This paper proposed an ultra-low power bandage-type ECG sensor (the size: 76 × 34 × 3 (mm(3)) and the power consumption: 1 mW) which allows for a continuous and real-time monitoring of a user's ECG signals over 24h during daily activities. For its compact size and lower power consumption, we designed the analog front-end, the SRP (Samsung Reconfigurable Processor) based DSP of 30 uW/MHz, and the ULP wireless RF of 1 nJ/bit. Also, to tackle motion artifacts(MA), a MA monitoring technique based on the HCP (Half-cell Potential) is proposed which resulted in the high correlation between the MA and the HCP, the correlation coefficient of 0.75 ± 0.18. To assess its feasibility and validity as a wearable health monitor, we performed the comparison of two ECG signals recorded form it and a conventional Holter device. As a result, the performance of the former is a little lower as compared with the latter, although showing no statistical significant difference (the quality of the signal: 94.3% vs 99.4%; the accuracy of arrhythmia detection: 93.7% vs 98.7%). With those results, it has been confirmed that it can be used as a wearable health monitor due to its comfortability, its long operation lifetime and the good quality of the measured ECG signal.

No abstract

The authors present a practical design process that considers the power noise problem in CPU blocks for application processors used in smart TVs. The target impedance is determined by modelling the RLC circuit of a system-on-chip power net. The target impedance of a power delivery network is then determined by applying the extracted chip current profile for finalising the design budget. The authors modelled the on-chip power net by combining vector network analyser measurements with an on-chip model for power integrity analysis. The authors demonstrated the optimisation and design strategy by using a ball grid array ball interconnection and case studies on the placement of multilayer ceramic capacitors. The simulation results showed good agreement with the measurement results. The error in the minimum value (negative direction) by voltage droop was less than 8.6%, while the difference in voltage noise ripple was 2.69% for a criterion of 1.1 V assuming a worst-case condition of 1.2 V.

No abstract

No abstract