A. Ibanez

A digital envelope detection filter for real-time operation

A time-domain method to extract the envelope of an amplitude modulated signal at high speed is presented. This method, the envelope detection filter (EDF), is based on a nonlinear function of two consecutive samples of the input sequence. In spite of its simplicity, EDF provides a quite good approximation to the analytical signal magnitude for a wide range of signals and conditions. The paper presents the basis of the technique and analyzes the sources of error and its sensitivity to noise and signal bandwidth. Results of experiments with simulated signals are in good agreement with the theoretical model. Furthermore, EDF allows a straightforward implementation using a pipeline of two registers and a memory look-up table. The hardware implementation of EDP has been tested at a 10 MS/s, although current technology could achieve several times this throughput rate. Results obtained with EDF for real data of nondestructive testing experiments, are compared with those provided by the full precision analytical signal magnitude obtained by a software Hilbert transform. In all cases, no significant differences are appreciated between the analytical signal magnitude and the EDF output, which is computed in much less time.

The progressive focusing correction technique for ultrasound beamforming

This work presents a novel method for digital ultrasound beamforming based on programmable table look-ups, in which vectors containing coded focusing information are efficiently stored, achieving an information density of a fraction of bit per acquired sample. Timing errors at the foci are within half the period of a master clock of arbitrarily high frequency to improve imaging quality with low resource requirements. The technique is applicable with conventional as well as with DeltaSigma converters. The bit-width of the focusing code and the number of samples per focus can be defined to improve both memory size and F# with controlled timing errors. In the static mode, the number of samples per focus is fixed, and in the dynamic approach that figure grows progressively, taking advantage of the increasing depth of focus. Furthermore, the latter has the lowest memory requirements. The technique is well suited for research purposes as well as for real-world applications, offering a degree of freedom not available with other approaches. It allows, for example, modifying the sampling instants to phase aberration correction, beamforming in layered structures, etc. The described modular and scalable prototype has been built using low-cost field programmable gate arrays (FPGAs). Experimental measurements are in good agreement with the theoretically expected errors

Dynamic focusing through arbitrary geometry interfaces

This paper introduces the fast focal law calculator (FFLC), a Newton-Raphson based algorithm that performs such task accurately at high speed. It is especially well suited for dynamic focusing through arbitrary geometry interfaces, where other algorithms are order of magnitude slower. In spite of the high speed of the FFLC, errors are kept very small, typically within a few tens of picoseconds. Besides a short background theory, the paper compares the results of the FFLC with regard to exact solutions (for planar interfaces) and those based on search algorithms. Field simulations are performed to assess the correctness of the method. Also, experiments are carried out with a curved interface showing the advantages of the FFLC for dynamic focusing to improve the image quality and the flaw detection and evaluation capabilities.

A new beamforming process based on the phase dispersion analysis

To improve the results of the Delay-And-Sum (DAS) beamformer a weighting factor based on a measurement of the phase dispersion of the signals in the image point under consideration has been proposed. With this objective, the spectral analysis of the phase dispersion is used here to obtain a new descriptor that can be easily computed and introduced as a ponderation factor in the beamforming process. Theoretical results show that, for a linear array of 32 elements operating in synthetic aperture technique with dynamic focussing in emission and reception (Total Focussing Method) the improvement can reach up to 24dB. In order to validate the theoretical hypothesis, experimental results obtained using a linear array of 64 elements (2,4MHz) and a CIRS 040GSE phantom also are presented.

Comparison of CW beam patterns from segmented annular arrays and squared arrays

Typically, 2D-squared (S) arrays have been proposed to generate ultrasonic volumetric images. This work presents segmented annular (SA) arrays as an alternative to S arrays for that purpose. In this sense, the beam patterns radiated by both apertures in continuous wave conditions are compared. The well-known spatial impulse response method is used for simulations and a far-field approximation for the array elements is employed. Comparisons are made assuming that both radiating apertures have equivalent characteristics. This means that, on the one hand, full radiating apertures of similar size (implying similar radiated energy) are considered and, on the other hand, both arrays are formed by equal number of emitting elements which have equivalent area (meaning similar electronic complexity). Side lobes and grating lobes at the focal plane are analysed in steering conditions for both apertures. CW beam patterns from S and SA arrays are also compared by varying the steering angle, the number of array elements, and the aperture size. The benefits of using segmented annular arrays are shown.

P2D-4 A Front-End Ultrasound Array Processor Based on LVDS Analog-to-Digital Converters

This work presents a beamformer based on standard LVDS 12-bit analog-to-digital converters, which deliver the samples serially over a differential pair. The bit stream is synchronized and converted to a sequence of words in a FPGA. Beamforming is then carried out following the progressive focusing correction technique described in a previous work. The low pin count of serial converters and the FPGA ball grid array package allowed a high level of board integration. A scalable architecture allows many different configurations. Some of the possible applications are: phased array, linear scan, multi-channel ultrasound systems, SAFT imaging and TOFD techniques. The configuration is changed by software or just installing more modules in a back-plane. This work describes the system architecture, analog and digital preprocessing and assesses the performance with experimental images from a test block and tissue phantoms

Ultrasonic imaging of solid railway wheels

This work presents an ultrasonic non-destructive technique for in service railway wheels inspection. To overcome the problem of accessibility to critical regions where cracks may appear this method uses phased-array with immersion coupling, which allows carrying out the inspection process on non-dismantled wheels sets. The acquired frames are subjected to compound imaging methods in order to improve the automatic detection of cracks. The proposed method provides low inspection time, high image quality and reliable crack detection capabilities.

Design method for large 2-D ultrasonic arrays with controlled grating lobes levels

A major problem in developing 2-D arrays is the large number of elements required for big apertures in order to produce high quality ultrasonic images. Recent advances in transducer technology have made possible to design arrays with fewer restrictions in the shape and layout of elements. In this paper we propose to use Evolutionary Algorithms to optimize these parameters to obtain large 2-D ultrasonic arrays with a low number of elements but maintaining a narrow mainlobe and low grating lobes level.

An Ethernet-Based Ultrasound Signal Processor

In many ultrasound non destructive evaluation (NDE) applications, the transducers are located far away from the evaluation site. Traditionally, low level analog signals are sent over long lines to the ultrasound instrument, which give rise to noise problems. In this work a different approach is followed by integrating in a single FPGA, digital processing functions together with an Ethernet link for several ultrasound channels. These functions include a set of digital signal processing functions to improve the information content of the output data stream: linear and non linear digital filtering, envelope extraction, multiple gates with programmable thresholds and data compression. Following the concept of system-on-a-chip, a processor is integrated in the FPGA. Although it is not involved in signal processing, it performs most of the internal and external communications and eases system upgrading. The Ethernet link implements the TCP/IP stack and allows networking for bidirectional communication with a host. Implementation in the FPGA hardware allows achieving high data rate transfers. Data packets are formatted in frames, which include information on status, position, peaks and processed data.

A pipelined architecture for high speed automated NDE

Automated ultrasonic inspection of large structures usually demands both high scanning speed and accuracy. Digital signal processing techniques help to achieve the latter, but a very high computing speed is required. This paper presents a pipelined architecture directed towards NDE applications which supports high data processing rates. At the same time, it is flexible for hard real-time NDE applications. It is built as a pipeline of specialized Processing Modules (PMs) which operate concurrently. Several PMs have been developed, each one implementing a specific digital processing task for NDE applications: digitalization, trace averaging, attenuation-distance compensation, multi-gate generation, multi-peak detection, deconvolution and split-spectrum processing. Most of these algorithms can be executed at pulse repetition frequencies above 8 kHz.