2007: 120Hz, Deep Color, Extreme CR and 1080p

[Graham]

For us in the world time forgot, sony's new D series seems to fit the bill quite nicely. It's not 1080, so it's actually a reasonable price, and it doesn't have ridiculous dynamic contrast claims (*cough* samsung *cough*). Plus it's 100hz.
I'm probably going to buy the 40" when it hits in a couple of weeks.

Toshiba are doing very well over here, but their contrast/black performance is simply pathetic. (we only get the lower end models). Seriously, 800:1 contrast on a $6000 highest-end TV?

There is a giant gap over here between 720 and 1080 sets, literally $3500->$6500 gap for 40". Absolutely nothing in between. Combined with the lack of content it makes 1080 an utter and total waste of money. Of course people still buy the damn things.

 

[pascal]

This is a confusing time. We are in a transition time with some good "old" LCDs (like my sony 2006 s-series), some new ones with 100+Hz and 1600+ contrast (Sony D), few with antireflective panel (Samsung M86/87), and some with 1080p support. Samsung announced new LED backlight HDTV. Sony announced new 1080p, 120Hz, 10bit color, 2000 contrast.

I visited this week a store were they had about 15 HDTVs in line with a single source and each remote available to the customer test it. All kinds XBR, M81, etc...
Its was customer nightmare of calibration. I will visit it again, but this time prepared with a calibration cookbook

 

[AlexV]

i don't know about lcds not having surpassed crts in quality.

tubes weaken rather fastly, and that alone is why i prefer lcds, even with slight bl bleeding.

what they need to do though, is move up to 48 bit color output. that would display the whole shader model 3 frame buffer and that also include the ps3's whole frame buffer.

we've been stuck at 24 bit forever.

the 3do and jaguar came out in 93 and they did 24 bit color output.

Are you mad?

 

[Mintmaster]

what they need to do though, is move up to 48 bit color output. that would display the whole shader model 3 frame buffer and that also include the ps3's whole frame buffer.

LCD's can't get even close to handling the range of 24-bit. 8 bits per channel means white has a value of 255 and the darkest non-black grey is 1. Throw in a gamma power of 2.2, and you need 200,000:1 contrast to display a greyscale correctly.

That means with proper gamma the best LCDs today (~1300:1 measured) can't resolve below a grey value of 10 on a 8-bit scale. There's no need for higher bit depth, especially for LCDs.

 

[pakotlar]

Are you mad?

Well with, a name like ballsweat, I don't know how much we can expect.

 

[pakotlar]

LCD's can't get even close to handling the range of 24-bit. 8 bits per channel means white has a value of 255 and the darkest non-black grey is 1. Throw in a gamma power of 2.2, and you need 200,000:1 contrast to display a greyscale correctly.

That means with proper gamma the best LCDs today (~1300:1 measured) can't resolve below a grey value of 10 on a 8-bit scale. There's no need for higher bit depth, especially for LCDs.

LCD's are somewhat poo for black levels though, so I don't think the best measure. Also, I'm not sure how it filters into ur calculations, but 1300:1 ANSI (which except for Sony Bravia XBR3, NO lcd has reached quite yet) on a Bravia XBR3 still reveals worse shadow detail than a Panny 8gen plasma with a ~1000:1 ANSI rating. Afaik.

edit: I realize that the above doesn't have much to do with theoretical resolution power. Do you have a link to info that correlates value scale (grey value 10 on what standard scale) with ANSI contrast ratio?

 

[mrcorbo]

It's been a while since I was last in this thread. I'm actually off the Toshiba now and onto the Samsung 81 series (pending reviews, of course). They had me at LED backlighting and local dimming. I'm looking at this in combination with the Onkyo 875 as my base HT platform for the next several years. It's been quite a while since this much cool tech was coming out or coming into its own at the same time. It's a good time to be an HT nut.

 

[Mintmaster]

LCD's are somewhat poo for black levels though, so I don't think the best measure. Also, I'm not sure how it filters into ur calculations, but 1300:1 ANSI (which except for Sony Bravia XBR3, NO lcd has reached quite yet) on a Bravia XBR3 still reveals worse shadow detail than a Panny 8gen plasma with a ~1000:1 ANSI rating. Afaik.

Yeah, when you have bad contrast you're sort of stuck between a rock and a hard place. If you want to maximize the accuracy of as many colours and shades as possible, then just going as black as possible for the shades you can't reach is optimal. Of course, that leads to black crush, so you have no difference in luminance between dark shades. If you want to maintain shadow detail (i.e. show a visible difference between shades), you have to boost the luminance of dark greys to form a ramp, forcing some colors further away from what you're capable of.

In an arbitrary example, imagine trying to get shades from 0-100, but your display can only go as low as 10. Now with inputs 0-9, you can output 10, so that 10-100 come out perfectly. Of course, now you see no difference between shades 9 and below. To fix this, you can map 0 --> 10, 1 --> 10.5, 2 --> 11, .... , 19 --> 19.5, and from 20 and above you map perfectly. Now you can see the difference between low inputs, giving you shadow detail, but inputs 10-19 are brighter than they should be even though your display can show them correctly.

Each manufacturer will place a different value on each solution. Crushing blacks and losing shadow detail gives you more subjective contrast in a bright image.

edit: I realize that the above doesn't have much to do with theoretical resolution power. Do you have a link to info that correlates value scale (grey value 10 on what standard scale) with ANSI contrast ratio?

Oh I was talking about 10 out of 255 to be consistent with my previous example. Usually they talk about IRE levels (0-100) which corresponds to the voltage going through the video signal. It gets kind of messy too with IRE7.5 being black for NTSC, and clamps on the 0-255 range too depending on how it's mapped to the video signal (i.e. 0-16 is black and 235-255 is black).

I like to look at it as a fraction to avoid this nonsense, with 0.0 being black and 1.0 being white. Gamma is just the relationship output=input^2.2. So the luminance of a signal at 0.04 should be 0.00084, i.e. 1/1200th of full output. If you use all 8-bits per channel in the active range of a display, then a value of 10 out of 255 is a display input fraction of 0.04. That's what I was trying to illustrate.

I'm not quite sure what you mean by a chart from grey scale to ANSI contrast, but hopefully this post helps. I was just saying that if you really want to accurately display 1-255 (i.e. 0.004 to 1.0) according to the standard of 2.2 gamma, then you need a display with 200,000:1 contrast ratio.

LCDs still have two orders of magnitude to go before they legitimately need a higher bit depth in the signal. I know HDR is all the rage, but 8-bit and 10-bit has a lot more range than people give it credit for, at least as a final image format.

 

[dizietsma]

It's been a while since I was last in this thread. I'm actually off the Toshiba now and onto the Samsung 81 series (pending reviews, of course). They had me at LED backlighting and local dimming. I'm looking at this in combination with the Onkyo 875 as my base HT platform for the next several years. It's been quite a while since this much cool tech was coming out or coming into its own at the same time. It's a good time to be an HT nut.

Just looked at the spec of those and they do look good with 1080p and 100/120Hz. I assume the 100 000 contrast ratio is some sort of marketing or "corrected" value to impress?

Shame they do not do a 37 inch version, 40 inches is a bit big for my smallish UK tv room, also the cost upsets me slightly, so my wife would be very upset. That's not even a UK price as well I have just realised.

 

[Graham]

OK.
Here is the question that is keeping me up at night:

I just ordered a 40" D-Series sony LCD. It will arrive in ~2 weeks. I paid $2900nz for it, plus $ for a cabinet/rack combo (retail on the tv is $3700). Now... The guy - the bastard - made me the following offer.
$1200 more for a 40" X-Series, which retails at $6000-6500.

As far as I can tell, the D-Series is better in most ways (especially 100hz), it's just not 1080 and not so pretty. And maybe doesn't have quite as good a processing engine...

So whats the recommendation? I'm leaning towards staying with the D.

 

[mrcorbo]

OK.
Here is the question that is keeping me up at night:

I just ordered a 40" D-Series sony LCD. It will arrive in ~2 weeks. I paid $2900nz for it, plus $ for a cabinet/rack combo (retail on the tv is $3700). Now... The guy - the bastard - made me the following offer.
$1200 more for a 40" X-Series, which retails at $6000-6500.

As far as I can tell, the D-Series is better in most ways (especially 100hz), it's just not 1080 and not so pretty. And maybe doesn't have quite as good a processing engine...

So whats the recommendation? I'm leaning towards staying with the D.

The first thing I'd suggest is have a look at this graph and match up your normal viewing distance with the size of your screen to determine whether you will even be able to resolve all of the detail in a 1080p display.

 

[Graham]

The first thing I'd suggest is have a look at this graph and match up your normal viewing distance with the size of your screen to determine whether you will even be able to resolve all of the detail in a 1080p display.

That was pretty much what I was expecting. I'm pretty much slap bang in the middle of the 720 range.

 

[pakotlar]

Yeah, when you have bad contrast you're sort of stuck between a rock and a hard place. If you want to maximize the accuracy of as many colours and shades as possible, then just going as black as possible for the shades you can't reach is optimal. Of course, that leads to black crush, so you have no difference in luminance between dark shades. If you want to maintain shadow detail (i.e. show a visible difference between shades), you have to boost the luminance of dark greys to form a ramp, forcing some colors further away from what you're capable of.

In an arbitrary example, imagine trying to get shades from 0-100, but your display can only go as low as 10. Now with inputs 0-9, you can output 10, so that 10-100 come out perfectly. Of course, now you see no difference between shades 9 and below. To fix this, you can map 0 --> 10, 1 --> 10.5, 2 --> 11, .... , 19 --> 19.5, and from 20 and above you map perfectly. Now you can see the difference between low inputs, giving you shadow detail, but inputs 10-19 are brighter than they should be even though your display can show them correctly.

Each manufacturer will place a different value on each solution. Crushing blacks and losing shadow detail gives you more subjective contrast in a bright image.

Oh I was talking about 10 out of 255 to be consistent with my previous example. Usually they talk about IRE levels (0-100) which corresponds to the voltage going through the video signal. It gets kind of messy too with IRE7.5 being black for NTSC, and clamps on the 0-255 range too depending on how it's mapped to the video signal (i.e. 0-16 is black and 235-255 is black).

I like to look at it as a fraction to avoid this nonsense, with 0.0 being black and 1.0 being white. Gamma is just the relationship output=input^2.2. So the luminance of a signal at 0.04 should be 0.00084, i.e. 1/1200th of full output. If you use all 8-bits per channel in the active range of a display, then a value of 10 out of 255 is a display input fraction of 0.04. That's what I was trying to illustrate.

I'm not quite sure what you mean by a chart from grey scale to ANSI contrast, but hopefully this post helps. I was just saying that if you really want to accurately display 1-255 (i.e. 0.004 to 1.0) according to the standard of 2.2 gamma, then you need a display with 200,000:1 contrast ratio.

LCDs still have two orders of magnitude to go before they legitimately need a higher bit depth in the signal. I know HDR is all the rage, but 8-bit and 10-bit has a lot more range than people give it credit for, at least as a final image format.

Thanks for the post, I learned a lot. If I'm getting this correctly, the reason that we get so much black crush on LCD's is because in order to display a palpitable image in dark scenes, the display has to map (as per your example) 0-19 to 10-19, therefore not only are you losing .5 bits of information (is that correct), but your image is going to obviously be brighter as well. Now my guess is that most LCD manufacturers are ok sacrificing that .5 step of detail for darker blacks. In that case they can claim a higher contrast ratio because they can reach down to their min luminance state for coordinate 10. If they mapped 0-9 to 10-19, then their whites wouldn't come out as white, and their blacks, though they would retain full shadow detail would appear significantly bleached. So they are making a tradeoff, which portion of the spectrum to compress.

Of course this becomes easier as their min luminance states decrease.

I have a question, why couldn't they expand their range by adding say 2 additional bits for luminance? This would mean that they could display whiter whites if they mapped 80-100 to 81-90? Wouldn't that mean that they still avoid "white crush", map out their 0-9 range to 10-19, and the only thing they are sacrificing is luminance on the upper range, and making their blacks appear more "grey" for the prize of perfect shadow detail? The appearance of the blacks would get better over time as min luminance decreased, and brightness has never been an issue to begin with. Basically I'm saying why not increase dynamic range for the inputs, so that the display has more choice as far as mapping its voltage steps to input^2.2. So instead of input being 0-9, it would be say 0-19 effectively, and 100 would end up at 200, and the manufacturer would have finer control about how much it has to sacrifice shadow detail for blacker blacks. I'm guessing that this combined with LED local dimming would produce some pretty nice results.

This is all thought experiment, I've never researched this so quite possibly I'm horribly off. It's pretty interesting to me.

A lot of this is guesswork. I'm confused on the relationship between input dynamic range, gamma and how it affects output curves in relationship to voltage steps, aka why can't you just lower gamma until your black information is perfectly displayed, and how much would that crush white, and how it would effect luminance.

I'm starting to go in circles. What's IRE?

 

[Graham]

Yay my new TV finally arrives tomorrow! about time too!

 

[Color me Dan]

Will "Deep Color" enhance colors on all sources or just sources supporting those bit-depths? Are all sources capable of those bit-depths and hindered by the hardware?

 

[Mintmaster]

pakotlar, for the most part you're understanding my correctly. A few of your points look a little off, though.

therefore not only are you losing .5 bits of information (is that correct)

No, I never made any inferences about the precision. The LCD should be able to produce any light level between its max and min output. The driving mechanism is analog.

Now my guess is that most LCD manufacturers are ok sacrificing that .5 step of detail for darker blacks. In that case they can claim a higher contrast ratio because they can reach down to their min luminance state for coordinate 10. If they mapped 0-9 to 10-19, then their whites wouldn't come out as white, and their blacks, though they would retain full shadow detail would appear significantly bleached. So they are making a tradeoff, which portion of the spectrum to compress.

Your last sentence is right, but I think you're off on the rest. Contrast ratio in my example is 10:1, regardless of mapping (unless one does something silly).

If you mapped 0-9 to 10-19, then all the other shades would be a bit brighter than they should be, resulting in a bleached look. An input of 10 should, on a perfect display, have an output of 10. While this LCD can indeed output 10 correctly, the problem is that the input 10 must be brighter than input 9, which you already mapped to output 19. Whites would still be white, though, as there's no reason not to map 100 input to 100 output.

I have a question, why couldn't they expand their range by adding say 2 additional bits for luminance? This would mean that they could display whiter whites if they mapped 80-100 to 81-90? Wouldn't that mean that they still avoid "white crush", map out their 0-9 range to 10-19, and the only thing they are sacrificing is luminance on the upper range, and making their blacks appear more "grey" for the prize of perfect shadow detail?

Bits don't matter, it's the total range that does. The LCD in my example can output any level -- even fractional -- between 10 and 100. I don't know what you accomplish by mapping 80-100 to 81-90. Now an input of 100 is less bright than it could be, and your contrast ratio goes down to 9:1.

A lot of this is guesswork. I'm confused on the relationship between input dynamic range, gamma and how it affects output curves in relationship to voltage steps, aka why can't you just lower gamma until your black information is perfectly displayed, and how much would that crush white, and how it would effect luminance.

The problem is that the LCD display just can't shut off light completely. In my example, that 10 is simply the minimum light it can let through. A local LED backlight can help if that part of the picture doesn't need anything bright. If that part of the image had a maximum brightness of 20, then you could lower the LED to 20% and your minimum output level would be 2.

BTW, I ignored gamma in my example, and just assumed the ideal output was a linear mapping of input to output.

 

[Graham]

I finally got my TV last night.
Sony's new D-Series, 40" model.

It's been interesting...
First I'll say I was very badly disappointed by the picture when I first fired it up. It looked terrible.

So a word of warning, this set appears to need a 'burn in' before it looks good.

When I first plugged it in, the ghosting (black ghosting in particular) was so bad, that the grid in geometry wars would dull by over 50% when you moved about...
Far worse, darker black areas would ghost for about half a second... It looks utterly horrible, the guide looked something like this:

And tire walls in forza appears to shrink and expand with vertical movement.

But long story short, it seems now I've given it about 8 hours run-in - and a lot of tweaking, the response time is *way* better, still not CRT-level, but good none the less. The geometry wars grid dulls by maybe 5% at most. Although PinballFX is still a tad blurry though due to it's rapid panning movement.

Anywho. Otherwise, it's very sharp, definitely bright enough and the colours are excellent. All up it's a very good TV. The blacks accuracy seems really good too.

And it's utterly impossible to take a photo of it

Even after only ~8 hours, I'd say it gets a definite recommended from me Not that I expect that to mean much

 

[pascal]

Congratz Graham :smile:

 

[Color me Dan]

I wonder why they need the "run in". Great find on the TV btw, looks awesome.

 

[Gubbi]

I wonder why they need the "run in". Great find on the TV btw, looks awesome.

LCDs don't need it. I think the ghosting Graham experienced is from the noise reduction modes that are generally enabled (to a high degree) by default when you buy a new set.

Spatial and temporal blur really isn't needed when playing games.

For PDPs, the phosphors apparently need to be aged/burned at a similar rate until some point or you risk permanent burn-in on the screen. However most PDP manufactures do burn-in at the factory (50 hours for Pioneer and Panasonic I believe).

Cheers