Colorimeter spreadsheet for SpyderTV

Tried on 3 different computers - no luck downloading the file.

It downloaded and opened, OK on my PC.
Using MS XP Pro and Mozilla.

I would not use this spreadsheet in its current form for the following reasons.

The spreadsheets RGB calculations are not very useful for calibration to D65. The RGB error displayed in this spreadsheet decreases to zero at 100 IRE no matter what the color measurement is. What you want to see is the color error in relative RGB levels at each IRE level to D65 not to the measured 100 IRE value as found in the spreadsheet.

This spreadsheet references only color temperature not D65.

Assuming zero light level at zero signal input to normalize the gamma curve is flawed.

The input signal level is assuming Avia levels for test patterns. This will be incorrect for many other test sources.

I also have other issues with these calculations, but they are less serious.

Gary's spreadsheet was one to which I referred in the STV review. As Jeff says, it has several errors, some of greater and lesser significance.


Later,
Bill

D65 is a specific color while correlated color temperature is a line in xy colorspace. You must calibrate to D65 for gray scale to achieve the desired results. This link shows these lines. http://www.ecse.rpi.edu/~schubert/Li...%20diagram.jpg


Your display must have very low black levels. This is also one of the problems with using Avia. The difference between the 7.5 and 10 IRE patterns is not very large. However, all displays will not have levels so low that the SpyderTV cannot measure them. You are in a bind though if it can't.

The problem with the calculation method in this spreadsheet will show up with displays that have higher black levels. Your normalized Y values (P55-P64) are wrong because of this. You are dividing by the maximum value, but you should also subtract the minimum value because your ideal curve is always starting at zero. All displays are not capable of zero black levels and many are fairly high. The error in your calculated gamma will increase with black level.

Here are examples based on simply modifying the spreadsheet to include a variable black level. This is simulated by adding a constant to the measured Y values. The calculated gamma should not change because of a shift in light level up in your form of gamma calculation. Black levels up to 0.4 cd/m2 are not uncommon. Here are the results.

Black Level (cd/m2) Calculated Gamma

0.00 2.33
0.05 2.11
0.10 2.01
0.15 1.95
0.20 1.90
0.30 1.82
0.40 1.77

There are other problems with the gamma calculation because it is not the Rec. 709 equation. The real gamma compensation being used in the majority of displays today is including the linear section found in the specification for gamma compensation. Here is a link to more information on this. Without this linear segment you will be calculating that the gamma compensation is much lower than it really is.

http://people.ee.ethz.ch/~buc/brechb...Q.html#RTFToC5

Bill, I noted in that thread that you have your own spreadsheet that (from the sound of it) is quite comprehensive. If that's true is it available for use?

Tony - it will be soon. However, I'll have a dedicated thread for it when it's released.

thanks rader,

for updating it. I was going to calibrate my pj tonight.

I read somewhere that the AVIA IRE greyscale is a little bit off and DVE is better. I have both. should I use DVE instead, even thought the patterns is only in 20IRE steps.

I am also curious as to why you changed to using the windowed pattern.

Thanks again for the spreadsheet

Thanks for the vote rader. It just turns my stomach to see people using calculations that are so flawed. Your calculations look reasonable now.

I would make one more change so Avia or non-Avia gray windows can be used with the spreadsheet. To do this you need to add a variable for 0 or 7.5 IRE black level patterns and calculate the normalized input level with either 100 or 92.5 as the range and 0 or 7.5 as the offset. This is not dependent on what the player outputs, but the percentage level of the pattern itself.

I would also consider adding a color tolerancing function. Delta E 1976 is the most commonly used. You would need to add Lab color calculations to do this. The following is a link to this calculation. To use it so it is color only you need to reference each measurement to the actual measured L value. This will result in L being 100 for each value.

http://www.brucelindbloom.com/index....taE_CIE76.html

No problem, you were just following what others had done before. I am just glad to see that you have something very useful now.

You will need to add more complexity to your calculated curves and gamma if you make the change to include Avia and non-Avia test patterns.

I have posted some correction factors for Avia on my web site. These measurements are based on the gray scale step pattern and show the significance of the errors at low input levels.

http://accucal.org/documents/Avia_Co...on_Factors.doc

will this also work with the Spyder2Pro, I have one on order so I can cal all my screens including 2 dlp projectors

thanks for putting this together

another suggestion is to also support the calibration disk by GetGray being developed http://www.avsforum.com/avs-vb/showt...9&page=1&pp=30

These bundled would make a great low cost cal system

That is what I am suggesting with altering the black level point.

Great is overstating the accuracy of the Spyder. I would rate it as the beginning of a good low cost calibration system. You need to add an optical comparator to your equipment list to touchup the color errors that the Spyder will likely induce in your display to have a system that is adequate.

Jeff - Here you rag on your PVM-96:
http://www.avsforum.com/avs-vb/showthread.php?p=6373309

...which lists for $900. Given how frequently you post telling people to add an optical comparator to the Spyder, this seems somewhat inconsistent. If one is too cynical, it might even be somewhat disingenuous given your software interest.

Later,
Bill

It is not inconsistent at all.

My PVM-96 or any other comparator I have used are not perfect but it is frequently much more accurate than my filter based colorimeters. I recommend a spectroradiometer if you want the most accurate color measurements. What I am saying is if you are targeting something that has a reasonable chance of being good then you should try and use a comparator to touch up the colors.

People here seem to want advice on how to come relatively close at minimal expense. I am offering what I believe would be the best approach for most people with the Spyder if they want to do a decent job. There are other simpler techniques that I and others on this forum have described before that can work as well at approximating D65 that are much less expensive, but those will not work on all displays.

Personally, I just want people to have decent colors and using a filter based instrument alone is not likely to give it. I can't do anything about your cynisim. Most people are buying software other than mine and I only started selling it because you and Ken asked me to. I never intended to offer it for sale to the general public. I could discontinue selling it if that would make you feel better.

Bill,

The following quote captures more of my thoughts on the issue of inexpensive gray scale calibration tools.

http://www.avsforum.com/avs-vb/showt...&&#post6007324

Your Delta E calculations look correct to me. You should round them to whole numbers. Additional precision is not realistic. The graph should not have negative values for Delta E since it can only be positive.

I don't know the answer to many questions, but I have pondered many of the ones you have asked.

Jeff - This does represent more thought-out advice, especially with the acknowledgment that the filter-based tools have differential accuracy between CRTs and lamp-based displays. However, what I was responding to are the far more common posts from you that do not offer such a nuanced opinion. I respect that you have issues with the filter-based colorimeters, but to state this opinion as an authority as frequently as you do without the background information seems counter to your stated purpose (e.g., several CRT owners thinking that the Spyder is not good enough for them).

What I have never seen you respond to is how to have confidence in the EyeOne at high and low light levels. Per your statements, it can get x and y right compared to a Lightspex. In my own limited testing, Y has been suspect at best at the high and low ends (e.g., readings of 1.2 cd/m^2 in one run, then 0.2 in another -- which matches what my Spyder returns consistently). As you know, x and y are merely transforms of XYZ, so getting x and y correct without getting X and Y right is a violation of basic algebra - unless they are off in exactly the right proportions to one another to make the aggregate ratios perfect (unlikely).

You had chided me previously for using averages in my own testing protocols, but that is in fact what you should do in cases like this. You should also look at standard deviations and standard errors across multiple testing runs. To do otherwise is incorrect. Please note: yes, I believe that the EyeOne is a better meter (it costs twice as much before taking into account your software, so it better be!). However, I have yet to see a good comparison of the Spyder against reference grade equipment. So, could you either a) be more complete in posting your advice, or b) participate in a valid colorimeter shoot-out? The point being not to prove that one is good and the other must stink, but to show where the relative strengths and weaknesses are.

As for your software, if you read my posts (search using the term "Jeff's software" with my userid), you will see that I have been very positive in recommending your software. I think it is a good thing for the market.

Later,
Bill

Bill,

I have seen and appreciate your support.

I am completely into statistics as well. I am not saying taking averages or calculating population statistics is worthless. I was quibbling with a statement that implied using an average value guaranteed accuracy. All it is likely to do is increase repeatability. My software actually uses some statistics and averages behind the scenes in a more complex manner than simply averaging xyY values.

The Y reported by the Eye-One is different than the XYZ coordinate system. The xy and Y values are calculated from the SPD of the radiometric measurements. I have not seen the variance you are describing seeing at 1 cd/m2 on the Eye-One.

I have done my own shoot out with my SpyderTV versus my Eye-One. You are welcome to come to my house and see how it performs on the variety of displays that I have (CRT direct view, LCD direct view. LCD RPTV, LCoS RPTV). I am not saying the SpyderTV stinks completely. I am just trying to add some balance to peoples excitement about these instruments.

Here is my take on the SpyderTV based on my own experiments...

It is a useful light meter, but no more sensitive than the Eye-One. The diffuser mounted on it makes it unreliable for luminance measurements. This means it cannot measure screen gain and is less useful for calibrating an RPTV's light levels. It is not calibrated for illumination, but since it has a good diffuser it should take measurements that are relatively linear, but the units will be in error. Therefore, it cannot be used to calculate a projectors lumens.

As a colorimeter it is a mixed bag. On some of my displays it is very accurate on others it is over 0.01 in x or y off. It seems to do well on LCD direct view and CRT direct view sets. It seems to be less reliable on other technologies. As long as its use is limited to those display types my experience with it is good. Other display types are more problematic. Filter based devices also tend to be hygroscopic causing the calibration to drift after several months.

I also find the SpyderTV to be slow and cumbersome to use.

I just think people should understand the caveats of what they are purchasing. The reason I was posting on this thread was to help correct some of the errors people were likely to make using the spreadsheet. I was only suggesting using an optical comparator because that is what I would do if I was attempting to use the SpyderTV to calibrate my display. I can easily live with it if you or anyone else disagrees.

Bill,

Here is a post from glenned on the Eye-One that you may find useful.

http://www.avsforum.com/avs-vb/showt...&&#post6243600

Bill,

Here is a post where I posted some low light repeatability results with my Eye-One.

http://www.avsforum.com/avs-vb/showt...&&#post5997199

I would be willing to send my Milori Spyder2 to just such a shootout - as the only access I have to reference spectrometer is what Milori told me when they sent it - 0.001xy accuracy - which is 3x the marketed spec and 40x the ISF calibration requirements. This self promoting interest in this thread really has no place.

umr, this is a thread about the SpyderTV spreadsheets - exactly why do you need to bring up the EyeOne at all? Stick to the advance on colormath without disparaging peoples choice in equipment.

Sorry. I was just trying to be helpful.

I only mentioned it as a quote in a previous post when Bill questioned my advice on using a comparator. I would never have mentioned it otherwise.

I think I will leave this discussion. A simple defense of my position seems to have generated much angst. Please continue without me and my biased perspectives.

Rader - Your XYZ->RGB matrix is transposed (compare your xy values against the CIE Gamut vs. what your RGB graph is telling you - e.g., 0.372 in y is noticeably green).

Ran though my first set of measurements using the Spyder in support mode and your spreadsheet. Thanks again for posting it and keeping it updated.

It was very cool to plug in the data see all these things about ones display. I can't speak to the veracity of the math and the approach but given the feedback from many of the forums experts it seems like you did a yeomen's job.

I have a Sony 34HS420 and a BK Instruments HD Pattern Generator. What's nice about the BK is I can easily change display modes and am not tethered to my DVD at 480i. It can display 0 to 100 IRE at 2.5 increments which is certainly overkill. I did have some issues with the Spyder at 0 IRE but from what I have gleaned it may be better at 7.5. I hesitated pumping up the brightness for the sake of an accurate reading from the Spyder because I have the basic parameters set to what pleases my eye. I did not like the Spyder Wizards suggestions at all.

Most of the information is rather self explanatory but this deltaE thing is still a bit of a mystery, my green channel is really high relative to R and B. Guess I am gonna have to bone up on this. I will be taking a lot more measurements over the weekend. However, collecting data is one thing but interpretation and implementation, well that's a different kettle of fish.