• View Cart
  • Contact Us

Sign-up for our Newsletter

Adaptive Optics


Content on this page requires a newer version of Adobe Flash Player.

Get Adobe Flash player

THE PROOF OF CONCEPT

  • Determine in "real time" (hundred of µs) all the aberrations in an incoming laser wavefront and close a regulation loop controlling the displacement of an array of 11x11 mirrors in order to correct the wavefront distorsion.
  • Replace an expensive, time-consuming and power-hungry computer-sensor system with a small, embedded and low cost smart sensor so the solution can become scalable for installations composed of a large array of mirrors (up to 1024 e.g. VLT Very Large)

THE IMPLEMENTATION

  • Recognize Zernike polynomials (one per specific aberration). These polynomials are not so easily computed on a PC, but a CogniMem neural network properly taught can swiftly identify the N first polynomial orders.
  • When the polynomials are identified a counter-pattern is drawn on the set of micro-mirrors to correct the recognized aberrations
  • The neurons operate in KNN mode and output the corrective displacement in 10 usec (800 Hz for the fastest estimated regulation loop). Thanks to the parallel architecture of the CogniMem neural network, the response time is constant.

THE SOLUTION

  • One V1KU board equipped with an array of lenses can perform the identification task. The regulation time of the system does not depend on the number of mirrors to drive nor on the size of the micro-lens array of the smart sensor.

NEXT STEP

  • This proof of concept has established that the CM1K chip is well suited for Adaptive Optics. A network composed of multiple chips will increase the number of orders of the identified Zernike polynomials. The resolution for a specific order is presently limited by the neuron memory size equal to 256 bytes (1024 x 16 bits would be ideal).