AVS Forum banner
Status
Not open for further replies.
1 - 20 of 26 Posts

·
Registered
Joined
·
23,188 Posts
Discussion Starter · #1 ·
There has been quite a bit of talk about irises around here for a while and I figured I would try to give some information on them.


Note: This got kind of long and I'm guessing a little confusing with the style I used (and didn't clean up), so I'll first give the main conclusion here and then people who want to know more can read on for the reasons. With single chip DLPs and non-dynamic irises, one iris can help CR somewhat, but two two irises that are well matched with each other is generally what it takes to get the really high on/off CRs. Now the rest of the story that I originally wrote:


I believe I've seen some mention here of just putting an iris some place in the light path and getting better on/off CR, but it really doesn't generally work this way. The placement is pretty important. And the real kicker is that the main way that Sharp leapfrogged the competition for on/off CR when they came out with the Sharp 12k was to use a dual system and I'm not sure many realized this. A single iris could not have gotten them the CR that they achieved. Anybody with an 11k or 12k can watch the iris in the lens and see that while there are 3 iris choices for this projector, one of them increases the CR even though the iris in the lens doesn't budge. I haven't seen far enough inside the projector to see exactly what kind of contraption they have deep inside, but my guess is that it is also an iris with 2 positions.


The dual irises working together are very important to getting the best CR, as BenQ seems to have found with their 8720 design (although it hasn't shipped). If that projector does ship then I believe that people will find that the best approach for getting the best tradeoff if they want the best CR at medium lumens is to move the irises together, for reasons I will try to explain. Just as one review found that going from a medium CR choice to a high CR choice on a projector did not lower the lumens a whole lot, which I believe is because the irises aren't really synced up well with each other in the medium case.


I'm going to step back a little bit and start with the fully open case of even no irises (probably close to what InFocus has with the 72xx line). TI has a document that has some diagrams that would probably help here, but I don't have link now and don't have any diagrams, so I will try to just explain it. Let's picture looking through the front of the lens and seeing what is going on about halfway back in there. We could picture the cross section of the lens here as represented by a quarter. Then when the DLP chips are told to do 100 IRE what we would see could be represented by two dimes layed on top of the quarter and overlapping each other. One of the dimes represents the light when the mirrors are on and I believe the other represents the light from the mirrors when they are at a midpoint between on and off or someplace in the transition. If the DLP has 10 degree mirrors then the dimes will overlap a fair amount, where if they have the newer 12 degree mirrors then there will just be a small amount of overlap (I should mention that this assumes the common single chip DLP projector design). The dime that represents the on case will be brighter brighter than the one the represents the transitions states also.


So, now we have what things look like at 100 IRE for the open lens design. At 0 IRE, lets assume that the quarter represents the light coming through the lens and that it is uniform intensity (for simplicity I am going to assume uniform intensity for 100 IRE, or the dimes, also). Therefore, the light that makes it to the screen is the light density for each of the coins times the area that the coin represents. The overall white level is both the mirrors in the on state and the transition state and the black level is the uniform (my assumption) light coming through when the mirrors are held in the off state. I should mention here that at least in one state I've looked at, the light coming through during the mirror transition state is discolored. Now if we want to increase the CR, what can we do? We can look at the dime that represents the on state at 100 IRE and make an iris where this portion is let through, but everything else is blocked. In theory this blocks all of that light from the mirror transitions, part of the light from the on state, but a lot of the light from the off state (because we've assumed it is uniform and we just decreased the area of the opening). In reality, a close to D-shaped iris is recommended here that does pretty much what I described, with the flatter part (but not necessarily totally flat) of the D being on the side where the light from the mirror transitions would come through. The D-shaped iris could be called the post-imager iris and there could also be a pre-imager iris to work in conjunction with it (the 2 iris system I mentioned earlier).


So, now we are at a place where we have an opening in the lens where 100 IRE light comes through and is uniform and 0 IRE light comes through and is uniform (simplifying assumptions again). How can we increase the CR further (even at the expense of lumens)? Would closing the iris further help? Not in this case. Just for the sake of discussion, let's say that we have 500 lumens and 3000:1 on/off CR. If we cut the iris in half the white level will go to half, but the black level will go down by a half also and now we are at 250 lumens and 3000:1 on/off CR. But here is where the assumptions come in. The white level is pretty controlled coming off the mirrors, but the black level is very random. So, if we can make half as much light fall on the mirrors, then we can get the black level to go down to half. If we look at that 250 lumens 3000:1 on/off CR case and use that iris in the lens, then we can go to the preimager and figure out exactly how to close it down to only let half the light through, but the half that will still go through the new opening in the lens. Basically, make the pre-imager and post-imager work together. Now all of a sudden we still have 250 lumens, but now the CR has gone to 6000:1 (because the white level stayed the same while the black level went to half). Cut both irises in half again while keeping them completely in sync and we are at 125 lumens and 12,000:1.


But, as you have probably guessed, these simplifying assumptions are not perfectly followed and so it doesn't work this well in real life. But, with DC3 chips it works a whole lot better than many people would think. I believe that things top out at about 11k:1 (at least on one I've tested), but at very low lumens and there is another factor here to look at that would be a major problem at 11k:1. Bulbs do not put out uniform light. So, part of the projector design is to randomize this light a certain amount with a light pipe. Closing the irises down like I have described can then cause uniformity issues with intensity and/or color shifts. So, it is something that needs to be kept into account. But over 5k:1 with pretty good uniformity is definitely achievable with some care, from my testing.


I should mention that this "halving two irises for close to double the on/off CR" theory is really just for DLPs and I really don't have a great grasp of how irises affect LCDs.


--Darin
 

·
Registered
Joined
·
3,679 Posts
Great reading Darin, congratulations on your success. This is the first time I realized what a small angle 12 degrees is.


Maybe you can explain something that puzzles me. You use the word transition, but the two dimes are really the on "beam" and the "off" beam.


Since to turn off each mirror must go from 0 degrees to 1 degrees to 2 degrees ... to 12 degrees to be fully "off". What is the effect of these interim positions smearing light across all 12 degrees during a flip.


Obviously its so fast as to be negligible. But still it seems like it would be measurable with the smear being 0 on one side of the screen and non-0 toward the other dime.


Thanks for the lesson. I hope all the manufacturers put in dual irises.


Ken
 

·
Registered
Joined
·
3,679 Posts
Darin,


One other thing. I take it this is what you were talking to Optoma about? Does this mean it went well? Or didn't go well?


I'm sure you can't say anything about it, but just in case you can I thought I would ask.


I love it when staffs of life long engineers miss something that a lone soul stumbles upon.


Ken
 

·
Registered
Joined
·
23,188 Posts
Discussion Starter · #4 ·
Quote:
Originally Posted by KenLand
Maybe you can explain something that puzzles me. You use the word transition, but the two dimes are really the on "beam" and the "off" beam.
Actually, the "off" beam goes toward the light sync and so we don't see that. What we see for "off" is really represented by the quarter. I know it got a little complicated in there and I may not have explained it that well.

Quote:
Originally Posted by KenLand
Since to turn off each mirror must go from 0 degrees to 1 degrees to 2 degrees ... to 12 degrees to be fully "off". What is the effect of these interim positions smearing light across all 12 degrees during a flip.
There is the transition like you said, but what I see with no iris is basically the two dimes (one is kind of reddish) and not really one dime plus something smeared like would be expected if the mirrors were moving very slowly. I think the reddish dime is mostly some middle state, like maybe during the spoke time between color segments.

Quote:
Originally Posted by KenLand
Obviously its so fast as to be negligible. But still it seems like it would be measurable with the smear being 0 on one side of the screen and non-0 toward the other dime.
One thing to remember is that even if there was a smear at the mirrors, it wouldn't look that way on the screen. The reason is that the lens is optically focusing the DMD onto the screen. So, if a mirror reflects some light straight into the lens and some at an angle into the lens, they both end up at the same spot on the screen. When I mentioned basically considering only half the lens and what is happening there, these two lines would be at different spots there, but after passing the rest of the way through the lens they would end up focused together.


This really gets into another level of complication where all the light off of a mirror even in the "on" state doesn't really take the same path through the lens. My dime thing was what I see, but only some percentage of the light takes those main paths. In fact, if you close an iris down far enough you will find that some pixels look like they are close to dead and I believe this is caused by them actually having a slightly different "on" angle that the rest of the mirrors. Because of lens focus TI can actually get away with not all the mirrors pointing in exactly the same direction for "on". They just have to be close enough, but with an iris closed down they have to be closer to keep from being blank spots on the screen than if there is no closed down iris. At least that is what I perceive.


--Darin
 

·
Registered
Joined
·
23,188 Posts
Discussion Starter · #5 ·
Quote:
Originally Posted by KenLand
One other thing. I take it this is what you were talking to Optoma about? Does this mean it went well? Or didn't go well?
The next step after the InFocus design is what Optoma has done IMO (higher CR, but less lumens, but still a pretty bright). I think my stuff has gone very well. I got better results than I imagined.

Quote:
Originally Posted by KenLand
I love it when staffs of life long engineers miss something that a lone soul stumbles upon.
I actually don't think I've said anything that many of the engineers working on these things don't already know. Since lumens and CR both matter I think Optoma took a good approach with their design. I just change things because I can use a high gain screen in a very dark room and live with less lumens. Kind of like with the Mocom 20 gain screen that I bought which allows going to a pretty high contrast ratio with images that are still fairly bright. I've listed that combo which includes an H79 modified with the principles I've laid out here in the classifieds here for local sale and also have another one I've modified for myself.


I do have some stuff that I haven't seen others do, but this isn't it (I'm not giving away any family jewels here). :)


--Darin
 

·
Registered
Joined
·
3,679 Posts
"One thing to remember is that even if there was a smear at the mirrors, it wouldn't look that way on the screen. The reason is that the lens is optically focusing the DMD onto the screen"


I see. No matter the angle of light leaving a single mirror, as long as it hits the lens it is focused onto the same pixel on the screen. So the "off" state has to swing the light completely out of the path of the mirror.


With all this light flying around in places and angles where you don't really want it, it's easy to see how setting up baffles or irises in the right places would help CR tremendously.


It would be fun to block all stray light at every possible position in the light path to see the ultimate results.


Ken
 

·
Registered
Joined
·
23,188 Posts
Discussion Starter · #7 ·
Quote:
Originally Posted by KenLand
With all this light flying around in places and angles where you don't really want it, it's easy to see how setting up baffles or irises in the right places would help CR tremendously.
And in some ways it just amazes me that it is possible to get such high CRs with these tiny mirrors and no extra dynamic system.

Quote:
Originally Posted by KenLand
It would be fun to block all stray light at every possible position in the light path to see the ultimate results.
This just reminded me of something that is one reason I started this thread. Many of us have thought of irises as blocking "stray light", but in truth what I found is that to get the best CR we aren't just blocking "stray light", we have to block light that is perfectly legitimate and not stray to take advantage of this halve-both-irises phenomena. I knew that irises blocked some good light, but just not to this level of synchronization. I guess that is one reason I was kind of exciting to figure that out. It almost feels like magic that halving one does very little useful stuff for CR, but halving both in unison kicks the CR up so much.


I should correct one little thing that I think I wrote earlier. I had a suspicion that I might have been wrong and when I looked I found that I was. The flatter part of the D is actually away from the dime that represents between "on" and "off" and not on that side. But things do get kind of reflected around inside the lens and so that stray light gets a little bit complicated to follow.


--Darin
 

·
Registered
Joined
·
23,188 Posts
Discussion Starter · #9 ·
Quote:
Originally Posted by drpp
For those who are intersted in one of the whitepapers, here is a link to one:

http://www.filmimaging.com/2002/06_j...whitepaper.htm
Thanks. I hadn't seen that one. On the 3rd page they said something that goes along with this:
Quote:
Note the dramatic increase in contrast when both the apertures are used, and that the addition of an aperture that matches the shape and orientation of the aperture used in the projection lens does not have a significant impact on light throughput.


It is clear from the experiments that the use of both illumination and projection pupil-shaping apertures is recommended for highest contrast.
I don't see anywhere that it goes on to address the effect of then cutting both in half, but maybe it is implied or I missed it, as I've only skimmed it so far.


Also, it seems to me that if you know that you are going to close an iris way down, then the dimple in the middle (via) doesn't really have to be flat with the rest of the mirror, it just has to be at an angle that will shoot to a spot that the iris would block. I have no idea whether it would be easier to put that via at an angle than to completely flatten it out though. If it is a normal indent shape then I think that would be bad, but see why closing the lens iris down would decrease the reflection from that one quite a bit, since it is only the part with just the right angle to make it through the iris that would matter.


Also, I noticed that the transitional state I didn't know the name for is called the "Flat" state there.


--Darin
 

·
Registered
Joined
·
9,977 Posts
I'm a little concerned in the dependency of using an iris, dynamic or not, to gain contrast. Its one thing to control transient light (light coming from the mirrors during their state transitions), but using it to raise/lower light output is a globally-effecting condition, and scenes in which the average light in the source is near-constant eliminates the DI's performance.


Also, I am having a little difficulty in understand your descriptive analogy. I get the quarter representing the lens area for which light 'can' pass through, the first dime representing the light from the mirrors in a fixed state, and the second dime for light due to transient mirror activity, but are not all mirrors in DMDs always in a transient state in order to dither color that is not pure red, green, and blue? I'm showing my ignorance here for DLP, so treat me with kid gloves. :)


Also, my concern with using the 2nd dime for the representation of the transient light seems inappropriate, because that is light that is being swept along an axis like a flashlight being whipped across a horizon. This is bound to have more than simplistic consequences that will affect the design of the iris that is implemented to control transient light, no?
 

·
Registered
Joined
·
855 Posts
Quote:
... scenes in which the average light in the source is near-constant eliminates the DI's performance.
From a contrast improvement perspective, this is correct. However, there are other possible benefits from a DI.


According to the IMAX patent, when the DI is open, the video data flows through the video path as normal. When the DI closes, the video data has a gain applied to it to compensate and keep the relative brightness constant. That is, something that has a brightness of 10 IRE must continue to have a brightness of 10 IRE. If the light is reduced by 10% by the DI then the data must be gained by some amount to get the light back.


On a system that has limited low level detail ability (think dithering DLPs :) ), this will move more of the video data OUT of the dither region and allow real bits to be used to represent that detail.


Sorry for the side track. :D
 

·
Registered
Joined
·
9,977 Posts
Huck, that went completely over my head. I take it you are suggesting that less of the processing bandwidth be utilized in the dithering and better utilized elsewhere with the addition of the DI?
 

·
Registered
Joined
·
286 Posts
Quote:
Originally Posted by TheFerret
Huck, that went completely over my head. I take it you are suggesting that less of the processing bandwidth be utilized in the dithering and better utilized elsewhere with the addition of the DI?


What he means, more simply, is if you decrease the light from the lamp reaching the imager, you have to boost the contrast to compensate for the loss of light. So if you cut the lamp output in half, you have to double the contrast to compensate. The downside to this is that you end up crushing peak whites as a result.


(Although if is is done properly, you'll never really know it.)


Mike
 

·
Registered
Joined
·
23,188 Posts
Discussion Starter · #15 ·
Quote:
Originally Posted by TheFerret
I'm a little concerned in the dependency of using an iris, dynamic or not, to gain contrast. Its one thing to control transient light (light coming from the mirrors during their state transitions), but using it to raise/lower light output is a globally-effecting condition, and scenes in which the average light in the source is near-constant eliminates the DI's performance.
First you say that you are concerned about any irises and then you mention a case that you thinks eliminates the DI's performance. I'll address that one first. If the scene is fairly dark with some black then it definitely hasn't eliminated the DI's performance just because it isn't moving. A 20IRE/0IRE checkerboard would be an example where the DI would not be moving, but would show obvious improvement. But even so, a feature doesn't have to help in every scene, just where things are weak. For instance, the fact that your CRT can do full screen blackout doesn't mean anything in the vast majority of scenes, but that doesn't mean it isn't a useful feature to have.


As far as fixed irises concerning you, why? If the ftL are good, what is the problem you see? You could get the ftL for white there with a darker screen or an ND filter instead, but with less on/off CR in light controlled conditions.

Quote:
Originally Posted by TheFerret
Also, I am having a little difficulty in understand your descriptive analogy. I get the quarter representing the lens area for which light 'can' pass through, the first dime representing the light from the mirrors in a fixed state, and the second dime for light due to transient mirror activity, but are not all mirrors in DMDs always in a transient state in order to dither color that is not pure red, green, and blue? I'm showing my ignorance here for DLP, so treat me with kid gloves. :)


Also, my concern with using the 2nd dime for the representation of the transient light seems inappropriate, because that is light that is being swept along an axis like a flashlight being whipped across a horizon. This is bound to have more than simplistic consequences that will affect the design of the iris that is implemented to control transient light, no?
While it would seem like they are sweeping between states like a flashlight, I think that in reality if you look at the light coming through the lens you will see that the light from the "on" state and the "flat" state dominate. Also, I'm starting to think that they may spend more time in the "flat" state than I had realized while going for non-black (just staying in the "off" state) and one reason I can think of for this is that for speed they may toggle between the "on" state and the "flat" state under some conditions instead of going all the way over to the "off" state. But, whatever reasons they may have, the "on" and "flat" states seem to be the dominant things coming through the lens for 100 IRE.


--Darin
 

·
Registered
Joined
·
166 Posts
All this is great, how do we get our projectors modified? There has to be a business here for people who are looking to get the most out of a particular setup?
 

·
Registered
Joined
·
9,977 Posts
Darin, fair enough about the low-IRE scenes, but I was thinking more of those ANSI-friendly scenes and certainly not comparing this to another display technology. I keep thinking about scenes like in the diner of MIB where they are wearing black suits and its a bright day outside. But maybe perceived blacks will resolve this for many, but I keep getting the fog-sensation with a DI-implemented condition. Now, this is due to my 'exposure' of DI technology as it is implemented in a low-ANSI performing LCD projector and I can see how this would be overcome with something that already has a nice ANSI performance to begin with.
 

·
Registered
Joined
·
457 Posts
Quote:
Originally Posted by TheFerret
Darin, fair enough about the low-IRE scenes, but I was thinking more of those ANSI-friendly scenes and certainly not comparing this to another display technology. I keep thinking about scenes like in the diner of MIB where they are wearing black suits and its a bright day outside. But maybe perceived blacks will resolve this for many, but I keep getting the fog-sensation with a DI-implemented condition. Now, this is due to my 'exposure' of DI technology as it is implemented in a low-ANSI performing LCD projector and I can see how this would be overcome with something that already has a nice ANSI performance to begin with.
I think what you may be missing is that DLP has a considerable advantage in ANSI contrast vs every other display tech(LCD, LCOS, CRT). I've seen you write in the past that the Sony wasn't capable of producing 6000:1 ANSI contrast........NO technology can do that........not even a 10K:1 CRT. I believe that Darin or someone else had posted a spreadsheet that could help you understand when on/off dominates a scene and when ANSI dominates........on/off is still where digitals can improve the most and (DLP)ANSI is truly reaching a point of diminishing returns. An iris can help ANSI some......but an iris(dynamic OR fixed) is most beneficial to on/off which is where we need the most help.
 

·
Registered
Joined
·
23,188 Posts
Discussion Starter · #20 ·
Quote:
Originally Posted by Erik Garci
Single-Panel DLP Projection System Optics (PDF)
Thanks. I think that is the one I didn't have the link to. The bulb not being uniform is shown on page 17 and stuff where I was using dimes to represent the "flat" state and the "on" state is pretty well covered on pages 28-29.


--Darin
 
1 - 20 of 26 Posts
Status
Not open for further replies.
Top