AVS Forum banner

Four Color Matrix Meter Profling Thread

46K views 327 replies 34 participants last post by  zielin 
#1 ·
Four Color Matrix Meter Correction Table Method

The best choice to achieve the most color accurate results by measuring any display/projector technology will require to have both a spectrophotometer/spectroradiometer and a colorimeter.

Meter profiling is a process for improving the accuracy of a filter-based colorimeter, for a specific display. Profiling creates a calibration profile for the colorimeter, for the display currently being measured, based on display measurement data from a reference spectrophotometer/spectroradiometer.

When you profile a colorimeter, you measure the unique spectral characteristics of a display with a spectrophotometer/spectroradiometer and basically you transfer that accuracy to your colorimeter, for that display.

The process is known as the Four-Color Matrix Method (FCMM) for Correction of Tristimulus Colorimeter developed by Ohno and Hardis at NIST (National Institute of Standards and Technology).

The accuracy of this method has been studied further from NIST for spectral variations of displays including CRTs, LCDs, and OLED displays and it works well.

The procedure starts using a reference spectrophotometer/spectroradiometer which measures 4 color patches, the primary colors (red, green, blue) and white of a certain display technology and then using a colorimeter it will measure these 4 color patches (W,R,G,B) of the same display also.

From these 8 in total measurements (4 per each meter), a 3x3 XYZ matrix correction table will be created and it will transform the colorimeter results as close as possible into the results obtained by the spectrophotometer/spectroradiometer used. The correction matrix is not a simple x,y subtraction/offset, which would result in unpredictable and invalid results.

That generated matrix solution can then be used to transform any other color measured by the colorimeter on that specific display into accurate and corrected chromaticity coordinates and luminance values for any color within the gamut of the display primaries.

A matrix developed for one type of display will not work on another display of a different technology because the spectra are different, so the error correction will not be effective.

How often a new meter profiling is required?

Because the filter-photodiode combinations for each colorimeter may be a little different than another of the same brand and model, a set of matrices needs to be developed for each colorimeter and for each display technology. For that reason, sharing correction matrices between the same colorimeter users will not work.

It's recommended to create a new meter profiling on every new display you need to calibrate. Even if it's the same display model that you have calibrated before, manufacturers occasionally change the characteristics of a display screen or backlight in the middle of a manufacturing run, without changing the model number.

Color performance with aging

It is less widely known that the color performance of LED is actually more susceptible to temperature changes, driving current and aging than traditional displays.

While the CCFL and UHP, as well as the CRT phosphors mainly decrease in brightness and change chromaticity only slightly (UHP, CCFL) or not at all (CRT), with LED the colors both decrease in brightness and considerably change the chromaticity.

The CRT contains red, green and blue phosphors that emit light when a guided electron beam hits them. These phosphors do not change color with time, but do decrease in brightness along usage, which results in both a darker display.

Because different phosphors can decrease brightness with a different pace, the white balance will also be offset with time. So, the main challenge in CRTs was to maintain the color temperature (white balance).

Lamps (CCFL or UHP) with aging or increasing temperature, the different lamp spectral peaks decrease in intensity. Their peak wavelength and peak width also shift; however, the brightness decrease of the primaries is predominant so for practical purposes we can neglect the chromaticity change.

The LED spectrum has a different behavior than both phosphors and lamps. With aging and/or higher temperature, the LED spectral peaks: a) lower; b) widen; and c) shift to different wavelengths, all at the same time. The lowering is connected to decreasing brightness. But the shift and widening means that the color also changes in chromaticity.

Moreover, the only way to control brightness of an LED is with changing the current through it. But a different current means a different local temperature, so this again changes the spectrum.

The end result of LED aging is thus: lower brightness, offset white balance, and changed chromaticity of primary colors and all other colors in between.

Because of the color performance shifting/drifting from aging, the spectra of the RGB colors can change from display to display and technology to technology, a new meter correction table will be needed for the most accurate profiling results before each calibration run of the same display.

After a meter correction profile is created, the calibration process would then be performed using the colorimeter, which performs faster throughout and has higher sensitivity at low light levels.
 
See less See more
#2 ·
Pre-Calibration Display & Meter Setup for Accurate Meter Profiling Procedure

There 5 initial pre-calibration steps to provide the most accurate meter profile creation and calibration measurements:

Display Warmup

Warmup times of an hour may be required before a display reaches a stability that will guarantee that the colors and luminances don’t change relative to one another during the time while the set of measurement results are obtained.

The luminance and chromaticity characteristics of displays change over the period of time that a display warms up after its powered on.

The change is most significant in the first 30 minutes, becoming less significant in the next 15 minutes. After 45 minutes of warmup, most displays, while not fully stable, continue to change only a small amount due to thermal changes.

If some of the profile measurements are made before the display becomes thermally stable, those measurements won’t accurately represent the luminance and chromaticity performance of the stabilized display.

That will cause the final calibration data to be less accurate than it could have been. Any display technology acts differently the first minutes after its powered on. CRT's/LED's are brighter and become darker, projectors do the opposite, OLED's have very low gamma initially; but after about one hour the most of displays/projectors have stable output.

Meter Warmup

Meters will require to be connected/powered from their USB plug for about 20-30 minutes before starting any measurement. The exact time depends from the temperature conditions. It can be longer if the device came from outside in winter time and will be shorter if stored at room temperature.

Usually users confuse warmup procedure and leave the meter on-contact with the display panel to become warmer from display's panel heat, which is a bad idea.

The meter must be at the room temperature, when the meter has stable temperature it will provide stable readings over the time, heat is not good for the sensors, when the meter is becoming warm, more noise is added to the sensor and this reduce the instrument S/N ratio.

Light emitted from displays is a electromagnetic radiation which will alter the temperature of the detection system of a meter, so the sensors themselves will heat up, one reason why meters like i1PRO require dark readings; to lower the drift caused by heat.

i1PRO2 has better thermal isolation due to its aluminum at the side. i1Display PRO using robust filters against prolonged exposure to heat.

basICColor DISCUS has aluminum unibody for added robustness and thermal stability and also features internal thermal compensating circuitry which will ensure that changes in the physical temperature around the device will not influence measurement results.

Konica-Minolta CS-2000 features a peltier thermo-electric cooling system to reduce dark current and improve S/N ratio.

Meter Placement

Because of the inherent accuracy of the Four-Color-Matrix-Method, alignment of the colorimeter and spectroradiometer with the normal of the display surface can be critical. This is particularly important for displays that have significant viewing-angle changes of their properties.

Depending upon the technology, it can be beneficial to assure alignment of the measurement system to a small fraction of a degree. Related to this viewing-angle sensitivity is the subtense of the detector.

If the angular aperture is too large or if the measurement field is too large, then it may be collecting light from too wide an area where by viewing-angle deviations can influence the measurement. Similarly, for screens that are influenced by mechanical pressure such as many LCDs, contact colorimeters that place too much force on the surface can influence the measurement.

If such contact devices are used, it can be important to assure that they are touching the screen lightly enough to avoid any changes in the measurement results from mechanical contact. Additionally, any stray light entering the colorimeter from the surround must be eliminated.

When you are taking on-contact measurements of CRT displays with X-Rite i1Display PRO, a small magnet inside the ambient light cover which X-Rite is using to signaling using a Hall effect sensor what's is the cover position, it can interfere the response of the CRT display, particularly if the ambient cover is in its natural position at 180-degree away from the measuring lens.

One way of minimizing this is to swing the cover down so that it touches the display adjacent to the lens, thereby moving the magnet away from the display surface. A more thorough but inconvenient way of avoiding this problem is to unclip the ambient light cover and slide it down the cable. Another solution is to take non-contact measurements with CRT displays.

EBU TECH.3325 Publication (Methods for the Measurement of the performance of Studio Monitors) recommends the distance measuring geometry rather than the contact method.

For projectors, we have two main concerns when deciding where to place the meter. We need the meter to be far enough back that it doesn’t cast a shadow on the area of the screen we’re reading from. But we also want the meter to be as close to the screen as possible so that we’re not adding inaccuracies such as off angle light or gain differential.

SMPTE recommends to take readings from the viewing position, but that's is not possible using consumer probes, it will require high-end spectroradiomenters with very narrow (~2.0°) field of view equipped with viewing optics (PhotoResearch/Konica-Minolta) or with laser aiming (JETI).

Meter Field Of View (FOV)

When you are measuring projectors, or displays with uniformity issues it's better to measure the same spot during the meter profiling procedure. Each meter has its own Field Of View (FOV) so their optics see different diameter area of pixels during a measurement.

Meter Profile Verification

When you create a four-color meter matrix correction table, it's recommended to always check if the calculated correction table is accurate and then start to use that correction table for all your other display calibration measurements.

A lot of users very familiar with the meter profiling procedure will be surprised when they will perform that verification test for their first time; they may be found that their generated meter correction table is not so accurate.

NIST's accepted tolerances are below xy +-0.001 and +-1.5% Luminance when you will compare the 4 color (W,R,G,B) xyY chromaticity and luminance readings between the reference meter used (spectrophotometer/spectroradiometer) and profiled colorimeter (colorimeter with four color matrix correction loaded).

Sometimes it will be required to re-perform the meter profiling procedure a second time to get better verification results with tighter tolerances.

A lot of CalMAN 5 users of Ted's LightSpace CMS Calibration Disk, they are using the Meter Profiling Check layout page to perform a meter profiling accuracy verification:

 
#3 ·
Lets dedicate this thread talking about different meter setup configuration, experiences using meter profiling of various different displays and ways to improve final results.
 
#8 ·
Ted, thanks for creating this thread. Profiling is so important and often overlooked so it's nice to have a single point to come for a reference.

To continue a thought we were discussing, the Red Y luminance profile errors with the OLEDs, assume for a minute that the CM profiler is working "OK". Thinking about it, I remember noticing this Red Luminance error happening on the 2017 LG OLEDs when, if I recall correctly, LG started using a larger Red LED. Is it possible the larger Red LED is heating up more and taking longer to cool than the other colors causing the problem because i'll have to double check, but I don't remember seeing this discrepancy with the LCDs.
 
#9 ·
A good way to avoid different drift between the reference and target meter readings is to display the same content for the same amount of time before any set of measurements. For example: 5 minutes of a movie, WRGB measurements with the spectro, again the exact same 5 minutes of the movie, WRGB measurements with the colorimeter. Same workflow for verification. This is the most practical way I know to replicate the condition of a display. If red luminance on LG OLED panels it is still over the NIST tolerance than, probably, it’s not a drift related issue.
 
#10 ·
Yes I put up a full screen 10% gray patch as a reference point.
Also I thought of another thing I can try. I'm going to take a spectrum scan of the Klein and the Jeti and compare them to see if they are both "seeing" the same.

But yes for profiling, a good technique is to have a stable reference point on the screen as you describe to let the panel stabilize.
 
#12 ·
Four Color Matrix Meter Correction Table Method
The process is known as the Four-Color Matrix Method (FCMM) for Correction of Tristimulus Colorimeter developed by Ohno and Hardis at NIST (National Institute of Standards and Technology).
While this is a way of creating a colorimeter correction, it is not actually the best way, since it is not a perceptually weighted method.
 
#15 · (Edited)
Interesting! Would you be so kind to explicit it more please? Does it apply to oled/led displays only?
It sounds like a manual FCCM tweaking, which I was discussing lately with two guys here.
 
  • Like
Reactions: jrref
#16 ·
Im trying to profile my K-10A using Chromasurf (5.53) with Ted`s suggestion,using manual entry in chromasurf and using meassurements from teds workflow page,but it doesnt work out for me,cant figure out what i do wrong,i looked at the Klein video (Link) and followed that one but no luck when accepting my manual written numbers (see pic) anyone know what im doing wrong?

Link:
 

Attachments

#18 ·
Im trying to profile my K-10A using Chromasurf (5.53) with Ted`s suggestion,using manual entry in chromasurf and using meassurements from teds workflow page,but it doesnt work out for me,cant figure out what i do wrong,i looked at the Klein video (Link) and followed that one but no luck when accepting my manual written numbers (see pic) anyone know what im doing wrong?

Link: https://www.youtube.com/watch?v=dYO29chM__4
Write the complete coordinates (with zero) and dots (not commas) and you will be fine ;)
 
#20 · (Edited)
I have reserved the first posts, to update them with instructions about how perform to meter profiling with LiVAL-ChromaSurf, or CalMAN, or LightSpace etc. (using as reference different software etc.)

I will update when I will have time. ;)
 
#21 ·
Ted, I'd really enjoy if you could talk in depth about the use of a tube frustum mask also for FCCM creation, as foreseen in the EBU TECH 3325 and IDMS publications.
 
#23 ·
...or when you calibrate during daylight like me ;)
 
#24 ·
Great thread Ted thanks maybe we can add some tips just like which pattern size for different TV kids maybe help to improve such as the right stimulus for creating a good meter profile
 
#25 ·
Do any AVS'rs here have a Jeti, Klein K10 and an OLED TV and CM?

I was doing some more experimenting profiling the Klein with the Jeti to my A9F/A1E OLED and the xy coordinates match almost perfectly but i'm still getting a 3% red luminance error between the two. I have the Jeti at 5'9" and the Klein with the hood in contact mode so the FOV should be the same. I can verify that the Jeti profile is exact and consistend vs the i1Pro2 which works but was harder to get the same accuracy and consistency in the profile probably in part due to the difficulty aiming the i1. The Jeti laser makes this task easy.

I'm curious if someone else can replicate this finding and compare it with a profile using Chromasurf. I don't have the time right now to do this experiment but it needs to be done.
Thanks:)
 
#29 ·
Hi John,

Remember what Ted posted in the other thread.

Klein K10A FOV in contact mode:

43mm = No Attachment
48mm = Rubber Hood
55mm = Extension Tube
60mm = Extension Tube + Rubber Hood..

At 5'9" distance, the JETI actual measuring diameter is 56.08mm. That's a big difference.
 
#26 · (Edited)
What do you guys think of circular test patterns instead of square windows, come to thinking about this with the laser pointer of my klein instrument,circular meassurement field,so why have square test patterns? also if we have circular patterns you could match your meter fov with the size of the circle,just hit continious reading and move meter backwards until it start to missread (hitting black) then you know the fov of the meter,cant think of any dissadvantage of circular patterns or am i missing something?

Would like to see:

* Circle pattern single adjustable in size.

* Circle patterns 2ea side by side for simultanious reading of 2 meters.

* Using circles with fov match also works as a perfect aim spot so meassurements are at same place.
 
#27 ·
Yesterday I had a little time to experiment and as many have said over and over, there are so many little things that can introduce errors or inaccuracies when calibrating making the final PQ less than optimal. But i'm finding profiling is probably more important than most people might think to get an accurate picture. For example, if your meter is seeing less red than is actually there, when you calibrate you may increase red more than is really needed to get to D65. You will then notice memory colors like skin tones having a more reddish tint that there should be. I started noticing this when comparing the i1Pro2 to the Jeti. I understand the two devices are in different leagues but although the i1Pro2 works well, if I had to comment on changes using the Jeti, if you are super critical, memory colors such as skin tones look more accurate to me profiling with the Jeti. Interesting.:)
 
#28 ·
For 6000$ I would kill if I’d not get better accuracy than a 1100$ probe lol
 
  • Like
Reactions: Miki1971 and jrref
#31 · (Edited)
Got it so the Jeti needs to be 4ft 9 inches away from the panel to match the Klein with the rubber hood in contact mode for 48 mm. For my A9F in my living room that's approximately where I had the Jeti because my couch was in the way of moving it back further and those are the scans I posted with the 3%ish red luminance errors. Something else going on because the rest of the colors and white are right on. I'll try it again with the exact FOV but it still would be good to see if anyone else can replicate the behavior. With your LCD are you using CM to profile and if so what are you seeing?
 
#32 · (Edited)
Incidentally, the Bosch laser device measures from the back of the case. The case is 4" or 100 cm long. If you have it side by side with the JETI, like I do - on a rail) you must add those 4" to the distance you are measuring. Because I have my 1501 six feet from the display, I have to set the laser to read 6'4". Get it?

If you place the laser in front of the JETI to measure distance, then you don't have to add the length of the laser device.


Edit: that's a typo, the case is 100 mm long
 
#33 ·
Incidentally, the Bosch laser device measures from the back of the case. The case is 4" or 100 cm long. If you have it side by side with the JETI, like I do - on a rail) you must add those 4" to the distance you are measuring. Because I have my 1501 six feet from the display, I have to set the laser to read 6'4". Get it?

If you place the laser in front of the JETI to measure distance, then you don't have to add the length of the laser device.
Yes I saw that in the instructions. I have a laser measuring device from many years ago that I use for construction projects but I like the Bosch because it's so small that it fits in my calibration pack and has a nice backlit display. Great choice.
 
  • Like
Reactions: Pres2play
#38 ·
I know your LCD shouldn't have this problem since it's an OLED behavior but after you profile, use Ted's profile verification workflow and post it so we can see your values.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top