Google offers at least $880M to LG display for OLED investment

Not only that, but OLED screens have perceptible low response times and burn-in (my old S3 screen has the Candy Crush UI burned in, and is also yellowish because blue LEDs burn in faster than green/red ones).

That said I suppose newer OLED screens do not have as many problems, but for a long time OLED screens were inferior to LCD screens, and were used only because they were capable of higher contrast ratios which made them much more appealing to the eye.

I had a GS3 and hated that I couldn't use it in the sun, but my GS7 is usable in the sun.

The high-end LCDs can go very bright nowadays, and HDR content means that there can be a great contrast between the bright parts and dark parts of an image.

The Samsung you point at is one of the best of last years TVs, so yes, a very good picture. Only downside is that it is edge-lit, meaning that the brightness of light objects can bleed into the surrounding areas.

Where OLED shines though is the low-end, the blacks/dark scenes, where you can go very dark right next to very light.

source: I'm looking for a new TV, so of course have to do extensive research...

I'm considering getting one of those LG 65" OLED TVs, after seeing one in-store and being completely blown away by the picture quality. Much like the GP I can't recall the last time I saw a display that actually wowed me, but watching a space shuttle take off, and bright stars as dots against a pitch black sky without the typical blue-ish backlight bleed, was just amazing. I stood there watching the same 4-5 minute video loop for what must have been half an hour or so.

But your comment now gives me pause – the picture warps over time? Do you have any sources on this so I can learn more?

Certainly it's nothing like degaussing; it's a panel of individually addressable OLEDs that do not move, not a vacuum tube.

My guess is that this is an artifact of some sort of 'motion control' or other processing the TV applies.

As with everything that should be trivial this becomes fairly non-trivial upon contact with reality.

At the most fundamental level you need three lookup tables with one entry per sub-pixel. On a 1920 x 1080 display that means a little over two million entries for each, R, G and B. In other words, six million entries.

Pixel period at 60Hz is going to be in the sub-10 ns range. This means you would need lookup table memory that can be randomly indexed and respond within less than 10 ns in order to be able to grab a correction value for each and every subpixel being displayed. Due to latency and other realities of DDR SDRAM this is impossible. DDR SDRAM does far better for burst access of sequential blocks of data within a row. It does not do well for fully random access of single memory locations. DDR4 can't do full random single location access at these data rates.

Wikipedia article on SDRAM basics:

https://en.wikipedia.org/wiki/Synchronous_dynamic_random-acc...

Other memory technologies (QDR SRAM and RLD RAM) do exist that can accomplish the above requirement. They are extremely expensive and so they are not commonly seen in products such as televisions.

Yes, custom chips or an FPGA would be required in order to manage the process.

Even then, the problem isn't resolved.

The next issue is one of differential quantization. Each subpixel is either going to be driven by an 8 or 10 bit data stream. I'll use 8 bits for this example.

Say a blue sub-pixel is 25% "worn out". Assuming a linear transfer function (it isn't) this means that in order for all other blue sub-pixels on the entire screen need to be limited to a max of 75% output in order to not be brighter than that one blue sub-pixel.

Yet, the problem doesn't end there.

We also need to knock down all red and green pixels to limit them to a maximum of 75% output. Why? Because, in order to make white light you now have to compensate all color channels, otherwise you'd have too much red and green against a weak blue output.

It very quickly becomes a lowest common denominator problem.

Going back to quantization. When we started with a fresh OLED display we had 256 levels from black to white. Now we are down to 192 due to the necessary compensation.

But, that's not all! There's more!

Let's say green sub-pixels on the opposite side of the screen have now "worn out" down to 60% max output. Now the entire screen needs to be brought down to this new limit. And total quantization levels are now down to 153.

The practical result of losing quantization levels is image degradation due to an inability to smoothly map input to output.

On first inspection this might seem to indicate that a high speed three channel lookup table isn't really necessary. If we trim the output of all pixels down to the worst case pixel, well, just adjust the entire display. Except this won't work. If you bring all pixels down to 60% you are also bringing down the bad pixels that could not produce full light output. The end results is that, yes, you do need a high speed lookup table in order to leave bad pixels alone (not quite) and adjust good pixels to match the bad pixels.

And yet, we are not done with this. There's a fundamental issue that hasn't been addressed, even after throwing all that technology at it: Calibration.

In order to be able to populate three lookup tables (red, green and blue) we need to be able to measure wear at the sub-pixel level. This is extremely time consuming and most definitely well outside the realm of consumers. At the factory? Sure. At home? Not likely. I can see gross measurement. For example, manually measure sixteen zones and accept the fact that the image might have visible color differences due to this. Or, maybe, use a high resolution sensor to measure a few hundred zones. And this is when, to reach for the popular saying, you start to find yourself up to your ass in alligators when all you wanted to do was drain the swamp.

Funny how things work.

EDIT: Forgot to add. A lifetime "on" time counter won't work. The degradation of OLED pixels isn't purely a function of time. You can end-up with pixels that have been active for exactly the same time and one is significantly brighter than the other. You have to measure.

What is special about this TV?

That's interesting because I have an Oculus DK2 with an OLED panel (Galaxy Note 3 panel) and I've experienced 0 issues in what many consider the most sensitive display environment.

800Hz PWM can cause nausea?

Is there any objective research supporting this? I haven't heard of anyone doing simple blind testing or controlling for other factors.

Please specify what you are measuring, because the frequencies you are quoting are far outside human perception limits. These also don't match the numbers being quoted by the VR guys.

I'm extremely sensitive to refresh rates. I used to have two gigantic ViewSonic CRT monitors on my desk along with special Matrox video cards so that I could have high refresh rates.

60Hz gave me whacking migraines. 72Hz was fatiguing but didn't give me migraines. 80Hz+ and I was just fine.

I haven't had a migraine due to refresh rates on digital displays ever. Not even for incredibly slow displays.

Apple designs their own SoCs, but they do not manufacture the chips themselves. Fabrication is contracted out to Samsung and TSMC.

But designing and manufacturing are different.

Best-in-class or near best displays are supporting and necessary but not sufficient for Apple. So acquisition doesn't make sense. The money is to guarantee supply for their huge product volumes and also sometimes a competitive measure to lock up supply so competitors can't get at it.

On the other hand, I would not expect Apple to farm out their software, OS, marketing, industrial design any time soon. Unlike the supporting value of LCDs these are all differentiators.