Bipolar vs. CMOS: Selecting the right IC for medical designs - Embedded.com

Bipolar vs. CMOS: Selecting the right IC for medical designs

Implantable, ingestible, interactive, interoperable, Internet enabled. The unique demands of medical devices, now and in the future, require the right IC process technology and packaging. Here's a comparison between bipolar and CMOS for medical devices, along with some packaging issues to consider.

Developers of medical applications have to make the right tradeoffs among competing requirements for power, noise, linearity, reliability, and cost. They, like designers of other electrical systems, need to carefully select process and design architectures based on these requirements.

This article compares bipolar and CMOS devices and will help you decide where each is applicable. Using a high-performance ultrasound device as an example, we explore the balance among noise, power, chip area, and integration.

Power dissipation is critical in several battery-operated applications; in such applications CMOS process technology is a great choice. However, the balance between leakage and performance is critical and gates the choice of the technology. Mixed-signal integration is also a key requirement for such applications.

Some packaging technologies can be used effectively when a broad range of functions needs to be performed in a single integrated circuit, such as when low noise is required along with dense digital functions. Such disparate needs, however, are sometimes conveniently met by using multichip modules.

We'll also explore future trends of medical devices, including direct measurements of biological signals and self-powered devices. These trends will drive the modification of existing process technology to include energy harvesting features and other non-standard sensor capabilities.

Analog performance
Let's first explore the analog performance needs using an ultrasound device as an example. With this example, we illustrate typical trade-offs among performance, power dissipation, size, and integration levels and examine the applicability of both bipolar and CMOS process technologies. Figure 1 shows a system block diagram of a typical ultrasound machine illustrating both the transmit and receive sections. These sections drive the transducer along with the digital processing sections (not shown), completing the ultrasound machine.1


Click on image to enlarge.


Figure 2
offers a more detailed block diagram of sections performing the receive functions in Figure 1.


Click on image to enlarge.

Some design considerations for this type of receive block are input referred noise, linearity, gain, and power dissipation, to achieve a given performance. Typical performance parameters can be seen in datasheet downloadable from Texas Instruments.2 Additionally, the number of receive channels in a given package size sets the integration levels. The received signals from the transducer can vary over 100 dB in amplitude. Hence, the input referred noise on one end for low-level signals (~10u V) and the linearity of the device for large input signals (~1 V) are critical performance parameters. In order to accommodate this large dynamic range, the channel gain is adjustable through a voltage controlled attenuator (VCA) and a programmable gain amplifier (PGA). Figure 3 shows the total gain through the device as a function of applied voltage on the VCA for several PGA settings.


Click on image to enlarge.

Now, let's consider the performance of both a bipolar and CMOS amplifier. Open loop amplifier blocks with bias current of 4 mA (total) designed for 20-dB gain are designed using both bipolar and CMOS devices. The target process technology is a TI internal BiCMOS process.

Table 1 compares the sizes of bipolar and CMOS devices being used in the amplifier. The larger size of the CMOS device and the associated input capacitance plays a strong role in limiting the amplifier's input bandwidth. In this particular case, lower noise at lower bias current is achieved using the bipolar amplifier. A possible limitation for the bipolar devices is additional noise contributions from the base current negligible in the CMOS devices. The exact extent of this base current noise depends on the impedance of the transducer and the system implementation details.


Click on image to enlarge.


Mixed-signal and low-power applications

Earlier, we observed that for certain medical applications the analog performance of bipolar devices is preferred over those of CMOS devices. However, a class of devices exists requiring mixed-signal performance in both analog and digital processing capabilities. These mixed-signal needs can be further classified into applications that need extremely low-power operational capabilities.

For example, implantable devices like cardiac pacemakers placed inside a human body are required to operate for long periods of time with a limited power supply. Such devices need both low-power analog circuits for detecting physiological signals from the body and low-power digital and memory functions for interpreting and storing those signals. Advanced implantable devices also need low-power radio communications for transmitting information to base units outside the body.

A deeper analysis of the signal types and operational modes show that these devices typically have a low operational duty cycle. For example, they are active only for very short periods making measurements or processing, and otherwise are asleep for long periods. Duty cycles smaller than 1% are not uncommon in such applications. Another feature is that most body signals are low frequency in nature. Thereby bandwidths and sampling frequencies for the data converters can be limited to few tens of kilohertz or less. Several external battery-operated devices used in consumer applications also have similar performance and power criteria.

For the devices with adequate active performance, the above requirements dictate a process with low off-state leakage current. An obvious trade off must be made between active performance and leakage in such process technologies. Typically, these technologies have gate lengths between 130 nm and 350 nm. In the future, however, it may be possible to scale to 90-nm feature size as well. For implantable devices the change in leakage current performance with either process, temperature, or supply, is a critical parameter as it directly affects battery life. Figure 4 shows the NMOS change of leakage current (Ioff) and drive current (Idrive) of one such process technology as a function of temperature. While Idrive does not show a strong temperature dependence, Ioff does.3 Figure 5 shows a similar plot for PMOS devices. Since the body temperature doesn't vary over a large range, changes of Ioff with temperature can be tolerated. Figure 6 shows the ring-oscillator frequency, a typical figure of merit for these devices as a function of supply voltage. This can be used as a guideline to trade off leakage current and performance in a real application.


Click on image to enlarge.


Click on image to enlarge.


Click on image to enlarge.

Another critical component in the design of low-power, mixed-signal devices is the availability of a reliable, compact and low-power, nonvolatile memory. Ferroelectric memory (FRAM) offers unique capabilities that make it an attractive nonvolatile memory choice for many applications. Distinguishing features include fast, RAM-like, write speeds, low voltage, low-power write operation, high cycling endurance lifetime, and architectural flexibility. This memory is integrated into the low-power digital process technology described earlier.4

The FRAM operates at 1.5 V, and unlike floating-gate devices, doesn't require charge pumps. As with all nonvolatile memories, the reliability challenges include write/read cycling endurance, data retention, and high temperature operating life. FRAM performance for noncycled and cycled bits shows excellent performance, even after a large number of operations.4

Packaging technologies

In applications needing different performance metrics in the same IC, packaging technologies can be used effectively. For example, in applications needing low-noise and low-power digital applications silicon (Si) dies from two different processes can be placed within the same package to achieve the required performance. Si dies can be stacked on top of each other to save board foot print. Advances in package technology allow for the integration of certain passive elements such as inductors and capacitors inside the package. Chip on board technologies enable entire ICs to be embedded inside the printed circuit board saving valuable area in dense applications.
Future trends
Medical electronics represents a very diverse field and as such, there are many different areas where innovations in process and packaging can help with creating new solutions. Just to mention a few: measuring physiological signals with sensors on the surface or even inserted under the skin is driving improvements on flexible substrates and special adhesives. ICs swallowed with every pill to track medication compliance, or in contact with the blood stream to measure or deliver drugs, represent some of the challenges on digestible electronics, coatings and how to deal with foreign body response.

High voltage (~100 V) process improvements translate proportionally into ultrasound transmit channel density increments. Innovations in micro machining promise the miniaturization, mass production, and massive channel count in ultrasound probes (CMUTs, capacitive micro-machined ultrasonic transducers) and of full analysis laboratories (lab-on-a-chip or LOCs).

Energy harvesting is another emerging field that can improve device lifetimes by assisting or replacing the battery completely. Several technologies being considered are thermal, vibrational, and solar power. These harvesting technologies will drive additional needs in areas of circuit design and process requirements.

The tip of the iceberg
The area of medical electronics is expanding and has unique needs in terms of performance, power, and integration levels. We've illustrated some of these requirements and future trends here, but there's much more to explore.

Karthik Vasanth is the general manager of the Medical and High Reliability business unit at Texas Instruments. He received his Ph.D from Princeton University, Princeton, New Jersey.

Acknowledgements
The authors wish to acknowledge the assistance, support and expertise of Xiaochen Xu, Harish Venkataraman, and Rajni Aggarwal, all from Texas Instruments.

Endnotes
1. Learn more about ultrasound solutions here: www.ti.com/ultrasound-ca .
2. Download a datasheet on the afe5805, which was used in the above example: www.ti.com/afe5805-ca .
3. R. Aggarwal, personal communications.
4. J. Rodriguez, et al., “Reliability Characterization of a Ferroelectric Random Access Memory Embedded within 130nm CMOS,” Proceedings of the IEEE Int. Symposium on the Applications of Ferroelectrics , Vol 1, pp 1-2, Feb 2008.

Leave a Reply

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