Originally Posted by ptchristensen
- Did you get the magnet work done under the sony warranty?
The initial factory tech visit (perhaps 1 week after the set was delivered to me) was arranged by Sony. I don't actually know where he came from. Free visit to me.
The second visit, from a local authorized service place, was also free to me. His follow-up visit (because he'd run out of magnets on the first trip) was also free to me, but he simply might not have told Sony and done it on his own because of the incomplete first trip. Very sincere.
|- Were your problems only convergence, or did you also have geometry issues?
As would be expected, I had both. And of course this makes geometry placement quite an art. A magnet affects not only convergence but the entire electron beam, which means even geometry can be affected (adversely or beneficially) while trying to cure convergence problems.
As it turns out, we were able to use magnets to reduce some of the more gross bow/curvature artifacts along with pincusion and trapezoid problems (especially obvious in the corners of course) and even some S-curves more in the middle of straight lines, working very carefully to not disturb the convergence corrections we had also implemented. As would be expected, the results were not absolutely 100% perfect, and some compromises (from me, really) were necessary. But outside of the test patterns where you can really see the problems and effects of the magnets, most actual live HDTV content looks perfect.
So once the magnet work was complete and we didn't want to fool with it anymore for fear of tampering with "success", I proceeded to work (on my own) with the full set of 2170D-1 and 2170D-2 groups, moving little-by-little all around the screen and trying to get "rectangularity" and "linearity" where previously I might have seen distortions. I am extremely pleased with my final results.
Just recently, in response to the Oct.1 premiere of "Saturday Night Live" in HD and the misplacement of the upconverted SD bug, I became aware that I also had misadjusted my overscan. I think I must have done this using a DVD-provided test image at 480p (from Digital Video Essentials) and INPUT5 on the set. Whatever the explanation, it was clearly off as the misplaced NBC bug was 50% cut off, partially because of their misplacement and partially because of my wrong overscan adjustment.
So last week I worked on re-sizing and re-centering, and reducing horizontal and vertical overscan. This time I connected my XBR960 to my PC (DVI output of ATI Radeon 9800 Pro to component video INPUT6 of the TV via ATI adapter, running at 1080i resolution using ATI 5.8 Catalyst drivers) and used a 1920x1080 convergence/overscan test pattern from DisplayMate for Windows Video Edition (attached to this post). Preliminary adjustments were entirely in the 2170D-1 (0-1, for vertical position and size of the raster) and 2170D-2 (1-2, for horizontal position and size of the raster) groups, followed by overscan adjustments within the now adjusted raster using MID3 (0-3, for horizontal and vertical position and size). I'm quite surprised at how much better things now look, and how much broadcast content I was previously losing before this recent tuning effort.
Much to my surprise, some previous residual geometry and linearity distortions around the extreme perimiter of the screen (which I never could quite tweak out) now seem to have been greatly reduced. To my eye, permiter lines and corners seem much more linear and rectangular than before. Perhaps I was previously "stretching" things unnaturally and now, with proper re-sizing and re-centering and much reduced overscan, things are being displayed with proper proportions and H/V dimension ratios, and thus appear much more linear.
Whatever the explanation, adjusting size/center and overscan using 2170D-1 (0-1), 2170D-2 (1-2), and finally MID3 (0-3) has seemingly had a positive effect on geometry and lineary as well, but that may be partially optical illusion since everything is just a tiny bit smaller now (with reduced overscan, more picture is now being displayed in the same 16x9 34" screen area).
NOTE: I think a real test of your adjustments is on ESPN-HD, when they're showing 4x3 content and have those gray vertical areas on the left and right to fill out the 16x9. The gray areas are striped, with thin black stripes against the gray background. And they're supposed to be truly vertical stripes, and black. Depending on your convergence settings and other geometry/linearity settings, anything not perfect will be quite visible in these two gray areas. On my set, the black stripes left side were initially way out (and almost rainbow-like) while the stripes on the right side were virtually pure black. Now, they're both very very similar (although the left side black stripes are not quite pure black as the right side stripes still are, but they're very very close and at least they're not chromatic and rainbow any longer). I think this content is the real proof of your overall convergence tuning and geometry tuning, whether by magnets or service menu tweaks. I know I always fall back to it as my "real world" test.
|- I'm also in Marina Del Rey...if you want to share the name of the magnet guy, please PM contact information.
No need for secrecy or privacy. The guy deserves some promotion.
Ask for Andy, at Audio Video Shoppe (on Teale St. in Culver City). 310-306-4600. They're authorized service providers for lots of brands, including Sony.
overscan.zip 9.4423828125k . file