Everyone ready? It's a BIG post, so watch out. In 5... 4... 3... 2... 1...
THE CRITERIA
Alright... I'm a little sleep deprived, but otherwise okay, so let's get to the report.
The only current-crop LCD sets Robert had available for testing were the edge-lit Samsung, the LH90, the Regza 670, and the XBR9. Performance on the XBR9 remains so bad that we didn't even bother trying to make it work, so the three LCDs on which I can report are the Samsung, LG and Toshiba. The set we used for reference was a Pioneer KRP500M plasma because, well, because those things are reference monitors. I went there to look for specific flaws that I know to be common, if not ubiquitous, in locally dimmed LCDs and in LCDs generally. I looked for:
(F1) Halos: these are glowing halos of light that appear around a bright object set on a dark background. They result from the local dimming grids these sets use. All LD sets have them, but they definitely come in many different degrees of badness.
(F2) Floating Black Levels: these are nasty little artifacts that crop up in dark and/or high contrast scenes where the black level starts out quite deep, but then when some content is displayed, the black levels “bounce†or float to a visibly higher level, either in the vicinity of the content or sometimes across the entire screen.
(F3) Dimming Plate “Cross-talkâ€: (I owe this term to the exceedingly knowledgeable xrox from AVS) this occurs when the local dimming function between two adjacent plates/clusters calls for different levels of light output, which is much of the time, of course. In order to avoid having one cluster look noticeably brighter than the other and in order to keep the bright bits bright and the dark bits dark within a single cluster, the sets have a kind of “smoothing function†whereby the light levels between the two plates will get equalized a bit. This yields “cross-talk†between the two plates, where they sort of strike a compromise with each other in their light levels. One very unpleasant result of this is that shadow detail in very dark scenes and dim stars in a star field will get dimmed more than they should and, in extreme cases, disappear altogether. I spent MONTHS fixing this on my 81F. It was an enormous fight (from which I learned most of what I now know). Another result of the cross-talk problem is that an image with lots of shadow information will look muddy and indistinct when it ought to be sharp and clear. At least on my 81F, that problem has proven to be largely uncorrectable. Dommage.
(F4) Off-axis Viewing: All LCDs have trouble maintaining a faithful, accurate image when viewed from off-axis, but the LED sets are much worse than the CCFLs in this regard. It’s one of the biggest enduring hurdles the manufacturers have faced. I have not seen anyone overcome it.
(F5) Motion Resolution: I think we all know what this is.
(F6) Shiny vs. Matte Screens: See F5.
(F7) Color Accuracy: See F5 and F6.
(F8) Cute Bezels: Watch me not care. The screens are all I ever look at. (I'm a guy! Waddaya want??)
THE TESTS
We looked at a bunch of calibration patterns, of course, plus I got Ed to put in a DVD of The Empire Strikes Back because that disc is great for finding floating black levels, dimming plate cross-talk, contrast problems, and halos. In no particular order, here is what I saw:
(1a) SAMSUNG GOOD: The Samsung does a consistently good job. Motion resolution on the Samsung was excellent. To me, it did the best of all the LCDs in this category. Off-axis viewing was much better than on the LG (but see below). It also calibrated to pretty respectable numbers, especially in Movie mode, which is the only mode I use on my 81F. Finally, the Samsung upscaled standard DVDs substantially better than the other two. None of them were rock stars for this, but the Samsung certainly was more respectable. It never approached the capabilities of a good DVD and/or BluRay player, though. I think the lesson here is that upscaling should never be done on the t.v. unless absolutely necessary. Do all upscaling on the input source. Of course, this also means, then, that Samsung’s stronger performance in this category becomes moot.
(1b) SAMSUNG BAD: It seems a bit hard to calibrate, it does have noticeable flashlighting/uneven backlighting (I hate that), and if you let Samsung control your black levels, you get floating blacks. Also, the colors were a bit off and I guess Ed was never able to bring them fully in line. Finally, I have a really big bone to pick with Samsung regarding the build quality of all their televisions. They put in all sorts of sexy electronics and fancy new technology, to be sure, but they seem to put these things together with toothpaste and prayers to the Virgin Mary. Example: on my 81F, I can press on the screen lightly (like when cleaning it), move my hand back and forth, and the screen will move back and forth a small amount with my hand. THE STUPID BASTARDS DID NOT EVEN PROPERLY BOLT OR GLUE THE SCREEN TO THE REST OF THE SET!!! Similar complaints about the build quality of these edge-lit sets are in the air, as apparently the bezels and screens are again separated from each other (as mine are) and pulling on the bezel a bit can be disastrous (like, as in LEDs and/or the little reflecting mirrors start falling out of place). I have not confirmed this last bit, mind you—it’s just something a guy visiting the shop mentioned. Independent confirmation or disconfirmation from trustworthy sources would be good here. I guess my biggest turn-off was the flashlights in the corners on the t.v. For whatever reason, those things really annoy me (plus it’s another clear build quality issue).
(2a) TOSHIBA GOOD: The Toshiba has some remarkable strengths that it brings to the table. The Dynalight feature seems very gimmicky when first enabled, but once Ed calibrated around it, it really did perform surprisingly well. As with all things, though, there is a price (see 2b). Black levels on the Toshiba were extraordinary when calibrated with Dynalight, especially during daylight, as this set has a clear shiny panel, not a matte one. Also, motion resolution on the Toshiba is discernibly better than on the LG. The difference is not huge, or anything, but it definitely is there. I checked this when both had all their motion correcting functions on and when both had all such functions off. The Toshiba just held up better here. Also, halos on the Toshiba were less noticeable than on the LG, especially in dark scenes and when viewed from off axis. This surprised me because my hope had been that the larger number of zones would cause the LG to shine in this category (no pun intended). Not so. Minimizing halos is a function of BOTH hardware AND software, and I guess the Toshiba’s software saved its butt here.
(2b) TOSHIBA BAD: Stupid Dynalight…guhh. It uncorrectably screwed up the picture at bright levels and in mixed scenes that were predominantly dark. Specifically, the colors on the “Dynalight-enriched†Toshiba got badly washed out in all bright scenes (this is why I hate dynamic contrast—it seriously screws up your image…). Turning off the Dynalight immediately brought the colors back in line, but then, of course, its black levels shot back up to where there were during the original shootout. Also, the Dynalight feature punched the contrast way too high on the opening star field in The Empire Strikes Back. With the Dynalight on, the Toshiba simply was not able to deliver an accurate image here. Also, there were noticeable halos on the Toshiba, though they were not pronounced and certainly far from offensive to me. I saw more of them when the set was displaying very high-contrast information, whereas in dark scenes with less contrast, they were pretty tolerable. My biggest quarrel was with all the problems generated by the Dynalight. Ed mentioned that the Dynalight function might be integral to local dimming on the Regzas, which would suck. One should be able to enable local dimming without also enabling a bunch of dynamic contrast crap. I can do that very easily on my 81F. Finally, calibration on it did not yield the sorts of numbers I like to see, particularly in the gray scale and the gamma. I can get the gray scale delta errors to below two across the board on my 81F, plus I can get the mean RGB gamma to sit consistently within 2.2 and 2.3 (this is with 20-point measures I took with HCFR). The Regza, after calibration, had some deltas over six. Grrrr. Work on this, Toshiba! Also, upscaling standard DVD material on this set was an unpleasant experience. Moving text shimmered and jumped around like it was on amphetamines. As most good BluRay, HDDVD and DVD players have considerably stronger upscaling capabilities, I would do all upscaling there. My biggest turn-off with the Regza was how the Dynalight feature uncorrectably screwed up contrast and colors. Were I to get the Regza, I’m not sure I would use Dynalight.
(3a) LG GOOD: I honestly think LG designed these sets with the conscious goal of freaking out some calibrators, because it just calibrates so well. The gray scale, the colors in general, and the gamma all did very, very well in my opinion. Of all the LCDs I have seen (anywhere), the LH90 has the best color accuracy and gamma (for those who don’t know, accurate gamma levels are very important for getting a nice, three-dimensional, lifelike image). It also seems to hold its colors better than other LED sets when viewed off axis (I suspect LG’s decision to use S-IPS panels has something to do with this). I saw almost no information loss from cross-talk and shadow details were about as good as they can be on an LCD (this was my happiest surprise of all). I also never saw any local dimming zone flash the way someone here reported for the LG90 (was that you, Flyers?). On all the material I saw, the zones behaved responsibly. When viewed in an environment with controlled lighting, the black levels on the LG really were very, very good. This took a hit when the room was bright, but I think that is to be expected when dealing with matte screens (see below). It remains to be seen if that can be fixed by generating a different list of settings for daytime viewing in a many-windowed room. When viewing the set from dead on, there are discernible halos, especially in dark scenes with low contrast, but they are not pronounced. The story changes the instant one goes at all off axis (see below). I was able to catch just a few floating blacks at times, too, but those, like the halos, were quite muted. I’m confident they can be calibrated out. Motion resolution on the LG worked just fine after all the frame interpolation and backlight strobing software was turned off. I did not see excessive motion blur or judder even when those functions were all off, and given the reputation of IPS panels, I kind of expected to see at least some trouble there. So that was a nice surprise.
(3b) LG BAD: First and foremost, I would never, ever use the frame interpolation and backlight strobing functions on this television. They, um…they suck. {EDIT: As I only saw the set with both functions on or both off, I had better point out that it may well be that Real Cinema can work alright without the TruMotion function. At any rate, I would never use the TruMotion function, though. Blecch!} They really suck. My personal advice is not to use them at all. If motion correction were a high priority for me, I’d go with the Samsung for sure. I suspect LG’s decision to use S-IPS panels also has something to do with what I saw here, as motion is reputed to be one of their relative weaknesses. Second, because the screen has a matte finish the image does visually wash out in a bright room (though the actual measured numbers stay rock solid). If calibration can correct for this without too much violence to image fidelity, though, then things will be easy, as the set has two distinct fully customizable ISF settings that the user can switch between at will. Still, I do wish they would have gone with a clear panel, not matte. The clear panels just work better, and hang the worries about reflections. Upscaling was not something the LG enjoyed having to do, much as I saw with the Toshiba. It was pretty cheap stuff. Like I said above, I also ran into several important issues with the local dimming, with the halos being the worst offenders (I’ll get to them in a minute). Floating blacks piss me off, so when I finally saw a few, my eyes locked onto them hard. Especially if one is going to watch this set in a dark room, I strongly advocate calibrating these things out. Doing it is time-consuming, but not especially harmful to the picture (if it’s done properly). As one of you noted, I spent many months figuring out how to do just that on my Samsung 81F over on a calibration thread I started at AVS. I might archive some of that information and repost bits of it on a calibration thread here if I get a new locally dimmed set sometime soon. Now, about those halos…I’m really pissed off to see them. 240 zones and ~3,000 LEDs really ought to be enough to get rid of these things at least at the immediately visible level. Unfortunately, it looks like LG got inexplicably cheap when it came to the software for this. (Or maybe they just used cheap-o, low-bit LEDs??? Idunno.) Because the set has so many zones, 99% of the time (that is totally a made-up number, so don’t get on my case about it) the halos are actually not bad at all when viewing the set from dead on, but going just a little bit off axis reveals surprisingly large and pronounced ones. This means LG got stingy when it came to the implementation of the local dimming and I think they easily could have improved this. I know every manufacturer has to cut costs somewhere, but in my opinion this was the wrong place to do it. I would much rather have gotten rid of a bunch of the unnecessary inputs that no one ever uses anymore. To my mind, that would be the ideal place to trim some fat and cut costs. Assuming the frame interpolation et al is kept off, the halos were what pissed me off the most about the LH90.
GENERAL STUFF:
Both locally dimmed sets are capable of getting absurdly bright without giving up their dark blacks; this is a basic strength of local dimming technology and one of the first things that drew me to it. Because of this, they make great choices for people who want to watch t.v. in a bright room. For those of us who watch in a more carefully light-controlled environment, as I do, they still bring an awful lot to the table. I think the Toshiba would be the best choice for buyers who aren’t pathological about absolute image fidelity and who just want a really good, strong picture that they can watch even in a very brightly lit room. In an environment like that with criteria like that, the Toshiba will best the LG for sure. I think the LG would be the best choice for someone who is quite fussy about image fidelity and is willing to do what is necessary to control ambient light (personally, I’ve found curtains do a great job here—it's not rocket science). There is no question in my mind that the Samsung is the best LCD for people who are very sensitive to motion blur. Samsung’s software here is miles beyond what Toshiba and, especially, LG are offering.
When all is said and done, I still think the LH90 is the best of the three LCDs. Its color accuracy and its calibration potential in general have to be seen to be believed. LG has done an outstanding job on this. The black levels, too, are just fantastic and I sorely wish LG had chosen to go with a nice clear panel that could properly show off those extraordinary black levels (and the colors). Perhaps in their next iteration, they will. Were it not for the matte screen and the halos, I would have recommended the LG without reservation. As it is, I still think it’s the best of the three, but not by leaps and bounds.
So…am I going to buy an LH90? Not yet, no. Samsung’s new inheritor to the A950 is coming out in a few weeks and if it has the 240 zones that I regard as the needed minimum for local dimming to work as it should, I’ve little doubt it will beat the socks off Toshiba and LG. If, however, Samsung remains cheap, stingy and lame about this, as they have been so far, then I suspect their claim to the performance crown in locally dimmed sets will be in serious jeopardy. One thing is for sure: with the kind of calibration controls I can get from an LH90 now, Samsung’s new 8500s had better have at least as impressive of a set of controls if they want to get my money from me. Now that I’ve seen what proper controls look like and what they can do, there’s no way I’m settling for less (especially at the price points Samsung generally sets).
Last point: the prices on the Toshibas and LGs just plain beat the crap out of Samsung and Sony. If one is on a budget that is at all tight, these two televisions deserve very, very careful consideration. Once the new Samsung comes out, there’s a good chance prices on these two will drop even further, too, so I’d keep an eye on them.
As soon as Robert gets his first new 8500s, I’ll be driving like a man possessed right back up to Scarsdale to get a look at one (and hopefully Ed will bring all his outrageously cool gear, too).
Yours,
David
.
