E14 Measurement, Resolution and Physical Limits

Topic

The resolution of any sensor—whether physical, biological, or artificial—is limited by fundamental laws of physics, not merely by the specific technology used to construct it. Consequently, the Rayleigh criterion determines the minimum angular resolution achievable by any optical sensor, based on the size of its aperture and the wavelength of the light it captures. According to this criterion, the human eye achieves a minimum angular resolution that remarkably matches the actual spacing of the cones in the fovea; this demonstrates that the biology of the human eye has, in practice, already reached this fundamental physical limit. Increasing the aperture of an optical sensor, or decreasing the wavelength of the captured light, results—under the Rayleigh criterion—in an improvement of the sensor's minimum angular resolution.

The Nyquist theorem establishes that correctly sampling a signal with a specific maximum frequency requires a sampling rate of at least twice that maximum frequency. The human ear detects sounds up to a certain maximum frequency, and the cochlea consequently samples at an effective rate close to double that maximum detectable frequency. If a biological sensor samples a signal below the threshold required by the Nyquist theorem, aliasing occurs—a phenomenon in which a high frequency within the signal is incorrectly perceived as a much lower frequency than it actually is. Weber's law states that the difference threshold of any sensor—the minimum difference in stimulus intensity that the sensor can still distinguish—is proportional to the background level of that stimulus; however, the constant ratio between this difference threshold and the background level varies depending on the sensory modality involved, differing for visual brightness, auditory sound intensity, the weight held in a hand, or a taste perceived in the mouth. Ultimately, Weber's law provides the empirical basis for the logarithmic Weber-Fechner law—another model in this set—implying that these biological sensors are optimized for signals that vary on a logarithmic scale rather than a linear one. Because the difference threshold of a biological sensor is a constant fraction of the background level (according to Weber's law), the sensor can discriminate between a weak signal and an intense signal with equal ease, provided both vary by the same relative proportion.

Meanwhile, the Cramér-Rao bound establishes the theoretical minimum variance achievable by any unbiased statistical estimator; an organism's brain acts—in response to an environmental stimulus—as a statistical estimator that, under well-controlled laboratory conditions, approaches the optimal limit set by the Cramér-Rao bound. Fisher information quantifies precisely how much information about the stimulus—the same information addressed by another model in this set—is contained within the complete distribution of neural responses elicited by that stimulus within the organism. The fact that the distribution of these neural responses contains more Fisher information regarding the stimulus implies—based on the aforementioned Cramér-Rao bound—that the brain can estimate the stimulus with even lower variance.

Consequently, this model—framed by the Rayleigh criterion and the Cramér-Rao bound—dictates that the spatial resolution of any sensor (whether in vision, radar, or biological sonar) depends on the distance to the observed object, the sensor's aperture, and the wavelength employed, exactly as established by the Rayleigh criterion. Thus, improving spatial resolution at a fixed distance requires either increasing the sensor's aperture (such as with a larger eye or a longer antenna) or reducing the wavelength used. In contrast, an insect's compound eye trades angular resolution for a much wider field of view, adopting a significantly smaller aperture than an equivalent camera-type eye; this represents a distinct evolutionary solution that sacrifices fine resolution in exchange for a much broader field of view.

ID:2383