The move from analog to digital design in medical devices enables smaller size, lower power, greater noise immunity and lower parts count for powerful, portable solutions in the health care sector.

The burgeoning medical device industry stands to make significant advances with a new generation of microcontrollers that boasts high performance and low power consumption. These microcontrollers are integrated with a full complement of peripheral devices that meets the noise and accuracy requirements of medical devices. Traditionally, medical designs have relied on discrete analog circuit blocks, but digital microcontrollers are now powerful enough to assume the functions of their analog counterparts; sacrificing nothing in terms of speed and accuracy, while gaining reliability and smaller system volume. By moving to predominantly digital designs, cost minimization, design flexibility and time-to-market are improved, since software alterations are trivial in comparison to hardware redesigns. Additionally, there is a wide spectrum of available microcontrollers to meet the needs of next-generation, digital medical devices.

An Example Application

A pulse oximeter is an excellent application to demonstrate the shift to digital design. This noninvasive medical device shines red and infrared (IR) light through a patient’s finger or ear, and measures the absorption at each wavelength to determine blood-oxygen saturation. In addition, the pulsation of a patient’s heart is detectable, allowing the heart rate to be calculated (see “What is Pulse Oximetry?” p.xx). Portable oximeters have several critical requirements: low power dissipation for maximum battery life, small size that does not encumber the user, and high accuracy and repeatability. The last requirement is particularly important, since incorrect blood-oxygen saturation readings could endanger the health of the user.

As with any analog-based product, a gamut of factors affects the performance and design. Semiconductor products, such as operational amplifiers (op amps), are sensitive to temperature variations. Specifications, such as offset voltage and input offset current, drift with temperature and lead to measurement variations. 1/f and broadband noise also play a part in corrupting the accuracy of measurements. Another issue to consider is system size. The dramatic, ongoing reductions in the surface area of digital integrated circuits have not been matched by analog chips, giving digital-based devices an advantage in reducing system volume. Designers usually need many discrete components for analog designs, which could introduce problems with reliability and increased cost.

The transition to digital alleviates many of the problems that analog implementations incur. Generally speaking, digital designs are inherently smaller than analog designs. For example, analog signal processing, which commonly involves one or more op amps and a number of passive components, can be completely translated into software routines. Not only does this transition reduce the number of parts and system volume, it also frees more design time, since revisions only require code alternations. Digital circuitry has much better noise immunity compared to analog circuits and is not as susceptible to temperature drift. For wireless medical devices, the superior noise immunity of digital circuitry also improves the rejection of EMI noise.

Microchip’s dsPIC33F digital signal controller (DSC) is well suited to realize a digital pulse oximeter, because the digital signal processing (DSP) capabilities of the dsPIC33F eliminate the need for substantial analog signal-processing stages. Many resources available in books and journals are dated, and recommend the heavy use of analog circuitry. The only book available on oximeter design, Design of Pulse Oximeters edited by J.G. Webster, was published in 1997 and recommends the use of discrete sample-and-hold amplifiers, multiplexers and analog filters. Microcontrollers and DSCs have advanced tremendously in the 13 years since, rendering these solutions obsolete.

Figure 1 illustrates two different signal-flow paths in a pulse oximeter: the traditional, analog-centric design and a new, digital approach. In the traditional design, sample-and-holds separate and demodulate the red and IR signals, which bandpass filters (BPFs) then condition (shown in the top of Figure 1). Programmable gain amplifiers (PGAs) amplify the signals for the microcontroller A/D converter to sample. The microcontroller D/A converter also applies DC offsets to ensure that the signals are within the measurable range of the A/D converter. The signals are sampled before the bandpass filters, to determine their DC levels and adjust the red and IR LED brightness to equalize the DC levels. Measuring the DC level is important in order to set LED brightness and to simplify calculations. Since the red and IR DC levels are due to tissues that have constant absorption, if the two DC levels are equalized, they can be eliminated from the oxygen saturation calculation; only AC levels matter. Traditional analog designs rely on bulky DC pass filters with cutoff frequencies around 0.5 Hz. In digital designs, however, red and IR data can be sampled and filtered quickly with a moving average filter or other digital DC pass filter.

The Benefits of Digital Signal Processing

Utilizing a microcontroller with integrated DSP capabilities, known as a digital signal controller (DSC), reduces the entire stage into a single PGA. The DSC performs the signal separation, demodulation and filtering (bottom of Figure 1). The compression of this stage into the digital domain has the greatest impact on system volume, reliability and cost, since multiple discrete ICs, op amps and passive components are eliminated. The elimination of these sample-and-holds and analog filters helps reduce variation in the measurements that stem from temperature drift in the components, and also improves overall system reliability by reducing the number of components that could eventually fail.

Moving to digital has an impact on power consumption that is dependent on several factors. If the analog filters, sample-and-holds and PGAs use tens of miliamperes, shifting to digital will likely improve power usage for a microcontroller operating at a few MIPS. Many microcontrollers also have advanced power-saving modes that can reduce the power consumption of the microcontroller to mere nanoamps while idle.

Analog vs. Digital Filtering

Sample-and-hold circuits are a composite of op amps, switches and capacitors that require a nontrivial amount of system volume, if they are built discretely. Most modern microcontrollers have sample-and-hold amplifiers integrated with their A/D converters, which obviate discrete sample-and-holds. A designer might make the case that analog filtering, such as in the top of Figure 1, is worth the volume, power and reliability trade-offs, because the aliasing incurred by an A/D converter can be avoided; with the filtering occurring before the signals are digitized. However, this is not true; the sample-and-holds still alias noise. Rather than the aliased signals appearing in the digital domain after digitization, they appear in the analog domain after the sample-and-holds.

Since oximeter bandpass filters have very low cutoff frequencies (0.5 Hz and 5 Hz), large-valued components may be required. In addition, narrow stopband width and strong stopband attenuation require cascading op-amp stages. Because analog filters require a large number of components and significant system volume, the trade-offs that digital filtering incurs are minimal. Pulse oximeter signal processing has several flavors, and if a designer needs a DC-reject, DC-pass, low-pass, or high-pass filter, Microchip’s free DSP software libraries support the Finite Impulse Response and Infinite Impulse Response realizations of these filters.

Data Acquisition Options

Designers have several choices regarding data acquisition. If there is minimal high-frequency noise, sampling at the end of each pulse of red or IR light is sufficient, as in the top of Figure 2. However, if the accuracy and precision requirements are stringent, and aliased high-frequency noise is a problem, a designer can increase the sampling frequency to mitigate aliasing, as in the bottom of Figure 2. Of course, the trade-offs are higher clock frequency and higher power consumption, which are required to support the larger data set and higher order filters. Pulse oximeters commonly have a blood-oxygen saturation accuracy of ±2%, which a 10-bit A/D—often integrated with a COTS microcontroller—can support. For higher precision, designers may want to use a 12-bit A/D, which are also available in COTS microcontrollers.

As a result of a new generation of microcontrollers and DSCs, medical devices are trending toward digital signal processing and the elimination of analog signal conditioning. As the healthcare market shifts toward early diagnosis and personal health monitoring, reliable, cost-effective medical devices will be necessary. The transition to digital design from traditional analog will enable consumer medical devices to proliferate, and predominantly digital designs will allow in-the-field upgradability and low-cost redesigns. As the market for consumer medical devices expands, making the leap to digital will be necessary to deliver high-performance products at a low cost with expedient time-to-market.

Microchip Technology
Chandler, AZ.
(480) 792-7200.
[www.microchip.com].