Sony X900F Image Quality Review
By Chris Heinonen on
We’re going to start looking at the image quality of a number of TVs for 2018 starting with the Sony X900F. This year I’m doing these in a different way than before as well. They aren’t a traditional review, where we care about the design, the user interface, the price, and so on. I cover that at Wirecutter and need to make this different. Instead, I’m just going to focus on the image quality in SDR and HDR, before and after calibration. I’ll answer any questions that come up outside of that if needed, but I feel most readers here want to know how the TV looks and performs, not about the audio quality or streaming services.
Sony X900F SDR
For the Sony X900F, I used the Cinema Pro preset with a target of a BT.1886 gamma. This is my preferred dark room setting but for a living room, I might set the gamma to 2.2 instead to avoid having shadow detail be obscured by ambient light. Since the X900F has a gamma slider that does a good job, you can adjust it to match your room quite easily.
Straight out of the box, the X900F is very good. I’ve tested Sony displays directly from Sony and those bought from stores when I do calibrations, and they all exhibit very similar performance so I don’t feel this was a cooked model. The RGB balance and gamma track near perfect, leading to an average dE2000 grayscale error of 0.71. This is effectively perfect, so for most people, there will be no reason to go into the menus and adjust this further. You can look at the chart and see that the highest measurement on the grayscale is 0.95, so I wouldn’t bother adjusting this.
Color gamut coverage is good, at 97-98% of the Rec709 standard. It is almost impossible to measure at 100% in CalMAN without going beyond the standard, since unlike some other programs, CalMAN only calculates what is inside the triangle. Looking at the saturations chart, we see that there are more errors here than on the grayscale chart, but they all remain below the level of 3.0 that we desire. Green has the hue shifted a bit, and red is under-saturated at points but then almost fully saturated at 100%. Magenta is also off in hue a bit, while blue is under-saturated. Overall they’re just fine, especially out of the box.
Luminances are similar in performance to the saturations, though there are errors that pass the 3.0 boundary at 100%. If you have a 100% red sample then you might see a slight bit of color error, but it isn’t anything I would worry about at all. The color checker shows that the issues with the Sony X900F are in the oranges, with the red saturation lacking a bit on them and some rising above the 3.0 level you see here. It isn’t a major issue, but it is something to be aware of.
The contrast ratio on the Sony with the backlight set high (you can measure faster with the backlight set high, but I should have it set to 150 to be equal to all TVs) is only 3376:1. I measured twice to be sure of this, but there is no local dimming engaged for this. For the post-calibration result,s I engaged local dimming, which is how you are likely to watch this TV anyway, and those numbers will be different.
Overall the Sony X900F looks fantastic before calibration, provided you use the correct setting of Cinema Pro. Post-calibration things improve, but only from small adjustments that we made to the color controls. The X900F, like every recent Sony TV, has no CMS access, so you have to deal with color and tint controls to get it right. Here we’ve reduced the color controls a bit, which also reduces the gamut coverage slightly to 94-96%. In doing so, we’ve managed to lower all the saturation, luminance, and color checker errors as well. The downside here is that you’re losing some saturation with pure colors, but you’re gaining accuracy with those that aren’t 100% saturated. Myself, I’d take the trade-off and have lower error levels, but it isn’t a situation where one is certainly correct. You can adjust this yourself and see what you think if you get an X900F.
We see the contrast ratio has also improved from 3376:1 to over 108,000:1 with the local dimming enabled. The big change on the X900F this year is the black frame insertion method. Where you usually lose a huge amount of light output with any BFI, on the Sony it is ramping up the backlight to compensate for the BFI, giving you an image with better motion but also without losing as much brightness. I found adjusting the controls to have just a touch of motion compensation enabled gave me an image that was clear and clean, but not soap-opera-y.
On the first episode of the White Rabbit Project on Netflix (thanks to a reader that told me about this), there’s a panning shot of a schematic that is black on a white background. This is great to use for testing motion as TVs that are not 120Hz will flicker and judder like crazy on it. The X900F does a good job of showing the motion here with less judder, but without artifacts or looking fake. It did a very good job watching sports with this enabled, giving you clean motion but also letting you keep the settings for movies without making it look fake.
Local dimming on the X900F works very well, with nearly invisible transitions between bright and dark scenes. On the Harry Potter hill scene in the final film, it manages to show all the details of everyone on the hill well but then ramps up the full array backlighting zones when they attack Hogwarts to make the explosions quite bright. This does introduce some slight blooming, but it’s very mild and hidden away by most of the explosions.
For SDR, the Sony X900F has near flawless performance. The grayscale is ideal, and colors get very close if you back the controls down slightly and reduce the gamut. I wish Sony would offer CMS access, or even direct LUT access like LG, but I’m even happier that they ship TVs that are this accurate without any work.
Sony X900F HDR
For HDR with the Sony X900F, I again used Cinema Pro as it was the most accurate out of the box. Looking at the data, what we see is that the RGB balance here drifts off past 50%. This is somewhat expected as Sony uses a single CMS across both HDR and SDR content, and if we adjusted it for SDR based then 100% SDR would line up to 50% of HDR. We didn’t adjust the RGB balance as it was accurate out of the box, but it’s not surprising to see it only drift after 50%. However, you can look at the Grayscale error charts without luminance and see that overall it is quite accurate.
Looking at the EOTF, we see that the image is too bright compared to what it should be. The chart data backs this up, with 50% coming in at 131 nits instead of 94 nits. From the charts, we see that the error levels with luminance are quite high (this uses the dEICtCp method from Dolby, which is meant for HDR, and not the traditional dE2000 formula) with an average over 10 compared to < 0.5 for errors without luminance. Colors are the same, with high error levels once you factor in luminance. Color gamut coverage for DCI/P3 is fair, with 85-92% coverage of the gamut while many other TVs are closer to 95-98% now.
As mentioned briefly earlier, Sony does not have separate memories for SDR and HDR for certain controls. If you adjust the white balance in HDR, it affects it for SDR, and vice versa. Since the RGB balance is already good, we don’t have to worry about that here and can focus on fixing the EOTF. We are targeting 1000 nits of brightness, not relative brightness as some other reviews do. The reasoning here is that content is mastered at 1000 nits or 4000 nits and not relative, so we want to evaluate how a TV does against real world numbers. Also, if you use relative than a TV that only does 400 nits might score as well as one that can do 2000 nits, and that just isn’t how it looks in the real world.
Post-calibration we have adjusted the gamma and now we find that the EOTF tracks almost perfectly. You can see that the RGB balance is better after a small tweak there, which didn’t cause much harm in SDR, and the grayscale errors with luminance included have fallen a lot. Now instead of an average error over 10 we are down to 2.61. The errors for colors are improved, but 100% saturations for DCI/P3 are higher as the gamut comes up short of the full range. Looking at a color checker chart, the error levels are almost all below 3.0 as well. We use dEICtCp for these as well, since they are HDR and not SDR, and these results are very good for that formula.
Since I created this workflow after doing measurements the color volume data is mostly missing as CalMAN doesn’t save it. For the Dolby method, the X900F measured 375 million distinguishable colors with 79.49% coverage of the DCI 1000 nit L*a*b colorspace and 53.73% coverage of the Rec2020 1000 nit L*a*b colorspace. Below you can find a chart with other color volume data we have measured to see where the Sony stacks up. It does quite well for the Dolby method, but behind some other TVs with L*a*b and some of this might play into the tone mapping that I’ll discuss next.
Sony uses a fairly hard clip in tone mapping HDR. It tries to keep the average picture level high, which means bright highlights in Pan or Batman vs. Superman wind up clipped and missing details, or lacking color. Other TVs from Vizio and Samsung reduce the brightness of peak highlights to show more detail. This means images on-screen might be dimmer than on the Sony to show those highlights, which some prefer and some don’t. I personally prefer other tone mapping options to the Sony, since it causes it to look more washed-out in those peak highlights to maintain brightness, but it’s also most noticeable on still frames. With content that uses a 1000 nit cap, which is half of the disc based content (or more) and virtually all streaming, this isn’t an issue as the X900F can hit those peaks.
I measured 1015 nits on the X900F post-calibration, which is enough for most HDR content. I used a 55” version for testing, and with HDR you can find different sizes produce different nit output levels. Usually the larger the TV the brighter it goes, thanks to more zones, but it isn’t always the case. The smaller 55” TCL R617 does highlights that are 150-200 nits brighter than the 65” version consistently, but I don’t know if the X900F is the same. I know people that have tested different sizes of the Z9D found the larger versions were brighter, so Sony might be the same here with the X900F.
With HDR content, the Sony X900F does a bang-up job once you calibrate it. The initial settings are too bright so you’ll wind up with extra highlight clipping, but it also makes it easier to watch it in a brighter room. With HDR I disabled any of the black frame insertion to avoid any drop in brightness as well, so the motion isn’t quite as good as it is with SDR. With a 120Hz panel, it still looks good, but it isn’t as good as in SDR.
With the X900F, Sony has produced a set with very high image quality. The SDR image is almost perfect out of the box, with just some slight changes to the color necessary once in Cinema Pro mode. I also adjusted the MotionFlow settings to have a bit of black frame insertion to improve the motion quality without adding too much interpolation.
For HDR, the EOTF had to be adjusted to be correct, which you can do with tools, but is hard to do without them. If you can do this, then you’ll wind up with an image that is very accurate to the HDR standards. Without these adjustments, it tracks close for RGB balance and colors, but the luminance levels are off causing high error levels.
The X900F added support for Dolby Vision, but Sony has not provided a way to calibrate it at this point, so we did not measure this. It also uses a different method for Dolby Vision and we are not certain that our test pattern generator even supports it at this point as well.
In the end, the X900F from Sony is a very safe recommendation. It is very good in almost every area, and has no major failings that would cause you to not like it. I wish Sony would implement dynamic tone mapping much like Samsung and LG have done to preserve more highlight detail, but don’t expect them to ever do that. They also could open up CMS or LUT access for better calibrations in the future. But as it is, the X900F is very accurate, does well with movies, video games, and sports, and comes in sizes ranging from 49” to 85” to fit any room.