Highly Integrated RF and Digital Architectures: Challenges, Benefits and Acceptance
by Lorne Graves, Chief Technologist, Mercury Systems
Moving from raw data to actionable insights requires deep analysis to differentiate the truly valuable signals from the baseline noise. In other words, we need highly sophisticated processing techniques to transform raw data into tangible, meaningful information we can readily act upon to improve the quality of our lives. This processing may occur locally at the point of data collection, or it may occur in a cloud-based processing architecture.
In this article, we consider the practical implications of this highly interconnected world of smart sensors and cloud computing we have created. The complexity of this new world cannot be sustained using the methodologies of yesterday, where key vendors manufacturing discrete components of a complex system are selected based on specialization and expertise. In this new world of interconnectedness, integrated design across national borders and functional boundaries has become the norm. The companies producing these smart sensors require new connections and deep integration between functional organizations that normally exist in isolation. Although this sounds simple in theory, putting this into practice requires a unique company culture capable of engaging in interdisciplinary communication and collaboration towards a common goal.
This Interconnected World
As the electronic devices around us become more sophisticated, we find ourselves surrounded by an ever-increasing number of sensors continuously collecting data: location, time, temperature, humidity, heart rate, proximity, flow, speed, pressure, level, motion and chemical to name but a few. Some are wired while others are wireless, transferring data to local storage devices or uploading data to a cloud-based storage and processing system.
With no universal sensor communication standards, sensor manufacturers are left on their own to design and integrate all necessary functionality. What type of interface protocol is best? One sensor may benefit from I2C, while another selects CAN bus, and yet another is most appropriate for a wired Ethernet connection. As volumes of data from multiple parallel sensor streams are collected for deep analytical processing, each rich data stream must be translated into a common baseline architecture. Without doing so, insights from one sensor stream are only viewed in isolation, thereby potentially missing a bigger picture correlation to other segments of data collected elsewhere or at different points in time. The data from our interconnected world is not ready for value-added processing until it is translated into a common language.
The design of smart sensors does not originate from a single engineering discipline. It requires a truly cross-functional engineering effort, with a broad range of skills spanning the RF domain, digital microelectronics, mechanical design, software engineering, and electrical engineering, among others. If we think beyond the smart sensor design phase and consider practical aspects of manufacturability of these smart sensors, a wider array of disciplines is required. For example, process integration expertise is needed to understand how one step in the manufacturing process impacts another step. LEAN/Six Sigma methodology is required to drive out inefficiencies from the overall device manufacturing process that potentially span countries if not continents. Our interconnected world requires an unprecedented level of collaboration.
Another phenomenon is happening in our interconnected world. In the defense community where I work, we refer to this as SWaP-C – size, weight, power and cost – demands. Our sensors are growing increasingly smaller and more lightweight, allowing us to pack more sensing capability into a smaller space than ever before. Particularly for applications requiring battery power for operation, consumers demand long durations between battery charges. Without a disruptive breakthrough in battery technology, a reduction in the size and weight of our devices dictates an evolution to more efficient methods of power consumption. Finally, widespread sensor adoption cannot occur unless it is affordable for the masses. The single factor that all sensors have in common today is the trend towards lower size, lower weight, lower power and lower cost. And the commonality stops there.
The ideal smart sensor wirelessly transmits its data stream for post-collection data mining and analysis. In order to transmit this data, the smart sensor must incorporate an antenna design. As mentioned earlier, our smart sensors are trending towards lower SWaP-C, yet this introduces a practical challenge as we consider shrinking the antenna. The required wavelength determines the size of the antenna; the laws of physics simply will not allow a further reduction in antenna size. To surmount this limitation, many designers today are turning to metamaterials. These metamaterials are precision engineered to have unique physical properties such that the antenna behaves as if it were larger in size than it really is.
As the number of wireless sensors in our world explodes, we find ourselves immersed in spectrum fratricide. That is, there is so much signal around us that we cannot possibly discern the specific signals of interest from those background signals of no importance. Highly integrated tunable filters embedded within our microelectronics allow us to quickly move between frequencies of interest. Yet this is not our biggest challenge. When a single antenna is used for simultaneous signal transmit and receive, we must develop some very unique isolation techniques in the antennas and the corresponding analog circuitry.
Analog to Digital Challenges
Before signals collected from these smart sensors can be processed, they must first be converted from the analog domain to the digital domain. Analog to digital converter technology has matured at a breathtaking pace to meet bandwidth demands. However, this introduces a new challenge of filtering data to the digital processor. Nonetheless, at a certain point the digital processing architecture struggles to keep up with the rate of analog to digital data conversion. One solution may be to use highly integrated tunable filters on the analog architecture. However, only a certain amount of information can be extracted at a given frequency, thereby shifting the bottleneck and not removing it.
Moore’s Law has clearly demonstrated for many years that the number of transistors per square inch on an integrated circuit will double every 18 months. Although the predictions for the end of Moore’s Law continue, we know that processing power will continue to evolve, even if that rate of evolution slows. However, processing power requires more than just a microprocessor. The performance of memory devices, historically the slowest components in the sensor processing chain, must continue to scale. We see evidence of this requirement with the emergence of new memory technologies, such as 3D XPoint and High Bandwidth Memory (HBM). Yet it is unclear what the ideal type of memory for smart sensors will be. One smart sensor application may benefit from volatile memory, while another may be more appropriate for non-volatile memory. No single next-generation memory technology has emerged as a universal panacea for smart sensor memory demands.
Three-dimensional packaging technologies have made considerable progress towards commercialization over the last 10 years. Through silicon via technology can now produce ultra-compact system-in-package devices with performance no longer limited by the length of wire bonds interconnecting the chips. 3D packaging techniques need not be limited to digital devices; analog devices can also be integrated using the same technologies to create highly integrated devices in the smallest possible footprint. For an even higher level of integration, advanced packaging techniques can incorporate power controllers, inductors, capacitors and heat sinks.
Power and Thermal Challenges
Heat sink technology has made tremendous advances forward to keep pace with developments in packaging techniques that are embedding more processing capability in ever-shrinking form factors. Performance and long-term reliability of these advanced devices are quickly limited by how efficiently heat can dissipated away from the device. To address this challenge, next-generation materials must be developed to most efficiently dissipate this generated heat. Man-made diamond materials, produced from chemical vapor deposition of hydrocarbon-based precursors, provide thermal conductivity 3 times higher than copper and at least 5 times that of heat-spreading ceramics. Beyond material advances for heat dissipation, efforts to modernize fan-cooled heat sinks, both in size and efficiency, have also been largely successful.
Given that smart sensors are functioning around the clock, there may be times when it is desirable to operate in a reduced-power mode. For example, a low-power mode can be activated when an accelerometer has not detected activity for a specified amount of time. There may even be circumstances where the smart sensor can turn off completely, thereafter waking at a predetermined time. Alternatively, dedicated low-power processors can take the simplest sensor data processing tasks from more power-hungry processors. When needed, processing demands can shift from the low-power processor and return to 100% high-power processing conditions when needed.
Consider the experience of a company seeking to commercialize a new smart sensor today. Product requirements are defined, and in doing so, the communication architecture is established. Analog and digital components are selected. Advanced modeling has been completed, and a number of challenges have been identified. To achieve the absolute smallest possible form factor, hybrid printed circuit boards are required. They include a variety of analog components and typically have multiple laminations to provide the best possible RF performance. To minimize space, bare die components are selected whenever possible. However, not every component is available as bare die; some components must be selected as surface mount. Designers of smart sensors may select multi-chip modules (MCM) to further reduce size and weight. These MCM devices in turn present manufacturing challenges of their own, including wafer thinning, through silicon via formation, bonding and test. As the form factor of the sensor becomes small enough, soon the RF components are in proximity of the noisier digital components. Finally, the sensor itself must be tested and debugged to address manufacturing anomalies.
Considering the manufacturing requirements outlined above, it becomes readily apparent that the broad range of capabilities are rarely housed in a single facility. Semiconductor foundries and outsourced semiconductor assembly and test (OSAT) companies have the expertise to create advanced 3-D multi-chip modules. Organizations with RF expertise rarely include the skill set required for integration with digital technologies, particularly when the smallest of form factors restricts the overall design space. There is no shortage of contract manufacturers with surface mount assembly specialization, yet few contract manufacturers are capable of producing high-reliability devices with hybrid manufacturing techniques.
Producing a truly integrated, state-of-the-art smart sensor demands a new type of contract manufacturer with deep vertical integration knowledge and practical manufacturing experience in the following areas:
- RF manufacturing
- RF/digital integration
- Hybrid manufacturing including wire bond, flip chip and surface mount technologies
- 3-D packaging technologies
- State-of-the-art test capabilities, including custom software capability
Even with all of the above design skills in the same company, one additional critical element is needed: co-located engineering. Given the wide breadth of manufacturing steps, inevitably there will be interactions between individual manufacturing steps in the smart sensor manufacturing process flow. To optimize yield, reduce cycle time and minimize rework, participants from various engineering disciplines must partner closely with manufacturing to understand subtle but significant complexities that impact device yield and possibly reliability.
Cost-effective high-volume production of smart sensor technologies to interconnect the devices and people in our lives will push the boundaries of what we perceive to be a contract manufacturer. Only entities with broad, yet tightly knit, engineering and manufacturing teams can rise to effectively address this challenge.
Having thoroughly reviewed the challenges associated with creating a world of interconnected smart sensors, we can now turn to a discussion on benefits. Small, lightweight sensors are integrated into the electronics we carry, the cars we drive, the planes we fly, the offices where we work, the homes where we live and even the factories producing the smart sensors. To minimize any potential disruption to our lives, adaptive technology determines when the sensors operate at full power and when low-power mode is permissible. This incredible technology is small enough for us to wear on our wrists in the form of a smart watch, for example, yet affordable for large numbers of people. With components designed for modularity, multiple sensors are integrated into the same device. When a performance enhancement is required or if an additional sensor capability is required, design modularity of these smart sensors enables rapid, cost-effective upgrades to move from initial design through volume manufacturing.
Consumer electronics have been a driving force behind smart sensor proliferation. No longer limited to phones, smart functionality has taken over our watches, our lighting, our thermostats, even our household appliances like refrigerators and washing machines. Smart sensors on commercial aircraft are providing real-time data on critical components to detect and diagnose issues that will impact future operational efficiency. Our appetite for smart sensor adoption shows no sign of being satiated in the near future.
Yet smart sensors need not operate only to make our lives more convenient. As military engagements continue to transition away from direct warfighter to warfighter combat, military battles have now become a contest to determine who can collect and exploit data against their enemy more effectively. Precision guidance, navigation and control systems for smart munitions, for example, enable enemy targeting while minimizing collateral damage.
Consider another example, developed by the Wyss Institute from the Harvard School of Engineering and Applied Sciences. Researchers have developed autonomous flying microrobots, or RoboBees, capable of operating in coordinated groups or swarms. Applications include search and rescue operations, environmental monitoring and even crop pollination, where RoboBees replace organic species of bees.
The Internet of Things is driving a remarkable adoption of small, lightweight and incredibly powerful sensors generating volumes of data that must be mined for significance. With more smart sensors weaving connections in our lives, we are experiencing RF fracticide, a sense of being overwhelmed by the congestion of the electromagnetic spectrum we have created. Our smart sensors must grow increasingly intelligent, finding ways to work around this congestion while becoming more power efficient.
As memory capabilities and device integration become more complex, thermal and power management design considerations are becoming more significant. These design challenges cannot be viewed in isolation without considering manufacturability. No longer can a series of disconnected specialty manufacturing groups, optimized only for cost, coexist. The expertise required for cost-effective, high-volume production of smart sensors requires a new breed of contract manufacturers with highly integrated, cross-functional engineering and manufacturing resources. By choosing to embrace this new model, the limits of future smart sensors are constrained only by our imaginations.
For more information about Mercury Systems visit www.mrcy.com