Arxiv.org paper "Determining Image Sensor Temperature Using Dark Current" by Richard Matthews, Matthew Sorrel, Nickolas Falkner from University of Adelaide, Australia claims that the authors were able to observe pixel variations in SEM:
I am always dubious of papers on Arxiv, because I think most of the time, peer review validates and improves paper quality. For this work, I would recommend that the authors seek peer reviewed publication of this work, and I especially would recommend that the device related aspects, which I think are weak here, get strengthened. I note one of the authors did attempt to contact me on Christmas Day for more information about radiation effects on dark current and how that might impact the forensics of camera identification. (In my mind, I replied, but now I can find no trace - sorry RM). BTW, if subtle pixel-to-pixel non-uniformity causes could be readily discerned by SEM imaging, it is possible it would have been done already. Just sayin'
The "features" in Figure 1 annotated as dielectric isolation likely aren't features at all. They appear to be sample preparation artifacts common on FIB prepared SEM samples.
The features in Figure 1 were remeasured in the lab and we were able to consistently replicate. Still, the concern remained, as you suggest, that it could merely be a scanning artifact of the FIB beam. We followed up these results with the onsite technician who is the certified expert on our helios regarding our concerns. The Adelaide Microscopy technician advised, that since it was repeatable, the measurement settings we were using it was unlikly to be an artifact and more likely to be a feature.
I want to revist this study on a wider sample of sensors when funding permits. Currently, I don't have the research funds to allow such a further study.
I am always dubious of papers on Arxiv, because I think most of the time, peer review validates and improves paper quality. For this work, I would recommend that the authors seek peer reviewed publication of this work, and I especially would recommend that the device related aspects, which I think are weak here, get strengthened. I note one of the authors did attempt to contact me on Christmas Day for more information about radiation effects on dark current and how that might impact the forensics of camera identification. (In my mind, I replied, but now I can find no trace - sorry RM).
ReplyDeleteBTW, if subtle pixel-to-pixel non-uniformity causes could be readily discerned by SEM imaging, it is possible it would have been done already. Just sayin'
The "features" in Figure 1 annotated as dielectric isolation likely aren't features at all. They appear to be sample preparation artifacts common on FIB prepared SEM samples.
ReplyDeleteThe features in Figure 1 were remeasured in the lab and we were able to consistently replicate. Still, the concern remained, as you suggest, that it could merely be a scanning artifact of the FIB beam. We followed up these results with the onsite technician who is the certified expert on our helios regarding our concerns. The Adelaide Microscopy technician advised, that since it was repeatable, the measurement settings we were using it was unlikly to be an artifact and more likely to be a feature.
DeleteI want to revist this study on a wider sample of sensors when funding permits. Currently, I don't have the research funds to allow such a further study.