Is a curved screen really neccesary?

A curved screen is nice to help alleviate the pincushion effect an A lens adds to the image, but many use a flat screen and overscan the image a little to hide the curve. In your case I would get a flat screen and overscan.

I had a 2.35:1 screen and overscanned the pincushion with very little loss of the image.

If you're not using an A lens, then you don't need a curved screen. If you use a curved screen when zooming, you'll have barrel distortion to deal with and will have to overscan the image to hide it.

Gary
Edited by Gary Lightfoot - 8/23/12 at 12:39pm

What about when you watch 16x9 material on a curve screen? Apart from moving the lens, Is there anything you need to adjust the image?

When you watch 16:9 material on a curved screen and remove the lens to do it (some leave the lens in place all the time and scale for 16:9), you will get a little barrel distortion, but it's usually small because the curve is less. You may have to zoom a little but you'll need to experiment to see how much.

I haven't used a curved screen so I couldn't say how much overscan you'd need.

Gary

The biggest plus for a curved screen is the geometry correction it gives (assuming it is matched to the throw ratio). I explained the 'why' with an image HERE

I'm not sure he has the room for a 36' radius.

Why not? I'm not sure you understand. He's not talking screen size but, the radius of which the screen is curved at.

Example: You could have a 10ft. wide 2.35:1 ratio curved screen with a 40ft. radius.

I understand perfectly. I was twisting things around a bit. Apparently it was too early in the morning.

FWIW I'm another lens user that has a flat electric screen, so I just overscan the image slightly but have a very long throw (projector is at minimum zoom) so the pincushion is minimal...it all depends on the throw really, but it's 100% not an issue on my setup.

However, I have noticed that the new JVC projectors have a pincushion control for when you use the projector's own vertical stretch mode. I think it could be a bit like keystone, where it introduces more issues than it solves with reduces resolution, but I'm going to try it out when I receive my X35(RS46). However, I found that my Lumagen did a better job of V stretch than my existing JVC HD350(RS10) so that will also have some bearing on the matter.

The pincushion is consistent across the screen I believe, but just that in the middle the image comes inside the screen area and at the outer edges it bends just beyond. It's just a matter of equalising the overscan and careful set up of the lens so that the pincushion is equal top and bottom. It's possible to get the bottom perfectly flat for example, but then the top will be twice as bad, so it's not a good choice. There doesn't seem to be any issue with the vertical edges left and right as they seem straight to me. Hope this helps.

The amount of overlap increases as you move to the edges of the screen. It will be zero in the middle with no distortion at all, and at its most severe at the edges.

I developed an accurate formula for pincushion, which works right down to about a throw ratio of 1.5 (and even then it isn't out by much).

Throw ratio is entirely a function of Throw Ratio ("Throw" divided by "16:9 width" see below for a formula). People who tell you "at this throw, pincushion is such-and-such" are wrong, if they do not factor screen size into it. Calculating pincushion must ALWAYS take account of BOTH throw and screen size.... in other words "throw RATIO".

This is a quote from a recent post I made at http://www.avsforum.com/t/1436661/calculate-pincushion#post_22545142



What's the formula based on?
Some research I did into pincushion when designing my Xeitoptics cylinder lens.

I found that pincushion was a straightforward function of throw Ratio (TR) for my lens design. The formula was derived from establishing a whole bunch of accurately calculated pincushion results in a spreadsheet table and then performing a statistical regression on them. The formula above in (2) gave a "best fit" regression.

I wondered whether it worked for other designs, even prism systems.

So I then tried it with models of other designs that I had, including prism lenses and old-fashioned designs like Isco. The results from all types of designs conformed pretty well (within 1% of the modelling results).

Adjusting the tilt of your lens does indeed change the amount of pincushion between top and bottom of the lens.

You tilt your lens in the direction of the excess you want to reduce.

So, say you have too much pincushion at the bottom of the screen (and the top has virtually none), tilt the lens DOWN i.e. towards the bottom of the screen.

You will see pincushion at the bottom edge reduce, while pincushion at the top increases. Keep adjusting until both are either the same or where you aesthetically want them to be.

I usually put up the JVC's green grid focus or zoom pattern as it has horizontal and vertical lines. I then adjust the tilt so that the pincushion is equal top and bottom as you say. It's quite a harsh test of the lens too because it's more obvious that the lines aren't perfectly straight, yet watching ordinary video I never seem to notice.

The horizontal lines look straightest across the centre of the screen height (for all intents and purposes they are flat), but the lines at the top and bottom of the screen seem to be effectively very large radius semi circles, so again I adjust the height and zoom such that the top and bottom of the image just touches the edge of the screen in the centre, so that the 'peaks' in each corner overspill the least amount. This way the white part of the screen doesn't have any area not covered with the image.

The hardest part for me is adjusting the astigmatism of my Isco II lens: I can't turn the adjustment and be stood by the screen to see the effect to best advantage. I've tried getting my son to stand by the screen to tell me when the lines look the sharpest in both directions, but being 13 I don't think he really seems to be as picky as I am. I end up turning the adjustment to one extreme and noting the number on the lens, then back the other way so I can split the difference. Or making many small adjustments and walking back and forth trying to tell which looks sharpest. I bought some cheap binoculars to try and they didn't help as they weren't sharp enough.

Setting a lens up certainly isn't a 5 minute job I find...though it is worth it once done.

I've set up plenty of lenses and it never gets easier.

It can take 2 minutes or two hours, depending on what you had to drink the night before.

Try getting the rotation of three cylinder lenses perfect to +/- 1/120th of a degree, not only internally to each other, but to an external horizon reference. I use a microscope I put together and it can still take an age.

Glad it's not only me. I think I had mine about two weeks before I finally got everything really dialed in. Even then it is on a stand so I have to make sure I put it in the same spot in front of my projector each time I watch a 2.35/2.40:1 film, though I've got that down to a fine art now.

I'm moving my projector slightly to a new shelf higher up. I know this will entail tweaking the lens again, so not looking forward to that. I may add some small 'stops' to the top of the shelf to help align the lens stand to help reduce set up time as well.

That's exactly what I do.

You won't need to once you get that VW1000 though.

Ha ha!

I'm a Yorshire man Kelvin. Say no more....

No. They can be in carrot sticks if you like, as long as both use the same units.

Correct.

You should also add in some extra overscan for 2.40:1 movies.

2.40:1 movies occupy 800 pixels out of 1080 pixels of image height. When the image is vertically stretched, however, the middle 810 pixels will be stretched to fill the height of the image (810*4/3=1080). So between the 800 pixels of a 2.40:1 movie and the top (and bottom) edges of the image there will be a 10 pixel black band.

To get rid of this thin black band you need to optically zoom the height of the image a little more than you might have thought at first. The ratio is 810/800 = 1.0125, one and a quarter per cent.

So, to fill your 60" high screen height you need to zoom it so that the 16:9 grid pattern (or whatever test pattern you are using) enlarges to 60*1.0125 = 60.75" high, an overlap (dividing the extra 3/4" equally between top and bottom), around 3/8ths of an inch top and bottom, say 1/2 an inch to make sure.

To prove it:

800 pixels represents 800/1080 = 0.740740741 of screen height.

Now multiply 60" by that number and you get 60*0.740740741 = 44.4444" BEFORE vertical stretch of 4/3 times.

Putting it all together 44.4444*1.0125*4/3 = 60". Your 2.40:1 movie now matches your exact screen height. 16:9 material will overshoot by a little.

I always recommend adding a little to this to make sure. Adding some extra also accounts for 1.85:1 Academy Aperture films, which are slightly shorter (at 1040 pixels high) than 2.40:1 scope presentations are after vertical stretch.

I know it sounds complicated, but I've seen some unfortunate installations where everything is bolted into place, only to find that the image doesn't quite fit the screen and, because everything else is maxed out, can never fit the screen. You need to be aware of the issues.

A1920 x 1080 test pattern showing all the common movie apertures is at my web site: http://www.xeitopticscom/test-patterns/ (scroll down).
Edited by Aussie Bob II - 11/19/12 at 2:51pm