There are numerous posts on this site regarding the merits of analog ASICs for various markets, applications, and functions. Clearly, there are a lot of variables to consider before embarking on a mixed-signal ASIC development, and the decision should not be taken lightly. In my experience, implantable medical devices (IMDs) very commonly justify mixed-signal ASIC integration.
IMD products include pacemakers, defibrillators, neuro-stimulators, and drug infusion pumps. While the functions of these devices vary, they all typically have goals of size and power efficiency. Due to their implantable nature, it is desirable to make these devices as small as practical for patient comfort and mobility, and for ease of implant. Long battery life is also an imperative, since IMDs require explant surgery after their batteries are depleted. The combination of very strict size and power limitations often justifies mixed-signal ASIC integration for IMDs.
If we conclude that a mixed-signal ASIC is justified, we then have to decide on the extent of the integration. It is very tempting, and common, for IMD designers to reach the conclusion that they need to maximize integration in order to gain the size and power benefits of the customization and integration. A maximized integration is represented by a system-on-chip (SoC) ASIC. A typical SoC ASIC might include all required analog sensors and drivers, data converters, and power management. The SoC will also include control logic in the form of an embedded microcontroller unit (MCU) or a custom digital controller.
Many IMDs also include wireless communication to support communication to the IMD after implant. This is often an RF link using the medical implant communication service (MICS) band (~400MHz). Finally, the IMDs usually require some non-volatile memory (NVM) to store calibration constants, user settings, and diagnostic data. A true SOC would include all of these sub-systems in a single ASIC, and would result in the maximum level of integration. A resulting IMD could conceivably consist of nothing more than a battery, the ASIC, and some passive components.
Despite its obvious attraction, in my opinion, maximizing ASIC integration is rarely the best overall solution for IMDs. Here are a few ideas about what not to integrate for most IMD ASICs and why.
MCU: MCUs are incredibly versatile tools for IMD designs. They provide highly flexible platforms on which a wide range of systems can be built. MCUs typically include a wide range of peripheral functions, like oscillators, timers, communication interfaces, and data converters, and these peripherals are seamlessly integrated with each other on the MCU chip. The argument against a standard MCU and in favor of integration into the ASIC is that MCUs often contain far more features than are necessary for any given IMD.
If the primary goals are to minimize size and power, it seems obvious that we should identify the minimum set of required features and integrate them into a mixed-signal ASIC. However, there are some very attractive MCUs available that directly contradict that argument. These MCUs offer a range of low-power modes and are incredibly small (under 5mm2 ). With these impressive characteristics readily available in a relatively inexpensive standard MCU, it is difficult to justify the development time and cost required to integrate the equivalent MCU functions into a mixed-signal ASIC.
RF Communication: When RF communication is required, complete ASIC integration is difficult to justify. One reason is that, like MCUs, there are some attractive solutions readily available. These include standalone transceiver chips and MCU/transceiver combo chips.
Other arguments against RF integration are that the RF circuit design is complex and often requires a technology with specific RF capabilities. This limits the technology options for the ASIC and often conflicts with requirements of various ASIC functions. Finally, committing to the integration of the MICS communication locks the device designer in to that particular communication protocol indefinitely. Any future change to a communication protocol would thus require a new ASIC development.
NVM: Many available technologies support the integration of NVM into a mixed-signal ASIC, but, in my experience, it is rarely justified. The primary reason for this is simply that most MCUs include NVM. If we choose to include a standard MCU, we get NVM for free. In cases where we need more NVM than the MCU offers, we need to consider available options for standalone NVM chips, which include decent memory arrays (1 Mbit/s) in extremely small form factors (< 5mm2 ).
When defining an IMD device, it is important to keep our focus on optimizing rather than maximizing ASIC integration. While the technologies and tools are all available to accomplish the maximum integration level of a mixed-signal SoC, we need to strive to find the optimal solution. With that approach, I have found that ASICs for IMDs almost never include the MCU, RF, or NVM functions.
Tell us about your experiences with medical electronics and with ASICs or other customizable mixed-signal parts.