International Journal of Terahertz Science and Technology
  TST >> Vol.11, No.4, December 2018: PP. 102-123

(Invited Paper)Characterization of the noise behavior in lens-integrated CMOS terahertz video cameras

Robin Zatta *, Ritesh Jain and Ullrich Pfeiffer
Institute for High-frequency and Communications Technology, University of Wuppertal, Germany
* Email:

View Full Text: PDF

Abstract: This paper reports on the detailed experimental investigations of the NEP behavior in 1k-pixel terahertz video camera module implemented in 65nm CMOS technology which can act as THz imager or power meter. The NEP is characterized for both operation modes, imaging and power metering, for different frame rates, averages, and clipping windows, to understand the trade-off impacts of different camera operation modes on the camera sensitivity for both modes and their limitations. Mean offset and noise trends are also investigated over time in a temperature-controlled environment to analyze the camera performance under long integration periods which show that a warm-up time of 60 minutes should be respected for very accurate measurements. Moreover, our results verify that the averaged single pixel real-time NEP is relatively stable over different readout frame rates and shows a minimum typical NEP at 822 GHz of 17 nW for different frame rates, within the wide 3-dB bandwidth between 740 and 930 GHz and hence, the minimum NEP for 1 second integration time at a frame rate of 30 fps is 3.1 nW. This means that a 100 pW THz signal can be detected for an integration time of 31 seconds, respectively. The single pixel NEP can be further improved through averaging over K frames due to a noise reduction by , which is helpful by using the camera module as an imager. Thereby, the RMS pixel noise is integrated over the whole video bandwidth which is limited by a LPF with a cut-off frequency of 150 kHz. For determining NEP, the  is measured through a 26-dBi illumination source from 0.6 to 1.1 THz. Using the camera in power meter mode the NEP increases by  for selecting  pixels at the readout and hence, the typical minimum NEP for the whole frame is 544 nW for different frame rates. This can also be improved through integration and the single pixel performance can be reached again after 60 seconds which corresponds to an average of 1800 frames. Thereby, the fixed pattern noise has to be measured periodically due to thermal drifts for long integration periods because otherwise the noise and hence, NEP saturates. Furthermore, the directivity of the pixels is analyzed and shows that it decreases from center to outer pixels due to the change in effective aperture at offset locations of the optical axis but most of them are still in a 3-dB range. For best performance the signal should be centered with respect to the image.

Keywords: Sub-millimeter imaging, Terahertz direct detection, Terahertz imaging, Noise, NEP, Silicon lens

Received: 2018-10-3

Published: 2018-12-31

Acknowledgments: The authors would like to thank TicWave GmbH, Germany, for support and for providing the camera module for the performance characterization in this paper.

Cite this article:
Robin Zatta, Ritesh Jain and Ullrich Pfeiffe (Invited Paper) Characterization of the noise behavior in lens-integrated CMOS terahertz video cameras [J]. International Journal of Terahertz Science and Technology, 2018, Vol.11, No.4: 102-123.  DOI:10.11906/TST.102-123.2018.12.10




Print | close

Copyright© 2008 Scinco Inc. All Rights Reserved
P.O.Box 6982, Williamsburg, VA 23188, USA