Saturday, May 18, 2013

Rambus Binary Pixel Patent Applications

WIPO published two Rambus PCT applications talking about its Binary Pixel idea:

WO2013070932 Conditional-Reset, Temporally Oversampled Image Sensor (111 pages)
by VOGELSANG Thomas, STORK David Geoffrey, LINSTADT John Eric, HARRIS James

WO2013070942 Image Sensor Sampled at Non-Uniform Intervals (74 pages)
by VOGELSANG Thomas, STORK David Geoffrey, SHEN Jie, WANG Yueyong, ALEKSIC Marko

The applications are huge and describe a lot of different pixel versions. Here are few quotes and figures giving just a taste of the Rambus' ideas:

"In various digital image sensor embodiments disclosed herein, numerous sub-diffraction-limit (SDL) pixels are clustered about an embedded sense amplifier and sensed with respect to a binary threshold to yield a collection of single-bit pixel values that can be combined to form the output of an image pixel. Because integrated-circuit (IC) process geometries permit pixel sizes below the diffraction limit of visible light, a photon striking within a SDL cluster of such "binary" pixels may activate (i.e., be detected by and exceed the threshold of) one binary pixel but not others illuminated by the same blurred source, thus decreasing the number of non-activated pixels available to detect subsequent photons from the source. Consequently, when exposed to a stationary photon flux, the rate of pixel activation slows as the exposure interval transpires (i.e., due to photons striking activated pixels), thereby yielding a logarithmic rather than linear sensitivity profile and thus a potentially much higher dynamic range than achieved through prior-art techniques (in other embodiments disclosed herein this behavior is modified while still improving dynamic range). Further, because the pixels can be scaled with decreasing process geometries, pixel densities can increase freely with process improvement, thus overcoming physical impediments faced by conventional architectures and enabling dramatically higher performance in future generations of imaging devices."

"Although theorized, hardware implementations of SDL pixels have long been deemed impractical in part because their small output voltage falls well below the noise floor imposed by the relatively long, high-capacitance path (i.e., bit line) to sense circuitry at the edge of the array. By embedding sense amplifiers within respective clusters of SDL pixels, however, and sampling the pixels as digital rather than analog data sources (i.e., sampling the pixels with respect to a binary threshold and thus as binary pixels), such noise impediments can be overcome, enabling construction of image sensors having pixel densities that scale with process improvements and superlative dynamic range."

4 comments:

  1. doesn't EPFL's gigavision sensor (http://lcav.epfl.ch/research/gigavision) constitute prior art for this?

    ReplyDelete
  2. I only skimmed the zillion pages but it clearly builds on work done at EPFL which in turn builds (in part) on my own QIS work. Nothing wrong with advancing the state of the art. There are a number of people working in this area now. I am happy to say, and there will be a few papers on the topic at IISW in just a few weeks. The results are certainly interesting and different from conventional image sensor behavior.

    ReplyDelete
  3. The description contains more words than 90% of PhD thesis in USA. It should be a hard job for examinators to figure out all the possible inventivities. This old european patent could be interesting for them: http://www.freepatentsonline.com/EP0211702B1.pdf

    ReplyDelete
  4. Feng Yang’s 2012 PhD Thesis at EPFL mentioned his work on binary image was acquired by Rambus. See p.5 below:

    http://www.kutter-fonds.ethz.ch/App_Themes/default/datalinks/YangPhDDiss2012.pdf

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.