Block diagram of the DSP board. 

Context in source publication

Context 1

... development and test of new methods for obtaining T valuable information by means of ultrasound (US) un- avoidably pass through in vitro and in vivo experimental phases. Experimental tests are typically based on the use of commercial US equipment and/or classic instruments such as waveform generators, power amplifiers, digital oscillo- scopes, and acquisition boards. Main drawbacks to such approaches are the lack of flexibility and portability, which limits the number of feasible tests and the locations where these can be performed. Unfortunately, the only assistance offered to the researcher by some commercial scanner consists of the possibility of saving a few seconds of raw radio-frequency (RF) data for possible postprocessing purposes. This feature is implemented, for example, in the Sonoline Antares TM through the Axius Direct Ultrasound Research Interface (Siemens Corporation, New York, NY), in Technos systems (Esaote, Florence, Italy), in Logiq scanners (General Electrics Medical System Company, Fairfield, CT), in the Ultramark scanners (Philips/ATL, Seattle, WA) and in a few others. A major flexibility characterizes the Sonix RP (Ultrasonix Medical Corporation, Burnaby, BC, Canada), a commercial computer-based system in which a software development kit (SDK) allows the researcher to build his own applications in the Microsoft Visual Studio envi- ronment.[1] A few development systems specifically dedicated to the US research have been implemented so far [2]–[4]. A significant example is represented by the RASMUS system [2], which aims to acquire multichannel data from multi- element transducers and to enable real-time or near real- time processing of the acquired data. Such high flexibility is obtained through four distinct modules implemented through a suitable high number of boards, including dedicated transmitter-receiver (TX-RX) boards and single- board computers. This paper describes an integrated system addressed to control one or two US probes according to highly flexible transmission and reception strategies. The novel system consists of a single proprietary board connected via USB 2.0 to a host computer in which a specific software runs. The board includes all the electronics for the transducers front end, as well as a programmable digital architecture the user can easily operate on to fit a large range of applications. The typical needs of US research have been taken into account in the board design (see Table I). Arbitrary waveforms can be transmitted to each transducer and arbi- trarily changed between consecutive pulse repetition inter- vals (PRIs). For example, it would be possible to transmit coded pulses [5] or special waveforms capable of compen- sating the transducer response through an “inverse filter” approach [6]. The received echo signals are directly sampled at RF with 14-bit analog-to-digital (A/D) converters to preserve their original features over a high dynamic range, capable of accommodating both strong signals reflected from tissue interfaces and weak echoes backscattered from blood. In cases in which baseband signals (I/Q) are preferred, the demodulation can be performed digitally by using al- most ideal multipliers and programmable low pass (LP) filters to maintain the original signal integrity [7]. Four demodulators have been implemented onboard, to make simultaneous demodulation possible by different frequencies, to obtain data from the fundamental frequency and harmonics, as could be useful, for example, in contrast agents applications [8]. A large class of processing algorithms can be applied to the input data, both in the slow-time and in the fast-time domain, including auto-correlation [9]–[11], cross-correlation [12], pulse compression [5], autoregressive methods [13], [14] and fast Fourier transform (FFT) [15]. The received echo data can be stored in a huge memory (64 Mbyte) and downloaded into a file on the host computer. The next section describes the overall system concept and provides details of each main block. An example of an application to the investigation of hemodynamic and mechanic features of human arteries is reported in Section III. Possible further fields of application are discussed in Section IV. The novel system includes all the signal conditioning and processing electronics requested in a complete US echographic equipment in a single board. The board is housed in a 180 × 200 × 55 mm box. Two connectors are available where probes with single element transducers of frequency in the range 1–16 MHz can be inserted. A USB 2.0 connection links the board to a laptop or desktop computer (Fig. 1). As highlighted in Fig. 2, the electronics included in the board may be roughly divided into four functional sections: TX-RX, processing, buffer memory, and USB interface. The TX section can independently drive two single- element transducers with digitally synthesized waveforms. The power amplifiers of the front end are designed to al- low driving a large range of transducers with different impedance, central frequency, and bandwidth. The two TX/RX channels can be arranged to work according to multiple configurations. In fact, they can be used: to con- currently investigate two separate, noninterfering regions; to investigate the same region of interest (ROI), by using each TX-RX every second PRI; in a pitch-catch configu- ration. In the RX sections, after suitable analog signal conditioning, two A/D converters capable of 64 million samples per second (MSPS) are used. The processing section is the heart of the board and is based on two powerful digital devices: a field programmable gate array (FPGA) from the Stratix family (Altera, San Jose, CA) and a digital signal processor (DSP) from the TMS320C67 family (Texas Instruments, Austin, TX). This section is fully programmable and, granting a calculation power of more than 4600 million operations per second (MOPS) and 600 million floating point operations per second (MFLOPS), is suitable to sustain intensive real-time processing of RF echo signals. The board is equipped with a 64 Mbyte SDRAM buffer in which input RF and demodulated I/Q data is stored to be downloaded, on demand, into a file in the host computer. The interface section manages the communication toward the host computer through the USB 2.0 high- speed channel that guarantees a bandwidth of at least 20 Mbyte/s with currently available computers. The system sends the results of the elaboration to the computer software for real-time visualization and receives commands from the user interface for the acquisition and elaboration setup. An audio codec (PCM3003 from Texas Instruments) also is present onboard. The stereo output channel is used to reproduce Doppler signals, obtained by suitably processing the acquired data. The transmitters embedded in this system are designed to independently drive each of the two available transducers with a fully programmable sequence of arbitrary bursts. This is obtained by using a concatenated chain of look-up tables housed in the FPGA memory (Fig. 3). The user fills each table with the samples of a specific burst. Then, by chaining the tables, the reproduction sequence can be programmed. A state machine in the FPGA produces the transducer excitation waveform by moving the samples in each table to a 64 MSPS digital-to-analog (D/A) converter. The look-up tables are scanned in the order assigned by the concatenation, one look-up for each PRI. The pulse repeti- tion frequency can be programmed from 100 Hz to 20 kHz. Whenever the reference contained in the last table of the desired sequence links to the first one, a periodical burst sequence is produced. The simple sinusoidal pulses that in standard Doppler applications are transmitted at all PRIs can be produced by involving only one table that links to itself. The software running in the computer helps the user to synthesize the transmission sequence through the graphical interface displayed on the left of Fig. 4. The open drop-down list shows the catalogue of predefined modulation functions available for each of the two transmitters. It also is possible to upload the digital samples of the waveform to be transmitted directly from a user-designed file. Depending on the chosen modulation, the appropriate parameters on the right column are specified. In the example shown in Fig. 4, the selection of “Step FM” enables the transmission of four consecutive bursts of four cycles each, with frequency stepping from 2 MHz to 8 MHz and normalized amplitude of 100. The software calculates the waveform samples and allows their preview as shown in Fig. 4 (waveform “a”). With a similar procedure, the second transmitter is programmed in “step AM” to produce four Hanning weighted bursts of increasing amplitude, each of three cycles at 3 MHz (see waveform “b” in Fig. 4). After the D/A conversion, the bursts are sent to a linear power amplifier that produces up to 100 Vpp. The front- end is designed to drive both low and high impedance elements at frequencies ranging from 1 MHz to more than 16 MHz. The transducer front end includes only minimal analog signal conditioning, the A/D conversion being performed directly after the front-end (RF sampling). For each channel, a low noise amplifier (LNA) (MAX4107, Maxim In- tegrated Products, Inc., Sunnyvale, CA) processes the received echo signal, while a programmable gain amplifier (PGA) (AD603, Analog Devices, Norwood, MA) is used to match the dynamic range of the following 14-bit A/D converter, which operates at 64 MSPS. The input chain features a programmable gain ranging from √ 20 to 60 dB with an ...