Best DIY method to compensate for meter drift

The problem you run into which you pointed out and why this has not been solved easily is the reference. Validating with a reference is straight forward. So the answer is you need is a reference you can trust over time to not drift. Well the short answer is they all do drift to some degree over time. In our NIST calibration lab we do use reference displays and they are checked and adjusted with our CS-2000 before each meter calibration run. Even the high-end reference displays we use drift and take time to warm up usually 30 minutes even for LCD CCFL before you can get a stable reading. You also have many environmental factors, ambient light conditions, room temperature and humidity, mechanical alignment with the meter and display to minimize angler issues. We have over $50k in jigs just to deal with mechanical alignment.

Now this is not to say it is impossible to do just not very cost effective at this time. When I first start with video calibration many years ago this was one of the first projects I worked on but never came up with an affordable solution I’m still looking for the answer.

What are you going to light the tile with?

Nothing has color until light is bounced off of it.
A stable light source in a controlled enviroment would be required to even begin the process.

The other thing would be that a simple white measurement isn't good enough to create a correction table. While that would give you an x,y offset, the Colorimeter is measuring colors in X,Y,Z space where each value is measuring an overlapping area of the spectrum, so you need measure segments of the spectrum that roughly correlate to the X, Y and Z areas which is why when you profile a meter in CalMAN we do an RGB and W reading.

You need a way better light source than the sun.

A neutral Gray card, RGB cards and a jig to align them and you'd be close.

The other issue is of course the reason we have individual profiles is that displays can have very different spectral signatures.

A colorimeter has sensors that are suppose to match the XYZ curves, X roughly aproximates red, Y green and Z blue, but it's much more nuanced than that.

Both of these monitors are D65

Dell u2410


Dell 2408wfp



What you are doing with creating a correction table is learning where the filters don't match the curves in those graphs. The corrections therefore are specific to the spectral output of the display.

Even if you created profiles off those cards, the profile would really only be good for those cards, since the corrections wouldn't correctly hint the software for the changes in the sensitivity of the filter for the corresponding XYZ curves.

Don't get me wrong, it's probably better than nothing, but you're talking about buying 4x $70 card + a $70 lamp, all of which are going to drift and fade, for $300+ dollars.

After a year your D3 won't have drifted far, after 2 it might be a good idea to re-run the thing, and you might get a decent correction. After another 2, the lamp and the cards will likely have drifted faded as much that you won't be able to get a reliable 2nd correction.

So you're talking about spending $300 to get an incremental improvement in 2-3 years from now, when the cost of a new meter now is $300 and likely to be less 4yrs from now.

Right, as Sotti said you would have an offset for how it has drifted when using the sun as a light source, but not when using an LED, CCFL, plasma, CRT, etc...

You don't have to buy a new meter. If you have ChromaPure, you can recalibrate/convert the D3 into the PRO version for $150 (same fee for both services). This service gives you enhanced accuracy and allows you to maintain that accuracy by redoing the service every 2 years or so. I believe SpectraCal might offer a similar service for their C6, but if cost is an issue the D3 PRO is about $300 cheaper.

We do offer NIST re-certs on our C6 and just about everything else we sell. The re-certs are not just recalibrate and check but a proper NIST cert. SpectraCal has one of the few labs for providing NIST cert's for every popular tristim made.

The C6 costs more because we do more and it can be field upgraded. So when we add more display types you don't have to send the meter back in for the update, CalMAN will update it for you. We just added LED RGBY, so for those that already have a C6 it is now calibrated for the Sharp Elite and other quattron's. You only need to send the C6 back in for NIST re-cert's.

Could you go into more detail why/how you can add more display types without sending the meter back in? If you have a new method of calibrating meters that is truly superior to the traditional method why keep this info secret? More people would buy the C6 if they fully understood what differentiates it from other calibration solutions like the D3 PRO.

Great question and one that I wish I could go in to extreme detail on how it is done because X-Rite has truly came up with a very advanced way of calibrating the D3 family. We are under NDA with X-Rite for much of the details but I will try and cover as much as I can without violating that.

We started working with X-Rite just about two years ago on what is now the D3 family. We went to the design team at X-Rite and had meetings with their color scientist team including some of the guys from the original Sequel Imaging group. We had a long laundry list of requests we felt the industry was going to need in the near future as display technologies were changing so fast. Some of those requests were: The ability to read at very low light with repeatable results because displays and projectors are getting so dark. The meter needed to have a tripod mount embedded in its body because displays are getting much bigger and hanging the meter cord over the top of the display is not a good option anymore. Accessories or components that can’t be lost or misplaced like with the D2 and C5 so a attached diffuser with detector to tell us if it is properly in place. A design that dealt with highly polarized and angular light that LCD manufacturers seem to be doing to focus more light into a narrow vertical but wider horizontal viewing area. Sealed optics and sensors to minimize drift over time. The ability to add display characterization data to the meter in the field to help make the meter future proof over a longer period of time.

What are the two things that make a meter obsolete? Sensor drift over time and display technology changes. Both of these problems are interrelated. You need a very stable and accurate filter array to solve the drift over time issue and once that was solved you can take advantage of it. With a stable and accurate filter/sensor array you can characterize each meter at the factory with a high degree of accuracy.

The second issue that can make a meter obsolete is changing display technology specifically the spectral response curve. Here is an example of a spectrum sweep with the 1931 2 deg standard observer curve on it.



A tristimulus meter through its filters and sensors try’s to mimic the standard observer curve in this case the 1931 2 deg CMF. Which if all display spectral response curves match the 1931 2 deg observer and the filters in the meter did the same we would not need to calibrate or characterize the meter. But in real life the meters filters are never perfect and have deviations. I have not seen a display that perfectly matches the 1931 2 deg standard observer. So in the past we had to create a calibration table for each major type of spectral response curve to cover our bases. As part of that processes of creating additional tables for each spectral response curve it also took into account the deviation in the meters filters. This is how it’s been done for a very long time and is the recommend method by NIST for calibrating a tristimulus device. But now that X-Rite has eliminated the need to deal with the filter/sensor deviation we can solely concentrate on just the display spectral response not matching the standard observer curve. So with a highly stable and factory characterized meter and the ability to add display characterization data outside the factory you have a field upgradable tristimulus meter.

As part of the work with X-Rite we worked out a deal to be able to provide these field upgrades with the SpectraCal C6. Our inhouse lab that does all of this work is quite extensive starting with our Konica Minolta CS-2000 which is only one of two devices certified by X-Rite for this work the other being a CS-1000. The lab is temperature and humidity controlled. Each display device has went through a very rigorous process of testing before being used in the lab and is calibrated before each meter or research run. The mechanical jigs used to line everything up for calibration or testing is controlling tolerance down to or even some times under 1mm.

I hope this helps everyone out and I wish I could go into a lot more detail because it is really very cool color science X-Rite really did their homework.

Derek,

I'm surprised you didn't mention requirements for new meters including being able to take readings through 3D glasses and from light reflected from a projection screen... most puck-type meters can't do readings through 3D glasses or of light reflected from projection screens. A DIY person looking for a meter today might be more likely to go for a meter that could measure light from a projection screen if they have an idea they will be getting a projector in 3 or 5 years or whatever... they'd know they could use the same meter. This would be especially attractive if the meter could be kept in calibration over many years.

And of course, taking readings through 3D glasses makes the requirement of measuring very low light levels even MORE important/critical.

Again good points:

The D3 family has the optics, filters and sensors to be able to read from a direct light source or reflected as in front projection off of a screen. The requirements for this are a narrow field of view normally less than 15 degrees. As you point out most puck style meters don’t have any form of optics and as a result have a very wide FOV they see light from all angles not what you want when reading off of a screen. The C6 has a FOV of around 10 degrees so more than suited for front projection screen work. Heck it even has a tripod mount built in. You would be surprised at how much push back we got when we said the meter had to have a tripod mount.

As for calibrating through the glasses for 3D the only meter we currently recommend is the i1Pro or other spectro device. As I pointed out in my previous post the accuracy for a tristim is all about its ability to mimic the standard observer curve with filters and based on known display spectral response curves. Well when you shoot through a pair of 3D glasses it does change the spectral response curve so that becomes an unknown. With a spectro they mimic the standard observer response curve using a high sample rate 40 to 1000 samples and math. As a result they are very accurate but are also much slower and don’t read down to the low light most displays are capable of today.

Is there a perfect meter no. All have their tradeoffs. Tristim’s are fast, read low light but the accuracy subject to the calibration tables they are based on. Spectro’s are very accurate but do poorly at lower light and are slow. Our CS-2000 at very low light can take up to 4 minutes for a single reading. That’s ok for doing research and lab work but not very practical if you used it to calibrate with.

So what do we do to get the best of both worlds you profile a target meter to a reference meter when needed. This involves setting both a spectro as the reference and a tristim as the target up to take readings from the exact same spot and field of view on the display. CalMAN will show a series of patterns White, Red, Green and Blue for each meter and then run those 8 readings through some math to come up with a corrective matrix. Now you have a tristim that is as accurate as the spectro for the display you just profiled with.

If you wish to only use one meter you can buy a stock D3 and replace it when it drifts too much or have it converted into the D3 PRO (assuming you have ChromaPure) OR Get the C6 and have it re-certified/recalibrated when the time comes. You can also get the D3 PRO now and have it recalibrated when the time comes. Those are the only single meter options you really have for the D3.

Not the only options you also have the X-Rite i1Display Retail which CalMAN v4.4 also supports. So the best option would be to find a X-Rite i1Display Retail locally for the best price. That way you don't have to pay for any additional shipping or duties. The filter design in the D3 family should be good for a couple of years if taken care of. After a couple of years sell it off and buy a new one.

Another option is you can rent a ColorMunki Spectro from us once a year but I think the shipping cost both ways to Australia would be prohibitive. This option works out really well when shipping costs or duties are not a factor.

I have a problem with this statement....
How and when do you know that the meter has drifted too much?
Only way to know, is to have it check against a reference meter.
I rememeber way back when, I first got into calibartion I had an i1D2 that I thought was correct cause it was brand new....sad truth was...it was reading wrong but I never knew that till I compaired it to a CS-200.

Just a word to the wise......ALL METERS ARE CORRECT in your mind..till they are compaired to a reference meter....then you see the errors.

Just my $.02
RayJr

well if you know the typical lifespan of the meter to be 2-3 years (for example) and you store it properly, you should be able to make a good decision as to when to get it re-calibrated/re-certified or replaced

having a reference meter to check the D3 against would be more precise but if that's not an option you have to do the best you can to determine when the meter is no longer returning reliable readings

You are missing the meaning behind my post.....The meaning is...even a NEW meter can be wrong..but you don't know it unless it is checked against a reference meter.

RayJr

Yes, that is the case with the stock D3, retail or OEM. However, the D3 PRO and the C6 are both checked against reference meters. This is exactly why I decided to upgrade my OEM D3 into the PRO version.

Assuming someone bought a meter primarily to calibrate one display, here was my question on the subject http://www.avsforum.com/avs-vb/showt...1#post19191171 Really the difficult thing about this line of thinking is that there's not much hard data to use as a reference for forming an opinion related to expected drift of different devices over time. I mean that I personally am not aware of examples of people explicitly documenting drift of colorimeters and spectros over time, in order to get an idea of worst-case scenarios. Personally my few year old colorimeter didn't vary gray by any more against my new i1pro than it had varied against an identical model, but I cannot say how Colorado climate might affect a colorimeter drift differently than other regions. My personal opinion is that I seem to find differences in relative light output more noticeable than variation in absolute color. Since gamma is essentially not standardized, and no one really seems to care about practical gamma differences, for my own use I'm just not aware of any clearly-documented information to worry about spectro or camera drift over time.

Based on your above statement, can I profile the i1D3 through 3D glasses against i1Pro through 3D glasses and get similar result compare to my i1Pro?

My i1Pro doesn't drift much after 3 years as I compare to my friend's i1Pro that had been recertified recently.

That is not something we have tried but I will ask our lab to check it out for us and see.

I have done a little experiment as to see if an alternate source could be used instead of a white tile. My alternate source was a piece of HP Photo paper that one would print an image on utilizing a printer. In every incident there was no problem. When testing the results where by a video display image was used as a standard, the error factor was <0.002. (please no replies on what complies as a standard unless you want to get into a long and boring discussion of the Heisenberg effect).

I have two i1Pro spectrophotometers which were used in this test and I also utilized a Xrite Color i7 bench-top spectrophotometer as well as a HP Photo Paper. All white tiles where measured and recorded as well as the photo paper and here are the results;

i7 Color calibration tile:
X: 83.30 L* 95.33
Y: 88.40 a* -1.00
Z: 94.45 b* 0.27


i1Pro (#1)
X: 83.38 L* 95.35
Y: 88.45 a* -0.93
Z: 94.58 b* 0.22


i1Pro (#2)
X: 83.35 L* 95.32
Y: 88.41 a* -0.97
Z: 94.41 b* 0.24


HP Photo Paper (chosen for its reflective properties)
X: 83.40 L* 95.43
Y: 88.64 a* -1.23
Z: 95.54 b* -0.29

For the record upon conducting a color analysis of the tile sets and the photo paper, it was noted that the photo paper contained a small portion of the color black where upon the white tiles did not.

One can easily see the values of all the samples were very close to each other. In a pinch it would be possible to utilize the photo paper for an emergency calibration. I am not suggesting to eliminate the white tiles for photo paper but one might consider testing the white tile for its ability to comply to standards. Since these tiles appear to be a form of plastic, it is not uncommon for air pollutants and age to discolor them. In the aforementioning that the photo paper is a very close in values to a white tiles within the test set, recording the value of the white tile that one might own and over time comparing them periodically to a piece of photo paper one could evaluate color drift of the tile itself. One can also record the values of the white tile in question before sending out the equipment for re-calibration to see if the tile had been replaced and if so what is the the value of the new tile. I have read many of the threads concerning the topic of color drift and noted that nobody seems concerned with the standard upon which the spectrometer resets itself to.



I am always interested in hearing the wisdom of others.

The issue with your research is the hardware you have access to.

Those meters, use tungsten lamps to produce a constant predictable, known spectrum.

Because they know with relative certain the spectrum of light that is being reflected, they can accurately measure the paper.

The original post was about measuring the paper or tile with a colorimeter and an unknown light source. It's the light source that makes all the difference.

The problem with DIY testing is what standards, the methodology and are used and the issue of consistency. First are you using spot checking as a procedure (determining the color of a tile or paper), which requires an illumination of the sample (either from an internal or external source)? What is the final usage (Video or Print)?

If video is the final usage, then the testing should be geared to that format. Since a video display is self illuminating the tungsten light source is not an issue and is only part of the equation when testing articles such as printed matter. Unfortunately, evaluating a piece of equipment over time introduces a number of issues. The best is to have a set of standards which would maintain a constant over time. Utilizing a television as a standard is a very poor choice. One might suggest that a small light box be constructed that has an internal light source of 6500k (D65). In a windowed area on the box have a frosted piece of glass (to diffuse the light), where by colored standards can be placed upon to evaluate a devices accuracy. I would purpose that one could use colored glass filters (which the x,y,Y values would be known), but this would be an expensive approach. An alternative would be to use colored gel filters laminated and sealed between two sheets of glass. Not only would you know the x,y,Y values but the cost would be significantly less.


As we all know spectrometers and colorimeters are somewhat adjusted via the USB port and unless you have access to the software your attempts are in vain. Theoretically, one could adjust the color matrix function that are found in packages such as HCFR and Calman v4. Measure the equipment against a perfect source and extrapolate the difference and adjust the matrix to compensate. Then again, can perfect values be placed into the aforementioned software packages and equipment profiled against those values? In doing so one would remove the question of how accurate is the unit a profile is being base on.

No matter what any laboratory claims, no one can ever claim consistency even if the tests were repeated one after the other. Don't we just hate that Heisenberg effect?

If anyone attempts this hypothesis, please let me know or at least post it, I would be very interested in the out come.