or Connect
AVS › AVS Forum › Display Devices › 2.35:1 Constant Image Height Chat › How much curve is needed to compensate lens distortion?
New Posts  All Forums:Forum Nav:

# How much curve is needed to compensate lens distortion? - Page 3

Does that mean that there is a non HT specific lens that does the same job without the HT specific markup or does generic just mean not ISCO?

Quote:
Originally Posted by Aussie Bob

About 0.4 inches (i.e. virtually none)

I modelled your set-up - ISCO, throw, screen-width - in my Zemax (professional optics) software and got this result.

I also have a spreadsheet that, instead of working things out iteratively with optical ray tracing like Zemax does, does the same thing using basic geometry and simple optics... and comes out with practically identical results (<.5% different) without needing to invest \$000s in Zemax.

The spreadsheet works out the pincushion without a curved screen to +/- 0.5mm, and then from that curves the screen until the pincushion is <0.1mm .

Results are:

*** Pincushion (as I said) = 0.4 inches.

*** Your screen curvature to eliminate pincushion has a radius = 72.1 feet,

*** The dip in the middle = 4 inches
(or alternatively, you pull the edges of the screen 4 inches out from the wall towards the projector).

I know no-one's going to believe me, but comparing ISCOs by analysing published performance against other generic cylindrical lenses there seems to be little difference in pincushion performance.

### AVS Top Picks

VBB,

"Generic" means a computer model of a lens that expands a 16x9 image by exactly 4/3 (1.33:1).

From that expansion factor, throw length and screen height (or width) you can work out a very good approximation of what the pincushion will be, and from that how much you will need to curve the screen to eliminate pincushion. Straightforward geometry plus one or two "trick" calculations.

I designed my spreadsheet to help out a couple of installers I know here in Australia. They put it on their laptops and take it to sites. I've had no complaints.

I tested my spreadsheet by inputting a variety of anamorphic setups into both my spreadsheet and the same setups also into my Zemax software. The spreadsheet uses a different, much simpler method than Zemax. Zemax actually "optimizes" the setup by iterative solving (i.e. trying lots of solutions until it satisfies the parameters you have specified, like Excel's "Solver" function) and tracing actual rays, based on lens shape, glass types etc.

Upon comparison, both sets of results - one from Zemax and one from the spreadsheet - both have been well within about 1% of each other. For practical purposes (even for the anal retentives in this forum) they are identical and therefore equally useful.

I've done a lot of these calculations now and the spreadsheet seems to work quite well, agreeing with Zemax (within that 1% tolerance) 100% of the time. A 1% tolerance boils down to a couple of millimetres here and there... nothing to fret about out there in the big wide world of 14 foot screens.

Zemax is, of course, indispensible for many, MANY other things which a mere spreadsheet couldn't do in a pink fit. But for these types of quick calculations the spreadsheet is OK.

I have never measured an ISCO in real life, but as I said in my last post, I have seen many reports on these pages of throw length, screen size and pincushion measurement using ISCOs. When, out of idle curiosity, I put these into my Zemax or spreadsheet programs, assuming the ISCO is a "generic" lens, I get the same results (with that 1% error).

Therefore I assume the ISCO, despite its reputation, is as good or as bad on pincushion as any other reasonably-well designed lens. I think it's price has led a lot here to believe it can work miracles. Well, I don't think so, at least as far as pincushion is concerned. Perhaps it's sharper or lighter or something else, but from what I've seen its pincushion performance is pretty much par for the course.

You mentioned "HT markup". If you mean that anamorphics specifically designed for HT have an extra price added, it's probably because they have to have large optics to accommodate wide beams typical of HT. The ISCO optics have a diameter greater than a DVD disk (> 120mm). Large optics mean big prices. Then there's the low manufacturing numbers. They need to amortize their cost over only a few units. Lastly there's the "Made In Germany" factor. That's worth at least another 25% on top. All those precise guys in white lab coats driving BMWs and Mercedes... they don't come cheap (especially in Euros). I've seen a lot of information that ISCO don't publish specifications, or even provide an installation sheet with their expensive lenses. I think the reason why is so that people like a lot of forum members here would panic if they did. Better to leave certain things (like pincushion specifications) unsaid.

Pincushion is pretty much a "given" with the small throw ratios of HT. Wind your throw ratio out past about 3 and you can get rid of pincushion for practical purposes (<0.2%, which is peanuts when you view it). Unfortunately not many HT projectors are capable of throw ratios of even 3. My Panasonic can only manage 2.75, and that's a pretty good performer among projectors.

To get completely rid of pincushion you need a really complex lens, with very complex mechanics (or a long throw ratio). Projector lenses have up to 20 elements in them and they still exhibit some pincushion and other distortions despite this. Anamorphics usually have only three or four elements at the most. You just can't correct for pincushion with so few elements, for anything like a reasonable price, especially considering the multiplicity of throws, screen sizes and projector lenses you could be feeding into your anamorphic expander.
FWIW, our hybrid prism-cylindrical HE lenses produce about 25% more distortion than the Isco III. Consequently, if you have 12mm of sag at the top and bottom of the image with the Isco III you'll have 15mm of sag with the UH380. Yes, the ultimate amount depends on throw ratio. Someday we plan to run experiments for some empirical-based way to determine a best screen curvature but we do not get that much demand at this point.

Also, the Isco III is a four element lens of two groups (assumed as cemented achromatic doublets). The UH380 produces higher image quality at optimum throw distances but on average is about the same when looking at pixel-level quality at 1080 or above. Much of the higher cost of the Isco III seems to be the additional difficulty in making so many cylindrical surfaces to proper specifications. Of course it could also be that none of our folks drive a BMW or Mercedes (at least that I see).

p.s. Bob, congrats on your lens.
Quote:

Much of the higher cost of the Isco III seems to be the additional difficulty in making so many cylindrical surfaces to proper specifications.

I had compounding problems with striae at first. My design is 6 elements in 3 groups (2 doublets). The aim was to keep the aspect ratio at 2.37 (+/- 0.75%) over a wide throw ratio range. The first prototype I put together didn't perform exactly as modelled. It wasn't until I detected the striae that I figured out why. They're hard to see, but have a big effect on even focus. The manufacturing company (in China) re-did the whole thing with new tooling and better glass (they cheated a little bit the first time with Indian glass) and it suddenly all worked "as advertised". I have the "good" version out on evaluation at the moment. I kept the dud. It's still fine for "private" use.

My point about pincushion was that it's pretty predictable, and most design strategies fall into a narrow range for the same set of of parameters. A 3mm difference between your prism design and the ISCO wouldn't be particularly noticeable.

I'm surprised the ISCO-III is only two groups. Two groups usually means you have aspect ratio problems at different focus positions, especially with HT throw ratios. Two groups would also make life easier for me, with mechanics (it's my middle group that moves). Oh well, they're the geniuses with the patents... maybe they're using asphericals, which would also add a lot to the cost?
I don't think they're aspherics, which would be a huge cost issue. But you are right - the aspect ratio changes with the spacing as the astigmatism is corrected. It's definitely an older technology but for a time it was the best available. I also would find it hard to believe there are any patents still valid on basic cylindrical designs although Schneider has some relatively new IP on some of their modifications.
Reading through some the patents is fun. There are a few from the 1990s I think, maybe one or two from the 2000s.

My favourite was the Delrama patent (1951 I think). What a clever idea! Using TIR to get the mirrors parallel to each other. I modelled a Delrama exactly as specified in the patent (as far as I could with the basic version of Zemax) and it gave beautiful results, but the thing - for HT throw ratios especially - was bigger than two house bricks. Would've weighed a ton. I have a neat little Super 8 projection version, 1.5x, that's tiny (two match boxes in size), but of course it's designed for a long throw. I also have a full size one (2x) and it's a monster. Both from eBay. Amazing what you can find for \$10!

I never see any complaints about aspect ratio overshoot on these pages. To me it's a bummer... all those elastic faces as you shift your projector closer to the screen.
Quote:
Originally Posted by Aussie Bob

I never see any complaints about aspect ratio overshoot on these pages. To me it's a bummer... all those elastic faces as you shift your projector closer to the screen.

I have a theory here, but would like to hear from an expert in this field, so could please elaborate?

Mark
It's all to do with the ratio of the distance a ray travels to the corners or sides of the screen against the distance a ray travels to the centre of the screen.

The wider the beam (i.e. the smaller the throw ratio) the greater this ratio becomes.

The greater this ratio becomes the greater becomes the magnification at the edges.

This goes a long way towards explaining pincushion and distortion, although not all the way.

Basically the focusing rays have further to travel to the edges of the screen. Therefore they'll be magnified more than the centre rays when they reach it. Screen height will be progressively higher from centre to edge (pincushion), and any grid squares will be wider and a little higher (grid distortion). This grid square size differential was what I was talking about as "distortion". I should have used the term "grid distortion".

As I said it's more complicated than that and specific anamorphic lens design can ameliorate it some. The projector lens also has an ameliorating effect as they all pre-distort the image so that it will be rectangular. The real calculations are hideously complicated (and I, for one, never tried to figure them out) and are best done with a ray-tracing program loaded with an exact model of the lens (that's why ray tracing software has taken over optical design). Applying simple Pythagoras' Theorem calculations will get you close, though.

To those who think that an anamorph should just pass on all the good flat-screen aspects of a projection lens, that's not quite true. A projection lens has many elements in it for the specific purpose of delivering an orthogonal (rectangular) image. As Shawn pointed out, even the ISCO has only two groups, and they are mostly bent in only one direction (horizontal). Same effect with prisms (although they are not bent, usually). You can't expect miracles from just two elements.

You can lower grid distortion with appropriate cylindrical design, and extra elements (by attempting to balance edge grid distortion against centre grid distortion), but not pincushion (or not by much anyhow).

To lower pincushion you have to add elements and include some cross-rotated ones (curved in the vertical direction). This will then affect screen height so you lose some of your CH effect (a few per cent), and you may also lose your constant focus (you may have to refocus the projector when you remove the anamorphic lens).

And, if this lens is to be a commercial product, it must work with a plethora of random user-determined throw ratios and varying projector lens designs. Too expensive to manufacture for a reasonable price in small quantities!

Hope this helps.
Aussie Bob - Would you mind posting your spreadsheet for others to use? Also is it set up for specific lens types or would it work for a Prismasonic as well?

I'm curious what results it will give for my setup that's in the works:
Throw - 13'
2.37 Screen Width - 111"
Lens - Prismasonic H-FE1500M

Also the projector will be offset to the left of the center of the screen by about 3'.

I'm currently planning to just project the pincushion onto the border but I could still put a little curve on the screen.
Quote:
Originally Posted by Aussie Bob

The wider the beam (i.e. the smaller the throw ratio) the greater this ratio becomes.

The greater this ratio becomes the greater becomes the magnification at the edges.

This goes a long way towards explaining pincushion and distortion, although not all the way.

Basically the focusing rays have further to travel to the edges of the screen. Therefore they'll be magnified more than the centre rays when they reach it. Screen height will be progressively higher from centre to edge (pincushion), and any grid squares will be wider and a little higher (grid distortion). This grid square size differential was what I was talking about as "distortion". I should have used the term "grid distortion".

What I thought, so thank you Aussie Bob, that does help to see it explained by someone with topic knowledge

Mark
So I'm guessing from this discussion that the fact that my throw is 26' is the reason I have no more than 1/2" of pincushion over 14' width.

Art
Quote:

Throw - 13'
2.37 Screen Width - 111"

Pincushion before curving screen: 0.6in
(Pincushion/height)%: 1.24%
Dip at Middle of screen (arc height): 3.8in
Chord Width of Screen: 9.14 ft
Curve Length of screen: 9.17 ft
16x9 Throw ratio: 1.874

Pincushion is defined as the difference in height between the top corner of the screen and the centre of the top boundary. Pincushion percentage is the pincushion figure divided by the full centre height of the screen (which I calculated to be around 47 inches).

Also checked your inputs on my Zemax program and your curvature came out to 32.7 feet, and your pre-curved-screen pincushion to 1.19%... both negligible differences. For example, your dip in the middle of the screen (arc height) does not change from 3.8in, whichever figures you use, despite the slight difference in screen curvature radius. You need a big difference in radius of curvature to notice it.

As I said above, this is for a "generic" lens which operates only on ordinary geometric principles (e.g. Pythagoras' Theorem). There will be some differences between designs, but not enough to matter, I would think.

Re. your projection position offset... that's a whopper! 3ft! You might get some skewing of your image - lengthening of image width and warping of the corner - on the opposite side to where you're mounting your projector. An anamorphic lens will magnify this somewhat, but I'm not prepared to put a figure on it.
Quote:
Originally Posted by Art Sonneborn

So I'm guessing from this discussion that the fact that my throw is 26' is the reason I have no more than 1/2" of pincushion over 14' width.

Art

Yes and that the 0.5" is nothing to be alarmed about...

Mark
Quote:
Originally Posted by CAVX

Yes and that the 0.5" is nothing to be alarmed about...

Mark

Mark,
Honestly ,I'm so much more than pleased I went CIH. You guys were certainly right. I had my reservations and until we had HDM to put up there I don't believe I would have pulled the trigger but it just presents a spectacular look !

Art
And again Art, that is really something given you have moved from twin CRTs

Mark
Quote:
Originally Posted by CAVX

And again Art, that is really something given you have moved from twin CRTs

Mark

I had a few reservations about giving up the CRTs but have not regretted the move, in fact, I now have a better picture and by some masrgin IMO.

Art
New Posts  All Forums:Forum Nav:
Return Home
• How much curve is needed to compensate lens distortion?
AVS › AVS Forum › Display Devices › 2.35:1 Constant Image Height Chat › How much curve is needed to compensate lens distortion?