Challenges of AFEs in ultrasound apps

The ultrasound imaging market is undergoing a period of rapid equipment evolution and expansion. InMedica, a market research company that focuses on the medical devices industry, forecasts worldwide ultrasound equipment revenues will grow to $5.7 billion in 2010, representing a compound annual growth rate of 6.5 percent. Driving this growth is ultrasound's unique position as not only the lowest-cost, most approachable medical imaging modality but the only truly benign, non-invasive technique. As the ultrasound market continues to expand, it will divide into segments, each of which will place unique demands on the suppliers of analog front-end (AFE) semiconductors to optimize performance and power consumption.

At the high end of the market are the leading-edge console machines. These fixed-installation systems provide the widest range of imaging modes and are typically the first to adopt new imaging modes because of their higher price and power budget. The latest trends in console machines are enhanced four-dimensional imaging and new imaging modalities such as elastography. Console designers deliver this premium image quality by increasing the number of analog channels, taking advantage of beam-forming processing gain, and using the highest-performance AFEs available.

Front-end sensitivity and dynamic range, typically superior in bipolar and BiCMOS processes, dominate this market segment. Nothing comes free, however, and designers are still faced with tough choices. As the number of channels increases, so do the complexity and stiffness of the cable assembly connecting the probe head to the machine electronics. Furthermore, even though these systems are wall-powered, power is not unlimited, as current is typically constrained to the capacity of a standard 30-A breaker.

The hand-carried ultrasound (HCU) market represents the fastest and perhaps most exciting segment, boasting a staggering 42 percent growth rate in 2007, according to Klein Biomedical Consultants (KBC), a strategic marketing consulting firm specializing in the medical diagnostic imaging field. HCU machines are defined to weigh less than 11 pounds and are typically battery-powered during operation. Because of their portability, hand-carried units find many applications outside the radiology department, including cardiology, anesthesiology, obstetrics/gynecology and emergency medicine. Driven by these applications, KBC predicts that the HCU market worldwide will grow an average of 17 percent over the next five years, to $1.2 billion.

Historically, portable ultrasound machines were developed primarily as lower-cost, lower-performance alternatives to console machines. These early models provided only basic B-mode or echo-mode scanning and, therefore, could trade off analog performance for significant power savings at the front end.

However, as the popularity of portable ultrasound machines has increased, so has the demand to provide the same enhanced imaging modes and channel density of higher-end console machines. As a result, the analog signal chain is often one of the dominant power consumers in the digital signal chain and has successfully leveraged Moore's Law and the power, size and performance gains of the mobile PC platforms on which almost all portable machines are based.

As HCU machines increase in capabilities and performance, the market will follow the same trend as the PC business, further cannibalizing sales of conventional models until, eventually, it dominates. Cart-based machines feature nearly every capability of high-end console machines, but with the advantage of portability. This portability is limited, however, in that, unlike HCU machines, cart-based models must be plugged into a wall outlet during operation.

Recent introductions of convertible ultrasound machines may signal a trend for this segment, and a blurring between the hand-carried and portable cart segments. Convertible ultrasound machines consist of a hand-carried unit that can perform scans in battery-powered mode, and may also be docked on a portable cart. The requirements on the AFEs for convertibles, compared with fully hand-carried units, may be only their higher channel counts—hey may maintain the same power profile on a per-channel basis.

Handhelds represent a particularly intriguing development in the ultrasound market. Weighing around 1 pound, these machines have been called the “stethoscope of the 21st century,” and might be the closest we will come in our lifetime to the medical tricorder of “Star Trek” fame. To achieve this vision, this machine must have a battery that will last a full 8-hour shift (longer for interns!). Ultra-portable solutions will place the highest burden on the power consumption of the analog signal chain, for battery life will be determined predominately by the efficiency of the front-end amplifiers and converters.

What's especially interesting is that each segment faces the challenge of maximizing performance of its AFE for a given level of power consumption. Each segment has a different power-consumption limit, above which the machine's usefulness is severely degraded. This is obvious for handheld and hand-carried ultrasound machines, but it is also true for cart-based and console systems. Meeting AFE performance isn't necessarily easy. Though the transmit section can generate up to 1 A at ±100 V, the receive section represents 80 percent of the AFE's power consumption.

Each transmit channel generates only a short pulse of less than 100 ns, while each receive channel processes echoes lasting more than 100 micros. The receive section of the AFE consists of a low-noise amplifier (LNA), voltage-controlled amplifier (VCA), anti-alias filter and analog-to-digital converter (ADC). Each of these components has its own impact on maximizing analog performance while minimizing power consumption.

Discussion of analog performance begins with the useful dynamic range, which translates to the range of scan depths within the body for B-mode imaging. The maximum scan depth is limited by the system noise voltage. The dynamic range depends on the useful input voltage level and noise in the system.

With the maximum input voltage limited by ultrasound physics, improving the system's noise level is the only way to improve dynamic range. Because the LNA is the first amplifier in the receive chain, it's the dominant source of system noise. Because the ADC's noise voltage is reduced by the gain of the LNA and VCA, its contribution to overall dynamic range is surprisingly limited. Bipolar or BiCMOS technology achieves the best noise performance for the LNA for a given power level, but CMOS technology is the dominant technology for ADCs.

To reduce size for hand-carried units as well as to improve channel density for console machines, a trend among semiconductor providers to the ultrasound market is to integrate the LNA and VCA with the ADC. Since CMOS is the best process technology for ADCs, combining the LNA and VGA into a monolithic die with the ADC will necessarily come at a sacrifice in power consumption compared with a discrete bipolar LNA. Alternatively, one could package a bipolar LNA with a CMOS ADC, but this approach adversely impacts yields and packaging costs.

In contrast to the compromises of vertical integration in the signal chain, higher channel density can be achieved through horizontal integration across analog channels. For example, the recently introduced SAM1610 is the first true 16-channel 12-bit ADC for the ultrasound market. Unlike pseudo 16-channel ADCs, which suffer from crosstalk due to time sharing the data converter across multiple analog input channels, this solution provides an independent ADC for each analog input.

The key to packaging 16 channels into a single 12 x 12 mm BGA package is the low-power consumption. At a mere 44 mW per channel, the SAM1610, when combined with a best-in-breed discrete LNA/VGA, achieves lower power consumption than vertically integrated devices targeted at the hand-carried segment. A best-in-breed bipolar LNA/VGA combined with the SAM1610, achieves both a higher dynamic range and lower power consumption compared to other integrated solutions.

Looking to the future, 4D imaging and new imaging modes, like elastography, will exceed the capacity interfaces to and from the AFEs. With 4D imaging (3D plus time), the ultrasound probe requires 2D transducer arrays to form a static 3D image, causing a quadratic increase in transducer channel count. Since it is impractical to run thousands of wires in the cable bundle, many OEMs are moving the AFE electronics into the probe head to concentrate the data onto fewer physical lines.

However, existing ADC solutions require a single LVDS output for each analog channel, which would double the number of wires required. In contrast, current integrated solutions with signal-compression technology can reduce both the total output data rate and the number of LVDS outputs. For a typical cardiology application using a 3.5-MHz transducer center frequency and a 16-MHz sample rate, an integrated solution with signal compression can reduce the data from 16 analog channels so it can be carried on a single LVDS pair.

In a typical ultrasound application, following the AFE is a beam-forming array that is integrated in the design. This array is usually performed in FPGAs. However, new imaging modes such as elastography use beam-forming topologies that cannot easily be mapped onto FPGA hardware architectures, and instead will be performed in software on Intel CPUs, graphical processing units, or cell processors. For a 256-channel machine with 12-bit ADCs operating at 65 Msample/sec, the raw data rate is 200 Gbps, creating a memory bandwidth bottleneck of epic proportions.

This memory bandwidth bottleneck will be a limiting factor in the frame rates in these new imaging modes. Again, using an ADC with integrated signal compression can cut the memory bandwidth by a factor of 3 to 4, translating into a commensurate increase in image frame rates with equivalent image quality.

Allan Evans is vice president of marketing at Samplify Systems, a fabless mixed-signal semiconductor company. Evans holds an MSEE from the University of California, San Diego, and an MBA from Santa Clara University.

0 comments on “Challenges of AFEs in ultrasound apps

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.