Powering medical haute couture

In spite of media excitement about wearable medical devices, industry experts say that this segment only now is starting to gain momentum: We have yet to see these products in volume. According to Broadcom’s president and CEO Scott McGregor, “…when we get enough functionality, this market will take off. It’s like a wave rolling in that won’t stop. There will be a knee in the curve in the next couple of years.”¹

The article in which McGregor and several other industry leaders were quoted was largely concerned with the power limitations currently holding back wearable devices. The consensus was that power consumption needed to be significantly reduced, and this would involve all aspects of product design from the battery to the sensors and processing ICs to the software routines.

“We tend to think about the system architecture and then the gate and implementation level and all the techniques that go along with that,” said Mark Milligan, vice president of marketing at Calypto. “The wearable market, in particular, has to include all of the above. You need to have a plan for everything from system architecture to micro architecture to gate at the same time. And we’ve seen some really novel circuit architecture being created, along with system partitioning for what you need to keep the power down and selection of IP like a lower power processor than what you might normally use. Power is first, and all the techniques are being deployed.”¹

Among the novel approaches being developed is one that uses almost no power. MC10, a venture-backed company dating from 2008, is commercializing Professor John Rogers’ previous research on stretchable/flexible inorganic electronics. The MC10 stretchable product aimed at large-volume use is the Biostamp (Figure 1).

Figure 1. Biostamp prototype
Courtesy of MC10

As described on the company’s website, “…The Biostamp [is] a seamless sensing sticker that by stretching, flexing, and moving with the body, is redefining the interface between humans and electronics. MC10 is developing a number of products in multiple markets, including fitness, consumer health, and medical devices. The Biostamp can measure a variety of physiological functions: data from the brain, muscles, heart, body temperature, and body movement.”

This device uses near-field communications much like an RFID tag and contains a very small, low-capacity thin-film battery that also is charged through the near-field link. The website description continued, “…[We] adapt conventional electronic devices with novel mechanics to enable new generations of thin and conformal electronic systems. Our devices incorporate silicon devices thinned to a fraction of the width of a human hair. These chips, combined with stretchable metallic interconnects, are further combined with elastic rubberlike polymers to form complete powered systems that sense, measure, analyze, and communicate information.”

As stated on the website, the battery can power the patch for “many hours,” considerably less than the ideal of a month or even a year mentioned by one of the reference 1 industry leaders. However, for an adhesive patch, a realistic power target might be a few days to a week or two—a day isn’t long enough, and a month is overkill. The MC10 website added that by being virtually invisible, the new sensor technology should encourage people to proactively monitor their health. The company saw the best opportunities among athletes, expectant and new moms, and the elderly.

Although none of MC10’s products are yet released, several developments are progressing. Enabling many of the medical devices the company is developing is a so-called nanomembrane (NM) sensor. In an interview, the company co-founder Roozbeh Ghaffari said, “We have established partnerships with leading medical device companies, including Medtronic, and institutions such as the Massachusetts General Hospital and the University of Arizona to deliver our unique NM-sensing platform onboard existing balloon catheters. We have additional product[s]… underway in the areas of complex arrhythmias that require high-resolution mapping technology.”

In one application, 128 of the NM sensors on the surface of a balloon catheter are used to map electrical activity. Ghaffari explained, “This involves taking existing FDA-approved catheters and wrapping our novel NM sensors around them.”²

Components for medical wearables

MCUs

At the processing level, Analog Devices’ ADuCM350 mixed-signal meter-on-a-chip integrates an analog front end (AFE) with a 16-MHz ARM Cortex-M3 processor core. The datasheet describes the IC as a complete, high-precision meter-on-a-chip that makes measurements based on current, voltage, and impedance. A 16-bit ADC and precision voltage reference in the AFE ensure accurate conversion of sensor signals that are routed via an on-chip reconfigurable switch matrix.

In addition to the IC’s extensive functional capabilities, it also provides a selection of four power management modes. As described in the ADuCM350 datasheet:

Active mode—all peripherals can be enabled. Active power is managed by optimized clock management.
Core sleep—the core is clock gated but the remainder of the system is active. No instructions can be executed in this mode, but DMA transfers can continue between peripherals and memory.
System sleep—in system sleep, most peripherals are clock gated and are no longer user programmable; the interrupt controller remains active, and the NVIC processes wake-up events for a limited number of sources.
Hibernate mode—some limited state retention, limited number of wake-up interrupts, and the RTC is active.”

The ARM Cortex M3 processor power isn’t specifically stated on the Analog Devices ADuCM350 datasheet. However, the figure of 175 µA/MHz is given in an article comparing this processor to a Renesas Electronics RX100 processor that draws 110 µA/MHz. So, at the specified 16-MHz clock rate, the M3 processing core would draw about 2.8 mA. In contrast, the same article quotes the sleep mode as requiring only 250 nA for the ADuCM350’s M3 core and 350 nA for the RX100 in standby mode.³

Yet another semiconductor developer, Ambiq Micro, is squarely focused on the wearable market with the company’s Apollo MCUs based on the ARM Cortex-M4F core. According to data presented in an Ambiq white paper, the Apollo MCU’s active current is only 35 µA/MHz compared to STMicroelectronics’ STM32F401 at 355 µA/MHz. However, on the STMicroelectronics website, the figure is listed as 128 µA/MHz, making it comparable to the Analog Devices and Renesas processors.

One specification not in dispute is the sleep- or standby-mode current consumption. The Ambiq white paper quotes 2.8 µA for the STM32F401 while the STMicro datasheet gives numbers from 2.4 µA to 42 µA depending on the exact mode chosen. Which number is accurate makes little difference compared to the Apollo’s 100-nA sleep-mode current.⁴

Battery

Low current consumption obviously is the key to long battery life. A typical CR2032 lithium coin cell provides a nominal 3.0-V output and 240-mAh capacity to 2.0 V with a 190-µA load. With the same load but a cutoff at 2.5 V, the amp-hour rating reduces to about 220 mAh. An Energizer datasheet shows a couple of pulse test conditions that further reduce the delivered capacity. For the ADuCM350, taking into account pulsed operation, a CR2032 could provide 180-mAh to 200-mAh.

It’s all about duty cycle. In sleep mode, 200 mAh/0.250 µA = 800,000 hours—considerably longer than the battery’s 1%/year self-discharge would support. Assuming a 3% duty cycle, (3 x 2.8 mA + 97 x 0.250 µA)/100 = 84 µA, corresponding to 2,375 hours or 99 days. The Ambiq white paper states a six-month battery life for a system using the Apollo MCU, so perhaps a 3% duty cycle for the ADuCM350
is reasonable.

Peripherals

Figure 2. Simplified EPIC schematic
Courtesy of Plessey Semiconductors

Of course, a wearable device requires inputs for the MCU to work with. Plessey Semiconductors has developed the electric potential integrated circuit (EPIC) ultra-high impedance sensor that is optimized for ECG applications. As shown in Figure 2 and described in the device’s datasheet, the PS25251 ECG sensor “… uses active feedback techniques to both lower the effective input capacitance of the sensing element (Cin) and boost the input resistance (Rin).” This technique results in typical Cin and Rin of 15 pF and 20 GΩ, respectively.

Compared to MC10’s thin and conformable Biostamp, the Plessey sensor is large, measuring 10 mm on each side and 2 mm thick. And, it’s not exactly power friendly either, requiring about 2 mA from a bipolar supply. Nevertheless, being able to obtain high-fidelity ECG signals without using traditional wet electrodes is a major advantage. The sensor is so sensitive that it even works well through a couple of layers of clothing. EPIC technology is available in the imPulse product, a handheld ECG monitor (Figure 3), and in the inCite ECG wristband.

Figure 3. imPulse ECG sensor
Courtesy of Plessey Semiconductors

The photoplethysmogram (PPG) is another type of sensor that the ADuCM350 might interface to. A PPG measures heart rate by monitoring how much light is reflected from the skin. According to reference 5, “PPG has mainly been used to measure blood oxygen saturation, but can provide cardiac information without a biopotential measurement…. In optical systems, light is transmitted through the surface of the skin. Light absorbed by the red blood cells is measured with a photosensor. As the heart beats, the changing blood volume scatters the amount of light received.” The Analog Devices ADPD142 optical module has a complete photometric front end including current sources, LEDs, and an integrated photosensor.

A related heartbeat sensor has been demonstrated by a UC Berkeley team. Instead of the conventional red and infrared LEDs, the experimental device uses organic red and green LEDs and sensors. According to an article describing the work, “By switching from silicon to an organic, or carbon-based, design, the researchers were able to create a device that could ultimately be thin, cheap, and flexible enough to be slapped on like a Band-Aid during that jog around the track or hike up the hill.”⁶

In addition to touting the advantages of the company’s Apollo MCU, the Ambiq white paper lists the current consumption for many types of sensors, making the point that sensing often can consume more power than the subsequent data processing. Accelerometers typically use little power. The Analog Devices ADXL362 three-axis device requires 1.5 µA when active, and this agrees well with the company’s datasheet. It states <2 µA with a 100-Hz output data rate, increasing to about 3 µA at 400 Hz. Acceleration is converted to a 12-bit value (11 bits plus sign).

The 16-bit resolution STMicroelectronics LIS3DH, also a three-axis device, needs 11 µA for full resolution and a 50-Hz output data rate. This reduces to 2 µA at a 1-Hz rate. The Analog Devices and STMicroelectronics devices can achieve even lower power consumption by outputting only 8-bit data. Both devices feature very low standby current.

In contrast, gyroscopes run on milliamps rather than microamps and even powered down still consume microamps rather than tens or hundreds of nanoamps. The white paper recommends choosing a low-power gyroscope, but don’t use it to sense initial motion. Instead, you should place the gyroscope in suspended or power-down mode and use an accelerometer to activate it.

The InvenSense MPU-6000 and STMicroelectronics ASM330LXH combine a three-axis accelerometer and gyroscope into a single package. Both devices draw about 4 mA with the accelerometer alone taking approximately 10%. Pressure sensors, microphones, and magnetometers also are discussed in the white paper.

When sensors, memory, and Bluetooth communications are considered, the MCU may account for less than half of the device’s total power consumption. Reducing the current drawn by any one part certainly helps, but as reference 1 emphasized, truly low power design means rethinking all aspects of the product.

References