LCoS, 3LCD, DLP (though unlikely in this forum!) owners—how is your projector's convergence?

What would you say is acceptable convergence?

I ask because my fairly new Sony HW10 seems pretty far out, though the panel alignment function does seem to help a lot. (but it causes there to be a slight haze over fine details, and can discolour very high frequency detail)

Taken up close from the very centre of the screen, before/after:

http://sr-388.net/images/ht/projector/_MG_1534.jpg http://sr-388.net/images/ht/projector/_MG_1535.jpg

Is this particularly bad? The lines are (obviously) 2px thick, and convergence seems off quite a lot. I'm having to use:

_____|Horiz.|Vert.
Red|+6|0
Blue|+4|-9

I believe the adjustments are in 0.1 pixel increments, so blue is almost a whole pixel out vertically.

It would be nice to see how far out other people's projectors are for reference, and how much things can be improved on them, if at all.


Things aren't quite this good edge-to-edge when the adjustment has been made (the zone adjustments of the VW80/VW200 would be great) but nowhere on the screen is it as bad as it was to begin with.


EDIT: I should also mention that even with the misconvergence, there's less fringing around white objects than with my old 480p DLP projector once I've set up the panel alignment due to the much better optics.

Hard to say...that is only 1 small grid. How is it on a whole across the screen?

Hard to say...that is only 1 small grid. How is it on a whole across the screen?
Ok, using this (http://sr-388.net/images/ht/projector/convergence/Convergence.jpg) pattern, I've taken some more photos showing how convergence is on my HW10 with panel alignment on/off.

Some of the shots are a bit blurred due to camera shake as the corners were awkward to get to, but should be sharp enough to show off the convergence.

Top Left:
http://sr-388.net/images/ht/projector/convergence/_MG_1539.jpg http://sr-388.net/images/ht/projector/convergence/_MG_1542.jpg

Top Right:
http://sr-388.net/images/ht/projector/convergence/_MG_1547.jpg http://sr-388.net/images/ht/projector/convergence/_MG_1548.jpg

Centre:
http://sr-388.net/images/ht/projector/convergence/_MG_1551.jpg http://sr-388.net/images/ht/projector/convergence/_MG_1556.jpg

Bottom Left:
http://sr-388.net/images/ht/projector/convergence/_MG_1559.jpg http://sr-388.net/images/ht/projector/convergence/_MG_1561.jpg

Bottom Right:
http://sr-388.net/images/ht/projector/convergence/_MG_1565.jpg http://sr-388.net/images/ht/projector/convergence/_MG_1568.jpg


While it looks pretty good after the convergence adjustment, is the physical panel convergence acceptable, or should I expect better?

That's tricky. Technically with the convergence adjustment, Sony can say that since it can be "fixed" it is within spec. If a projector didn't have the adjustment, it would pretty much be a toss up (if you complained enough they probably would fix it). Convergence is a tough thing to really gaged when it is similar to this.

Does the Sony have full one pixel shift like the JVC does or does it have to do the upscaling/downscaling business?

Later.... .1 pixel increments, which is beneficial from the fine tuning standpoint, but obviously there are inherent issues with this system as well.

Does the Sony have full one pixel shift like the JVC does or does it have to do the upscaling/downscaling business?
The adjustments are in increments of 0.1 pixels.

If you use less than a full pixel adjustment, there is a slight "haze" around fine details (as shown in the photos above) and single pixel wide, high contrast details may suffer from discolouration. If you use +/- 10 with the adjustments, you don't get this.

I have only been able to see it with test patterns though, and not actual image content. (where you are almost never going to have such high frequency information)

From where I'm currently seated (approx 12/13ft from a 120" screen) I can see the misconvergence, even without my glasses on. I don't notice this "haze" at all though.

How long have you had the projector?

How long have you had the projector?
It's been on around 100 hours now.

I would shift blue down one full pixel and experiment moving it over to the right one full pixel. Full pixel shifts should leave you no more than .5 pixel out in the center and you won't have the problems the .1 pixel shifts cause. I've read that the .1 pixel shift increments can cause more problems than they address but I haven't seen this in action so can't comment from personal experience.

You might want to check with your dealer on possible options since it is bothering you. At 100 hours they are not obligated to do anything, but you never know.

You might want to check with your dealer on possible options since it is bothering you. At 100 hours they are not obligated to do anything, but you never know.
Well it's less than a month I've had the projector, so I'd probably still be ok to make a complaint. I've just been told on another forum that Sony's convergence tolerance is 2 pixels out though, so it's within spec.


I've only just put the projector on, so it's a bit out of focus as it hasn't properly warmed up (that's another issue I have with the projector—it takes at least 20/30 minutes before it's back in focus) but here's the discolouration I was talking about.

Black/white alternating pixels, no convergence adjustment:
http://sr-388.net/images/ht/projector/convergence/_MG_1572.jpg

1px convergence adjustments:
http://sr-388.net/images/ht/projector/convergence/_MG_1573.jpg

0.1px convergence adjustments:
http://sr-388.net/images/ht/projector/convergence/_MG_1571.jpg


I have yet to ever encounter anything like this with actual video/game content, only test patterns. Anything wider than 1 pixel, or 1 pixel lines on their own are fine though. The improvement from using the 0.1px adjustments rather than shifting a full pixel are more noticeable with real-world content.

In order to adjust the R, G or B by a fraction of a pixel (0.1 in this case) you need to use an interpolation filter. The "haze" you are seeing with your grid test pattern is the composite impulse responses of the R, G and B filters in both vertical and horizontal directions. You would probably never see this with a correctly transferred Blu-ray since it would have spatial frequencies limited, which would work correctly with the spatial frequency limitation of the filters. In other words, the test pattern you are using could never appear in a transferred film. If you use the projector with your computer or a video game machine, then you might want to follow mrlittlejeans advice and adjust to the nearest integer position for each of the R, G and B. This would be optimum for computer/game use, but I would think you should be better off with fractional positioning for films. Just for the record, my current projector does not have this feature so I've never seen this in action. But I did sleep in a Holiday Inn Express last night;)

I would shift blue down one full pixel and experiment moving it over to the right one full pixel. Full pixel shifts should leave you no more than .5 pixel out in the center and you won't have the problems the .1 pixel shifts cause. I've read that the .1 pixel shift increments can cause more problems than they address but I haven't seen this in action so can't comment from personal experience.

I can, as raised here...

http://www.avsforum.com/avs-vb/showthread.php?p=15765051#post15765051

Well, I've put the projector off for tonight, but the next time it's on, I'll try to get photos of this pattern (http://sr-388.net/images/ht/projector/convergence/Convergence.jpg) again to compare 1px and 0.1px adjustments to see how it affects sharpness.

I can't say I've noticed a significant change in sharpness with actual images so far (rather than test patterns) but I have found misconvergence to be less noticeable when using 0.1px adjustments rather than 1px.

Well, I've put the projector off for tonight, but the next time it's on, I'll try to get photos of this pattern (http://sr-388.net/images/ht/projector/convergence/Convergence.jpg) again to compare 1px and 0.1px adjustments to see how it affects sharpness.

I can't say I've noticed a significant change in sharpness with actual images so far (rather than test patterns) but I have found misconvergence to be less noticeable when using 0.1px adjustments rather than 1px.Please try a still frame from a Blu-ray film and compare close-up screen shots to see if you can tell a difference while you are at it. Sounds like the fractional adjustment works best (as it should) for "actual images" (Blu-ray film?), or maybe I read your last sentence wrong:confused: Also, it might be beneficial for those of us without convergence adjustments to show screen shots without any adjustment, i.e., OOTB.

I'm sure that I speak for more than just myself when I say that your efforts are appreciated here!:)

Well firstly, another shot of that test pattern from the centre of the image. No adjustments, 1px adjustments, 0.1px adjustments:
http://sr-388.net/images/ht/projector/convergence/_MG_1587.jpghttp://sr-388.net/images/ht/projector/convergence/_MG_1589.jpghttp://sr-388.net/images/ht/projector/convergence/_MG_1588.jpg

You get that "haze" around things with the 0.1px adjustments, but it does not seem to affect detail—the number 3 is made up of single pixel-wide lines and is still fully resolved.

Having that uniform "haze" seems to be much less noticeable at a distance than having one edge with a stronger colour. That said, it doesn't seem to show up much on video content rather than test patterns.


I used some of the B&W scenes from Memento on Blu-Ray for "real-world" comparisons, as I figured B&W would show off misconvergence easier (as there shouldn't be any colour at all) and I seemed to remember them looking pretty sharp/detailed when I last watched. (actually not quite as good as I thought now that I'm viewing at 120")

Unfortunately there's a little bit of moiré in these photos, but these are the ones that turned out best.

First scene. (to show how tight a crop it is)
http://sr-388.net/images/ht/projector/convergence/_MG_1597.jpg

Crop of his collar, without, 1px, 0.1px:
http://sr-388.net/images/ht/projector/convergence/_MG_1599.jpghttp://sr-388.net/images/ht/projector/convergence/_MG_1598.jpghttp://sr-388.net/images/ht/projector/convergence/_MG_1597-2.jpg

Note the magenta/green fringing without adjustment, the predominantly green fringe on the left with 1px adjustments, and effectively none with 0.1px adjustments. (I should have really moved the camera closer/zoomed in to show it off better)

Second:
http://sr-388.net/images/ht/projector/convergence/_MG_1601.jpg

This time I did move the camera in closer. Again, no adjustment, 1px, 0.1px.
http://sr-388.net/images/ht/projector/convergence/_MG_1606.jpghttp://sr-388.net/images/ht/projector/convergence/_MG_1605.jpghttp://sr-388.net/images/ht/projector/convergence/_MG_1604.jpg


I've not done the best job capturing these shots as the difference between 1px and 0.1px adjustments is more noticeable by eye, with the reduction in colour fringing using 0.1px adjustments being much more apparent.


Nothing I've seen from "real-world" content has been able to produce the discolouration shown further up, and I haven't noticed any detail loss.



So while logic dictates that you should use 1px adjustments rather than 0.1px ones, and test-patterns agree. Real-world video/game content shows the most improvement with the 0.1px adjustments, and I have yet to notice any real downside to using it.

I would still prefer to have a projector that is either much better converged physically, or whole pixels out rather than fractions of a pixel, but I'm pretty happy with the results of the convergence adjustment now that I've done more testing with it.

Thanks for the continued testing, posting the photos and confirming what I'm seeing on my setup.

andrewfee:

Very nice analysis! It looks like the Sony engineers knew what they were doing;) It seems to be splitting hairs for computer use (based on the "3" in your test pattern) whether integer or fractional adjustment would be preferred, but there is no doubt that for movies the full fractional adjustment is best, just like theory would suggest. The only other problem right now is that it will be really difficult for me to purchase another projector without this feature:( Hopefully all 3-panel PJs include this feature ASAP.

Can the pixel adjustments be stored in individual profiles (or per input)?

Seems like I would prefer the 1px approach for computer use and .1px for video use. Fantastic to have this option though.

Nice pics and analysis - thanks.

Can the pixel adjustments be stored in individual profiles (or per input)?

Seems like I would prefer the 1px approach for computer use and .1px for video use. Fantastic to have this option though.

Nice pics and analysis - thanks.

Unfortunately no...it is a global setting.