Patent – 8,049,900 Apparatus and methods for uniform frequency sample clocking

This was an amazing project for many reasons. The swept source lasers we used for OCT were linear in K-Space but not in time. A Fourier Transform (FFT) assumes that the time domain signals are sampled uniformly in time. If there is no correction for the wavelength change over time, this can lead to effects that appear like dispersion. An approach that we pioneered with this patent is to sample the Analog-To-Digital Converter (ADC) with a signal derived from the laser sweep. This signal is not uniform in time, but uniform in K-Space and therefore must be re-sampled. If the signal is not, then the FFT signal will appear as if the light was dispersed, and not very useful.

Resampling can be done in a number of ways, but we proposed a method to have a 2nd interferometer that was sampled in real time by a circuit I designed that converted the sinusoidal, frequency modulated signal, to a digital clock that was used to control the sampling of the analog to digital converter.

The analog frequencies were on the order of 15 to 80 MHz, so the project was a good first mixed signal project.

What did it teach?

  • The sampling time can be optimized for the signal and need not be linear
  • The FFT assumes linear spacing and can exhibit artifacts if this is not taken into account.
  • Different devices have different “master clocks”
  • Lot’s of board layout tips and tricks:
    • Be careful of ground planes under chip pins, as this can cause stray capacitance
    • Analog and digital grounds need to be carefully planned
    • Hand soldering and hand modifying of PCB