Theoretically and Experimentally Demonstrated Enhanced-Speed Thermoreflectance Imaging

Charge coupled device (CCD) -based thermoreflectance imaging using a “4-bucket” algorithm to essentially provide pixel-by-pixel lock-in imaging is a well-established research tool for obtaining two-dimensional thermal maps of optoelectronic, electronic, and MEMs devices with high spatial and thermal resolution [1,2]. However, the technique is relatively slow, limiting broader commercial adoption. In this work, we show that the image acquisition speed is fundamentally limited by an underlying statistical bias in the 4-bucket imaging algorithm. Furthermore, the straightforward extension to an n-bucket technique by faster sampling fails to address this statistical bias and hence does not improve image acquisition time. Instead, we develop a modified “enhanced n-bucket” algorithm that halves the image acquisition time for every doubling of the number of buckets. We derive detailed statistical models of the algorithms and confirm both the models and the resulting speed enhancement experimentally. In addition, we investigate optimizing a stochastic resonance imaging enhancement and explore other, novel avenues to increase the speed of the imaging technique.
[1] M. Farzaneh, et. al, J. Phys. Appl. Phys. 42, 143001 (2009).
[2] P. M. Mayer, et. al, JOSA A 24, 1156 (2007).