AVS Forum banner
Status
Not open for further replies.
1 - 15 of 15 Posts

·
Registered
Joined
·
4,654 Posts
Discussion Starter · #1 ·
With Sony and Sanyo coming out with 1920 x 1080 projectors, I'm thinking there are a lot more pixels to go dead, so we should expect more dead pixels. But, they should be so small that you can't see them while viewing.


My HD20/PLV-70 has two dead pixels. One is stuck on blue and one is stuck on red. I only see them when I walk past a black screen on the way to the John. I presume the HD chips' pixels will be substantially smaller.


How much smaller? Wouldn't they be that % less visible? Perhaps we could tolerate more.


The answer of those with "perfection is required whether I can see it or not" attitudes goes without saying. But what about the rest of us?



This makes me hopeful that the extra cost of the higher resolution chip might be offset somewhat by fewer actual returns and discards.


Am I dreaming?
 

·
Registered
Joined
·
11,020 Posts
The smaller the pixels, the harder they will be to see. But we both know that won't make an iota of difference in how much they bother people Joe :).
 

·
Registered
Joined
·
32,172 Posts
The answer is: by and large, yes. Tiny dead pixels are very hard to even notice. Tiny stuck pixels might still be annoying, however.


Also, perfectionists here will never be happy with any dead pixels in their rig.
 

·
Premium Member
Joined
·
24,986 Posts
I agree with you Joe. This could increase yields tremendously. I wouldn't be surprised if they could actually sell 1920x1080 panels cheaper than 1280x720 to those willing to accept the(invisible) dead pixels.


Hmm, if they get up to 3840x2160, they could group pixels in 2x2 arrays (also saving on processing power), giving 1920x1080 addressable. Then when a pixel is bad, it's only 1/4 of a visible pixel. Yields could skyrocket - hundreds or even thousands of bad pixels would not be noticeable

(1000 = 0.012% of 3840x2160=8,294,400).
 

·
Registered
Joined
·
4,654 Posts
Discussion Starter · #5 ·
Noah, if your not pulling my leg, then I'm pretty excited.


Q, if they have to move closer than their viewing position to see them, then they should have their dosage increased.


And I presume this chip would be one to one mapping of HD with no scaling at all. Right? I think I'm getting a chubby.
 

·
Registered
Joined
·
6,092 Posts
Joe,


I think this is exactly what they are planning to do.


I'm with Rogo. They might stop covering stuck-off pixels because they are so hard to see. Stuck-on might still be visible and thus fair to whine about.


-Mr. Wigggles
 

·
Registered
Joined
·
1,501 Posts
Some time back Joe Kane wrote about this in a Widescreen Review article. I recall that his statement concerned not getting excited about an increase to 1080P, waiting for a chip density that had redundancy. Makes sense, but at the time I thoght those chips were too far away. Wrong again, Art.
 

·
Registered
Joined
·
1,474 Posts
Quote:
Hmm, if they get up to 3840x2160, they could group pixels in 2x2 arrays (also saving on processing power), giving 1920x1080 addressable. Then when a pixel is bad, it's only 1/4 of a visible pixel. Yields could skyrocket - hundreds or even thousands of bad pixels would not be noticeable

(1000 = 0.012% of 3840x2160=8,294,400). [/b]
Oh Noah,


you couldn't be more wrong on this point. To raise the number of pixels, you have to shrink the die. This creates two problems:


1. You increase screen door due to increased dead space which is typical at a minimal possible size now.


2. The theoretical yields of an die shrink, decrease exponentially. (see "Computer Organization and design" by Morgan Kaufmann.) So a die shrink of 50% (increase in pixels by 4 (2 x 2 in each direction) might yield 1/8th of good useable wafers.)


~D
 

·
Registered
Joined
·
365 Posts
Quote:
Originally posted by noah katz
I agree with you Joe. This could increase yields tremendously. I wouldn't be surprised if they could actually sell 1920x1080 panels cheaper than 1280x720 to those willing to accept the(invisible) dead pixels.


Hmm, if they get up to 3840x2160, they could group pixels in 2x2 arrays (also saving on processing power), giving 1920x1080 addressable. Then when a pixel is bad, it's only 1/4 of a visible pixel. Yields could skyrocket - hundreds or even thousands of bad pixels would not be noticeable

(1000 = 0.012% of 3840x2160=8,294,400).
This technique is called "binning" and is used by astronomers to increase sensitivity of CCD chips during imaging. The CCD cameras address full resolution, 2x2, and 3x3 binning through the control software. I do a bit of it, and it's quite fun.


Todd/Indy
 

·
Registered
Joined
·
1,180 Posts
"You increase screen door due to increased dead space which is typical at a minimal possible size now"


So you are saying that 1024x768 on a .7" chip has more screendoor than on a .9" chip?
 

·
Premium Member
Joined
·
24,986 Posts
D,


"To raise the number of pixels, you have to shrink the die."


Why is this? Why can't you leave the die (not sure about the terminology, I mean it as the size of the substrate on which the chip is built) the same size and make the pixels smaller?


Thanks
 

·
Registered
Joined
·
1,474 Posts
David,

It all depends on the manufacturing process and optics. If the .7 and the .9 are using the same process, but with smaller twisted crystal, then yes, the .7 will have greater screen door. There's a similar problem that arises when raising the resolution on computer LCD screens.


Noah,

Increasing the die increases production cost in terms of material. It also raises the possibility for more defects. Using a standard pouison distribution of large defects (impurities) in the matrix, the chance of hitting in impurity exponential increases with die size. On the flip size, shrinking gate size raises the risk of an impurity making a circuit path inoperable. (Smaller circuits are more subseptable to impurities.) There's also signal bleed issues, current rush etc...


(Also covered in chapters 1&2 of "Computer Organization and Design: The Hardware/Software Interface" by Morgan/Kaufmann.)


I believe another reasons why they don't make many LCD's larger than 1.2" as it has a tendency to slow down the matrix affecting response time.
 

·
Premium Member
Joined
·
24,986 Posts
D,


I was asking about keeping the die the SAME size, not making it smaller or bigger.


Point taken, though, about smaller features being more susceptible to impurites
 

·
Registered
Joined
·
692 Posts
I think it's interesting, because they're close to the lower limit for chip size for optical reasons. Any feature size reductions now will increase resolution. Screendoor is likely to increase on a denser chip, but would that actually be visible by a human at any reasonable viewing distance? While smaller features may be more susceptible to defects, the redundancy of the 2x2 matrices would probably totally negate that. I think that yields would improve. As the silicon real estate would stay the same, the final costs should decline.
 
1 - 15 of 15 Posts
Status
Not open for further replies.
Top