| Popis: |
The architecture used for driving linear micromirrors in raster scanning systems is typically composed of digital circuits, responsible for generating a sawtooth-like reference signal synchronized to the fast axis, and analog circuits responsible of driving the device. Bridging the two domains is the D/A converter, typically clocked in the MHz range, whose noise sources and distortion affect the accuracy of the scan line. With typical refresh rates in the order of 60 Hz, simulating transistor-level implementations requires up to days for a few operating cycles (i.e. 1.7 cycles/day). This drives the need for accurate models of the dominant noise sources and their impact on scan accuracy, able to achieve verification times compatible with typical design flow timelines. The architecture of this work is composed of a sigma-delta based current-steering D/A converter, which is modelled analytically and behaviorally with its white and flicker noise sources and non-idealities. Each current generator is modelled independently to capture time-variant effects. The developed model can accurately predict noise both from Cadence simulations and experimental measurements, while also reducing simulation times by three orders of magnitude (i.e. 5.7 cycles/minute). The model thus allows to optimize the design and quickly verify the possibility to achieve a tilt-angle rms noise within 1 m◦ in open-loop driving conditions. Experimental results also show a significant distortion, which is not predicted by the model: as hypotheses on its root causes are formulated, the model will enable their investigation within reasonable times. |
