A non-invasive imaging tool, portable ultrasound equipment provides a good way for emergency personnel to quickly diagnose the injuries of victims of road-side accidents and natural disasters such as earthquakes and hurricanes. Portable units vary in size from a laptop to a handheld weighing 10 lb or less, and can run off batteries.

The devices' key attributes are the same as any portable device: size, weight, battery life, cost, and performance. A main challenge in the development of a portable ultrasound system is that of ensuring excellent image quality.

A look at the architecture

Medical ultrasound systems work by focusing pulses of high-frequency acoustic energy into the body and processing the return signals to form an image of the tissues under the skin. Transducer arrays, together with beamforming circuitry, are responsible for focusing the ultrasound waves. Medical transducers are usually made up of eight to 512 elements. Each element usually corresponds to a transmit-receive channel.

Ultrasound image formation begins at the beamformer control unit. The transmit beamformer, high-voltage (HV) pulser and HV multiplexer form the transmit path that is responsible for the pulse-excitation of transducer elements. Transducer elements, made of a piezoelectric material, convert HV electrical pulses into a focused, high-frequency, acoustic wave ranging from 1 to 15 MHz. These acoustic waves penetrate the body and as they encounter a boundary between different tissues, are reflected back to the transducer and converted into electrical signals. These signals enter the receive path via the transmit-receive (T/R) switch. The T/R switch toggles from transmit to receive mode and prevents HV pulses from damaging the receive electronics.

The electrical signals are then amplified, filtered, and converted to digital format by the analog front end (AFE). The AFE consists of a low-noise amplifier (LNA), voltage controlled attenuator (VCA), programmable gain amplifier (PGA), anti-aliasing filter (AAF), and analog-to-digital converter (ADC). The LNA provides low-noise amplification for good sensitivity. The VCA and PGA are part of the time gain control (TGC) blocks and improve the system's dynamic range. They also allow the gain to be increased with time to compensate for the increased attenuation of the signal as it passes through the body. The amplified signal is then filtered to improve its signal-to-noise ratio (SNR). The resulting signal is converted into digital format through an ADC and is processed by the receive beamformer. The performance of the AFE significantly drives the ultrasound system's attributes in terms of size, weight, battery life, and image quality.

Ultrasound systems are differentiated by their functions and performance. Functional differences are mainly determined by the digital processing of the system. There are typically three major modes of ultrasound operation: B-mode imaging results in a gray-scale image used for examining tissue structures and organs; Color Doppler mode results in a color image that is super-imposed on the B-mode image — the color code denotes the direction and magnitude of blood flow; and Spectral Doppler processing, which produces a scrolling display of blood flow velocity distribution at a user-specified location.

In addition to algorithms specific to these modes of operation, there are essential signal processing functions that each of the conventional ultrasound systems perform. These include filtering, detection, log compression, and scan conversion. The filtering is typically band-pass filtering used to reduce noise and select whether imaging will be done on the fundamental frequency (for better penetration) or on the second harmonic (better resolution due to better tissue distinguishing properties). In ultrasound processing, detection is the envelope extraction process and usually involves either a Hilbert transform or uses a complex rotator to demodulate the signal followed by low-pass filtering. Log compression is used to fit the signal into the dynamic range used for display. Scan conversion is the processing step that performs the transformation from the coordinate system of the raw data to that used by the display, so the displayed data can be accurately represented.

Depending on the particular ultrasound machine, other algorithms are implemented in the digital processing path to extract a clearer image that can yield improved diagnosis. Among these algorithms are matched filtering, time frequency compensation, echo line averaging, speckle reduction, frame smoothing, and edge detection. The more digital processing capability a system has, and the more flexible the processing section is, the better equipped that system is for producing a high-quality ultrasound image.

Optimizing analog front end for portability

For portable ultrasound systems, low-power consumption and integration of AFE components are vital criteria for battery life and size optimization. These requirements are often in conflict with the performance of the system and compromises must be made.

One of the first tradeoffs comes when determining the number of channels in the AFE. Fewer channels support smaller, more compact systems with greater battery life. Unfortunately, image quality begins to suffer as the number of channels goes down. Systems using between 16 and 64 channels generally get the best balance between portability and image quality.

Power consumption of the AFE components is an important area that has a direct impact on battery life. The AFE's overall noise and linearity performance is related to power consumption. To obtain excellent noise performance, the LNA must dissipate certain amounts of power. High linearity or a large input signal range drives the need for higher supply voltages to avoid clipping the signal and to maintain the dynamic range performance. A higher supply voltage results in increased power consumption. Portable ultrasound system designers now have flexibility to design with different levels of configurability or new architectures, which optimize low power and low noise. The relation between performance and power consumption is best explained with examples.

AFE products, for example, provide several modes of operation for noise and power optimization. This lets system designers use different levels of configurability for the best power-noise tradeoff. One example is the AFE5804 from Texas Instruments (TI): It has eight channels and noise-performance and power-optimization options. Through the use of registers, system designers can configure their particular application's power and noise number. This AFE can be configured at 101 mW/channel for 1.23 nV/rtHz of full chain input referred noise (IRN) or 112 mW/channel for 0.89 nV/rtHz of IRN.

Another example comes from the use of new architectures for power optimization. The AFE5851 is a 16-channel AFE without the LNA integrated, a new architecture. The best solution is to integrate the LNA into the transducer. By doing this, the system's noise figure is significantly improved because signal losses in front of the LNA are minimized. The AFE5851 provides 39 mW/channel of power consumption. The noise performance at a full chain IRN of 5.5 nV/rtHz is counterbalanced with the integration of the LNA into the transducer. The result is an innovative architecture for portable systems that implements stringent requirements for low-power consumption while maintaining noise performance.

It is worth mentioning that there are opportunities for other components in the receive-transmit (T/R) signal path to reduce power and size. T/R switches consist of protection diode bridges and clamp diodes, and traditionally have been implemented discretely. Multichannel, fully integrated T/R switches are now available to minimize size. Note that trade-offs for T/R switch integration in high-end systems can include insertion loss and cross talk, but the switches work well in portable devices.

When an integrated T/R switch solution offers over 50% board space savings, the pros for portability outweigh the cons. The T/R switch usually remains on to protect the receiver path from HV transmit pulses. Bias currents related to the diode bridge are constantly being drawn, affecting power consumption. Programmable bias currents are an option to adjust power consumption. For example, the TX810 is an integrated, 8-channel, 6 mm X 6 mm T/R switch that contains a 3-bit interface used to program a 7 mA range of bias currents. With this combination, an ultrasound-system designer can program seven different current settings and also have a power-down mode to reduce power consumption.

Although the HV pulser is not generally an area of concern for a system's power consumption in B-Mode due to its low duty-cycle, there are options to reduce size and noise in the transmit path. Discrete pulsers are being integrated into multichannel ICs, reducing the size of the transmit path with smaller and fewer components. The output of the transmit path are high-voltage, positive-and-negative symmetric pulses preceded and followed by 0 V. The signal's ability to return to 0 V is critical to reduce the ringing induced in the system by the pulser. This is called the damping function. The TX734 is an integrated, quad-channel, +/-90 V, pulser packaged with active damping in a 9 X 9-mm QFN to reduce noise and minimize size in portable ultrasounds.

Digital processing

In the design of a portable system, designers often have the misconception that the fewer components, the better. They therefore try to find a single processor that can do everything. Actually, it is often better to break up the processing among a few processing elements rather than try to force a single processor to do tasks for which it is not ideally suited.

For example, while an FPGA is not an efficient solution in terms of power and space for the majority of the digital processing, a low-cost FPGA is still useful for getting data from the AFE, performing beamforming, and connecting to the back-end processing engine. After the beamformer, it is generally better to move the rest of the processing onto a digital signal processor (DSP), as its highly programmable, real-time architecture is better-suited for the remainder of the ultrasound processing, resulting in systems with better power efficiency, smaller footprints, and increased flexibility.

Depending on the ultrasound system, there are several DSP options. High-performance DSPs, such as TI's TMS320C6455, provide enough computational capability to efficiently perform all the back-end ultrasound processing, while advanced system-on-chips (SOCs) have highly integrated architectures that can run the operating system (OS), man-machine interface, and drive the display. The C6455 comes in a variety of processor speeds, including a 1.2 GHz version at the high end for systems that must implement additional algorithms to improve image quality or add features. For truly portable systems, a low-cost FPGA and a single SOC like TI's OMAP3530 device may be all the processing that is necessary. In this case, the TMS320C64x+ DSP core in the OMAP3530 does the filtering, detection, log compression, and scan conversion, while the ARM Cortex-A8 runs the OS, graphical user interface (GUI), and drives the display.