Implantable neural interfaces face a structural tension: richer neural information demands more power, yet safe chronic operation is bounded by tissue heating and by the efficiency of wireless power delivery. This review develops a quantitative framework for the coupled power–data problem. The framework decomposes the raw data rate into channel count, sampling rate, and bit depth; partitions implant power into six sub-budgets—analog front-end, digitization, processing, telemetry, stimulation, and overhead—and constrains operation by a calibrated thermal budget and a depth-, alignment- and load-dependent wireless-power link. Five co-design lever families are evaluated: data compression, spike and event detection, adaptive sampling and on-chip processing, telemetry architecture, and wireless power architecture. Two criteria are derived: a per-channel break-even condition for net power saving, and an architecture-level necessity threshold at which raw-data telemetry exceeds the thermal budget. A global sensitivity analysis over seven parameters shows that the threshold lies between roughly 10 and 103 channels across the plausible parameter space, and that on-chip processing raises the feasible channel count by at least a factor of five in 95% of that space. Applied to motor brain–machine interfaces, epilepsy monitoring, adaptive deep brain stimulation, and visual prostheses, the analysis shows that the optimal operating point is application-dependent, and that the thermal budget rather than a universal channel count determines when on-chip reduction becomes mandatory. Closed-loop latency is analyzed in two regimes: where telemetry transit time is comparable to the latency budget it independently forces on-implant computation, whereas where the biomarker estimation window dominates, power rather than latency is binding. Open challenges include standardized end-to-end latency benchmarks, stimulation artifact–recording trade-offs, and the drift of the power budget under chronic electrode–tissue evolution.
Hiding that a transmission occurs constrains how the transmitting aperture must be built, yet the adversary model in common use cannot express that constraint. This paper poses covert transmission from a continuous aperture as a realization problem: how many spatial degrees of freedom must the aperture carry, and how densely driven? When the warden occupies one known point its leakage is rank one, any aperture with two degrees of freedom nulls it exactly and the constraint never binds. Imposing covertness over a bounded region of warden positions removes this degeneracy. The optimal current is proved to span the legitimate and region responses, its dimension set by the adversary model, not the aperture sampling. A closed form follows from the region leakage covariance, reaching the convex optimum within 0.51 per cent at far lower cost. The region dimension grows linearly with its radius, its slope close to doubling at the feasibility boundary. Beyond it the covert rate is affine in worst-case decorrelation, with a parameter-free slope reproduced within 0.8 per cent over five geometries. One-wavelength spacing recovers 99 per cent of the continuous-aperture rate, four times sparser than the grating-lobe rule, once power is certified on a wavelength-referenced grid.
A recent paper by Katourani et al., (2022) reports a recycling folded cascode (RFC) operational transconductance amplifier (OTA) with a 100.61dB DC gain, an 18.1MHz open-loop bandwidth, a minimum slew rate of 21.1Vµs−1, and a 72.4µW quiescent power from a 1.8V supply while driving a 150pF load. This comment examines whether these metrics are compatible with the transconductance-boosting limits of the RFC topology and with the average power required to charge and discharge the stated load under the stated large-signal test conditions. A topology-specific analysis using the current-mirror ratio K=5 that the authors themselves report shows that the reported bandwidth exceeds the RFC boosting bound by a factor of approximately 2.6, and that reproducing the reported value would require K≈14.5, roughly three times the value used in the authors’ own design. The large-signal analysis shows that the average power required to drive the load is approximately 594µW, roughly eight times the reported quiescent power, and that the paper does not report the supply-current waveform needed to verify the energy balance. A dimensional error in the large-signal equations and several reporting inconsistencies are also identified. The conclusions are framed as severe, unresolved inconsistencies rather than as a proof of physical impossibility, because the original paper does not report the operating-point data needed to close every remaining ambiguity.
Electromagnetic methods are widely used for resource exploration and underwater target detection in complex marine environments. In this study, a three-dimensional finite-difference time-domain (FDTD) model is developed to calculate the radar cross section (RCS) of submerged targets and examine its dependence on incident frequency, observation angle, and target depth in a layered air–seawater environment. The model accounts for the dispersive and lossy properties of seawater and the interaction of target scattering with the air–seawater interface. The results show that the target-induced RCS enhancement is most pronounced in scattering directions nearly perpendicular to the incident wave. Quantitative analysis further shows that the angularly averaged absolute RCS-level difference between the target-present and target-absent cases decreases with target depth, while lower frequencies preserve greater separation at larger depths. By integrating angular-frequency mapping with this difference metric, the study offers a unified RCS characterization of underwater-target scattering. The resulting framework provides a basis for optimizing detection frequency and observation geometry in future underwater electromagnetic sensing systems.
