGB Ia mode, HUE -3. I used to use the Fotogenetics 0.3EV curve, but found it couldn't do as well with skin tones compared to those settings. I always found the full-strength White Wedding curve (now P&S) too bright for my taste.

Al,

Your point is valid and I repeated the analysis without reference
to ANSI, using gray card +0.5 EV for Nikon's normal curve and gray
card without compensation for the PS4 (0 EV). The PS4 curve places
the 18% card at 131 pixels and the +0.5EV places it at129. I think
photogenic was shooting for 127. Let me know what you think.

http://bjanes.smugmug.com/gallery/383257

The gray card reading +0.5 EV does give the correct pixel level for
sRBG of 117 pixels. To be frank, I am still confused. What does
Thom mean when he says to take a reading from an 18% gray card and
increase 0.5EV? When would you want to do that?

Bill Janes

 

[Bill Janes]

I haven't dug too deeply into the sRGB articles (and controversies)
-- and don't think I have the time and expertise to fully
understand them -- but it seems to me that your test results (and
premise) goes against the grain of the various folks who argue that
the pixel level is non-linear, including Norman Koren himself. In
particular, the various sources you quoted, including NK and the
W3C site, provide a formula for sRGB that suggests pixel level for
18% gray should be 117.6 as you concluded in the first post that
started this sub-thread. But yet, your latest revised results
suggest it should be 127.5 assuming the D70 needs that +.5EV.

BTW, didn't you say at some point that you did get close to 117
w/out any compensation?

Man

That was with the gray card reading plus 0.5EV using Adobe Camera Raw with default settings. On my repeat measurements, I got 128 with Nikon Capture's defualt settings and 121 with ACR's defualt settings. I made no exposure adjustment (to the white point) in either program but the value does change with gamma (affects midtone values) and contrast settings. The latter apply an s shaped curve and primarily affect the 2nd and 5th gray patches of the Macbeth card.

I'm still not sure about the ANSI calibration issue, but Fotogenic and Al Pasanda stressed that the P&S curve was based on the gray card reading with no exposure compensation, so that is what I used.

Seems to me that the problem remains, and we are back to square 1
here.

Hmmm... Since you've already established correspondence w/ NK,
could you perhaps ask him for his definitive word on what the pixel
level should be for 18% gray w/ backed up test results (perhaps w/
the 10D in his Imatest samples)? I ask for test results because we
already know he'll say 117.6 based on the formula. If the 10D
is also giving 127.5, maybe he will explain why there's a
discrepancy there. And maybe you can also ask him about the
controversy too while you're at it.

Yes, I'm confused also and after I study the Imitest results I will try to contact Norman. His thoughtful reply could induce some readers to buy his program.

--
Bill Janes

 

[Man-Fai Wong]

BTW, didn't you say at some point that you did get close to 117
w/out any compensation?

Seems to me that the problem remains, and we are back to square 1
here.

Actually, Bill's tests were with the P&S curve. With the EV3
curve, you'll get close to 117.

Actually, that was just a "BTW", not the actual problem I had w/ trying to sync his results w/ the various hypotheses. The actual problem was that the sRGB formula tells us we should get 117-118 for 18% gray, which means using +.5EV on a 13% meter. Yet, Bill is getting more like 128.

If 117-118 is truly 18% gray and +.5EV really should yield 128 on the D70 (and +.3EV should yield 117-118), then it sounds like Nikon's not really following the ANSI standard w/ the D70 -- not quite anyway -- as that should call for +.5EV.

I think that's basically what I was starting to conclude, except there's still the controversy between the scientific guys and people like Timo. Someone like Timo might jump in and say, "See, the 128 result w/ +.5EV adds evidence that I'm right since Nikon's probably just following the ANSI standard." If Nikon is indeed using ANSI standard, then I'm not sure how else to look at this.

Man

--



Just another amateur learning to paint w/ 'the light of the world.' (John 8:12)
Motto for the season: 'Cameras are for making photos, not war...'
See my profile for more + some basic photog resources.
As usual, YMMV + caveat emptor.
Contact me at [email protected]
Indulge my fancies at http://www.pbase.com/mandnwong

 

[Man-Fai Wong]

Man

That was with the gray card reading plus 0.5EV using Adobe Camera
Raw with default settings. On my repeat measurements, I got 128
with Nikon Capture's defualt settings and 121 with ACR's defualt
settings. I made no exposure adjustment (to the white point) in
either program but the value does change with gamma (affects
midtone values) and contrast settings. The latter apply an s shaped
curve and primarily affect the 2nd and 5th gray patches of the
Macbeth card.

Bill,

First, I'd forget about using ACR for these tests. You don't want to add anymore variables to the equation than you already have w/out using ACR. Personally, for these kinds of tests, I would just stick w/ in-camera JPEG results w/ accurate in-camera WB because we want to know the D70's normal behavior (and then adjust as needed). Once you start doing PP and whatever tweaks, things will fall apart. I suppose using Nikon View or Capture to do RAW conversion would be fine also, but it's probably still best to get the in-camera WB correct rather than fix it in post.

Anyway, read my reply to fotogenetic just above for more on my thoughts about all this.

http://forums.dpreview.com/forums/read.asp?forum=1034&message=12116823

I'm still not sure about the ANSI calibration issue, but Fotogenic
and Al Pasanda stressed that the P&S curve was based on the gray
card reading with no exposure compensation, so that is what I used.

Yep. That's my understanding too. But read my above reply to see more. Basically, I'm not so sure Nikon uses ANSI standard anymore unless the sRGB formula is wrong.

Yes, I'm confused also and after I study the Imitest results I will
try to contact Norman. His thoughtful reply could induce some
readers to buy his program.

Sounds good. Thanks again for all the efforts.

Hmmm... as I think about this some more, I think something just came to mind that might help address the issue I have. I will post it as a follow-up to the above reply to fotogenetic.

Man

--



Just another amateur learning to paint w/ 'the light of the world.' (John 8:12)
Motto for the season: 'Cameras are for making photos, not war...'
See my profile for more + some basic photog resources.
As usual, YMMV + caveat emptor.
Contact me at [email protected]
Indulge my fancies at http://www.pbase.com/mandnwong

 

[Man-Fai Wong]

Hmmm... as I think about this some more, I think something just
came to mind that might help address the issue I have. I will post
it as a follow-up to the above reply to fotogenetic.

--



Just another amateur learning to paint w/ 'the light of the world.' (John 8:12)
Motto for the season: 'Cameras are for making photos, not war...'
See my profile for more + some basic photog resources.
As usual, YMMV + caveat emptor.
Contact me at [email protected]
Indulge my fancies at http://www.pbase.com/mandnwong

 

[Chris Sweet]

I agree with Man here about using in camera JPEG's for these sorts of tests. One less thing to worry about when trying for reproducable results. It would still be interesting to see the data using Normal curve with +.5EV or +.3EV against PS4 or EV3 with + - 0EV.

Even though this thread got sidetracked discussing using all the right variables it is still a very good idea.

Chris

Man

That was with the gray card reading plus 0.5EV using Adobe Camera
Raw with default settings. On my repeat measurements, I got 128
with Nikon Capture's defualt settings and 121 with ACR's defualt
settings. I made no exposure adjustment (to the white point) in
either program but the value does change with gamma (affects
midtone values) and contrast settings. The latter apply an s shaped
curve and primarily affect the 2nd and 5th gray patches of the
Macbeth card.

Bill,

First, I'd forget about using ACR for these tests. You don't want
to add anymore variables to the equation than you already have
w/out using ACR. Personally, for these kinds of tests, I would
just stick w/ in-camera JPEG results w/ accurate in-camera WB
because we want to know the D70's normal behavior (and then adjust
as needed). Once you start doing PP and whatever tweaks, things
will fall apart. I suppose using Nikon View or Capture to do RAW
conversion would be fine also, but it's probably still best to get
the in-camera WB correct rather than fix it in post.

Anyway, read my reply to fotogenetic just above for more on my
thoughts about all this.

http://forums.dpreview.com/forums/read.asp?forum=1034&message=12116823

I'm still not sure about the ANSI calibration issue, but Fotogenic
and Al Pasanda stressed that the P&S curve was based on the gray
card reading with no exposure compensation, so that is what I used.

Yep. That's my understanding too. But read my above reply to see
more. Basically, I'm not so sure Nikon uses ANSI standard anymore
unless the sRGB formula is wrong.

Yes, I'm confused also and after I study the Imitest results I will
try to contact Norman. His thoughtful reply could induce some
readers to buy his program.

Sounds good. Thanks again for all the efforts.

Hmmm... as I think about this some more, I think something just
came to mind that might help address the issue I have. I will post
it as a follow-up to the above reply to fotogenetic.

Man

--



Just another amateur learning to paint w/ 'the light of the world.'
(John 8:12)
Motto for the season: 'Cameras are for making photos, not war...'
See my profile for more + some basic photog resources.
As usual, YMMV + caveat emptor.
Contact me at [email protected]
Indulge my fancies at http://www.pbase.com/mandnwong

 

[Bill Janes]

Al--

Something's wrong if +0.5 EV + P&S4 = 129 and 0 EV + P&S4 = 131.
What's your exact workflow? The reason I ask is because you
mentioned in a previous post that you use ACR. If that's the case,
you need to use Nikon Capture if you shoot RAW for the curve to be
applied, or else shoot JPEG.

Below is a cut and paste from my message. I think you misinterpreted the results. The PS4 curve without any compensation placed the gray card at 131 pixels. An exposure compensation of +0.5EV and no curve placed the gray at 129. This is exactly what the PS4 curve is designed to do.

"Your point is valid and I repeated the analysis without reference to ANSI, using gray card +0.5 EV for Nikon's normal curve and gray card without compensation for the PS4 (0 EV). The PS4 curve places the 18% card at 131 pixels and the +0.5EV places it at129. I think photogenic was shooting for 127. Let me know what you think."

All analyses were done in Nikon Capture and raw files, since as you say, ACR ignores the tone compensation and all other of the tags except white balance.

In real life, I shoot with the built-in Medium-Low tone curve and
+0.33EV , in sRGB Ia mode, HUE -3. I used to use the Fotogenetics
0.3EV curve, but found it couldn't do as well with skin tones
compared to those settings. I always found the full-strength White
Wedding curve (now P&S) too bright for my taste.

Al,

Your point is valid and I repeated the analysis without reference
to ANSI, using gray card +0.5 EV for Nikon's normal curve and gray
card without compensation for the PS4 (0 EV). The PS4 curve places
the 18% card at 131 pixels and the +0.5EV places it at129. I think
photogenic was shooting for 127. Let me know what you think.

http://bjanes.smugmug.com/gallery/383257

The gray card reading +0.5 EV does give the correct pixel level for
sRBG of 117 pixels. To be frank, I am still confused. What does
Thom mean when he says to take a reading from an 18% gray card and
increase 0.5EV? When would you want to do that?

Bill Janes

--
Bill Janes