Premium Single-Chip DLP
Samsung SP-A900B DLP Front Projector
JETI 1201 spectroradiometer
Minolta LS-100 luminance meter
Accupel HDG-4000 video signal generator
ChromaPure video calibration software
Joe Kane has been working for some time with Samsung to create a reference home theater projector. The latest incarnation in this effort is the SP-A900B, a 6-speed, 6-segment, single-chip DC4 DLP projector. This model supercedes the A800, with the primary difference being the inclusion of the DC4 DMD (the A800 used the DC2) and the substantially lowered black level and higher contrast that this brings. This was one of the major criticisms of the A800 so this upgrade is welcome. The A800 superceded the A700, the main difference being a greatly improved lamp.
The primary claim of this line of projectors is color accuracy and high-quality optics. You can immediately see that a substantial portion of the cost of the unit was devoted to the optical path. The lens is enormous and appears to show very high build quality. As for color accuracy, well, we'll get to that later.
The A900 is a premium product sold through select outlets (sorry, no Internet sales) who generally provide—and in may actually require—custom installation. This is, I suppose, intended to lend a certain cache to the product, emphasizing its premium pedigree. However, I tend to think that this is a mistake. Regardless of how well the projector is constructed, digital front projectors are quite easy to setup, even for amateurs. Ease-of-setup, for those of may have forgotten, is one of the primary advantages of digital projection over the CRT units they have replaced in most high-end home theaters. Those old CRT units did require a professional installer. Digital units don't, and the Samsung is no different. Just place the projector on a flat surface at an appropriate distance from the screen with the lens more or less centered (vertically and horizontally), and then adjust zoom and focus. That's it. All of the other adjustments in the user menu are no more or less complicated than what you might find on your average budget home theater projector.
The A900 also carries a premium MSRP of $9,999, which I understand is scheduled to go up shortly.
Throw and Lens Shift
Perhaps the first piece of information any prospective buyer wants to know is whether the projector will work in their installation. The A900 has a throw of 1.72-2.24/1 screen width. It also offers vertical lens shift for ease of installation, but you should use as little of this as possible to get maximum performance from the lens. Focus, zoom, and lens shift mechanisms are all manual.
The optics of the A900 are simply superb, perhaps the best I have seen. On a crosshatch pattern every pixel was clearly delineated with virtually NO chromatic aberration. On the left side of the image I saw perhaps one-quarter of a red pixel in the horizontal plane. Everywhere else I looked it was. . . .perfect. This is arguably the most appealing aspect of this projector.
The A900 is a very bright projector that is suitable for all but the largest screens. The powerful 300W UHP bulb puts out a lot of light. However, since there are two lamp modes and three iris modes, stating a lumen figure is a little complicated.
These figures were obtained with the A900 installed near the shortest limit of its throw range. In a longer throw installation the figures would be somewhat lower. Also, the light output is substantially affected by the internal CCA color calibration system, so higher considerably figures are easily available by adjustments in the factory menu.
The A900 offers a mechanical iris, which can be used to increase contrast and lower peak output. It has three settings: Off, On, and Auto. The On setting offers a slider that you can use to vary the aperture size of the iris. I found the manual iris useful for adjusting light output (the unit was initially too bright for the 110" Stewart StudioTek screen used for evaluation), but it did not appear to have much impact on contrast. Engaging the iris and setting it to 60 brought the output down to a more comfortable 15 ft-L from the eye-searing 25 ft-L we got from the Iris Off, Lamp Theater setting.
The auto iris had a significant impact on both measured on/off contrast and light output. Its operation was silent and did not create any obvious artifacts. However, it also did not obviously improve image quality. Joe Kane recommends leaving it off and I found the contrast quite satisfying without it. It also had a deleterious effect on the gamma response. In the end, I just decided to leave the iris in manual mode set at 60. Owners could gradually adjust this downward as the lamp aged to maintain a constant lumen output.
First a word about methodology. I have noticed that some home theater publications have stopped publishing raw data for contrast. This makes very little sense for flat panels, where measuring contrast is a straightforward process. However, there is some merit to this approach for front projectors. The reason for this is that measured contrast can vary quite a bit depending upon the methodology one uses. One obvious method for measuring contrast is to use an illuminance meter pointed directly at the lens. However, one probably gets a more accurate results by taking contrast readings directly off the screen. SMPTE recommends that all measurements should be taken off the screen, and I have always done so for color, grayscale, and gamma. The reason for this approach is obvious. Your eyes see the light that comes off the screen, rather than what comes directly out of the lens. But using this method for contrast readings on front projectors is problematic because the black level is so low it is often difficult to get a good reading.
I have recently started using a Minolta LS-100 luminance meter that reads down to 0.001 ft-L, which makes off-screen readings a real option for most projectors. All on/off contrast readings were taken this way and thus offer a very realistic assessment of what you are likely to see. The contrast measurements are summarized below.
Notice that engaging the manual iris yields almost exactly the same contrast as when it is in the off position, but with lower peak output. With this projector's strong light output, it would be interesting to see what sort of contrast performance would be possible with a more effective iris, such as what is included on the old Sharps, which could actually double the on/off contrast, albeit with a substantial hit to light output. I might have been able to squeeze out some additional contrast from the Samsung had I closed the manual iris down more, but that would have compromised the light output more than I wanted.
Putting the iris in Auto mode not surprisingly boosted the measured on/off contrast substantially to over 5000:1. However, during the time I spent watching actual program material I cannot say that this made a much of a difference to image quality, beyond the fact that it lowered the peak output to a more tolerable level.
Measuring checkerboard contrast is even more problematic. First, you really cannot take off-screen readings. When reading the dark squares, the luminance meter's sensors are contaminated by the light in the bright squares of the test pattern outside the viewing cone. Thus, you must take an illuminance reading. Unfortunately, the results of this reading can vary quite a bit depending upon the placement of sensor and the conditions in the room.
I got lower ANSI figures than I would have expected, 264:1 with the sensor about half way between the lamp and screen. This is a lower figure than I am accustomed to seeing with DLPs, but again I noticed no obvious problems in the image quality that I could attribute to this.
This display has an excellent reputation for grayscale and color accuracy, so I was especially interested in measuring it first hand. There are three white balance presets: 6500K, 5500K, and 9300K. Each of those presets is associated with a picture mode preset.
The Movie 2 preset is intended for older black-and-white films. The 9300K preset is a Japanese standard. Otherwise, there is no reason I can think of to use anything other than Movie 1 and the 6500K preset, which is reasonably close to D65.
The user can also manually modify any of the white balance presets by using RGB gains and offsets located conveniently in the user menu. Using these controls, I was able to achieve a nearly perfect grayscale in just a few minutes. The controls offer reasonably fine levels of adjustment and were quite effective.
Here is how the white balance presets measured.
As you can see, this preset provides a white balance that is much too blue and red deficient. The dE method is CIE94.
The profile in Standard Mode is similar to Dynamic, though somewhat less exaggerated.
Here's the Movie 2 mode.
Interestingly, this mode, which is intended for old B&W films, is achieves the lower color temperature by lowering the output of the blue channel. This would impart to the image a yellowish tinge. Here's how the 80% point plots on a CIE chart.
See how the white point is shifted towards yellow? My understanding of the 5500K standard is that the white balance should be shifted from D65 towards red or perhaps orange. In any case, this preset is of limited utility.
Next, the Movie 1 mode, which is what most people will use for critical viewing.
This is the most accurate of the presets, though as you can see it is red-deficient throughout the entire range. The dE average is 4.7, which exceeds SMPTE's DCI tolerance of 4.0 CIELAB units, so even this mode clearly benefits from custom calibration. This is good performance for a preset, but by way of comparison the out-of-the-box grayscale performance of the Planar PD-8150 was even better, returning an average grayscale error of only 3.3 dE.
Finally, here are the results of a custom grayscale calibration using only the controls provided in the user menu.
As you can see, the grayscale after custom calibration was as close to perfect as it is humanly possible to get. The RGB gain and bias controls in the user menu work very well. They offer sufficiently fine adjustments over a wide enough range to achieve the truly excellent performance you see here.
Joe Kane has always trumpeted the color accuracy of this line of projectors, so I was eager to put this claim to the test. Again, let's look at the projector's various preset modes—Dynamic, Standard, Movie 2, and Movie 1.
This is interesting. Even in Dynamic Mode, the color points (except for white, which is quite bluish due to the 9300K color temp preset) are quite accurate. However, the dE values are very high (again, CIE94 is used). The reason for this is evident on the luminance chart. All of the colors are excessively bright, and by a quite large margin. Unfortunately, this aspect of color performance is often ignored, but it is just as important as the chromaticities plotted on the CIE chart.
The Standard mode looks much the same as Dynamic Mode, except the excessive luminance is somewhat less intense.
Movie 1 and 2 Mode
We'll look at these two modes together. This is because their color performance is the same. They are distinguished only by the different white point used. Movie is traditionally the most color accurate mode for Samsung displays.
These results, using the HD preset, were a little disappointing, especially for a display in this price range whose claim to fame is color accuracy. As you can see, the chromaticities are reasonably accurate as before, but the color luminance is still too high, especially for red and blue. This performance does not equal even what you can get from Samsung plasma and LCD flat panel displays in the uncalibrated Movie mode.
The problem with the red primary on the A900 is particularly troublesome, because the excessive luminance is easily visible on-screen as overly rosy skin tones. We used to call this "red push" back in the day when manufacturers used an very bright red to cover up a bluish white point (Mitsubishi rear projectors were particularly bad offenders in this respect.). The Samsung engineers have obviously gone to considerably lengths to ensure that the xy chromaticities are very close to the standard for the selected color space (You can choose HD [Rec. 709], SMPTE-C, or EBU in the user menu). Clearly, less attention was paid to the brightness of the colors.
The next step was to custom calibrate the color performance to try to reduce the errors to below the threshold of perception. For those accustomed to the color management system (CMS) on the high-end JVC projectors or even what's available on the aforementioned Samsung flat panels, the A900's CMS is not as effective or nearly as user friendly. First, it is buried in the factory menu. Second, it offers control over only RGBW. The secondary colors may not be adjusted independently. Third, though using the CMS is not particularly difficult, working with it is somewhat labor intensive. I began by performing a full factory reset of the EEPROM. Then I measured the xyY values of the primary colors and white point in the user menu in their uncalibrated state. Next, I entered the factory menu and then manually entered the measured values into the color calibration area of the factory menu. The projector then automatically calculates how much correction is required to achieve the correct result. There is no way for the user to specify what target to use, so the projector has a calibration target hard-wired. This appears to be SMPTE-C. The EBU and HD selections in the user menu are simply offsets from that. I then turned off the projector and restarted in the regular operating mode, made some small adjustments to the white point using the RGB gains and biases, and then re-measured. The results are shown below.
Using a SMPTE-C target, you can see that the A900's color calibration method lowered most of the dE values, but not as low as I generally like to see. You can see why independent control over the secondaries would have been useful, especially with yellow. Although the yellow secondary should be derived from green and red, its luminance was too low, despite the fact that the luminance of red and green were both slightly too high. Nonetheless, the process improved the image noticeably. Red, in particular, was now much more accurate and the rosy skins tones I saw before went away after the process was complete.
For most projectors this would be excellent performance, but for one at this price point whose reputation is color accuracy I found myself wanting more. The arcane method of color calibration is a utility supplied by TI. All the manufacturer has to do is implement it in the firmware. Thus, it is a relatively inexpensive and easy way to offer a CMS (some of the BenQ projectors use a similar system, as do the Samsung DLP rear projectors). Nonetheless, the CMS offered on many current Samsung plasmas and LCD flat panels offer an approach that is more flexible, more powerful, and much easier to use. I also never found that the internal calibration system of the A900 resulted in a white point so accurate that it could not be improved by using the standard RGB Gains and Biases in the user menu.
The A900 offers three gamma presets, Film, Video, and Graphics. Just as with the white balance presets, these gamma presets are automatically selected with the picture modes. Film is used with Movie 1 and 2. Video is used with Standard. and Graphics is used with Dynamic.
By default, the Film gamma tracked just below 2.2 and the Video gamma tracked just above 2.35. After adjusting the Brightness control slightly (results not shown), these gamma values became very flat at their respective 2.2 and 2.35 points. The Graphics gamma is rather erratic, centering around 2.4 throughout much of the low end and then falling rapidly after 60% down to 2.05. Either the Film or Video gamma would be suitable for critical viewing, depending upon how the content was mastered, viewing environment, and the tastes of the audience.
Like the color, the gamma is adjustable in the factory menu. You can assign any one of a number of gamma compensation values to the three presets. Oddly, the factory default sets the Film gamma to 2.6, and the JKP calibration manual recommends setting Film at the 2.2 value, which must have been done at the distributor before shipping it to the dealer.
The two aspects of the Samsung's image that were immediately obvious to me were the same strengths revealed in the objective tests. The high light output and excellent optics lent the image an extremely vivid and sharp appearance. Its depth, color accuracy, and black level were not as good as the JVC RS20, but the Samsung's image did have a crispness and almost tactile quality that was quite appealing. I thought the image the A900 provided was somewhat, though not enormously, superior to the Planar I looked at recently. The Samsung surpassed the Planar in terms of clarity and sharpness, while simultaneously revealing less noise in the image. However, the Samsung is considerably more expensive and this comparison is based on sense memory, so should be taken with a grain of salt.
The Samsung runs almost completely silent, but it also runs quite hot. It can actually raise the temperature in a room a few degrees. This is, I suppose, one of the consequences of using a 300-watt bulb.
The A900 is a very nice projector, but this price point is very competitive. If deep black levels and nearly perfect color performance are what you want to see, then you should probably look elsewhere. But if bright, punchy, and vivid realism is your cup of tea, then it is hard to think of a better choice.
Thanks to Mark Haflich of Soundworks in Kensington, Maryland for his assistance.
There are some gotchas with the Samsung A900 color calibration system that I was unaware of. They are simple procedural issues, but if you don't follow the correct set of procedures in the right order, you get sub optimal results. Once I resolved these issues, I calibrated a couple of additional A900s.
The results have been fairly uniform. First, the ability of the Samsung's internal calibration tables to achieve correct xy chromaticities works as advertised. The results are almost scarily accurate—when finished, the primary colors measured within xy0.001 of their intended targets. The luminance values were not as accurate. On one of the units the red primary was considerably too bright, while on the remaining colors were reasonably accurate. On the other two units I tested all of the colors were too bright. This latter group was easy to fix by simply turning down the main Color control four or five ticks. Modest changes to the main Color control will generally reduce the intensity of the color without significantly altering its xy chromaticity. However, nothing could be done for the remaining unit. Turning down the Color control would have fixed red, but then the remaining colors would have been less accurate.
It seems that the RGB mixing matrices that the A900 uses are 2-dimensional. An oversaturated red, for example, is desaturated by adding amounts of green and blue to it. This moves the color towards the white point and creates a pseudo red with the correct xy chromaticity. However, moving a color towards the white point will generally increase its intensity. A 3-dimensional CMS would compensate for this by adjusting the amount of red in the mix to achieve the optimal result. Nonetheless, with small adjustments to the main Color control, very precise color calibration is easily achievable with the Samsung A900.