An emerging business philosophy in semiconductor design says that the way to prosper in the new global marketplace is to use your engineering skills to design your customers’ products for them—or at least the “hard parts.”
One corollary of this is that you have to keep beating your own previous personal-best benchmarks over and over again at the same old 18-month cycles, not just at some component level, but at the subsystem level. The reward is that you get to keep doing it. If you do it right and pick your markets wisely, you can also wind up owning a key component to a ubiquitous technology.
One such potentially ubiquitous technology is diagnostic ultrasound, which is poised to move far beyond the baby-picture business. As a diagnostic tool for cardiology, for assessing the condition of accident victims at the scene, and for patient monitoring in the operating theater, ultrasound is becoming indispensible across a spectrum of medical specialties beyond obstetrics. In recognition of this, companies in Europe, the United States, and Asia are pursuing the technology across a range of resolutions, form factors, power requirements, and price points.
So maybe it pays to gamble on ultrasound in much the same way that companies used to gamble on set-top boxes, personal media players, cars, or appliances. Maybe the margins offset the volumes. Maybe if we’re talking about equipping emergency medical technicians, the volumes aren’t so small. Not to mention, there are at least 128 analog transducer channels even in today’s simplest ultrasound probe head. That seems to double in every generation.
With those points in mind, last year, Analog Devices and Texas Instruments introduced eight-channel analog front ends (AFEs) for ultrasound equipment. ADI started shipping the AD9271 in April of 2007, and TI announced the AFE5805, which began sampling in April of this year (see “Monolithic Ultrasound AFEs Usurp Multiple Chips In New Designs” at www.electronicdesign.com, ED Online 18660).
Analog Devices has now updated the AD9271 with two new eight-channel chips that are almost pin-compatible but offer greater flexibility in terms of application targeting. To aid in using this flexibility, there is also an SPI-bus (serial port interface) interface and GUI for tweaking internal registers via that interface.
The basic architecture carries through (Fig. 1). There are still eight channels, each comprising a low-noise amplifier (LNA), a variable-gain amp (VGA) to adjust the gain of the channel over time, an anti-aliasing filter (AAF), a 12-bit analog-to-digital converter (ADC), and a serial low-voltage differential signaling (LVDS) output port.
The new AD9272 was designed to minimize terminated noise—that is, problems associated with noise generated by the ultrasound probe heads, a fundamental issue in making clearer images. It targets high- and mid-end cart-based ultrasound devices, which are marketed on the basis of topnotch image quality.
ACCORDING TO THE SPECS...
In terms of noise specifications, the datasheet lists a typical input-referred noise of 0.75 nV/√Hz at 5 MHz and a gain of 21.6 dB. (Other gains,15.6 and 18.1 dB, are programmable via the device’s SPI bus.) But the raw specs need to be understood in the context of the design concepts explained in the datasheet.
The anti-aliasing filter combines a single-pole high-pass filter and a second-order low-pass filter. The high-pass filter can be configured for dc coupling, or the filter can be configured at a ratio of the low-pass filter cutoff. This is selectable through the SPI.
The 12-bit ADC uses a three-stage pipelined architecture— a 4-bit first stage followed by eight 1.5-bit stages, and a 3-bit flash. Each stage provides sufficient overlap to correct flash errors in the preceding stages. The quantized outputs from each stage are combined into a 12-bit result in the digital correction logic.
The pipelined architecture lets the first stage operate on a new input sample and the remaining stages operate on preceding samples. Conversion rates from 10 to 80 Msamples/s are possible. The signal-to-noise ratio (SNR) is specified at 70 dB, with spurious-free dynamic range (SFDR) at 75 dB.
Where the AD9272 is intended for high-resolution plug-in applications, the AD9293 is for portable apps. It’s optimized for power efficiency, addressing the needs of those portable ultrasound systems that will be deployed in ambulances and rescue vehicles.
For that kind of ultrasound equipment, the AD9273 provides power dissipation of less than 100 mW per channel, still at 12 bits, but at no more than 40 Msamples/s. It can run 50, but at higher power consumption. At that, you’re talking about 1.2 nV/√Hz, compared to 0.75. On the other hand, you’re getting the same SNR and SFDR.
That GUI is the key to using the devices’ SPI to customize the noise and power performance for various imaging mode, probe, or power requirements (Fig. 2). By changing SPI registers, designers can optimize an ultrasound signal processing architecture for noise performance or battery life.
That, as much as the tweaking of the designs of these two chips, may be what signals that these devices represent a true second generation in ultrasound AFEs. It means that there’s growing market acceptance of the idea of a long-term partnership between integrated device manufacturers and original device manufacturers in medical products. DON TUITE
ANALOG DEVICES • www.analog.com