The combination of wafer-level chip-scale-packing (WLCSP) technology, a new generation of sensors, and DSP functionality will open areas of expansion for MCU applications in the Internet of Things (IoT). The most frequently cited application space lies in wearable devices.
Other design opportunities exist wherever remote sensing nodes are required, such as home automation systems, flow meters, and barcode scanners. Chip-scale packaging also puts implantable and even ingestible medical monitoring devices on the near horizon.
In some applications, it can be important for signal analysis to reside close to the sensor. This application space is particularly fertile ground for 32-bit, DSP-equipped MCUs. Integrating a chip-scale MCU into a module with a sensor, wireless-communications chip, and energy source would give designers more flexibility in creating small, self-contained systems. It also allows them to choose smaller, less sophisticated sensors and still address high-performance applications.
From a purely electrical design perspective, using MCUs with chip-scale-packing does not present significant challenges as long as designers use PCB layout tools that can handle the physical characteristics of WLCSP. Specifications vary from manufacturer to manufacturer, but the equivalent pin pitch (WLCSPs use solder balls) is 0.44 mm, trace widths are about 100 µm, and solder mask thickness is about 25 µm.
In addition to their smaller footprint and heights compared to plastic packages, WLCSPs offer other advantages, including lower inductance between the die and the printed circuit board, high-thermal conduction characteristics, and a short manufacturing cycle time.
The pins of conventional packages are replaced by solder balls (bumps) that are arranged in an array that has a bump-to-bump pitch that is compatible with state-of-the-art circuit-board assembly processes. To mount a WLCSP die on a printed circuit-board substrate, the die is placed bump-side down on the substrate metal landing spots. A solder reflow process is used to melt the solder and form a joint. Figure 1 illustrates this process.
Figure 1: WLCSP mounted in a printed circuit board. (Courtesy of Atmel Corp)
The die is attached to the substrate by the solder. Adding an insulating underfill is an option not shown in the figure that will improve the joint’s reliability.
Design teams may undertake to design the entire subsystem themselves using WLCSP parts or they can opt for a solution designed by a third party that integrates two or more functions. The decision depends primarily on three criteria: The available system footprint, energy requirements, and the uniqueness of the application.
Since leading-edge IoT designs are likely to include an MCU and communications capability, using third-party products has the advantage of pre-certification by the trade group that manages the standard. Wireless MCUs have become a popular choice for putting both MCU and communications functionally on the same die.
Nordic Semiconductor’s nRF51822, for example, offers an MCU and Bluetooth low-energy baseband functionality in a single package. The chip comes pre-certified by the Bluetooth SIG. In this case, the MCU portion is a 32-bit ARM Cortex-M0. A 10-bit ADC and several serial interfaces are included to make integration with sensors easier, and a 128-bit AES co-processor is there to provide secure data connections when needed.
One application in which chip-scale MCUs could take this concept a step further—and be particularly useful—is an activity monitor that senses heart rate that is embedded in headset earbuds. Since a typical use-case for headsets includes listening to music while exercising, it’s a logical extension to add a sensor and an MCU to the Bluetooth functionality that is already in the earbud. Products such as the LG Heart-Rate Monitor Earphones are approaching this concept, but they do not fit all the functionality into an earbud.
Sensor integration into the multi-chip module is still a challenge because sensors typically do not scale in the same way that logic chips do. A leading candidate for adding the sensor capability is—somewhat surprisingly—LEDs, which are fabricated with a semiconductor process. Briefly, the concept is that an LED transceiver can sense variations in fluid flow (i.e. a pulse) over time. In an optical heart-beat pulse sensor, light directed into the skin either bounces back to a light sensor, or is absorbed by blood cells. Whether it is reflected or absorbed at a particular measurement point depends on the state of the pulse beat. Continuous light-sensor readings can provide a surprisingly accurate heart-beat pulse reading.
Integrating heart-rate monitoring into an earbud is an application that would likely require a powerful compute engine. It all depends on the algorithms used to interpret the LED data into meaningful information about pulse rate.
Texas Instruments’ SimpleLink wireless MCUs support a range of wireless technologies, including standards-based 6LoWPAN, Bluetooth low-energy, Wi-Fi, and ZigBee, as well as proprietary sub-GHz and proprietary 2.4 GHz. The CC3200 family integrates TI’s portal into the chip family integrated into an ARM Cortex-M4 with a Wi-Fi transceiver and baseband functionality. TI has made the CC3200-LAUNCHXL development kit the portal into this solution.
An important subcategory of wireless MCUs use proprietary, sub-GHz transceivers, which are included in TI’s SimpleLink product offerings.
One example is Silicon Labs Si106x/Si108x wireless MCUs that operated in the 490 MHz band, the 868 MHz band, and the 915 MHz band. The company’s Si1060
Wireless MCU Development Kits provide an entry point for designers to familiarize themselves with these products.
Designing with WLCSP
Design teams may undertake to design the entire system themselves using WLCSP parts. This route obviously requires more design than using third-party solutions and potentially adds standard certification to costs and design cycle time. This may be worth doing in particularly high-volume products in which bill of materials costs are kept as low as possible. In other instances, the design team must work from scratch because none of the third-party solutions provide a good fit.
The challenges of designing from the ground up involve choosing the right contract manufacturer. WLCSP chips are more delicate because they are basically a piece of silicon. However, there are other considerations as well. Since the chips are not encased in plastic, silicon’s sensitivity to light may come into play, particularly with the ultra-low-power devices that would be typical of an IoT application. The effects of light occur at the device physics level.
Light shining on the device can provide enough energy to cause perturbations in the chip’s behavior. The problem can be solved by depositing an opaque material on the chip. Light sensitivity is one example of how an experienced contract manufacturer can identify potential problems that may not be obvious to designers.
Many of the design issues encountered with conventional packages become more important with WLCSP. Providing an energy source is probably the most important. Even a coin cell battery is large when compared to systems based on WLCSP devices but it is still usually the best choice.
Energy harvesting may be a possibility for some applications, but even photovoltaic solar harvesters provide only 100 mW/cm². Ultra-low-power MCUs and transceivers are starting to make the harvesting scenario a realistic possibility. Figure 2 shows the components of an energy-harvesting sensor node. Note that some energy storage is usually required because the application is likely to have activity spikes with a majority of time spent in sleep modes. A capacitor bank is the most likely form of energy.
Figure 2: Energy-harvesting sensor node. (Courtesy of Silicon Labs)
The MCU and the transceiver are worth a closer look. The requirements of an energy-harvesting application fit nicely with a simple, proprietary sub-GHz wireless solution. The two significant parameters for the transceiver are energy consumption in standby mode and in transmit mode.
An ideal candidate for such an implementation is an MCU built on the ARM Cortex-M4F core, which has a range of dedicated DSP functions enabling signal processing to be achieved in many fewer clock cycles than on a non-DSP-capable MCU.
Compared to the M3, the M4F core integrates an FPU hardware assist engine and DSP instruction set extensions. It therefore consumes more energy when operating at full throttle. The M3 must execute the algorithm in software, which means it is in active mode longer than the M4F. For many algorithms used to process sensor data, the M4F core consumes less energy.
Locating WLCSP parts
A number of MCU options are available for WLCSP solutions that are built from scratch. In the 32-bit MCU design space, Atmel Corp. offers its SAM4L and SAM4S families in 48 MHz and 120 MHz clock speeds. Both families are based on the ARM Cortex-M4 core. A typical part, the ATSAM4LS4BA-UUR
, offers 256 KB of Flash memory, I²C, IrDA, LIN, SPI, UART, and USB connectivity and a variety of peripherals.
Freescale Semiconductor has a range of WLCSP MCUs in its Kinetis family. The MKL02Z32CAF4R, for example, runs at 48 MHz. Connectivity includes I²C, SPI, and UART/USART.
NXP Semiconductors bases its WLCSP MCUs on ARM’s Cortex-M3 core. The LPC1768UKJ runs on a 100 MHz clock and has 512 KB of Flash memory. It offers CAN, Ethernet, I²C, IrDA, Microwire, SPI, SSI, UART/USART, USB OTG connectivity, and a variety of peripherals.
Silicon Labs offers WLCSP devices based on at least two ARM cores. EFM32LG360F128G-E-CSP81 is based on the Cortex-M3 core that runs at runs at 48 MHz and has 128 KB of Flash memory (64 KB and 512 KB versions are available). Connectivity options include UART/SPI/SmartCard, IrDA, I²C, USB with host and OTG support, and USB 2.0.
The EFM32WG360F64G-A-CSP81 is based on ARM’s Cortex-M4 core that runs at runs at 48 MHz and has 64 KB of Flash memory (128 KB and 256 KB versions are available). It has a similar selection of communications interfaces but also offers DSP instruction support and floating-point unit to execute compute-intensive algorithms. Its analog peripherals are also interesting and include a 12-bit 1 Msamples/s ADC, an on-chip temperature sensor, a 12-bit 500 ksamples/s DAC, and capacitive sensing with up to 16 inputs.
As the definition of a wearable, internet-connected device evolves, the size of systems that integrate MCUs, sensors, and communications is destined to shrink. MCU manufacturers are enabling this trend with WLCSP devices that can be integrated into modules smaller than the MCU itself would be in a conventional plastic package. There are significant challenges to this level of system integration, however, including the energy source and the size of the sensor, which is more difficult to shrink than semiconductor logic. Innovative engineers in all disciplines are coming up with solutions that include new sensing options and energy-harvesting techniques. There is still plenty of room for innovation, however, before the ultra-small wearables—as well as implantable and ingestible—systems move into the product mainstream.
For more information on the parts discussed in this article, use the links provided to access product information pages on the Hotenda website.