Display Calibration Sensors
As part of updating our review of monitor calibration tools, we evaluated most of the commonly available measurement hardware. Some monitor profiling packages work well with certain displays but not with others. Determining why requires investigating whether the measurement hardware is up to the task or if the software is the culprit. An inaccurate instrument will not produce an accurate monitor profile.
We compared the measurement accuracy on a range of commonly available display panels. These included both standard (sRGB compatible) and wide-gamut (Adobe RGB) displays and Cold Cathode Fluorescent (CCFL) and both RGB and White LED backlight technologies. The display models ranged from sub-$300 basic graphics-editing monitors to high-end (and high-dollar) state-of-the-art panels. We measured the accuracy in white, middle gray and darkest usable black.
For nearly a decade, the general trend in the most common display measurement hardware has been to trade off measurement accuracy and robust construction for lower manufacturing cost and increased measurement speed. Kudos to Datacolor for bucking this trend with their Spyder products. Display technology, however, has advanced considerably. Gone are CRT monitors, replaced by LCD screens not only offering standard color gamuts (roughly covering sRGB) but also wide-gamut monitors capable of displaying colors exceeding the Adobe RGB range. Until mid-2011, we were left trying to measure today's monitors with sensors designed for monitors last manufactured many years ago.
High-end monitor vendors offer sensors tuned to their particular displays. These customized sensors work admirably well, but only for the monitor they are paired with. Summer of 2011 brought two new measurement products to market that — finally — were designed to measure all typical display types and that could be software-programed for new monitor characteristics when they arise. The new products are, hands-down, the winners of our tests. Software support (Fall, 2011) is spotty at best, but will improve.
If cost is no object, the BasICColor Discus is an exceptional piece of equipment. It offers superb, consistent accuracy when measuring any display technology. It also costs $1300 (all prices are street). As of mid October 2011, monitor profiling support for the Discus is limited to one commercial software package, albeit a good one. Additional profiling software for the Discus is scheduled to arrive soon.For $250 , the X-Rite i1Display Pro is also a capable instrument. It comes close to the Discus in accuracy and intra-unit variability, and also is compatible with all display types. The i1Display Pro (aka i1 Display 3) requires more care to use than the Discus but certainly wins on the affordability front. X-Rite also offers the ColorMunki Display. At $170 it shares identical hardware and has the same measurement accuracy as the i1Display. The ColorMunki Display is limited to a measurement speeds five times slower than its more costly sibling and ships with a feature-limited software package. As with the Discus, software to make use of the i1Display Pro is limited, but more packages are due to support it. X-Rite mentions that the ColorMunki Display will only be officially supported by their own software. In our tests, monitor profiles generated by X-Rite i1Profiler, i1Display Pro, and ColorMunki Display software were very good but did not match the quality of profiles from the best of the competition. Once third party companies license the i1Display Pro and, one hopes, X-Rite improves their profiling code, the i1Display Pro will set the standard for all but the most demanding uses.
Cost aside, ability to run monitor profiling applications such as ColorEyes Display or BasICColor Display are the only reason to consider older measurement devices. For standard-gamut displays, the best of the rest was the X-Rite DTP-94. It offers a combination of accuracy, consistency across multiple units, and shadow resolution no other legacy sensor can match. X-Rite is no longer manufacturing the DTP-94, but it is still available from third parties.
For wide-gamut displays, the best legacy colorimeter option is the Datacolor Spyder 3. It is reasonably accurate, but unit-to-unit performance is not as consistent as could be desired.
The X-Rite Eye-One Pro spectrophotometer is a capable instrument that has no issues with measuring wide-gamut displays — as long as the output is bright enough. Spectrophotometers lack the resolution in dark shadows to measure accurately. All too often, this produces color artifacts in darker tones on high-end displays.
The popular Eye-One Display 2 showed high levels of unit-to-unit variation. Whether any given Eye-One Display 2 (i1D2) sensor measures accurately is largely a matter of luck; what is more certain is that two different i1D2 units will give widely different readings for the same color. As with the DTP-94, the i1D2 is no longer in production. The only versions of i1D2 we can recommend are those custom-tailored to a specific panel type (see below). The new i1Display Pro and ColorMunki Display are far and away superior options.
The Eye One Display 2 can be calibrated using a reference spectroradiometer. This is the approach used by at least some monitor OEM manufacturers and third party suppliers of monitor calibration hardware. Once calibrated individually, the i1D2 offers both consistent performance across units and good accuracy, albeit only for one particular type of display panel.
As a reference, we used a Photo Research PR-730 spectroradiometer. The PR-730 is one of the few instruments on the market capable of accurate measurements of polarized sources (e.g. LCD screens) at the low light levels the best current displays produce. It also is fairly fast and has modes that give quick and accurate readings of brighter values.
We measured as many samples of each sensor as we could. These included units we owned, borrowed, or that were supplied by the manufacturer or a calibration system vendor. Models tested were:
Colorimeters are noted for their low-light measurement capability, but need to be tuned for a particular light source. The oldest instruments, such as the DTP-94, were tuned to CRT displays, while newer legacy models are set for standard-gamut LCD displays. High-end monitor vendors use specially-tuned colorimeters to match the output characteristics of their particular displays. The new kids on the block, the Discus and i1Display Pro, have both internal calibration tables for all standard backlighting options and the capability to load additional tables if a new display technology requires it.
Spectrophotometers do not have the same limitations on measuring arbitrary light sources as do colorimeters, but they tend to have more measurement noise at low light levels. This makes them less able to accurately characterize the blackest levels good monitors can reach.
We characterized at least 10 of each instrument with the exception of the DISCUS and ColorMunki.
First, we measured how much variation we saw between instruments on a standard-gamut LCD monitor at a 6500K white point and luminance of 150cd/m2. Absolute accuracy was not measured, just how much variability we saw between units when measuring the same input color. Variation is expressed in terms of Delta-E 2000 (dE-2000). One dE-2000 approximates the minimum color difference human vision can detect.
|Instrument||Units Tested||Mean Delta-E 2000||Max DE-2000|
|i1Display Pro/CM Display||13||0.4||1.6|
|Eye-One Display 2||17||7.0||14.0|
|OEM Eye-One Display||11||3.7||7.0|
|Spyder 3 Elite||12||4.5||8.0|
Based on our tests, each Discus unit gives results visually identical to another on all panel types. Our sample size was limited, but we saw no indication of any outliers. The i1Display Pro/ColorMunki Display colorimeters also show good agreement between units, although not quite to Discus levels.
Amongst the older instruments, the DTP-94 and the spectrophotometers gave the ,ost consistent performance while the X-Rite Eye-One Display 2 shows comparatively high unit-to-unit differences. This means that different i1D2 devices give very different values for what should be the same color. If you are calibrating more than one monitor with an i1 Display 2, you should only use a single i1D2. Otherwise, there are likely to be visible color differences between displays. The Spyder 3 fares better, although there still is relatively high variability between units. The OEM i1 Display 2 models were individually calibrated by the vendor to a specific display. Our measurements were made on the specified display model.
We evaluated sensor accuracy (referenced to a PR-730 lab grade spectroradiometer) on a variety of display panels. The results from two of the displays illustrate the main points. One monitor was a standard-gamut (measured 96% sRGB coverage, 66% Adobe RGB), CCFL backlight, e-IPS panel — a basic, low cost 24-inch display capable of reasonable graphics editing. The second monitor was a wide-gamut model having an p-IPS panel and RGB LED backlighting. This monitor measured in at 115% Adobe RGB gamut coverage, aims at the pro-level market, and is priced accordingly.
Looking at the average accuracy at both a 6500K, 150cd/m2 white point and at a minimum usable black point (0.19cd/m2 for the standard-gamut display; 0.11cd/m2 for the wide gamut) shows the strengths and weaknesses of each sensor.
|Instrument||Mean sensor error dE-2000|
|Standard Gamut||Wide Gamut|
|X-Rite i1Display Pro/ColorMunki Display||1.1||2.0||1.7||2.8||X-Rite Eye-One Display 2||4.5||5.3||6.5||5.3|
|OEM Eye-One Display 2 (wide-gamut calibrated)||7.6||4.9||1.8||3.9|
|Datacolor Spyder 3 Elite||4.2||5.9||3.7||5.7|
|X-Rite ColorMunki Photo/Design||4.9||10.6||6.1||15.9|
|X-Rite Eye-One Pro||4.4||9.6||2.3||10.7|
The absolute error level for each sensor in part reflects what calibration standard was used by the manufacturer and how it corresponds to our reference spectroradiometer. The values also need to be viewed in light of the variability detailed above. Nonetheless, some general conclusions can be made:
Measured white point variation for X-Rite Eye-One Display 2 and DTP-94 on sRGB gamut display.
The BasICColor DISCUS is a no-holds-barred instrument for monitor profiling. The price tag is high; more than the basic Eye-One Pro plus any other colorimeter combined. Photographers requiring the utmost in monitor accuracy, however, should seriously consider the DISCUS. We saw exceptional performance on all backlight technologies. The DISCUS also has built-in thermal compensation, and we measured only minimal drift with over a 10°C change in temperature. That said, the DISCUS exhibits one noteworthy flaw. The sensor can measure ambient light. In contrast to the superb accuracy seen when measuring monitors directly, ambient light measurements of blackbody illuminants (e.g. natural daylight or tungsten bulbs) were consistently off. Reported color temperatures were low, by the equivalent of 6 - 8 dE. BasICColor states that the Discus is, for now, only calibrated for ambient light measurements of JUST Normlicht viewing booths. More options are coming; until they arrive, ambient light measurements are not a strong point of the Discus.
X-Rite has excellent products in the i1Display Pro and ColorMunki Display. The sensors offer equal, and equally outstanding, measurement accuracy. Unit-to-unit variability is higher than with the Discus, but at levels only a fraction of what any competing sensor offers. Overall measurement accuracy is also slightly below the Discus. When price is a major consideration, the X-Rite pucks are untouched by anything else on the market. Thermal compensation is also built into the i1Display Pro and ColorMunki Display. Unfortunately X-Rite's monitor calibration software is not up to the same standard as the hardware. Grayscales do not render with the same level of neutrality obtained from either BasICColor Display or ColorEyes Display Pro, more posterization and banding artifacts appear, and critical options such as L* gamma calibration are unavailable.
For legacy instruments, if you are only concerned with calibrating and profiling standard, sRGB gamut monitors, the X–Rite DTP–94 is an excellent choice. It is accurate and consistent. The consistency gives confidence that any given DTP-94 will be a good one. Unfortunately, the DTP-94 is only available from third party profiling software vendors and is no longer being manufactured. The DTP-94 performs poorly with wide-gamut displays.
Datacolor appears to have solved the worst of the manufacturing inconsistency problems seen with older Spyder 3 Elite models. On average, the Spyder 3 is a good performer. We still have concerns about the level of variability measured between units — almost 3 times what the DTP-94 shows. For wide-gamut monitors the Spyder 3 is the best choice among legacy colorimeters unless your monitor vendor offers a specially-tuned sensor.
After X-Rite purchased GretagMacbeth, the combined company settled on the Eye-One Display 2 instead of the DTP-94. The Eye-One is a simpler unit, and probably less costly to manufacture as it uses plastic filters rather than the glass in the DTP-94. Measurement speed of the Eye-One Display 2 is fast; only the new i1Display Pro is faster among colorimeters. Inter-instrument agreement is particularly poor with the Eye-One Display 2, however. Whether any given unit is accurate is unknown, but you can be fairly sure that two i1 sensors will not agree on the color of your display. Average accuracy was comparable to the Spyder 3 on sRGB-gamut displays and better than the DTP-94 on wide-gamut models. Particularly now that the i1D2 is no longer in production, we do not recommend it for monitor calibration and profiling.
The OEM-specific units of Eye-One Display 2 we tested performed commendably well on the monitor it was designed for. On all other panels, however, measured accuracy was worse than a generic i1 Display and the resulting calibration suffered visually. We do not know if vendor-specific i1 Display units show the same level of difference between units as do the generic models.
X-Rite's Eye-One Pro sets the industry-standard for spectrophotometers. There are several packages available, all offering monitor profiling with various printer profiling options. The i1-Pro comes in UV-cut and standard versions. These only affect whether any UV components of print paper are measured, not emissive display measurements. The Eye-One Pro capably measures all but the darkest colors and tones on any monitor, sRGB or Adobe RGB gamut. Where it has problems is in the darkest values. Any reasonably high-quality display designed for graphics work can show colors dimmer than the Eye-One Pro can accurately measure. Calibrating and profiling these displays with an Eye-One Pro often results in visible color shifts in the darkest levels. You can either sacrifice display contrast by setting the black level to higher than 0.3 cd/m2 or accept that dark shadows likely will not display as they should.
Achieving better results from the Eye-One Pro is also possible through software. The open-source ArgyllCMS increases measurement integration and averaging time when reading darker values. This approach improves the accuracy of Eye-One Pro measurements. Calibrations made with Argyll show reduced color shift problems in dark shadows than does any other package we have tested. Argyll is a mostly command line package, does not support DDC to optimize in-monitor adjustments, and has none of the user-friendly features of commercial packages. For those who like to get under the hood, the price is right and the performance good.
The Spectrolino is similar to the Eye-One Pro except that it trades off measurement speed for improved accuracy. X-Rite discontinued manufacture of the Spectrolino over 5 years ago. Although the Spectrolino offers exceptional capability, buying a used unit may not be the best option. Recertification, calibration, and servicing require the instrument be sent to Switzerland with a base cost of $600. Also, most state-of-the-art monitor profiling software does not support the Spectrolino.
We cannot recommend the ColorMunki for critical monitor calibration needs. It likely meets the needs of its intended market — advanced hobbyists wanting to profile both displays and papers. If your business success depends in no small part on how accurately your display renders colors, the ColorMunki is not your best option. Caveat: It is certainly possible that we obtained a particularly poor performing sample. Based on the good agreement between models of other spectrophotometers, however, it is unlikely that the ColorMunki we tested is that far off the baseline.
If cost and portability are not major considerations, the BasICColor DISCUS is easily the most capable sensor we tested. It provides very good accuracy when measuring RGB back-lit monitors and even better results on the CCFL panels found in the highest end Eizo and NEC displays. If your business depends on how accurately your monitors render color, the DISCUS is an attractive choice. The Discus also proved essentially immune to ambient temperature changes and has the most secure and foolproof attachment to a monitor screen of any measurement device available.
Coming in a close second — often by margins invisible to the human eye — is the X-Rite i1Display Pro. We have a detailed comparison of the i1Display Pro and Discus. Numerically, as measured by a reference grade spectroradiometer, the Discus is the more accurate and consistent meter on every display we characterized. Visually, the differences ranged from small to invisible. Until a commercial software package is released that drives both the i1Display Pro and BasICColor Discus equally well, any conclusions on whether the added accuracy of the Discus produces visually superior monitor profiles remains speculation. We expect to have further evaluations in early November, 2011.
Two points where the i1Display Pro comes out ahead of the Discus are price — purchasing five i1Display Pros costs less than a single Discus — and portability. The Discus is a beast, closer in size to its track and field namesake than a normal monitor measurement puck. The i1Display Pro is svelte, smaller even than the i1D2 it replaces. A drawback of the small size is that the i1Display Pro needs to be placed on the screen with care and, ideally, measurements should be performed in a darkened environment for best accuracy. The Discus is less fussy. Place it on the screen and measure away.
X-Rite's ColorMunki Display also deserves mention. From a hardware perspective it is identical to the 1Display Pro with the exception of 5x slower measurement speed and only measuring ambient light luminance rather than color temperature as well. The included software is also more limited, offering fewer options for calibration settings and validation. For photographers not earning their living based on how accurately images appear on their screens, the ColorMunki Display is a viable choice. One caveat is that X-Rite does not license other vendors to use the ColorMunki Display, so your are stuck with the X-Rite software. For hobbiests and those out to save $80, the ColorMunki provides a lower cost solution capable of profiling any type of display.
For high-end displays, many vendors offer their own pre-tuned colorimeter sensors. Our experience with a limited sample of such sensors (all X-Rite Eye-One Display 2 based) from HP and NEC has been good. The OEM calibration performed reduces the variability between units by half compared to a stock Eye-One Display, and greatly improves accuracy on the monitor it is tuned for. The drawback is that these customized sensors are only accurate for a single display type.
For legacy colorimeters, if all your monitors are standard, sRGB-gamut models, the X-Rite DTP-94 is an excellent overall choice of calibration sensor. It is supported by two of the best monitor profiling and calibration packages in our testing: ColorEyes Display Pro and BasICColor Display 4. It is no longer being manufactured, but the price for existing units is attractive.
The best wide-gamut choice among older meters is to be the Datacolor Spyder 3. The DTP-94 is inaccurate measuring wide-gamut displays while the stock Eye-One Display exhibits such high variability that any single sensor cannot be relied on. This advice applies to users with critical color matching needs; if your business does not depend upon it, any colorimeter is a vast improvement over none at all.
Spectrophotometers are a mixed blessing. They can accurately measure any display panel except for the darkest shadow areas. Monitors capable of resolving dark shadow levels cannot be measured accurately by a spectrophotometer. This leads to the curious situation where a $1000 spectrophotometer is an excellent choice for characterizing a low-end, $350 wide-gamut monitor.
Intelligent monitor profiling and calibration software can help here. The free ArgyllCMS package increases measurement time on dark samples, and being open-source, one can go code-diving to tweak the algorithm further. Of the commercially available packages, ColorEyes Display does the best job of taking more samples with an Eye-One Pro when measuring in shadow areas, but the number of measurements probably needs to increase by a factor of ten to be competitive with Argyll. No software wizardry will improve a spectrophotometer's accuracy at low light levels to be competitive with a colorimeter. Alternatively, spectrophotometer measurements of the monitor white values can be used to tune colorimeter measurements, allowing the strengths of each instrument type to be combined. ArgyllCMS is the only package that provides this capability.
We wish to thank a number of vendors for their support in providing access to more measurement sensors than we had in-house and for valuable discussions regarding hardware and software capabilities. These include: CHROMiX Inc., X-Rite Inc., Integrated Color Corp., BasICColor GmbH, SpectraCal Inc., and HP.
All other trademarks and copyrights are the property of their respective owners.
© 2013, Dry Creek Photo. All Rights Reserved.