This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Is there a reason nobody uses two dmds to boost contrast?

If a single-chip dlp projector employed a second dmd that followed after the first in a single light-path, and the second chip simply slaved a mirrored image of the first..wouldn't it multiply its contrast upon the first? You would lose light efficiency, but wouldn't need to complicate video processing since the dmds are performing the exact same actions, and you would easily gain an enormous boost in on/off AND ANSI contrast. It seems like a fantastic and inexpensive building solution for creating a better picture, but I've never seen the design mentioned in with the other TI dlp designs. I have a hard time believing something both so simple and with such potential would be left in the dust without a good reason. My question is; what am I missing? Is there a major flaw that kills the design or some other glaring reason this isn't being done?
  • Hi.

    it's an interesting idea. We do have design for having two controlers interface to DMDs, but our controller cannot talk to 2 DMDs at the same time. It's for bandswidth, mainly for cost issue.

  • So it would always require two controllers for bandwidth reasons, but would otherwise be very much possible? It just seems strange to me that manufactures would spend so much into small design tweaks, better optics, more specialized DMDs and other expensive parts/designs to squeeze out tiny contrast improvements when they could simply align a second inexpensive DMD+controller and receive a 100-TIMES contrast increase. A 1000X increase should be possible with the better DMDs. You could cheapen several aspects of any design and still come out quite a ways ahead in performance. Think of the improvement you could make for off-sets and zoom-range and lens-angles with high enough contrast that you can afford to throw some away for greater flexibility AND still have an impressive contrast ratio leftover! Is there anything TI can do to help this along? A single(double)chip dlp with the placement flexibility of an LCD, on/off contrast exceeding LCoS, with all the motion and 3D benefits of DLP..and it could be made with cheaper components and higher profit margins. A lower price moves more units and each unit moved sells TWICE as many TI DMDs. The consumer, PJ manufacturer and TI all profit..everybody wins! Is there anything holding this back aside from the comparatively small price of a second DMD/controller and a fairly straightforward design change?
  • The concept here is not new. Somebody has done it before. DLP Cinema has taken this approach before and indeed showed very impressive contrast. However, For mainstream applications, it is too expensive (2 dlp chipsets + optics) and too inefficient (~68% efficiency per dmd) to be practical.

  • Would a consumer be willing to trade a 32% brightness loss (minimally more than the loss from eco mode) in exchange for going from 1000:1-2000:1 up to 100,000:1-200,000:1? The price of high-end parts currently used to achieve 5000:1-8000:1 contrast at a solid profit margin dwarfs the comparatively small $800or less increase to make a 100,000:1 single-lightpath device. Hometheater screens up to 135" only require 800 lumens to reach 14ftL, and a 110" screen only requires 500lm. People seem more than willing to pay $3000 more for a 50% brightness loss and 10X contrast gain over a $1000-1500 DLP; it happens every time somebody chooses a JVC or Sony LCoS. Tell a consumer they can instead pay $800 more, lose 32% brightness and gain 100X contrast by staying with a DLP device and anyone would be pleased.
  • Thanks for the information so far. It's understandable that years ago such a design wouldn't make sense when brightness was hard to come by, DLPs led the pack for contrast, and a watchable unit already cost $8000-10,000. But now there's brightness to spare, $2500-4500 LCD and LCoS units are slaying DLP contrast levels, and parts prices are allowing for $500-800 DLPs that rival the high-end of old. If all it takes is an old design, $500-800 more parts on top of an $800machine, and the will the try it again to create a 800lumen, 100,000:1CR, $1,600 device.. It looks like the old design has become relevant. What other hurdles are there? I have to guess there are more reasons this hasn't popped back up.
  • The two DMD approach you mentioned is actually built into product by Zeiss planetarium http://www.zeiss.com/planetariums/en_de/products/products/velvet.html.

    Any customer would like to have this design, we certainly will support it.

  • Are there any particular licenses or rights that need to be bought/rented for others to use that design or a very similar one?
  • By all means, I want a consumer two-chip DLP to become a reality. As Dus Java pointed out it's a much more efficient upgrade to a DLP than basically everything else including a change of a light engine (LED, laser). Like was already pointed out it's a win-win situation for everybody. Frankly speaking, without this upgrade I can't see how DLP can compete with LCoS projectors in the middle price market (2500$-8000$) in the nearest future.

  • I first thought two DMD's in series would be too complex/expensive because of the off-axis light path, but might it be as simple as 45 deg bounces off of two DMD's that are 90 deg from each other, with the rest of the optical path the same as before?

    I'd love to have a DMD projector for the simplicity and robustness, but the contrast just doen't cut it; I'm on my third JVC for 

    The new JVC's with dynamic iris give real-world contrast >100k:1 which really puts the nail in the coffin for DLP for me.

    I'm surprised to hear DMD's have only 68% reflectivity; is that typical or does something about using two lower it?

  • Another thing: I wonder how close to squaring the CR (contrast ratio) two DMD's would actually give.

    I've long wondered how much of the CR is determined by diffraction off of the mirrors' edges, i.e. what would the CR be if the rest of the light were perfect.

    Frank, are you able to share that info?

  • Noah Katz said:

    I first thought two DMD's in series would be too complex/expensive because of the off-axis light path, but might it be as simple as 45 deg bounces off of two DMD's that are 90 deg from each other, with the rest of the optical path the same as before?

    I'm surprised to hear DMD's have only 68% reflectivity; is that typical or does something about using two lower it?

    A slightly different set of lenses would be required to keep the light-path between DMDs symmetrical and it appears there's no getting around needing a second controller rather than chaining the DMDs, but it doesn't sound terribly complex. A 3panel LCoS is still more complex. The 68% reflectivity is per chip. The gaps between pixels which are required for movement account for a large portion of unreflected light. The center also has a small point in front of the hinge where it isn't reflective. The spaces between pixels of LCD panels are much larger and make LCD panels even less efficient. Their advantage comes from the light efficiency of a 3panel design. This is why a 3chip dlp is so bright, that added 3panel/chip efficiency coupled with a 68% reflectivity is a powerful combination. 68% may sound low, but it is actually so high it allows inefficient single-lightpath DLPs to compete with the significantly more efficient 3panel/3-lightpath designs of other technologies. A second chip in a single-lightpath will steal another 32% of your light away, but keep in mind that many DLPs can easily afford this drop in a room dark enough to take advantage of the better contrast. An inexpensive 190-210 watt singlechip dlp can produce 1200lm. The second chip would force it down to 800lm. This is still brighter than several of the JVC models, and those looking for further brightness could easily step up to a 240-300watt lamp like the BenQ W1070 and W7000. This would allow a calibrated image of roughly 1000lm, which is a very high output using standard equipment that's already in mass use.
  • Yes, I'm all for 2-chip DLP, especially when combined with a solid-state light source like LED, which as I'm sure you know eliminate the colorwheel light loss and can be duty-cycled to give some of the benefit of a dynamic iris (a physical iris increases CR).

     The gaps between pixels which are required for movement account for a large portion of unreflected light. 

    I don't think so, given DMD's >90% fill factor.

  • Hi Noah,

    Thanks for your interest in two DMD design concept. Here is my reply to your previous questions.

    DMD mirrors have two digital positions, neither of which is flat. The arrangement of light angles into or out of a DMD is determined by the mirror angle positions, not some arbitrary folding or “bouncing” arrangement like the DMD was just another flat mirror. However, whatever that angle is would be no more or less difficult to arrange than a completely arbitrary 90 degree arrangement, which would not work by the way. We already have several professional and cinema customers doing this with no issues. This technique can produce contrast ratios in the range of millions to one, or to put it another way, beyond the capability of any measurement instrument to measure it. Likewise, there are dynamic iris solutions in many DLP projectors as well, and produce dynamic contrast similar to the JVC.

    We quote the DMD efficiency at about 68%, not the reflectivity, and it is a number that is valid only under a certain set of optical system conditions that are typical for a projection application using that device. Reflectivity is only one of the factors of efficiency, and reflectivity of the mirrors is much higher than that. There are other factors of physics that determine the efficiency in an optical system like a projector, and they are really no different from any other microdisplay type device. For example, there is diffraction caused by the regular periodicity of the small pixel array. We include that factor as it impacts a typical projection system using it, since the amount of diffracted energy that goes through the projector depends as much on the projector optics as it does the device. We just happen to be up front about it; other technologies may choose to ignore this factor when quoting their “efficiency”. If the device is not being used in a typical projection application, then the efficiency would have to be calculated based on that application and may be quite different. Other factors such as material spectral reflectivity, pixel fill factor, window transmission, etc. are all factors that impact all microdisplays in exactly the same ways as for DMDs.

     

    Another way to think about DMD efficiency and contrast is this: without any other optical system constraints around the use of the device, such as a projector, it can be said that the efficiency of the DMD alone is well over 90% in visible range, and the dynamic contrast of the DMD alone has NO MEANING. This is because all the light that comes into the DMD also leaves the DMD, less a small amount of absorption. Therefore, for the same reason, the DMD has no contrast. Contrast cannot be defined for a DMD without specifying the location and size of the illumination and collection pupils, i.e., the system optics. The DMD merely moves the light from one position to another; it does not absorb light to produce black, which also happens to be the main reason for the vast superiority of the DMD for brightness capability. And to quote “DMD contrast” is to quote the ability of the projector manufacturer to manage those light paths, not to reference any native capability of the DMD.

     

    If you are interested in learning more about these techniques as a producer of DLP technology products, we can help you understand these basic principles in great detail. Otherwise, the best advice is to take manufacurer’s specifications with a grain of healthy skepticism, since the conditions of measurements are rarely specified and often are given in values that are beyond capabilities for any instrument to measure. Also, specifications can be mutually exclusive in that they are NOT measured under the same condition. For example, the maximum brightness of a projector cannot be measured simultaneously with the dynamic iris condition which gives maximum dynamic contrast, and so on. As another example, there are projectors that can actually turn off the light source(s) in the projector when measuring dynamic contrast black levels. That produces a dynamic contrast ratio with as many zeroes after the 1 as you may care to type, but it is neither possible to measure nor practical for any picture content.

  • Hi Frank,

    Thanks so much for the comprehensive reply and straightening me out re efficiency/reflectivity (dumb mistake on my part).

    I understand that CR can't be determined w/o the rest of the optical system, so let me put it this way:

    If you illuminate the DMD with collimated light from a typical angle of incidence, roughly what would be the % of light reflected specularly, and what % scattered by diffraction?

    Another tack is, let's say a DLP projector has 5k:1 native CR (by which I mean on/off CR with a fixed iris); if there were no mirror diffraction, what kind of increase in CR might there be?

    If that's still not specific enough, what I'm trying to understand is how much diffraction off of the mirrors' edges contributes to DLP's not being able to advance past its current rather mediocre (sorry) native CR of <10k:1?

    > the best advice is to take manufacurer’s specifications with a grain of healthy skepticism

    I do, so I hope you won't be offended by the below.

     

    > Likewise, there are dynamic iris solutions in many DLP projectors as well, and produce dynamic contrast similar to the JVC

    The usable dynamic CR w/o overly bothersome visual artifacts like black level pumping and brightness compression seems to be about 5X the native CR.

    Whereas there may be DLP projectors claiming > 100k:1, the best measurements I've seen are ~20k:1 dynamic, which corresponds with their native CR of ~4k:1.

    JVC's do better than this natively, and the new ones with dynamic iris have measured >150k:1.

    Back to the main topic, if I've understood the discussion, a 2-chip projector w/series DMD's could be made for maybe twice the cost of a 1-chip.

    Even if it were 3X, it would still not cost more than the cheapest JVC's and should exceed their overall performance.

    So I'm mystified that a consumer projector mfgr hasn't yet done this.

    At a minimum I'd think a high end consumer mfgr like Sim2 would have done it, as they get many $k for their stuff.

  • Is there any chance of TI manufacturing a simplified reference build to nudge the design ahead and increase awareness among manufacturers?
  • I know I'm bumping an old post but this really interests me. From a price standpoint yes, it may boost the manufacturers cost up close to a $1000 more, but it seems like such a blissfully simple idea. I would imagine the light engine looking something like this:

    http://i.imgur.com/4AMrfTH.png