45
turned off (biased) and the rest turned on (zero bias).
5.3.2 Crosstalk
To investigate the amount of crosstalk among the pixels in our THz SLM, this ex-
periment does not use an optical chopper and instead modulate only certain pixel
elements directly by applying a 3-kHz square-wave ac voltage bias, alternating be-
tween 0 and 14 V. Using a lock-in amplifier referenced to this square wave, combining
the detected THz signal at every receiver position of the raster-scan produces a trans-
mission image (Figure 5.3 inset). Signals with the largest amplitudes are concentrated
at the two modulated pixels, with only a small amount of crosstalk in the surrounding
pixels. To measure system noise requires another raster-scan with all pixels unbiased
and unmodulated (while the lock-in amplifier is still referenced to the square wave
voltage frequency). From the first data set, the ratio of the signal power at
the surrounding un-modulated pixels (due to both crosstalk, C, and noise, TV) to the
signal power at the modulated pixel, S, is calculated for each frequency (dotted curve
in Figure 5.3). Then, the ratio of the noise power from the second data set to S
from the first data set is calculated (dashed curve). The difference between the two
ratios gives the crosstalk level, independent of the system noise. This procedure is
necessary because the crosstalk is so small as to be nearly indistinguishable from the
noise.
As shown in Figure 5.3, the crosstalk level (solid curve) is larger (-15 dB) at
lower frequencies due to diffraction effects, but drops to around -30 dB near 0.33
THz. Above this frequency, quantifying the amount of crosstalk becomes challenging
because the noise-to-signal ratio increases due to the decreasing THz spectral am-
plitude, making the crosstalk indistinguishable from system noise. As a result, this