Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Canon 5DMkII ISO usage and LR/ACR  (Read 30647 times)
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« on: January 17, 2009, 09:48:31 PM »
ReplyReply

After having received enough raw files to analyze this issue thoroughly, I can make some statements regarding the usability of ISO steps with the 5D2.


1. When shooting JPEG, using the 1/3 ISOs and 6400, 12800 and 25600 is sometimes justified (if the exposure can not be adjusted as necessary).

However, using ISO 50 is not justifiable.

2. when working with the raw data:

a. ISO 6400, 12800 and 25600 are wasting of the dynamic range by a full stop per ISO stop.

b. ISO 50 is identical to ISO 100, except for the metering. This means: if you meter for ISO 50, then the result is the same as metering for ISO 100 with +1 EV bias.

c. ISO 125, 250, 500, 1000, 2000 and 4000: these are numerical derivatives of the next lower stop. They cut off 1/3 stop from the highlights without enhancing the shadows, i.e. the DR is reduced by 1/3 EV; there is no reason to use these.

d. ISO 160, 320, 640, 1250, 2500, 5000: these are numerical derivatives from the next higher ISO. They do not reduce the DR; they do reduce the number of levels, but the leftover (about 11800 levels) is more than enough. However, when someone switches to 1/3 ISO steps, the "bad" steps are in the way.

Now, to ACR: ISO 160, 320, 640, 1250, 2500, 5000 are even more mistreated by ACR, than the other ISOs: ACR does not recognize the reduced numerical pixel value range, thus it does not notice pixel saturation. The consequence is, that clipped areas turn out with wrong colors. For example if the greens are saturated in the sky, ACR converts that in magenta.
Logged

Gabor
Daniel Browning
Full Member
***
Offline Offline

Posts: 142


« Reply #1 on: January 18, 2009, 12:40:08 AM »
ReplyReply

Thanks for posting your analysis. I had hoped Canon would use a more sensible implementation of "tweener" ISO settings (compared to previous non-1D cameras), but I'm not surprised.

Unfortunately, 1/3 ISO is one of the settings I have to enable for movie mode on my 5d2 in order to get more reasonable "control" (if you can call it that) over the exposure.
Logged

--Daniel
Ben Rubinstein
Sr. Member
****
Offline Offline

Posts: 1733


« Reply #2 on: January 18, 2009, 04:09:01 AM »
ReplyReply

Do you have an analysis of the original 5D, would be much appreciated especially re ACR.
Logged

Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #3 on: January 18, 2009, 10:30:01 AM »
ReplyReply

Quote from: pom
Do you have an analysis of the original 5D, would be much appreciated especially re ACR.
1. I have only one raw file with 1/3 ISO, at 1250; that appears to be genuine, i.e. not numerically derived.

2. ISO 3200 is fake with the 5D.

3. I can not determine with the available raw files how ISO 50 works. I need a pair of shot of anything fixed scenery with constant illumination, one shot @ 50, one @ 100, either with the identical exposure or ISO 50 with 1 stop higher exposed.

4. I don't think the same problem with ACR's interpretation occurs with the 5D.
Logged

Gabor
madmanchan
Sr. Member
****
Offline Offline

Posts: 2108


« Reply #4 on: January 18, 2009, 10:31:15 AM »
ReplyReply

Gabor, unfortunately the reason for this discrepancy is that the original unit that we received from Canon (obviously a preproduction model) had identical white points at all ISOs. We had no way of knowing at the time that the final release would be different in this regard.
Logged

Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #5 on: January 18, 2009, 10:42:29 AM »
ReplyReply

Quote from: madmanchan
the original unit that we received from Canon (obviously a preproduction model) had identical white points at all ISOs
This is extremely strange. I am not doubting what you are saying, but I would like to see such a raw file; I wonder, how for example ISO 320 has been realized at that time. The ISO 320 pixel values are simply the ISO 400 values, multiplied by 0.8 (after black level correction, and recorrected afterwards). Thus is is natural, that the saturation level goes down, just like with the 40D, 50D, etc.

Logged

Gabor
stever
Sr. Member
****
Offline Offline

Posts: 1065


« Reply #6 on: January 18, 2009, 10:54:26 AM »
ReplyReply

is there any practical reason for using "in-between" ISOs (or Canon supplying them in the first place)?
Logged
Daniel Browning
Full Member
***
Offline Offline

Posts: 142


« Reply #7 on: January 18, 2009, 11:02:57 AM »
ReplyReply

Quote from: stever
is there any practical reason for using "in-between" ISOs (or Canon supplying them in the first place)?

For JPEG shooters it allows you to get exposure *just* right for a given aperture/shutter and picture profile. For raw shooters, if the "tweener" ISO worked like normal ones, it would allow you to do the ITTR (ISO to the right) with more precision. For movie shooters, it gives you a slight advantage in "manual control" over the exposure. There are probably more reasons.
Logged

--Daniel
denoise
Newbie
*
Offline Offline

Posts: 10



WWW
« Reply #8 on: January 19, 2009, 07:39:12 AM »
ReplyReply

Quote from: Daniel Browning
For JPEG shooters it allows you to get exposure *just* right for a given aperture/shutter and picture profile. For raw shooters, if the "tweener" ISO worked like normal ones, it would allow you to do the ITTR (ISO to the right) with more precision. For movie shooters, it gives you a slight advantage in "manual control" over the exposure. There are probably more reasons.

it's not possible to control select ISO, shutter speed or aperture in movie mode.
Logged

Daniel Browning
Full Member
***
Offline Offline

Posts: 142


« Reply #9 on: January 19, 2009, 10:07:16 AM »
ReplyReply

Quote from: denoise
it's not possible to control select ISO, shutter speed or aperture in movie mode.

Exactly. The trick, obviously, is to cover the lens and slowly uncover it and watch as the camera chooses values and try to lock exposure at *just* the right time. After maxing all three values, the first it will change is 1/f shutter. Then ISO. With 1/3-stop ISO increments, it's easier to watch it slowly decrease so you can lock exposure at the right ISO. Otherwise the increments are too fast and crude to time correctly. (Especially since the procedure must be repeated for every shot.)
Logged

--Daniel
eronald
Sr. Member
****
Online Online

Posts: 3976



WWW
« Reply #10 on: January 19, 2009, 05:10:09 PM »
ReplyReply

I made several C1 ICC profiles for the original 5D, and these produced good color on the units I tested, but never worked for customers. I always assumed that it was a batch issue with respect to  the sensor, but now I wonder whether it wan't an ISO and conversion effect.

Edmund
Logged

Edmund Ronald, Ph.D. 
Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 558


WWW
« Reply #11 on: January 19, 2009, 05:38:58 PM »
ReplyReply

Now my 5D2 has come, and I tried to find out how to expose.

Hardware was a Zeiss Contax Sonnar 85, used via a CameraQuest adapter; 2 Solux® lamps, the ColorChecker card, a Kodak and a Wand gray card.

In-camera settings were "everything neutral". ACR settings were Brightness +50, Contrast +25, Tone Curve Moderate Contrast, everything else zero.

The in-camera lightmeter shows 2/3 EVs lower than the Asahi Pentax spotmeter, i.e. when the Asahi shows EV 9 ~ f/11, 1/8 second, the camera shows 1/5 second. (200 ISO) This is also the value that places the in-camera histogram in the middle of its window.

When opened in Rawnalyze, the DNGs, but not the CR2s, show "+1/3" left of the lightness slider. The Rawnalyze histograms of a gray card  however look exactly the same for both CR2s and DNGs.

On the 1/3 f-stop level, there is virtually no difference between the highlight warnings in aRGB and sRGB. It looks like the warning is triggered as soon as the rightmost part of the histogram touches the -1/3 EV mark (in the Rawnalyze histogram). So one can use the first highlight warning as ETTR exposure; no need to go back 1/3 EV.

When I opened the DNG image of the Kodak gray card in ACR (CS3), the peak of the histogram was exactly in the middle of the histogram window, if I chose Adobe Standard as the camera profile. The Camera Neutral profile moved the top 0,2 EVs to the left.

I tried to find out how these different exposures related to the density of the CC 4/4 patch. Once upon a time, there was a densitometer on the Mac. For now, I tried the Color Meter utility, and the color sampler pipette in ACR. Results were different from each other, and different for 2 camera profiles, Adobe Standard and Camera Neutral. (Camera Neutral was used when shooting).

The correct RGB values for ProPhoto according to BabelColor are
101 or 102/102/102. Here are mine:

--these were faulty. New values in post #19. Hening.

So...the only finding that I can immediately use is the highlight warning. The rest is a mess to me. In particular: How does the +1/3 come in, which is shown by the  Rawnalyze lightness slider? In which direction does it go? How does ACR handle this? ***How can I achieve an exposure that renders D 0.7 as D 0.7 ?***

Jeez it's certainly not easy to photograph if you have to reverse engineer everything you touch before you can use it...

And why are the RGB values so way off? I am aware of that they are for 8 bit, while I am working in 16 bit. But that could not change the overall density that much? A single set I compared made no difference. And what are the differences between the Mac Color Meter and the ACR pipette about?

Sigh...
« Last Edit: January 20, 2009, 06:22:36 PM by Hening » Logged

JDClements
Sr. Member
****
Offline Offline

Posts: 312



WWW
« Reply #12 on: January 19, 2009, 07:41:18 PM »
ReplyReply

I became confused at post #5: unsure whether it questions the original post, or the analysis of the original 5D.

Am I right in assuming the conclusions for the 5D MkII are to avoid 6400 if possible, and to set the ISO increment to 1-stop to avoid the bad 1/3 stops (or memorize the bad stops in your list c. and avoid them)?
Logged

Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #13 on: January 19, 2009, 08:24:56 PM »
ReplyReply

Some answers.

Quote from: Hening
When opened in Rawnalyze, the DNGs, but not the CR2s, show "+1/3" left of the lightness slider. The Rawnalyze histograms of a gray card  however look exactly the same for both CR2s and DNGs.

Rawnalyze blindly shows the exposure adjustment dictated by the creator of the raw file, in this case by the Adobe DNG converter. This adjustment usually reflects the difference between the camera maker's and Adobe's interpretation of the ISO standard.

The raw histogram displayed by Rawnalyze is absolute; it reflects the unadjusted raw pixel values. Only the scale underneath depends on the top saturation level, which is fixed for the camera and the ISO (like 16383).

NOTE: the mapped histogram is only for educational purpose, i.e. for the demonstration of what is happening there, particularly when discussing ETTR. Neither demosaicing, nor color space conversion (from the camera's color space to whatever) took place before calculating that histogram. Accordingly, the effects of the contrast and saturation adjustments are not reflected in the mapped histogram.

Quote
So...the only finding that I can immediately use is the highlight warning
You mean the clipping warning, right? That's how it should be.

The entire concept of metering and of headroom in highlights and footroom in shadows are the worthless leftovers of the film era, for the reassurance of dynosaurs. There is no such thing as headroom and footroom with raw processing. The raw data is absolute, and you can put the head and foot wherever you want to. This is particularly relevant with ETTR: you are striving to eliminate the headroom in order to lift out the shadows from the noisy region.

Quote
How does the +1/3 come in, which is shown by the  Rawnalyze lightness slider? In which direction does it go? How does ACR handle this?
The 1/3 reflects ACR's rendering. +1/3 indicates, that ACR will apply a +1/3 EV adjustment to the "exposure" slider (but it is not shown on the slider!!).

Quote
***How can I achieve an exposure that renders D 0.7 as D 0.7 ?***
I know it is not nice to say so, but the question itself is wrong. The goal of raw conversion is not to render D 0.7 as D 0.7 (if you are expecting that, then go with JPEG). Initially, it should show how the shot is on its own, i.e. without faking exposure, ISO, etc. From that point, you should decide what to do. Adobe's misunderstanding of ACR's role is becoming apparent here: they think they are supposed to "correct" the shot automatically, instead of helping you in processing it

If you underexposed the shot, then it is underexposed. Basta. If you have blown the highlights, then they are blown. Basta.

The raw processor is not supposed to make an experienced photographer from someone, who does not understand his DSLR. P&S cameras are supposed to do that.

Quote
And why are the RGB values so way off?
Pls describe it better, what you mean; this is an important issue.

Quote
Sigh...

The questions indicate, that this is not so bad.
Logged

Gabor
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #14 on: January 19, 2009, 08:53:01 PM »
ReplyReply

Quote from: JDClements
I became confused at post #5: unsure whether it questions the original post, or the analysis of the original 5D
Neither, nor. Eric gave an explanation for the discrepance.

Quote
Am I right in assuming the conclusions for the 5D MkII are to avoid 6400 if possible
To avoid generally. There is no reason to use it when working with raw: you can achieve the very same in PP without reducing the DR. ISO 6400 is for JPEG only.

Quote
and to set the ISO increment to 1-stop to avoid the bad 1/3 stops (or memorize the bad stops in your list c. and avoid them)?
No need to memorize. Half of the 1/3 EV stops are bad, the other half are not different from the full stop. Thus, there is no need to use any.

You did not mention ISO 50: drop it from the menu.
Logged

Gabor
eronald
Sr. Member
****
Online Online

Posts: 3976



WWW
« Reply #15 on: January 20, 2009, 02:05:31 PM »
ReplyReply

Panopeeper,

 The guys here are photographers, not scientists. So, exposure to them is A PRIORI exposure, meaning a decision which is taken on the ground prior to making a shot, not A POSTERIORI terms which you as a scientist use to analyze  the resultant pixel map.

 In this context your remarks are misleading the poor pragmatic men and women in this forum who are trying to earn a living with their cameras. Exposure for them is something to be done with respect to an expected reflectance in a given lighting, with precautions for headroom for speculars (forehead, nose), tableware  or in-image lights (lamps in architecture), and acceptable shadows.

Exposure is a term of the art that is very much alive. As a rule of thumb, I'd recommend  one stop headroom over the white of a Colorchecker.

Ronald

Quote from: Panopeeper
Some answers.

You mean the clipping warning, right? That's how it should be.

The entire concept of metering and of headroom in highlights and footroom in shadows are the worthless leftovers of the film era, for the reassurance of dynosaurs. There is no such thing as headroom and footroom with raw processing. The raw data is absolute, and you can put the head and foot wherever you want to. This is particularly relevant with ETTR: you are striving to eliminate the headroom in order to lift out the shadows from the noisy region.
« Last Edit: January 20, 2009, 02:09:19 PM by eronald » Logged

Edmund Ronald, Ph.D. 
Daniel Browning
Full Member
***
Offline Offline

Posts: 142


« Reply #16 on: January 20, 2009, 05:39:42 PM »
ReplyReply

Quote from: eronald
Exposure is a term of the art that is very much alive.
Agreed. The two most significant artistic choices in exposure are:

1. What portion of the dynamic range of the scene to capture.
2. How to display that dynamic range in a print.

#1 is deciding, for example, that everything outside the window will blow out in order to have enough usable range for the indoors.

#2 is the gamma, tone curve, and contrast adjustments made in post processing (or in-camera JPEG).

Maybe in #1 you decided to underexpose to capture some of the detail outside the window. Then in #2, use a film-like tone curve so the highlights blow out gradually and gracefully.

After and between the artistic choices there is a lot of room for technical excellence. After choosing how much dynamic range to capture, for example, a technical mistake could lead to capturing less than that, or the wrong part of the range. Or, a technical mistake in the camera settings or raw converter could cause the image to not produce the desired effect.

The way I see it, the art happens in the mind; technical excellence is just trying to get that from the mind into real life.

Quote from: eronald
The guys here are photographers, not scientists. So, exposure to them is A PRIORI exposure, meaning a decision which is taken on the ground prior to making a shot, not A POSTERIORI terms which you as a scientist use to analyze  the resultant pixel map.

 In this context your remarks are misleading the poor pragmatic men and women in this forum who are trying to earn a living with their cameras. Exposure for them is something to be done with respect to an expected reflectance in a given lighting, with precautions for headroom for speculars (forehead, nose), tableware  or in-image lights (lamps in architecture), and acceptable shadows.

In film, #1 and #2 above were always commingled, due to the nonlinear response. But now that response is linear and we can make more choices about #2 after the fact, I think it's beneficial to understand the full range of possibilities.

You can get great results with a simpler system (e.g. your white+1 stop is a great one), but with the time and inclination to learn more complicated possibilities (boring technical stuff), one is able to photograph with greater precision (less mistakes) and more ambition (subjects that were too heretofore too difficult to render in the way you desired).
Logged

--Daniel
JDClements
Sr. Member
****
Offline Offline

Posts: 312



WWW
« Reply #17 on: January 20, 2009, 06:15:38 PM »
ReplyReply

Quote from: Panopeeper
You did not mention ISO 50: drop it from the menu.
I never turned the ISO expansion on, so not an issue for me. On the 6400, that may explain why when using Auto ISO, it only runs up to 3200. I have set the camera to full stops to avoid the tweener ISOs.

Re: Art vs. Science comments, it is useful to know how something impacts your art, and if in-between ISO settings reduce the dynamic range, I want to know that.
Logged

Hening Bettermann
Sr. Member
****
Offline Offline

Posts: 558


WWW
« Reply #18 on: January 20, 2009, 06:21:21 PM »
ReplyReply

Hi Gabor. Thank you very much for your detailed reply. -

Yes, by "higlight warning" I meant "highlight clipping warning". I agree that it is like it should be. I just thought it might also have been so that when the warning starts, highlights *are* already clipped. In this case, I would have needed to "step back" 1/3 EV from the 1st warning. So I found it useful to see in Rawnalyze what actually is going on.

>The entire concept of metering and of headroom in highlights and footroom in shadows are the worthless leftovers of the film era, for the reassurance of dynosaurs. There is no such thing as headroom and footroom with raw processing. The raw data is absolute, and you can put the head and foot wherever you want to.

Well if the clipping warning starts at -1/3 EV from saturation, then at this point you have 1/3-x EV of "headroom" left. This is just semantics.

>This is particularly relevant with ETTR: you are striving to eliminate the headroom in order to lift out the shadows from the noisy region.

This is exactly my understanding of ETTR.
---
>>***How can I achieve an exposure that renders D 0.7 as D 0.7 ?***

>I know it is not nice to say so, but the question itself is wrong.

Well if the question is wrong it's wrong and it must be OK to say so, preferably in a nice way, as you did :-) For the remainder of the paragraph, we either misunderstand each other or disagree.
 
>The goal of raw conversion is not to render D 0.7 as D 0.7 [...]. Initially, it should show how the shot is on its own, i.e. without faking exposure, ISO, etc.

This is exactly my idea. But doesn't this imply that if I measure and shoot a gray card, the image should have D 0.7 after all *standard* processing is done? (Presupposed my light meter is correct) Isn't this exactly  "showing the shot on its own"? Well there could be other sources of error, the sensor itself,... and I agree that the raw conversion should not try to cover up anything. But  what I asked for was an *exposure*  with a predictable result, not for a raw processing that faked anything. What I actually meant was an overall *standard* procedure, of which exposure is the first step, and the raw conversion the next.  

>From that point, you should decide what to do. [...]

Yes. And in order to have a guideline for the subsequent processing, I make what I call my reference shot. In this shot, I blow on highlights and shadows ;-) It's based on the gray card (in the future to  be replaced by my palm, which turns out to have the same reflectance), and its only purpose is to aid my memory in remembering how dark/light the scene actually was; in technical terms: show correct mid-tones; which ETTR does not. In this sense, I do not think that metering is obsoleted by our being able to see the histogram at shooting time. (In case your phrasing  "The entire concept of *metering* and of headroom in highlights and footroom in shadows are the worthless leftovers of the film era" meant something like that).

My above color values are wrong, I'll try to delete them from the post. I made new ones, which show a somewhat more understandable pattern, yet leave some questions. DNGs opened in ACR with my standard settings as noted above.

RGB values for CC patch 4/4 in ProPhoto. Correct values according to BabelColor:
101 or 102/102/102
[attachment=11018:RGB_valu...ProPhoto.tiff]

So the answer to my question seems to be: middle gray is rendered as middle gray when using external light metering (or adding 2/3 EV to the camera metering) and the Adobe Standard camera profile - if the ACR pipette is correct. This is with ACRs 1/3 EV added. Question left: What causes the difference between the Mac Color Meter and the ACR pipette?

Next step, I will make my own DNG profile and figure out how I have to expose with that to achieve correct tone values.

Getting closer, bye and bye... :-)

Logged

Tim Lookingbill
Sr. Member
****
Offline Offline

Posts: 1153



WWW
« Reply #19 on: January 21, 2009, 03:24:27 AM »
ReplyReply

Hening,

When using Apple's DigitalColor Meter, you need to set it to Lab readouts instead of RGB which derives those numbers NOT from the color space of the file within the imaging software but from how the monitor profile is used to adjust the preview in a color managed system. You'ld have to convert to your monitor profile within Photoshop to get the same readouts as the ColorMeter. This of course will not work within ACR and Lightroom.

If you know the Lab measurements of your CC chart, you can trust the ColorMeter within ACR and Lightroom. I've used this procedure tweaking within ACR my own CC chart to calibrate my camera getting very accurate results once opened in Photoshop which will show the same Lab numbers as the meter.
Logged
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad