Sebastian Hoyos

State-of-the-art portable ultrasound imaging systems employ a small transducer array and a low carrier frequency to fit stringent constraints on power and form factor, and this tends to compromise the ultrasound imaging quality. In this paper, we present a low-complexity low-power asynchronous compressed beamformer (ACB) for portable diagnostic ultrasound. The proposed ACB integrates asynchronous sampling and compressive sensing (CS), and is capable of reducing data conversion power and handling a large data volume at the mixed-signal interface. A high-rate continuous-time ternary encoding (CT-TE) scheme eliminates the need for interpolation filters and coordinate rotation digital computer (CORDIC) units typically used in a conventional architecture. A split-projection least squares (SPLS) signal reconstruction algorithm is applied that replaces high-cost nonlinear signal recovery with a series of low-complexity and independent linear problems. Experiments with measured ultrasound data demonstrate the proposed ACB architecture, and the SPLS reconstruction algorithm achieves 9-fold data compression compared with Nyquist sampling.

Compressive sensing (CS) is a promising technique that enables sub-Nyquist sampling, while still guaranteeing the reliable signal recovery. However, existing mixed-signal CS front-end implementation schemes often suffer from high power consumption and nonlinearity. This paper presents a digital-assisted asynchronous compressive sensing (DACS) front-end which offers lower power and higher reconstruction performance relative to the conventional CS-based approaches. The front-end architecture leverages a continuous-time ternary encoding scheme which modulates amplitude variation to ternary timing information. Power is optimized by employing digital-assisted modules in the front-end circuit and a part-time operation strategy for high-power modules. An S-member Group-based Total Variation (S -GTV) algorithm is proposed for the sparse reconstruction of piecewise-constant signals. By including both the inter-group and intra-group total variation, the S-GTV scheme outperforms the conventional TV-based methods in terms of faster convergence rate and better sparse reconstruction performance. Analyses and simulations with a typical ECG recording system confirm that the proposed DACS front-end outperforms a conventional CS-based front-end using a random demodulator in terms of lower power consumption, higher recovery performance, and more system flexibility.

Wideband spectrum sensing for cognitive radios requires very demanding analog-to-digital conversion (ADC) speed and dynamic range. In this paper, a mixed-signal parallel compressive sensing architecture is developed to realize wideband spectrum sensing for cognitive radios at sub-Nqyuist rates by exploiting the sparsity in current frequency usage. Overlapping windowed integrators are used for analog basis expansion, that provides flexible filter nulls for clock leakage spur rejection. A low-speed experimental system, built with off-the-shelf components, is presented. The impact of circuit nonidealities is considered in detail, providing insight for a future integrated circuit implementation.

We consider a mixed-signal ultra wideband (UWB) direct-sequence spread-spectrum (DS-SS) receiver. The receiver expands the signal over a basis set, and then operates on the basis coefficients. An analog computation of the basis coefficients efficiently parallelizes the signal for digital processing, relaxing the sampling requirements and enabling parallel digital processing at a much lower rate. Bit error rate (BER) performance is evaluated, as well as distortion due to basis truncation. The parallel processing also facilitates efficient interference mitigation. Various models for interference including adjacent multicarrier interference are developed and analyzed. Several simulation results confirm the analysis and provide insight into quantization error, basis truncation distortion, and the impact of the UWB channel. We show that in the AWGN uncoded case for SNR < 10 dB, the receiver can be operated at almost half the Nyquist rate with only 1 dB of performance loss. Moreover, when a common convolutional code is employed, the aggregate sampling rate can be 37.5% lower than Nyquist rate with little loss in receiver performance.