Coffeehouse Post

Single Post Permalink

View Thread: Project Natal + Windows
  • User profile image

    BitFlipper said:
    aL_ said:

    Yea, this sounds very promising:

    "The technology performs superior depth imaging (depth resolution of millimeters) in real-time (60 fps or more), using little or no CPU"

    One thing I think people should realize:

    The actual 3D camera concept is pretty simple, if you look at what each pixel does. It is only slightly more complex than what a typical, cheap CCD does in a consumer camera. The biggest reason why it has been so expensive up to now is that our technology just wasn't fast enough to perform the shutter speeds required to make this work (the technical description I saw showed the actual "shutter" as a multiplexing electronic switch, one for each pixel).  Once we have the required speed, it becomes much more viable to mass produce this, resulting in much lower cost.

    That is why I don't believe it is so far fetched that something like this can end up costing $200. Once again, not much more than a set of Guiter Hero controls, and we all know what a "failure" that turmed out to be Wink


    EDIT: This is how I understand it to work:

    Each pixel in the 3D camera CCD consists of one light detector, and two capacitors (as opposed to one detector and one capacitor for a typical CCD). Initially, both capacitors start out discharged. An IR light pulse is sent out. The first capacitor it connected to the light detector.  Exactly halfway through the cycle, the light detector is switch over to the second capacitor. At the end of the cycle, the difference between the voltage of the 1st and 2nd capacitor describes the distance at that pixel.  Due to the fact that the difference between the two capacitor is used as the actual value to process, the level of ambient light is cancelled out since it will cause each of the two capacitors to have the same value.  Only the dynamic light is having an effect on the final output. If the object is close, both capacitors will have almost the same voltage. The further the object, the longer the first capacitor will be without the light from the pulse, resulting in it having a lower voltage at the end of the cycle.

    Yes this is an oversimplification, but this is how I understand the basic functionalty.  Not that much more complex than the current mass-produced CCDs we find in cheap cameras. The biggest challange seems to be to get the high speed of switching required to make it work. Since they have now reached that speed, it now becomes possible to mass-produce it much more cheaply.

    If it's true, this is bigger then a video game controller. There are many many many many many

    many many many

    applications that have been waiting for an inexpensive 3d range finder