Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: Calibrating The Spectraview Reference 241  (Read 2914 times)
adrian_m
Newbie
*
Offline Offline

Posts: 6


Photographer working in South East UK


« on: July 24, 2013, 06:04:20 AM »
ReplyReply

Hi folks,
I am struggling to get a calibrated system that I can feel 100% confident about; this forum was recommended as a place where sensible and non-inflammatory advice might be sought.
I shall endeavour to explain things clearly:

My Setup
---------
Spectraview Reference 241 monitor
DVI (monitor) to HDMI (laptop)
Toshiba Satellite Pro with ATI Mobility Radeon HD 5650, Intel Core i5 CPU
Windows 7 Professional SP1 (64bit)
Photoshop CS6 64bit (colour settings: Adobe RGB, grey gamma 2.2, dot gain 20%, Adobe Ace, Perceptual intent, black point comp, dither, compensate for scene-reffered profiles)
Spectraview Profiler 5 & Spyder 3 puck
I don't print my own work, I use professional printers.

The Problem
------------
I noticed dark tones and blacks looking slightly different in Photoshop and Bridge.
The Bridge preview was looking a little cooler/greener than the Photoshop image.
So I created a pure black image in PS and switched the PS background to black.
The image looked slightly red compared to the background.

The calibration settings used at this point were:
LCD hardware calibration
white point 5500
gamma 2.2
white luminance = 100
black luminance = min neutral
16-bit LUT
chromatic adaptation = CAT02
tick ICC v4


I re-calibrated with some different settings, thinking that it might be an issue with using icc v4 profiles that are not understood system-wide.
The calibration validation process kept failing, with the blacks or dark tones exceeding tolerance values.
In the end I changed the following settings:

chromatic adaptation = none
untick ICC v4

This seemed to balance the black in the PS image with the black in Bridge, and the black PS background (very very slight difference), although
the validation still failed with the RGB 32 32 32 greyscale range being slightly over the tolerance threshold (I can upload log image if that is useful).

So, am I making a mistake with settings? Could it be a hardware limitation or a software problem?
Any ideas from members experienced with calibratiing Windows systems would be appreciated as I am tearing my hair out with this.

Thanks.

Adrian
Logged
digitaldog
Sr. Member
****
Online Online

Posts: 8580



WWW
« Reply #1 on: July 24, 2013, 08:42:34 AM »
ReplyReply

Forget Bridge, you want to be evaluating at 100% using Photoshop (or LR in Develop). Forget V4 profiles, they do nothing useful here. And there's this too:

http://www.luminous-landscape.com/tutorials/why_are_my_prints_too_dark.shtml
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #2 on: July 24, 2013, 12:10:08 PM »
ReplyReply

I don't have Windows calibration experience, but here are some potential pitfalls that might be helpful:


1. ICC v4 redefines how blackpoints need to be interpreted. You don't want to open that can of worms, so leave it off.

2. While photoshop is properly colormanaged, the window that it is drawing in may not. This is particularly true when the profile creates different response curves for the individual channels, even though it also sets the hardware luts.

This used to be a problem on the mac, specifically for the photoshop background color. I don't know whether it still is a problem and whether windows even had that same problem to begin with, but it could also be a potential problem when bridge is drawing thumbnail previews in a window.

3. Where it says: blackpoint luminance = min neutral. What other options do you have? This is definitely a potential source of problems. Some video hardware may even dislike non-zero values for black. (Additionally check whether this is defined in the profile, or only in the video lut).

Logged

Regards,
Oscar Rysdyk
theimagingfactory
adrian_m
Newbie
*
Offline Offline

Posts: 6


Photographer working in South East UK


« Reply #3 on: July 24, 2013, 03:01:43 PM »
ReplyReply

Thanks for the responses folks. To answer your questions / points:

I do only use PS for image evaluation. However, noticing the difference in Bridge was disconcerting. Possibly it had always been there - it is only really evident with greyscale images as the colour shift is subtle.

ICC v4 - thanks for the comments on that, I shall not use going forward.

The other black point options are: min. native, or a specific value may be entered, or a measurement may be taken (presumably from a second monitor, if synchronising).
Min neutral, the option I have always used, sets the target to 0.00cd/m2 but the calibrations usually only achieve between 0.34cd/m2 and 0.37cd/m2.
How do I check whether this is defined in the profile or just the video LUT?

I've attached the 'log' from the most recent calibration verification effort too.




Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #4 on: July 24, 2013, 05:48:11 PM »
ReplyReply

Not sure whether this is in any way useful, it being rather dense and hardly visual, but the problem goes more or less like this:

A blackpoint is never 0.00cd/m2, in fact, the whole mathematics behind gamma compensation would fall apart if that were true. So it is correct to have some small value there, especially for LCD.

Now, in order to match your desired whitepoint, the response of the individual channels can be adjusted to match the selected target white (5500K). These adjustments are usually applied in the videolut which influences white and all the shades of gray.

If the native blackpoint is not equal to your selected whitepoint of 5500K it will obviously not match the lighter shades of gray, so we have to make a choice:

1. find the lowest shade of gray that still matches 5500K, which requires that at least one of R, G, or B is not zero, perhaps significantly so,

2. allow R=G=B=0 for deepest black, but its color appearance will not match the lighter shades of gray.

In the latter case we have again 2 options:

2.1. Match the entire grayscale to 5500K, and have only the darkest color deviate at R=G=B=0 (native),

2.2. Gradually match from white = 5500K, to black = native.

I am unfortunately not familiar with the software so I do not know what it uses if you make specific choices. But I thought that perhaps this can help you pin down the specifics. If I have to guess:
"min native" should mean "allow R=G=B=0" for black.
"min neutral" would mean that at least one of R, G, or B is not zero.

I once implemented something similar to "min neutral" on a mac in the videolut, and the video card decided to go crazy on me by for example inverting the cursor at irregular times. So, it could be that min neutral is implemented in the ICC profile, but not in the video luts.

That is okay if you use photoshop which would properly honor the profile curves, but outside photoshop this may not be the case.

tl;dr 1 ---> therefore, use "min native" if you don't see a significant deviation in blacks.

tl;dr 2 ---> therefore, if you use "min neutral" you may see slight differences between an image in photoshop and other software (including the operating system which is usually responsible for drawing the userinterface elements like the window background).


PS. if you can make the profile available for download, I am sure some of us can easily check for you what the TRC curves look like.
Logged

Regards,
Oscar Rysdyk
theimagingfactory
MarkM
Sr. Member
****
Offline Offline

Posts: 257



WWW
« Reply #5 on: July 24, 2013, 10:55:47 PM »
ReplyReply

A blackpoint is never 0.00cd/m2, in fact, the whole mathematics behind gamma compensation would fall apart if that were true.

A little off-topic—I apologize—but maybe you could explain this. Gamma compensation is simply raising values with an exponent. Zero works just fine here. In fact one of the nice things about gamma correction is that 0 maps to 0 and 1 maps to 1. What am I missing?
Logged

opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #6 on: July 25, 2013, 02:53:23 AM »
ReplyReply

A little off-topic—I apologize—but maybe you could explain this. Gamma compensation is simply raising values with an exponent. Zero works just fine here. In fact one of the nice things about gamma correction is that 0 maps to 0 and 1 maps to 1. What am I missing?

Yes, a power function can handle "real" values just fine.

However, gamma compensation is based on the idea that each successive step in a discrete distribution is some (fixed) ratio of neighbouring values. You obviously can't compute that ratio if the denominator is zero in the same way that you can't compute contrast ratio if the minimum value is zero.

Therefore we wouldn't be able to define the first step if the actual luminance of the first value was truly zero.

i.e. you want to answer the question what minimum luminance would be "barely" visible compared to an empty screen? (For the original CRT gamma response problem: what voltage would produce that barely visible luminance? Obviously, any value/voltage is infinitely larger than zero which is why the gamma function has infinite slope at zero.)

Logged

Regards,
Oscar Rysdyk
theimagingfactory
adrian_m
Newbie
*
Offline Offline

Posts: 6


Photographer working in South East UK


« Reply #7 on: July 26, 2013, 04:05:09 AM »
ReplyReply

Many thanks for that detailed response opgr. We are operating at the boundaries of my understanding of colour management here, but I believe I understand the main thrust of your post.
I will try the min native black lum & see how that works out.
Would I be correct in thinking that a different white point might also help matters? 5500 has always been my preferred WP, but if another setting made for a better calibration...

After removing the icc v4 and the CAT02, the difference between Photoshop image & Bridge preview / PS background is really minimal.
However, and this has been a problem for a while, there is a single green-ish band in the dark tones on a solid greyscale gradient (which otherwise looks fine) displayed in PS - even when calibration verification is 'successful'. Maybe linked to your black point observations.

Be great to stop these failed verifications as a job that should take half an hour ends up taking a couple of hours each time as I have to run the process several times to get a 'pass'.

As requested, I have attached my current profile. It failed verification (as per log above).
Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #8 on: July 26, 2013, 05:47:31 AM »
ReplyReply

Oh horror.

There should be an option somewhere to select "matrix" based profile instead of a "lut" based or "table" based profile. Make sure that you first try the matrix profile type option.

All of your problems, including the green band, the R=G=B=32 spike, and the difference between apps are a direct result of the profile.

A table based display profile, while theoretically superior, is a bit of an oddity. While Photoshop honors table based display profiles, most other apps including the OS usually don't support tables in display profiles and simply resort to the matrix form. Since we don't know which app does what, it is probably not a very useful option since it doesn't particularly increase consistency as you have found out.
Additionally, the tables in this profile look horrible and account for your green bump and R=G=B=32 outlyer. Now, this may be a direct result of the options selected for whitepoint and blackpoint, which could deviate too much. A 5500K whitepoint may be somewhat too warm for a modern LCD to emulate. However, first try the matrix profile option and leave the 5500K as your preference since you seem to be happy with the rendition of a grayscale apart from the band.

For informative purposes some additional info on whitepoint:
The wisdom used to be to select "native" whitepoint for LCD which basically means the color of the backlight becomes the reference. The advantage is that the use of the videolut is optimized and the LCD does not have to emulate color. It worked well because the backlight in the LCD was like the daylight tubes that are also used for viewing prints. If you get a reasonable match between the two, then your eyes will adapt quickly enough to translate between the two. etc…

Of course, these days LCDs have spiky LED backlights which use fluorescence for producing white. The led itself produces a blue light which excites the phosphor that produces white. Hence you get a relatively cold white, with a somewhat spiked spectrum.

Now, the objective is to synchronize the whitepoint of your monitor with the viewing conditions for your output. (In such a way that you can make the translation).

For example:
if you have a print viewingbooth with daylight, you might want to stick to the 5500K whitepoint if it matches well.

if you make a lot of prints for office space with halogen lights, you might want to reconsider your viewingbooth characteristics, and then also the whitepoint of your monitor.

etc etc.




Logged

Regards,
Oscar Rysdyk
theimagingfactory
adrian_m
Newbie
*
Offline Offline

Posts: 6


Photographer working in South East UK


« Reply #9 on: July 29, 2013, 11:23:58 AM »
ReplyReply

Hi opgr,

Thanks again for that detailed reply - very much appreciate indeed.

The reason I have previously selected 16 Bit LUT is this from the Spectraview Profiler user guide (spelling mistakes carried from pdf):

Matrix based
The siplest way to describe the color properties of a device is a color matrix. It contains the 3 primaries red, green and blue and a function that describes the tonal response curves
for the 3 channels.
The main advantage of this profile type is its small size (4-8 KBytes, depending from the way, yor hard drive has been formatted). matrix profiles are thus best suited for applications
where size matters, e.g. the internet.
The downside is that a not so perfect device cannot be described accurately.

16-bit LUT based
In this profile type, the gamut of a device will be described in a table of a defined number of points. All other color values will be interpolated. This allows to describe non-linear behavior of a device. The size of a LOUT profile can be somewhere between 200 KB and more than 2 MB.
With 16 bit encoding, the accuracy of a LUT profile will be increased dramatically (256 times more accurate), while the size only doubles. That’s why SpectraView Profiler offers 16 bit LUT profiles only.

i.e. the manual seems to strongly recommend the LUT. But then it also recommends an L* tonal response curve, which I do not employ...

What you say about the whitepoint may be an issue.
 I recently received this monitor as a replacement for my old one which had developed a fault. NEC were pretty good and did a straight swap. I noticed that when the calibration process starts, and the software resets the monitor prior to measuring, this screen goes very blue. So maybe I am pushing it too far with 5500.

One thing to note - most of my work is supplied to clients digitally, and I do not do my own printing. I am therefore not trying to match up to a specific viewing booth.

I will recalibrate tomorrow with the suggested matrix setting.
If that doesn't work, I'll try D65, one of the default whitepoints.
If that doesn't work, I'll try native blackpoint.

Regards,

Adrian
Logged
adrian_m
Newbie
*
Offline Offline

Posts: 6


Photographer working in South East UK


« Reply #10 on: July 30, 2013, 12:30:27 PM »
ReplyReply

Ok, progress of sorts....

I've tried a number of different calibration options today using the matrix profile option:

a: min neutral blackpoint, D5500 whitepoint - failed verification with a spike in 32 32 32

b: min native , D5500 - succeeded, but still cool band in dark end of greyscale gradient

c: calibration aborted

d: min native , D65 - succeeded, but still cool band in dark end of greyscale gradient

e: min neutral , D65 - succeeded, but still cool band in dark end of greyscale gradient

f: min neutral , blackbody 5500k - failed verification with a huge spike in 0 0 0

g: min neutral , D5500 - succeeded, but still cool band in dark end of greyscale gradient, plus a slight warm band in the lighter half

h: min native , D5500 - succeeded, but still cool band in dark end of greyscale gradient

The following options were kept consistent:

LCD hardware calibration, gamma 2.2, white luminance = 100, chromatic adaptation = none, not ICC v4

So, as things stand:

Positives:
Photoshop and Bridge now render the same as far as I can discern.
A pure black image merges perfectly with a black PS background.
Calibration verification succeeded.

Negatives:
The cool band in dark end of greyscale gradient.
Also bothering me: when switching to Monitor RGB proof view in PS (so the monitor's profile is not used), the gradient doesn't really change - it is still smooth with the single green band. I expected it to look awful without the profile - or at least different.

I have attached the latest profile... I'm hoping the profile looks better than it did before.

Apologies, I know that's a lot of info above - I'm desperate to nail this properly & feel confident about my display.
Logged
Stefan Ohlsson
Newbie
*
Offline Offline

Posts: 13


« Reply #11 on: July 31, 2013, 12:10:11 AM »
ReplyReply

Have you tried with another calibrator? In a recent test I did I found that about a third of the colorimeters I tested (about 20 colorimeters and 10 spectrophotometers) failed. Most of the calibrators that failed were older than three years.
Logged
opgr
Sr. Member
****
Offline Offline

Posts: 1125


WWW
« Reply #12 on: July 31, 2013, 01:19:14 AM »
ReplyReply

Negatives:
The cool band in dark end of greyscale gradient.
Also bothering me: when switching to Monitor RGB proof view in PS (so the monitor's profile is not used), the gradient doesn't really change - it is still smooth with the single green band. I expected it to look awful without the profile - or at least different.

Switching a monitor profile in general shouldn't change a thing in grayscale rendition except for gamma differences. Since most profiles have 2.2 gamma, you should never see a change in grayscale gradients.

Selecting a monitor profile for proofing purposes in PS is just selecting an intermediate step in colorconversion, your monitor profile will still be used to display the final result. It is however useful to see what happens if your carefully created wide-gamut image will be converted/displayed as sRGB. It is not useful to see what the effect is of different monitor profiles in a workflow, because videoluts aren't applied or changed when proofing monitor profiles.


The cool band is a point of concern:

Try white = native, and black = native

if that doesn't remove the band, then there is reason to contact the manufacturer, or perhaps Basiccolor.

if it does remove the band, then you could try to gradually lower the white target from native to where you start to get the band. Then see if that potentially high whitepoint is comfortable to look at.
(The native white point could be in the 9000K range.) 

Remember, for your situation, once your eyes/brain adapt to the new look, it will not affect your workflow since there is no outside reference. That is: there should be no outside reference. If your workenvironment includes windows where you can see clouds, or get diffuse outside light, then you might have a problem. But even then, only if part of your workflow involves "eyeballing" correction. That is: if you are requested to make an image look "warmer" and you don't do these sort of corrections by the numbers, then you obviously need some kind of universal and consistent reference. You can't do that consistently if you keep adapting your eyes to different whitepoints.

However, it is far more important to have some form of reference graybalance at capture time, than during viewing. For example, if you have a graycard in some of the captures of a capture session, it allows you to set the graybalance of all the images correctly, and hence you know that the images are correct. Your eyes will adapt while viewing the monitor, so that is not too much of a problem. Just make sure that it is consistent.

As a side note:

1. a consistent whitepoint to me means your whitepoint is the same each time you use your computer,

2. a reference whitepoint to me means your whitepoint is the same as some outside reference which include for example a viewing booth, or UK daylight.

In your case, a consistent whitepoint is more important than a reference whitepoint.




Logged

Regards,
Oscar Rysdyk
theimagingfactory
adrian_m
Newbie
*
Offline Offline

Posts: 6


Photographer working in South East UK


« Reply #13 on: July 31, 2013, 03:44:43 AM »
ReplyReply

Hi,
Thanks again for your input!

I shall try the native points together. Also, further research has suggested modern monitors may struggle with a white luminance value as low as 100; I selected this because my environment is dim, but not dark, and too extend the life of the monitor. So I may try lum = 120 as well.

I always use a greycard when shooting, and I do reference the numbers (LAB rather than RGB - more intuitive). So I can do colour corrections.

If no success I shall contact the manufacturer. It does seem strange that the verification should 'pass', but the cool band be present. It has been there on 2 monitors (both Spectraview 241s), so is not a fault with a specific unit.

Regards,

Adrian
Logged
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad