To reply on the wish about frank opinions from other scientists: We´ve evaluated at least 20 and up to 100 units of the following devices in the past 18 months to get an idea about the precision, repeatability and inter instrument agreement in real world usage for a scientific research project. The devices have been stock units - not especially selected ones to avoid any manufacturer selection. The Munki is missing because there is no real third party support for it.
- Eye One Display 2
- Spyder 3 Elite
- DTP94b
- EyeOne Pro Rev D
We´ve used a Minolta CS200, CS1000 and CS2000 to get reference values. The test have been done on stabilized displays with the following base specs:
- 72% NTSC display with LG S-IPS and CCFL Backlight
- 72% NTSC display with LG S-IPS and pseudo white LED (blue LED + yellow phosphor) Backlight
- 92% NTSC display with LG S-IPS and CCFL Backlight (WCG)
- 102% NTSC display with LG S-IPS and CCFL Backlight (extended WCG)
- 108% NTSC display with LG S-IPS and RGB LED backlight
I will summarize the findings we had so far:
1) Inter instrument agreement between at least 20 units:
- EyeOne Display 2 - max 18 dE, mean 8 dE
- Spyder 3 Elite - max 15 dE, mean 7 dE
- DTP94b - max 3 dE, mean 1,5 dE
- EyeOne Pro Rev D - max 3 dE, mean 2 dE
As recommended by others, one should only make use of one (1) EyeOne D2 or Spyder 3 unit for all installed displays, because the tolerances between these devices are much too high to get consistent results. Even the 3dE max tolerance of the DTP94 or EyeOne pro can be too high. However, both devices show a much better inter instrument agreement.
2) Measurement quality
We used the "best" set of the evaluated units to perform some quality tests. All selected colorimeters showed an acceptable performance on the 72% NTSC CCFL display. However, on a white LED unit, the readings differ a lot from the reference. The spectra of a white LED unit differs a lot from a standard CCFL it replaces and the colorimeters are trained on a specific panel type and spectra. As soon as the spectra differs too much, the device tends to report wrong readings. The Wide Gamut CCFL and RGB-LED readings showed - as expected- even larger deviations. When the colorimeters are trained on the specific panel/spectra, their readings again reach an acceptable quality. However, the filter set of the DTP94 proofed to be farer away from the CIE standard observer than the ones from the EyeOne Display 2 or the Spyder. It needs therefore a higher correction for wide gamut display. Although, the EyeOne Dispaly 2´s and Spyder 3´s filter may match better to the CIE curves, they are still far off. As a result, all colorimeters need a correction for wide gamut (and white LED according to our test). Some display vendors use specially trained sensors for their wide gamut displays, others use standard ones and implement the training on the specific panel/spectra in the calibration software. The result should be always the same. But, keeping the large deviations from (1) in mind , it´s at least questionable if the single EyeOne Display 2 or Spyder 3 Elite matches the correction curve (either in Hardware or in Software). For the sensors that come bundled with a display, one can only hope that the inter instrument agreement is better than with standard retail units.
3) Differences between Colorimeter and Spectro during the tests
The only visible deviations that we´ve seen was a quite noticeable difference in the dark tone represenation on the selected displays. The older 72% NTSC unit had a lowest black of 0.5cd/m2. Here, the EyeOne Pro Rev.D performed equal to the colorimeters. On the 72% white LED and 92% NTSC CCFL with only 0.3cd/m2, the EyeOne pro started to created more noise in the darks and had a lot of questionable readings. As far as I know, the Spectro is not able to set individual integration times with different luminance levels (in contrast to the colorimeters). The visible result can be described as drowning dark tones (no matter if they were real blacks or dark colors). This gets even worse with the RGB LED backlighted 108% NTSc unit. This unit had a blackpoint of only 0.1cd/m2 and here we´ve seen even more drowning and colorfull (noisy darks with visible colorshifts and high Kelvins) darks - to be fair, the EyeOne Pro is not specified for dark readings below 0.2cd/m2. However, even the EyeOne Display and the Spyder created more noise on this display than on the thers. The DTP94 created the best results - in terms of details and neutrality - in the darks but is also well known for being the slowest device at all. The dark readings of the DTP94b take even more time as the device adjusts the integration time (like the other colorimeters, too). The EyeOne pro uses - compared to the 1nm reporting of the CS2000 - a wider sampling interval. It has been often discussed, that the relatively small peaks of CCFL displays (discontinous spectral characteristic of CCFLs) may cause interpolations errors with the EyeOne pro, because they can be spectrally located between two 10nm measurement steps. This can be noticed on both, the 92% and 102% unit in comparison to the Minolta Spectroradiometers. However, there is no visible difference between the Colorimeters and the EyeOne pro in real world perception tests. Therefore - according to our tests - there is a limitation of the EyeOne regarding the spectral interval, but is has no or little effect on the color representation on the selected displays.
To summarize, the EyeOne Pro Rev D (the A/B Revs performed far less good) is well suited for today´s displays except the extreme dark readings. All Colorimeters need an additional correction matrix on wide gamut and white LED displays. Additionally, 2 of the 3 evaluated colorimeters suffer from poor inter instrument agreement. The DTP94 (not available from Xrite any more) performed close to the EyeOne Pro in regards of inter instrument agreement and with a correction for wide gamut, it showed the best results even on the displays with lower black luminances.
I hope this makes the image a bit clearer.
