According to TI the vast majority of early production failures were due to contamination during fabrication. As that issue was dealt with they moved on to the next item - contamination due to poor packaging by the OEM.

Other factors come into play much lower down the charts, but stuck or broken hinges are not, according to TI, a factor. I suspect the failures you saw were probably not hinge related just based on their published reports. If you looked at the dead mirror with a transmission electron microscope and determined the hinge was malfunctioning then obviously the above would not apply.

I've watched several dozen DLP systems and never saw a stuck mirror, in some cases on systems with thousands of hours on them. Perhaps your experience is not a representative sample? I really doubt the manufacturers (OEMs like Samsung, Mitsubishi and Zenith) would jump onto the worst, least reliable and most expensive technology with both feet. I expect the expert engineers and scientists in the appropriate fields at the companies in question have evaluated the competing technologies and have rational defensable reasons to recommend DMD to their employer. I have no basis to second guess that level of expertise. Do you?

Please understand - I don't intend any of the things I say regarding the various technologies as a personal slam. I mention this because you seem to take some of this very personally.

I do understand roughly how the GLV stuff as published by SL works and their literature even states that the scanning beam is moved across the screen in a linear fashion. I don't understand what you're so uptight about. It is obvious the spot has finite size and must be either modulated in the analog domain or switched. SL seems to recommend an analog approach, which would eliminate the potential issues with moire and sub pixelation but would reintroduce all the possible old analog linearity issues with the additional caveat of a dependency on the long term consisency of a mechanical device instead of glowing phosphor. This is probably OK, but is an area of concern at the very least.

The price numbers are obviously pulled from thin air, I was merely trying to illustrate that as MEMS get cheaper to make the distribution of costs in the 1D vs. 2D approaches means the 1D approach is likely to benefit less and thus not scale as well with the technology. I tried it once without an example and you didn't understand, so I tried to be helpful and make a simple example. Anyone with a background in manufacturing or engineering would understand this as almost second nature.

As for convergence I think it's pretty clear some 3 chip systems have need of adjustments and I suspect that includes theater 3 chip DLP. As for how that scales to HT 3 chip DLP I wouldn't care to guess. I suspect the LCD we have has inside a set of screws for the assembly and service folks to adjust. The fact that some LCDs actually need external or motorized adjustments indicates to me that in reality they all (or at least most) probably have internal adjustments. Now it is true that routine convergence adjusting in most cases seems to be a thing of the past, but it also seems clear it can be required from time to time in 3 chip one lens systems. If not, why would the cited systems have advertized adjustments?

Also DMDs (pixels) are smaller that the LCDs I was able to find specs on which would only make the alignment requirements tighter, by a factor of 2-3 in the cases I could easily find.

Of course the obvious thing no one has worried about is that the GLV ribbons work by flexing back and forth really really fast. In a big scale machine that would be suicide, but that doesn't worry me either, since as I said before this size object and movement is something we are not equipped by experience to intuitively understand. The implications of the rules change and are not familiar to us at that scale.

Have a good day.


[This message has been edited by charlie (edited December 01, 2002).]