TECH BRIEFS EBOOK 2020 Guide to Smart Sensor Technology, Connectivity and the IoT TECH BRIEFS EBOOK

3 Contents

3 The Expanding Role of Sensors in the IoT/IIoT

6 Smart Sensor Technology for the IoT

11 Networking the IoT with IEEE 802.15.4/6LoWPAN 11

14 The Sensors Behind the GEN II Wearable Device

18 Key Factors to Consider When Designing a Smart Factory

20 Executive Forum: Sensors & the IoT 14

18

ABOUT ON THE COVER

The increasing variety of Internet of The IoT, sensors, and machine vision Things (IoT) applications, whether for city are powerful tools many companies are infrastructures, factories, or wearable integrating into their manufacturing devices, is driving predictions that the techniques. Augmented reality can be sales volume for sensors will reach 75 used to connect human personnel with billion units by 2025. The 2020 Guide to the digital spectrum. Learn more in the Smart Sensor Technology, Connectivity article “Key Factors to Consider When and the IoT compiles the top resources in Designing a Smart Factory” on page 18. smart sensor technology from Tech Briefs and Sensor Technology magazines.

2 SMART SENSORS EBOOK Fig. 1 Dual optical image image stabilization (OIS) sensors.

The Expanding Role of Sensors in the IoT/IIoT New Applications Are Demanding New and More Sophisticated Sensors.

he increasing variety of applications for the Internet of more intelligent and to achieve ever higher levels of performance. Things is driving predictions that the sales volume for For the foreseeable future, this trend will be seen both for con- sensors will reach 75 billion units by 2025. Applications sumer applications (IoT) and industrial applications (IIoT), although unforeseen just five years ago, some of which are de- generally, the requirements for industrial applications are an order T scribed in this article, are requiring sensors to become of magnitude more demanding than for consumer applications.

Fig. 2 Checking on machine performance.

SMART SENSORS EBOOK 3 THE EXPANDING ROLE OF SENSORS

Fig. 3 Drone out for delivery.

Smart Phones noise, to provide the best possible user experience. For the very high Consumers are now looking for a level of photo quality in slim smart- precision required by VR and AR controllers, sensors also need high phones that had earlier only been possible with expensive and bulky resolution to be able to accurately handle the slow and precise move- DSLR cameras — and this while walking, if not running. To achieve such ments of a hand. image quality with the limited optics that can be accommodated in such a tight space, OEMs have deployed Optical Image Stabilization (OIS) Navigation and Electronic Image Stabilization (EIS) solutions. The performance of Dead-reckoning systems are used for navigation in both consumer the OIS/EIS components, especially the motion sensor, has a major im- and commercial applications. They provide location data in the ab- pact on overall performance. For example, the Google Pixel 2 device, sence of GNSS/GPS signals by means of inertial measurement units which has a camera with both OIS and EIS, received the highest DXO (IMU), which use linear acceleration information derived from ac- Mark score from DxOMark Image Labs ever given to a smartphone. The celerometers, and angular rotation rate from gyroscopes, to calculate technology that led to that high performance is a motion tracking sensor altitude, angular rate, linear velocity, and position. The accuracy of with extremely low gyroscope and accelerometer noise (<0.004 these measurements is critical because errors are additive — they in- dps/√Hz and <100 µg/√HZ respectively), sensitivity error within 0.5%, crease over time. For high-precision navigation, a gyroscope needs and very tight temperature stability (less than ±0.016%/°C). to have extremely low gyro offset for temperature (between 3 to 5 mdps/°C) and low gyro noise density (below 4 mdps/√Hz). The Variety of Realities: VR, AR, and XR It is not just cameras that require precise motion sensor perform- Location ance. Virtual reality (VR), augmented reality (AR), and mixed reality A promising consumer application for dead-reckoning navigation, (XR) all need low noise and extremely good temperature stability. is to sense one’s location inside a shopping mall. It is particularly chal- For these applications, the gyroscope noise requirement is similar lenging to sense the floor someone is on. It would allow users to know to that for OIS, (<0.004 dps/√Hz), but the accelerometer noise has which shops are near them. But more importantly, it could save time to be even lower, (<75 µg/√HZ). Other key parameters are a very low and lives by providing emergency services with the specific floor and sensitivity error (<0.5%) for both gyroscope and accelerometer, and location where the mobile caller requested help, rather than searching also good temperature stability (<10 mdps/~C) for the gyroscope. floor by floor. This can be achieved with high-precision, low-noise Another system-level feature that is beneficial for VR/AR/XR appli- pressure sensors that can measure individual stairs. This is now cations is the ability to synchronize the various sensors to a ppm-level achievable because pressure sensors are available that can go as low (0.0001%) accuracy system clock, so that timing errors stemming as one pascal, or about 10 cm, in precision. from less accurate internal oscillator clocks (1% clock error for PLL) can be avoided in applications that depend on precise timing. Factory Maintenance VR stand-alone head-mounted devices (HMD) especially, need Predictive maintenance (PdM) is one of the most important appli- good temperature linearity and stability, as well as low hysteresis and cations for sensors in the IIoT. It is a system in which machine and

4 SMART SENSORS EBOOK THE EXPANDING ROLE OF SENSORS

other processing parameters are continuously monitored. Managers lenging environments. These ultrasonic solutions can meet the can therefore have an ongoing picture of machine health throughout needs of access control: a false acceptance rate (FAR — giving ac- a factory, enabling them to schedule maintenance when data indi- cess to an unauthorized person) of at least one in 50,000 and false cates there might be an impending problem. This is an alternative to rejection rate (FRR — not giving access to an authorized person) of waiting for a failure to occur or shutting down at regular intervals less than one in 50. whether or not it is necessary. Continuous online monitoring can provide benefits such as reduc- What’s Ahead ing downtime by as much as 70%, extending motor lifetimes by up These applications are only the tip of the iceberg for sensor to 30%, and reducing energy consumption by up to 10%. Although applications in the rapidly growing world of interconnectivity. these types of sensors are now being integrated into new production There is a push-pull relationship between sensor performance equipment, they can also be added to older equipment. and applications. The proliferation of new and improving appli- cations drives the birth of new sensor technologies and the con- Agriculture stant improvement of existing sensor performance. At the same Another example is driverless agri cultural tractors. In addition to time the availability of new types of sensors and higher levels the other requirements for autonomous vehicles, tractors have to deal of performance can suggest applications that weren’t pos- with extreme vibration, which can cause navigational errors because sible before. These two trends will be the engine to of its effects on the motion tracking sensors. The sensors in the pre- drive the ever expanding IoT and IIoT connected cision navigation system therefore require vibration isolation — bias world. stability of below 2°/hr for gyroscopes and below 10 µg for ac- This article was written by Nicolas Sauvage, Sr. celerometers, an angle random walk (ARW) of 0.084°/√hr, and ve- Director Ecosystem, at InvenSense, a TDK Group locity random walk (VRW) of 0.016 m/s/√hr. It is also important to Company, San Jose, CA. For more information, have very small misalignment and mounting errors — less than 0.05° contact Nicolas Sauvage at nicolas.sauvage@ for both. invensense.com.

Home Microphones Microphones are increasingly de ployed in home automation. For example, smart speakers have multiple microphones — the Amazon Echo has seven. You can easily imagine a future where a family has more than 100 microphones inside their home. These would have to be high performance, with a signal-to-noise-ratio (SNR) of more than 70 dB, while having an acoustic overload point (AOP) of more than 120 dB. One device that meets these specs is the InvenSense ICS- 40730 microphone, which has a 74 dB SNR and 123 dB AOP. This allows the microphone to listen to everything happening near the smart speaker even in the presence of loud noises, like a door closing too hard, or a TV playing an action movie too close to the speaker.

Autonomous Vehicles Range-finding or time-of-flight (ToF) sensors with improved performance are starting to ad- dress the requirements of machines that move, such as self-driving cars, service robots, and in- dustrial monitoring drones. For them to be able to “see” their surroundings precisely and avoid obstacles, they need very precise sensors. For example, ultrasonic-based sensors are now able to measure distances up to five meters very precisely. However, different sensor parameters are im - portant for different applications. For example, drones care more about bias, time, and tem- perature stability, while house-cleaning robots care mostly about bias stability and noise.

Access Control Improved ultrasonic-based sensors can also enhance access control systems. For example, they enable fingerprint sensors that can work well even under metal or plastic, allowing for operation in chal- Fig.4 Fingerprint access.

SMART SENSORS EBOOK 5 Smart Sensor Technology for the IoT

nternet of Things (IoT) applications I — whether for city infrastructures, fac- tories, or wearable devices — use large arrays of sensors collecting data for transmission over the Internet to a cen- tral, cloud-based com- puting resource. Analytics software running on the cloud computers reduces the huge volumes of generated data into actionable informa- tion for users, and commands to actuators back out in the field. Sensors are one key factor in IoT suc- cess, but these are not conventional types that simply convert physical variables into

6 SMART SENSORS EBOOK electrical signals. They have needed to evolve into something more sophisticated to perform a technically and economically viable role within the IoT environment. This article reviews the IoT’s expectations of its sensors — what must be done to achieve the large sensor array’s characteristic of the IoT. Then it addresses how man- ufacturers have responded with improvements to fabrication, more integration, and built-in intelligence, culminating in the concept of the smart sensors now in wide use. It will become evident that sensor intelligence, apart from facilitating IoT connec- tivity, also creates many more benefits related to predictive maintenance, more flex- ible manufacturing, and improved productivity.

What Does the IoT Expect of its Sensors? Sensors have traditionally been functionally simple devices that convert physical variables into electrical signals or changes in electrical properties. While this func- Smart tionality is an essential starting point, sensors need to add the following properties to perform as IoT components: • Low cost, so they can be economically deployed in large numbers • Physically small, to “disappear” unobtrusively into any environment Sensor • Wireless, as a wired connection is typically not possible • Self-identification and self-validation • Very low power, so it can survive for years without a battery change, or manage with energy harvesting • Robust, to minimize or eliminate maintenance Technology • Self-diagnostic and self-healing • Self-calibrating, or accepts calibration commands via wireless link • Data pre-processing, to reduce load on gateways, PLCs, and cloud resources Information from multiple sensors can be combined and correlated to infer for the conclusions about latent problems; for example, temperature sensor and vibration sensor data can be used to detect the onset of mechanical failure. In some cases, the two sensor functions are available in one device; in oth- ers, the functions are combined in software to create a ‘soft’ sensor.

IoT The Manufacturers’ Response: Smart Sensor Solutions This section looks at the smart sensors that have been developed for IoT applica- tions in terms of both their building blocks and their fabrication, and then reviews some of the advantages that accrue from the sensors’ in-built intelligence, especially the possibilities for self-diagnostics and repair.

What’s in a Smart Sensor and What is it Capable of? We’ve reviewed the IoT’s expectations of a smart sensor, but how has the industry responded? What’s built into a modern smart sensor, and what is it capable of?

Sensing Unit Local User Interface

Signal Analog to Digital Application Communication Conditioning Conversion Algorithms Unit / Transceiver

Memory

Background Image: By a-image/Shutterstock.com Figure 1. Smart sensor building blocks. (Image: ©Premier Farnell Ltd.)

SMART SENSORS EBOOK 7 SMART SENSOR TECHNOLOGY

Solar Panel

Power Diode Management

Current Transducer Diode Storage Battery

Current Timer Sub 1 GHz Transducer OpAmp Comparator SPI Transceiver

GPIO LED Driver Temperature ADC CPU Sensor

RS485 Driver UART Power Line RAM (ground line) FRAM

MSP430 FRAM MCU

Figure 2. Functional block diagram of a smart fault indicator based on the MSP430 FRAM MCU. (Image: Texas Instruments)

Smart sensors are built as IoT components that convert the real- Smart Sensor: A Practical Example world variable that they’re measuring into a digital data stream for An application developed by Texas Instruments provides a practical transmission to a gateway. Figure 1 shows how they do this. The ap- example of a smart sensor, and how its building blocks work together plication algorithms are performed by a built-in microprocessor unit to generate useful information from analog current and temperature (MPU). These can run filtering, compensation, and any other process- measurement, as well as providing the intelligence for the other func- specific signal conditioning tasks. tions mentioned. The application uses a variant of their ultra-low- The MPU’s intelligence can be used for many other functions power MSP430 MCU range to build a smart fault indicator for electric as well to reduce the load on the IoT’s more central resources; power distribution networks. for example, calibration data can be sent to the MPU so the sen- When properly installed, fault indicators reduce operating costs sor is automatically set up for any production changes. The MPU and service interruptions by providing information about a failed sec- can also spot any production parameters that start to drift be- tion of the network. At the same time, the device increases safety and yond acceptable norms and generate warnings accordingly; op- reduces equipment damage by reducing the need for hazardous erators can then take preventative action before a catastrophic fault-diagnostic procedures. Fault indicators, due to their location, failure occurs. are primarily battery-powered, so low-power operation is also highly If appropriate, the sensor could work in “report by exception” desirable. mode, where it only transmits data if the measured variable value The fault indicators — which are installed on the junctions of the changes significantly from previous sample values. This reduces both overhead power-line network — send measurement data about the the load on the central computing resource and the smart sensor’s temperature and current in power transmission lines wirelessly to the power requirements — usually a critical benefit, as the sensor must concentrator/terminal units mounted on the poles. The concentrators rely on a battery or energy harvesting in the absence of connected use a GSM modem to pass the data to the cellular network to relay power. real-time information to the main station. The main station can also If the smart sensor includes two elements in the probe, sensor self- control and run diagnostics on the fault indicators through this same diagnostics can be built in. Any developing drift in one of the sensor data path. element outputs can be detected immediately. Addi tionally, if a sen- Continuous connection to the main station has several advantages. sor fails entirely — for example, due to a short-circuit — the process The first is the ability to remotely monitor fault conditions rather than can continue with the second measuring element. Alternatively, a searching for them in the field. A smart fault indicator can also con- probe can contain two sensors that work together for improved mon- stantly monitor temperature and current so that the controller at the itoring feedback. main station has real-time status information about the power distri-

8 SMART SENSORS EBOOK SMART SENSOR TECHNOLOGY

bution network. Accordingly, power utility providers can quickly identify the fault loca- tion, minimize power downtime, and even take action before a failure occurs. Workers at the Processor main station can run diagnostics PAM Battery on the fault indicators at re- quired intervals to check that they are working correctly. DC-DC Ram / Flash Figure 2 is a functional block Converter FPGA diagram of such a smart fault in- dicator based on the TI MSP430 ferroelectric random-access IAS memory (FRAM) microcontroller RTM (MCU). The current transducer produces an analog voltage pro- portional to power-line current. BUS An operational amplifier (op amp) amplifies and filters this voltage signal. The analog-to- digital converter (ADC) on the Figure 3. Hardware configuration of a wireless sensor node. (Image: ©Premier Farnell Ltd.) MCU samples the output of the op amp. The digital stream from the ADC is then analyzed by software throughput and reduces the central processor — or local PLC’s — pro- running on the CPU or accelerator. The op amp output is also con- cessing load. nected to a comparator on the MCU. The comparator generates a flag Manufacturing flexibility is improved — a vital advantage in today’s to the central processing unit (CPU) in the MCU if the input level trans- competitive environment. Intelligent sensors can be remotely pro- gresses a predetermined threshold. grammed with suitable parameters every time a product change is The MSP430’s computing power allows frequency-domain current required. Production, inspection, packaging, and dispatch can be set measurement analysis that provides a deeper insight into power line for even single-unit batch sizes at mass-production prices, so each status than previous time-domain methods. The fast FRAM read-and- consumer can receive a personalized, one-off product. write speeds enable the accumulation of data for pattern analysis, Feedback from linear position sensors has traditionally been ham- while the MCU’s ultra-low-power operating modes allow extended pered by problems relating to system noise, signal attenuation, and battery life operation. response dynamics. Each sensor needed tuning to overcome these problems. Honeywell offers a solution with their SPS-L075-HALS Fabrication Smart Position Sensors. These can self-calibrate by using a patented To realize the full potential of the IoT, sensor fabrication methods combination of an ASIC and an array of MR (magnetoresistive) sen- must continue to reduce the size, weight, power, and cost (SWaP-C) sors. This accurately and reliably determines the position of a magnet of the sensor component and system. The same trend needs to apply attached to moving objects such as elevators, valves, or machinery. to sensor packaging, which currently accounts for as much as 80% The MR array measures the output of the MR sensors mounted of the overall cost and form factor. along the magnet’s direction of travel. The output and the MR sensor Smart sensors form when microelectromechanical system (MEMS) sequence determine the nearest pair of sensors to the center of the sensor elements are closely integrated with CMOS integrated circuits magnet location. The output from this pair is then used to determine (ICs). These ICs provide device bias, signal amplification, and other the position of the magnet between them. This non-contacting tech- signal processing functions. Originally, the wafer-level vacuum pack- nology can provide enhanced product life and durability with less aging (WLVP) technology used included only discrete sensor devices, downtime. A self-diagnostics feature can further reduce downtime and smart sensors were realized by connecting discrete MEMS chips levels. to IC chips through the package or board substrate in an approach These sensors also tick other IoT smart sensor requirements. Their called multi-chip integration. An improved approach interconnects small size allows installation where space is at a premium, while IP67 the CMOS IC and sensor elements directly, without the use of routing and IP69K sealing options allow deployment in harsh environments. layers in the package or board, in a construction known as a system- They are smart enough to replace several sensor and switch compo- on-chip (SoC). When compared to the discrete multi-chip packaging nents together with the extra wiring, external components, and con- approach, SoC is typically more complex but leads to reduced para- nections also previously needed. The sensors are used in aerospace, sitics, smaller footprints, higher interconnect densities, and lower medical, and industrial applications. package costs. Smart Sensors with Self-Diagnostics and Repair Capabilities Other Advantages of Smart Sensor Intelligence Smart sensors can also be well-suited to safety-critical applica- Smart photoelectric sensors can detect patterns in an object struc- tions like detection of hazardous gas, fire, or intruders. Conditions ture and any changes in them. This happens autonomously in the sen- in these environments can be harsh, and the sensors can be difficult sor, not in any external computing element. This increases processing to access for maintenance or battery replacement, yet high reliabil-

SMART SENSORS EBOOK 9 SMART SENSOR TECHNOLOGY

Software and Hardware causes Energy causes

Problem Down Down Down Down Soft Low Energy Solution Processor Ram IAS RTM Bug Energy Depletion

X PAM enables FPGA processor to replace processor

X PAM enables FPGA memory to replace Dead Node RAM/Flash memory

X Wait for recharging battery

X PAM changes mode of operation to relay point

X PAM changes mode of operation to local processing

Malfunctioning Node X Processor reboots

X PAM selects consistent mode of operation and waits for battery recharge

Figure 4. Issues and corrective actions for a self-diagnostic sensor node. (Image: ©Premier Farnell Ltd.) ity is critical. A team at the Lab-STICC Research Center, University DC converters). The node also includes a Power and Availability Man- of South-Brittany, has been developing a solution that improves re- ager (PAM) combined with an FPGA-configurable zone. The first one liability by using dual probes and hardware that can self-diagnose is considered as the intelligent part for the best use of energy, auto- and repair itself. diagnosis, and fault-tolerance, while the other enhances the availabil- The ultimate goal of their project is to integrate all the elements ity of the sensor node. described into a single discrete device, suitable for applications such The table in Figure 4 shows how the sensor node can respond to as hazardous gas detection in areas such as harbors or warehouses. various node issues. The FPGA contains a soft-core 8051 CPU that is The project centers on a node that can pinpoint an internal failure and activated when performance enhancement is needed or to replace take corrective action to improve both reliability and energy effi- the main processor if it fails. The FPGA is an Actel type IGL00V2, cho- ciency. This reduces the node’s vulnerability and alleviates mainte- sen for its reliability and low power consumption. The remainder of nance costs. The design recognizes the limitations of such sensors: the node comprises a PIC processor, RAM memory, Miwi radio trans- restricted battery autonomy, energy harvesting subject to unreliable ceiver module, two Oldham OLCT 80 gas detectors, LM3100 and energy source behavior, limited processing and storage resources, MAX618 power switches, and a battery. and a need for wireless communications. The node is equipped with two sensors; during normal operation, Conclusion the first captures environmental data while the second is only acti- In this article, we have seen how chip manufacturers and re- vated by users to verify the obtained data. If the first sensor were to searchers have been responding to the IoT’s need for smart sensors. fail, the node’s reliability is downgraded, while battery power is being This has partly been a matter of adding intelligence and communica- wasted on supplying the non-functioning sensor. However, if the node tions capabilities to the basic transducer function, but it also involves disconnects the first sensor and switches to the second, no energy is improved fabrication. By integrating the MEMS sensor elements and wasted and node reliability is maintained. CMOS computing components onto a single substrate, smart sensors Accordingly, the project’s objective was to develop a novel self-di- can be implemented in small, low-cost packages that can be embed- agnostic based on functional and physical tests to detect a hardware ded in space-constrained applications with resilience to their envi- failure in any component of the wireless sensor node. This method ronmental conditions. can identify exactly which node component has failed and indicate Accordingly, IoT designers can source the sensors that they need suitable remedial action. —small, cheap, resilient, and low-power enough for ubiquitous de- Figure 3 shows the hardware configuration of the self-reconfig- ployment, while having the intelligence to deliver useful information urable sensor node. Its components include a processor, a as well as raw data. They also facilitate more flexible, granular au- RAM/FLASH memory, an Interface for Actuator and Sensors (IAS) to tomation, as they can accept incoming commands for recalibration interface with the environment, a Radio Transceiver Module (RTM) to to accommodate production changes. transmit and receive data, and a battery with power switches (DC- This article was contributed by Newark element 14, Chicago, IL.

10 SMART SENSORS EBOOK Networking the IoT with IEEE 802.15.4/6LoWPAN

he Industrial Internet of Things is predicated on large-scale, networks, specifically the Internet Engineering Task Force (IETF) IPv6 distributed sensor/control networks that can run unat- over Low-power Wireless Personal Area Networks (6LoWPAN) im- tended for months to years with very low power consump- plementation. This implementation supports both the cloud and fog Ttion. The characteristic behavior of this type of network models. entails very short bursts of message traffic over short distances using wireless technologies, often described as a low-rate, wireless IEEE 802.15.4 PHY Layer personal area network (LR-WPAN). We keep the data frames short The IEEE 802 standards family is broken out into a number of to lessen the possibility of radio interference forcing the need to re- task groups including 802.3 (Ethernet) and 802.11 (Wi-Fi), as well transmit. One such LR-WPAN approach uses the IEEE 802.15.4 stan- as 802.15 (Wireless PAN). In particular, IEEE 802.15.4 (15.4 for dard. This describes a physical layer and media access control that brevity) is the responsibility of Task Group 4, which is responsible are often used in the industrial control and automation applications for various characteristics of the protocol including RF spectrum referred to as Supervisory Control and Data Acquisition (SCADA). Bytes: 2 1 0 to 20 n 2 In the IoT, local “edge” de- Frame Control Data Frame Check Address vices, typically sensors, collect Field Sequence Sequence MAC Layer Information Frame Payload data and send it to a data cen- (FCF) Number (FCS) ter — “the cloud” — for pro- cessing. Getting the data to MAC Header (MHR) MAC Payload MAC Footer the cloud requires communi- (MFR) cating using the standard IP protocol stack. This can be done by directly connecting the edge devices via the Inter- net to the data centers — the Synch Header PHY Header MAC Protocol Data Unit (MPDU) “cloud model.” Or, we can (SHR) (PHR) PHY Service Data Unit (PSDU) communicate from the edge devices to a collection point known as a border gateway to Bytes: 4 1 1 2 1 2 2 2 n=2 2 have the data relayed from there to the data center — the Preamble Sequence SFDFrame Length FCF Data Sequence Number PAN ID Dest Address Source AddressPayload FCS “fog model.” This article will describe characteristics of IEEE 802.15.4 Figure 1. IEEE 802.15.4 Frame Format

SMART SENSORS EBOOK 11 NETWORKING THE IOT

and the physical layers. The 15.4 standard has been expanded to IEEE 802.15.4 MAC Layer include Radio Frequency identification (RFID) PHYs, ultra-wide- The IEEE 802.15.4 MAC layer (OSI Model layer two — data link layer) band (UWB) PHYs, and is also being discussed as a possible solu- is responsible for: tion for both car-to-car and car-to-curb communications. • Joining and leaving the PAN; 802.15.4 only addresses the physical (PHY) and media access con- • Carrier Sense Multiple Access with Collision Avoidance (CSMA-CA) trol (MAC) layers — in the OSI network model, layers one and two. It for channel access; leaves the upper layers to the implementer. At layer three and above, • Guaranteed Time Slot (GTS) transmissions; there are a plethora of offerings including Zigbee, Z-Wave, Thread, • Establishing a reliable link between two peer MAC entities; and 6LoWPAN. Each of these implements the remainder of the OSI • Beacon transmissions for a coordinator; protocol model to deliver services such as routing and discovery as • Synchronization to the beacons. well as APIs for user applications. In addition, the MAC layer supports the use of symmetric encryp- In general, 15.4 supports data transfer rates at 20 Kbit/s, 40 tion using the AES-128 encryption algorithm. There are also options Kbit/s, 100 Kbit/s (soon), and 250 Kbit/s. The basic framework as- for SHA-based hashes and access control lists to limit the transfer of sumes a 10-meter range at 250 Kbit/s. Even lower data rates are sensitive information to specific nodes or links. Finally, the MAC com- achievable to further limit power consumption. In spite of the 10- putes a freshness check between frame receptions to help minimize meter (32-foot) range specification, in the 2.4GHz ISM band, typical the potential for old frames that may have been traveling over a cir- achievable ranges for IEEE 802.15.4 radios are on the order of 100 cuitous path from being delivered late to the upper-layer protocols. feet indoors, and 200 – 300 feet outdoors. In the sub-GHz frequen- cies, practical implementations of the protocol have been demon- Node Types and Network Topologies strated at ranges of over 6.5 km (4 miles) with appropriate antennas IEEE 802.15.4 identifies two different types of network nodes: re- in the 900 MHz ISM band. duced function devices (RFD) and full-function devices (FFD). FFDs At the physical layer, IEEE 802.15.4 manages the RF transceiver can talk to other FFDs or to RFDs and can even create their own and channel selection, as well as energy and signal management fa- networks. However, RFDs can only talk to FFDs. This implies a hi- cilities. There are six PHYs currently defined, depending on the fre- erarchy that leads to two possible network topologies: a star topol- quency range and data performance required. Four of them use ogy or a peer-to-peer topology such as a mesh. These are depicted Direct Sequence Spread Spectrum (DSSS) frequency hopping tech- in Figure 2. niques. Chirp spread spectrum (CSS) is in use in the Ultra-Wide The star topology is the easiest and least expensive to implement, Band (UWB) and 2450 MHz frequency bands. Parallel Sequence it only requires a single FFD. The rest of the devices can be either Spread Spectrum (PSSS) is available only with the hybrid RFDs or FFDs, depending on the implementation. The downside of binary/amplitude shift keying modulation technique found in the the star topology is that the coordinator represents a single point of European 868 MHz band. failure. This can result in a total failure of the network and should be The frame size for 15.4 is 133 bytes including PHY, MAC, and the avoided in all but the simplest of applications. data payload. The format for this frame can be seen in Figure 1. By The use of a mesh topology provides for multiple, redundant, com- keeping the frame relatively short, we can limit the amount of time munication paths to ensure the delivery of messages. When running needed to transmit it while simultaneously limiting the probability in mesh mode, the network is essentially an ad hoc, self-organizing of radio interference due to the normal operation of industrial entity. Connectivity can therefore be continued despite changing RF equipment. propagation characteristics such as multi-path or effects from fo- liage. The use of a mesh topology also provides for moving nodes, Star Topology Peer-to-peer or Mesh Topology such as found in industrial robot- ics. A “lossy mesh” is one where not all links are reliable, so a higher-layer routing protocol is used to reroute message traffic based on connectivity at any given point in time.

IPv6 Owing to the exhaustion of the IPv4 address space, there is con- siderable interest in transitioning to IPv6, which provides layer three (network) and layer four Reduced Function Device (transport) and sits on top of the MAC layer. Normally, IPv6 uses a Full Function Device forty-byte header and provides 128 bits of address space, which Communication Flow can deal with even the largest es- timates for IoT-connected de- Figure 2. Topology Options vices.

12 SMART SENSORS EBOOK NETWORKING THE IOT

However, when coupled with AES-128 en- IEEE 802.15.4 Header – 22 bytes cryption overhead, the use of a default-sized IPv6 header would leave only thirty-three Typical Source Address DestinationAddress IEEE 802.15.4 FCF bytes for user payload in the frame. IP DSN (00-CA-FE-DE-AD-BE-EF-00-01) (00-EF-BE-AD-AD-DE-AF-02-03) Length header compression (IPHC) was introduced PAN ID to address the problem. This can reduce the size of the IPv6 header to a mere ten bytes, Compressed UDP/ipv6 Header (fe80::0314:3b33:1111:2222-> fe80::0314:3b33:3333:4444) including routing for Internet traversal. This Local PAN IPHC can be seen in Figure 3. Routing UDP NHC Only IPHC Dispatch

This combination of IPv6, IPHC, and UDP Port Checksum standard TCP/UDP sitting on top of the 15.4 PHY and MAC layers is what is known Compressed UDP/ipv6 Header (fe80::0314:3b33:1111:2222-> ff02::1) as 6LoWPAN. When coupled with the use Multicast for of POSIX-style sockets, the developer can Neighbor & UDP

Router NHC have end-to-end packet delivery any- IPHC Dispatch Discovery UDP Port McastGrp where in the world using normal Internet Checksum protocols. Compressed UDP/ipv6 Header (fe80::0314:3b33:1111:2222-> 2001:4328:b003::68) Implementing 6LoWPAN for the IoT Routing for There are many existing implementations Internet CID UDP NHC IPHC

Traversal DST ID Dispatch UDP Port of 6LoWPAN. One is sub-GHz 6LoWPAN for Hop Limit Checksum the advanced metering infrastructure (AMI) currently implemented in power meters for Figure 3. IP Header Compression of IPv6 Header residential use. These meters provide utilities a means of both reading and controlling power usage across the electrical grid. They rely on a lossy-mesh routing facility to ensure the delivery of meter measurements regard- less of multi-path or atmospheric effects such as rain or snow. 6LoWPAN code size is moderate. The typ- ical implementation is on the order of about 30KB and is often implemented directly in the radios from companies like Texas Instru- ments, Silicon Labs, and others. This ap- proach provides a UART-style interface between the sensor microcontroller and the radio, thereby offloading the protocol over- head to the radio unit. Alternatively, many operating systems such as Linux, already implement 6LoWPAN on a number of radio platforms. This pro- vides for the use of Linux-based border gate- ways to provide security for the edge devices using a fog model via hardened ker- nels, next-generation firewalls, and more. The border gateway can also be used to pro- Figure 4. Raspberry Pi Border Router with 6LoWPAN Module vide data filtering and compression to re- duce overall communication costs. Summary Since 6LoWPAN is compatible with normal Internet protocols, the The IoT is all about connectivity and the IEEE 802.15.4 standard developer is free to leverage higher-level protocols such as MQTT, provides a means that is ideal for implementing it — low-power op- CoAP, and HTTP for communications from application to application. eration across a lossy mesh. The use of 6LoWPAN on top of IEEE A border router that interfaces to 6LoWPAN on the southbound side 802.15.4 provides for secure, transparent connectivity with the cloud and standard IPv4 or IPv6 on the northbound side can easily provide and significantly reduces the burden on developers and system de- network address translation (NAT) automatic translations from the signers by providing standard IP-compatible protocols and readily internal 6LoWPAN packet format to standard IPv6 or via a NAT64 available libraries. to standard IPv4. This makes the edge device’s addressing com- This article was written by Mike Anderson, CTO/Chief Scientist, The pletely transparent to the cloud and to the developer. A Raspberry PTR Group (Ashburn, VA). Pi-based border router with a 6LoWPAN module attached is shown in Figure 4.

SMART SENSORS EBOOK 13 The Sensors Behind the GEN II Wearable Device

The population is aging and more people need healthcare support, which is having a big im pact on the overall cost of medical care. As a result, authorities and health in- surance companies are putting more emphasis on prevention, health awareness, and lifestyle including more interest in monitoring certain vital body pa- rameters. This is why companies in the smart watch and health watch business have seen their revenue grow over the past few years. Buy- ing a health watch and monitoring certain body parameters over a period of time gets the user familiar with these numbers and uses them to adapt day-to-day life for improvement. This article focuses on Analog Devices’ wearable VSM platform and the sensor technologies used (Figure 1). ADI is not a manu- facturer of final products; however, this platform has been de- signed as a reference to help the electronic designer and system Figure 1. ADI’s GEN II integrated wear- architect speed up the development process while designing smarter, more able device refer- accurate wearable de vices for the professional and medical market. ence design.

AVDD DVDD

Time Slot Switch Analog Block PDC AFE: PD1-2 Signal Conditioning ADPD105/ TIA BPF ADPD106/ ±1 Integrator ADPD107 WLCSP Versions

VBIAS

VREF AFE: Signal Conditioning Time Slot A 1 µF TIA BPF ±1 Integrator Data

V 14-Bit BIAS ADC

AFE: Time Slot B Data PD3-4 Signal Conditioning AFE Configuration GPIO0 TIA BPF ±1 Integrator A GPIO1 PD3-4 On B ADPD105/ VBIAS ADPD107 Slot Only Select Digital AFE: Datapath Signal Conditioning and Interface MOSI Control TIA BPF ±1 Integrator MISO ADPD106/ ADPD107 SCLK Only VBIAS CS

SDA ADPD105 LED3 LEDX3 SCL Only LED3 Driver LED3 Level and Timing Control LEDX2 LED2 LED2 Driver LED2 Level and Timing Control DGND LED1 LEDX1 LED1 Driver LED1 Level and Timing Control

VLED AGND LGND

Figure 2. Block diagram of ADPD105/ADPD106/ADPD107.

14 SMART SENSORS EBOOK What Are We Measuring? How and Where? appropriate. Light is sent into the tissue and its reflection, as a result A broad range of vital parameters can be measured with a wear- of blood flow in the arteries, is captured and measured. From this op- able device. Depending on the overall objective, certain parameters tically received signal, beat-to-beat information can be retrieved. This are more important to measure than others. The location of the wear- technology sounds rather straightforward; however, there are several able device on the body has a big impact on what can be measured challenges and influencers that can make the design difficult, such as and what cannot. The most obvious location is the wrist. We are ac- motion and ambient light. customed to wearing a device on our wrist, which is why so many Analog Devices’ GEN II wearable device reference platform has smart watches and wrist-worn devices are on the market. Besides most of the previously described technologies onboard. The device measuring on the wrist, the head is another good location for wear- is designed to be worn on the wrist but the soft belt can be removed ables; for example, headphones and earbuds are offered in different and the device attached to the chest to use it as a smart patch. The styles that contain embedded sensors to measure parameters such device includes technology to support biopotential measurement, op- as heart rate, oxygen saturation, and temperature. The third location tical heart rate measurement, bioimpedance measurement, motion for wearables on the body is the chest. First-generation heart rate tracking, and temperature measurement — all integrated in a tiny, monitors were designed around a chest strap and this biopotential battery-operated device. measurement method is still a very accurate technique. Today, we The goal for a system like this is to evaluate various sensing tech- tend to prefer a chest patch, as the strap is not very comfortable to nologies and to measure, in an easy way, several vital parameters on wear. Several manufacturers are involved in the design of smart the body. These measurements can be stored into flash memory or patches to monitor vital parameters. sent over a BLE wireless connection to a smart device. Since the Depending on body location, we are not just faced with the choice measurements are done simultaneously, it also can help to find cor- of which parameters can be measured but what technology should relation among several parameters. Biomedical engineers, algorithm be used. For heart rate measurement, biopotential measurement is providers, and entrepreneurs continuously are looking for new tech- one of the oldest technologies. Signals are strong and easy to retrieve nologies, applications, and use cases to detect diseases at an earlier from the body by utilizing two or more electrodes. For this ap- stage in order to minimize negative effects or damage to the body proach, integration of the circuitry in a chest strap or that might occur at a later stage. headphones is perfect; however, measuring biopo- tential signals at a single point like the wrist is Sensors Make the Device nearly impossible. You need to measure The device is designed around two PCBs that are stacked as a By Syda Productions/Shutterstock.com across the heart, where these elec- sandwich. The main board contains a low-power tric signals are being gener- processor, a BLE radio, and the complete ated. For single-spot power management section in- measurement, opti- cluding battery condi- cal technology tioning and is more

SMART SENSORS EBOOK 15 GEN II WEARABLE DEVICE

bility on the photodiode selections and the LED wavelengths as well as mechanical con- 80 straints such as spacing between LEDs and photodiodes. The GEN II device supports two green LEDs, one red LED, and one infrared 70 LED. For those without experience in design- ing optical systems, it might be easier to in- 60 tegrate a complete optical module. There are different options in terms of the number of photodiodes, their sizes, and the 50 selection of LED wavelengths. The latest modules have been developed in such a way 40 that they show a great optical performance Comp A even when they are mounted behind a plas- tic window. The first generation required a 30 split window to reject internal light pollution, Comp B which can be seen as optical crosstalk. A

Current Consumption (µA) 20 split window helped to reduce dc offsets from light coming directly from the LEDs without penetration into the body. Such a 10 split window is not easy to integrate, nor is ADXL362 it attractive from a cost point of view. The 0 latest families have been improved substan- 0 100 200 300 400 tially and even with just one complete win- dow, the ILP effects have been reduced to Output Data Rate (Hz) almost zero. Biopotential measurement is supported Figure 3. ADXL362 power consumption as a function of the output data rate. by two individual AD8233 analog front ends. The AD8233 is a single-lead ECG front end charging. A second board supports all sensing technologies. The op- with embedded right leg drive (RLD) capability, and has been de- tical system for PPG (photoplethysmogram) measurement is built signed to extract, amplify, and filter small biopotential signals in around the ADPD107, ADI’s second-generation optical analog front the presence of noisy environments. Focus applications for this end. The block diagram is shown in Figure 2. component are wearable devices, porta ble home care systems, and The analog front end operates as a complete transceiver, driving exercise equipment. The AD8233 operates in a dc coupled config- the LEDs in the system and measuring the return signal from the pho- uration. The input stage is divided over two gain stages. The first todiode(s). The objective is to measure photocurrent that is as high stage, with limited gain, is followed by a second-order, high-pass as possible for a given amount of LED current spent (current transfer filter and a second gain stage. The total gain of this input block is ratio). The input receive signal chain is designed around a config- 100 V/V, which includes the subtraction of the offset as a result of urable transimpedance amplifier where the gain can be programmed the electrode half-cell potential. The second stage is combined with in four steps up to 200k. The second stage is responsible for ambient a third-order, low-pass filter. It is second-order Sallen Key working light rejection. Ambient light interferers are a big issue, especially in unity followed by an additional low-pass filter. The objective of when the light is modulated, as with solid-state lighting systems with this filter is to reject all EMG-related signals coming from muscle LEDs or energy-saving lamps. The ambient light rejection block con- activity. tains a bandpass filter followed by an integrator to support synchro- The operating frequency of the biopotential front end de pends nous demodulation. This is a key function and rejects external light on the use case. For a normal heart rate monitor, where just QRS interferers very effectively. When the ambient light rejection stage is detection is needed, the operating frequency range is much less not needed, this block can be bypassed completely. compared to an ECG monitor where more information is required, The optical system makes use of light pulses. There are three LED such as timing and amplitude data from the P-wave vs. QRS-Com- current sources that are fully programmable. The maximum LED plex vs. T-wave. The band of interest can be configured by external currents are programmable and can be as high as 370 mA. Also, resistors and capacitors. To support flexibility, the GEN II wearable the pulse width is programmable and can be as narrow as 1 µs; device has the ECG front end connected to the embedded elec- however, for a good signal response, pulse width should be around trodes, configured in a sports bandwidth, supporting a band of in- 2 µs to 3 µs. Usually a series of LED pulses is given while the ana- terest from 7 Hz to 25 Hz. The second AD8233 that can be operated log-to-digital converter is sampling the photodiode receive signals in combination with external electrodes is configured to monitor related to the pulsed LED transmit pulses. The digital engine is able signals from 0.5 Hz up to 40 Hz. In principle, nearly any bandwidth to average multiple samples to increase the overall effective num- can be selected; however, this requires modifications of the hard- ber of bits. ware by changing R and C settings. Along with the optical system, mechanical design also has a major Depending on the required accuracy, output can be sent to the 12- impact on overall performance. In this GEN II device, the optical com- bit successive approximation register (SAR) ADC embedded in the ponents have been selected as discrete devices. This provides flexi- Cortex®-M3 processor on the sensor board, or digitization can be

16 SMART SENSORS EBOOK GEN II WEARABLE DEVICE

done by the standalone 16-bit SAR ADC. Tradeoffs can be made and Bringing it all Together depend on either accuracy or battery lifetime. The GEN II device makes use of two processors. This is not absolutely At the back side of the device are two electrodes. These have a needed but provides more flexibility. The interface board with BLE radio double function: in addition to ECG measurement, these also can be has one processor and the same device is used on the sensor board to used for electrodermal activity (EDA) or galvanic skin response be able to run auton omously. The ultra-low-power ADuCM 3029 has (GSR). This is related to the conductivity of the skin, which is mo- been integrated to collect sensor data and run the algorithms. mentarily changed by emotion, coming from either an internal or ex- The core is a 26-MHz Cortex-M3 with a rich peripheral set, on- ternal stimulus — skin impedance changes, for instance, as a result board memory, and an analog front end. There are four operating of stress or epilepsy. The GEN II device is able to detect this minute modes; in full operation, the chip consumes 38 µA per MHz. If pro- change in conductivity. The system is making use of an ac excitation cessing power is not needed, the device can run in flexi-mode in signal that is applied over the two dry electrodes. Wet electrodes which the analog front end is running, peripherals are active, and can be used as well and will be better; however, this device is just the measured signals can be stored in memory through DMA. This making use of two embedded dry stainless steel electrodes. The mode consumes 300 µA, making the chip very attractive for low- main advantage of using an ac excitation signal is that this will not power, battery-operated systems. There are several security features polarize the electrodes. embedded for code protection and a hardware accelerator for cryp- The receive signal chain represents a transimpedance amplifier, fol- tographic functions. lowed by the 16-bit, SAR ADC. The ADC sampling rate is much higher than the excitation rate for performance reasons. The ADC output is Selection of Use Cases followed by a discrete Fourier transform (DFT) engine, running on The GEN II wearable device can be used for many purposes. The the ADuCM3029 processor, to represent the complex impedance. The sensors can be integrated in smart watches but the range of func- measurement principle described above is capable of measuring skin tions, including accurate heart rate monitoring and activity measure- impedance or skin conductance at a high signal-to-noise ratio and a ment/ calorie burn, are also helpful for sport watches. The tradeoff very good suppression of 50 Hz/60 Hz environmental noise. The cir- between a smart watch and sport watch is mainly made between ac- cuit around this measurement principle is completely built with dis- curacy vs. battery lifetime. crete components. The main reason for this design decision is The device can be used to measure stress or emotional state. Usu- flexibility and accuracy at a rather low power dissipation. ally a combination of measurements is used to get a reliable reading such as skin impedance together with heart rate variability and tem- Vital Measurement Parameters perature. Blood pressure monitoring is another interesting use case. A wearable device is worthless for measuring vital parameters This is a very important parameter but most of the systems are cuff- without having a notion of what the human body is doing. For that based, which are hard to integrate in a wearable and continuous sys- reason, motion detection and profiling are important. Some use cases tem. There are certain techniques that can be used to measure blood like optical heart rate monitoring are very sensitive to motion, and pressure without the need for a cuff. One technology is by making motion can destroy the accuracy of the measurement completely. For use of the pulse-wave transmit time (PTT). This requires ECG meas- that reason, motion also needs to be tracked to compensate for arti- urement in combination with PPG measurement. The sensors inside facts. Motion sensors will help to track movement and, where needed, the GEN II wearable device can support this. motion can be compensated in the final outcome of the readings. The The last key market is related to elderly care and independent liv- ADXL362 low-power motion sensor has a 3-axis MEMS sensor with ing. There is huge need for systems that can help caregivers monitor an integrated 12-bit ADC to detect motion in the X-, Y-, and Z-axes. certain parameters remotely. This wearable device supports 95% of The output data rate (ODR) of the ADC represents the power dissi- the features needed. The system monitors several vital parameters. pation of the sensor, which is 3 µA at the full ODR of 400 Hz per axis. It can track if people are moving or walking but is also able to detect In Figure 3, a plot of the power dissipation as a function of the output falls. The missing piece in the wearable design is an emergency button data rate is shown. but this is a matter of connecting one I/O pin on the processor to a This sensor can also be used as a motion activation switch. There switch on top of the device. is a possibility to reduce the sampling rate to just 6 Hz. Every 150 ms, the sensor wakes up and measures the motion activity. Without mo- Conclusion tion, it goes straight back to sleep for another 150 ms. At the moment, The GEN II device has many high-performance sensors and fea- motion is being detected at a g-force equal to or higher than the pre- tures embedded in a small, wearable system. Besides the elec- programmed threshold level. For at least the minimum time pro- tronic design, many mechanical aspects have also been taken into grammed, the sensor generates an interrupt or enables a power consideration. This makes the platform very attractive to design switch to turn on the application. With this mode, the sensor is con- companies and device manufacturers focusing on the semi-pro- suming only 300 nA, and can run for years on a single coin cell bat- fessional sports market, the medical market, and companies in- tery. All the use cases summarized make the motion sensor a volved in systems for smart buildings, independent living, or must-have in a wearable device. elderly care. All parameters can be measured simultaneously but Temperature sensing is another vital parameter; the GEN II wear- algorithms need to complement the application to support the use able has two temperature sensors embedded. The wrist-worn device cases. Instead of building hardware before testing and validating uses NTCs to measure both skin temperature and the temperature in- the algorithms, this device will give developers and device manu- side the device — there are multiple methods to measure temperature facturers a quick start. via sensors contacting the body. The NTCs are powered and condi- This article was written by Jan-Hein Broeders, Healthcare Business tioned by discrete circuitry and the 16-bit ADC finally converts the Development Manager for Analog Devices’ Healthcare Business in Eu- signals into the digital domain. rope, the Middle East, and Africa.

SMART SENSORS EBOOK 17 Augmented reality can be used to connect human personnel with the digital spectrum, achieving the ultimate level of integration. (By Gorodenkoff/ Shutterstock.com)

Key Factors to Consider When Designing a Smart Factory

18 SMART SENSORS EBOOK SMART FACTORY

Key Features of a Smart Factory At its core, the smart factory utilizes a combination of technologies to mold the traditional automation process into a synergetic ecosys- tem that brings together all aspects of the business including opera- tions, process, enterprise, supply chain, and beyond. Ultimately, this results in a dynamic, optimized, and flexible system. None of this can be made possible without first understanding the key concepts that drive both the digital and physical mechanisms of the smart facility: connectivity and optimization. Connectivity – From the smallest node to the most complex robotic system, a smart factory relies on connectivity. It’s the bedrock on which everything else is built; it facilitates the seamless transfer of data across all levels of the factory, enterprise, and supply chain. Regardless of whether you’re building a new factory or upgrading an existing one, the design phase should be guided by the concepts Interconnectivity enables monitoring from a smart device, enabling manufac- of data collection and connectivity. With this in mind, manufacturers turers to identify issues without being physically present in the facility. (By Pop- need to focus on how to utilize technologies such as sensors, cloud Tika/Shutterstock.com) computing, and the IoT in their designs. Sensors can be used in a wide variety of applications including however, to reach peak efficiency, these digital technologies perform monitoring machinery, detecting hazardous conditions within the fa- best when coupled with physical technology. In a smart factory, this cility, identifying inefficient energy usage, and tracking inventory. generally refers to the use of robotics, additive manufacturing, and These sensors are designed to collect data 24/7 and deliver it to a vi- more recently, augmented reality. sualization system that helps users in assessing the data and drawing When designing a smart facility, manufacturers should consider subsequent conclusions on how and where to optimize their system. how these physical technologies can work best for them. Most com- With machines, for example, the sensors can detect whether or not panies already utilize robotics on the shop floor, for example, but a a piece of equipment is damaged, requires cleaning, or is lagging in smart factory would also use such machines in other capacities such production efficiency. This information informs manufacturers regard- as in warehouse operations. Most importantly, these robots would ing exactly what’s happening in their facility in real time, empowering be interconnected with other aspects of the facility. With the facility’s them to make better informed decisions that optimize production sensors collecting data, the IoT and cloud computing making the and decrease downtime. data accessible throughout the plant, and the various digital tech- When connected to digital technologies like the IoT and cloud com- nologies analyzing it and learning from it, these machines can per- puting, sensors can collect data from all corners of the facility and bring form routine processes with more accuracy and efficiency than they it together to provide manufacturers with a 360° view of the factory, ever could through traditional methods. identifying all issues at once at any given moment. Because such inter- Although robots are nothing new to industry, additive manufactur- connectivity can allow for monitoring from a smart device, manufacturers ing and augmented reality are more recent additions to smart man- can even identify issues without being physically present in the facility. ufacturing. Additive manufacturing is primarily being utilized for rapid Optimization – Optimization refers to the implementation of analytical prototyping applications. While this is useful in and of itself, additive digital tools in an effort to minimize the need for human intervention while manufacturing processes get a real boost when coupled with tech- simultaneously increasing productivity and efficiency. These tools com- nologies such as AI; for example, AI can be used to analyze a range pound the impact of the data collection and connectivity technologies. of variables, such as material characteristics or cost restrictions, and By considering the potential incorporation of machine learning use that analysis to generate a prototype design. platforms, AI, Big Data, the digital supply network, procurement 4.0, Meanwhile, augmented reality can be used to connect human per- and RPA, among others, manufacturers can automatically handle and sonnel with the digital spectrum, basically achieving the ultimate level maximize efficient production across a broad variety of processes of integration. When human workers collaborate with digital tech- without the need for human intervention. nologies, they combine the best of both worlds by pairing high levels When approaching the design phase of a smart factory, manufac- of automated precision with human nuance. turers must consider how these technologies can be leveraged for them specifically. Moreover, manufacturers need to think about how Next Steps process, operations, enterprise, and supply chain can all be stream- While designing a new smart facility or even incorporating smart lined into a singular, networked system. elements into an existing plant can seem like a daunting process, the Together, these technologies can seamlessly manage processes such advantages outweigh the drawbacks by creating lean, efficient sys- as advanced planning and scheduling, inventory analysis, increased supply tems that enhance production, lower costs, and improve business chain transparency and traceability, automated paperwork processing, practices. predictive maintenance, and more. Not only is such optimization cost-ef- Smart technologies do much more than just enhance automation fective, but it also dramatically decreases risks associated with human processes — when designed properly, they fuse together operations error while freeing up employees to tackle more sophisticated tasks. and information technologies to create highly interconnected, agile sys- tems. These systems are not only capable of improving all aspects of Beyond the Digital Sphere operation but can also be used to optimize all aspects of a business. Many of these technologies and practices are performed within the This article was written by Kristin Manganello of Thomas, New digital world and as such, are essentially invisible to the human eye; York, NY.

SMART SENSORS EBOOK 19 Benson Hougland Alan Grau Executive Forum: Sensors

Jen Gilburg & the IoT

Robbie Paul Dr. Boris Golubovic

o longer just a buzzword, the Internet of Tech Briefs: The IoT is changing everything, yet the path and Things (IoT) is rapidly taking hold in many scope of those changes seem very unsettled. Are companies Ndifferent industries, from aerospace and au- rushing to implement IoT initiatives before aligning the tech- tomotive, to medical and manufacturing. The IoT nology to their business? ecosystem incorporates Web-enabled smart de- vices that use processors, sensors, communication Hougland: Certainly, they are. But you have to understand the hardware, and software to collect, send, and act on technology before you can determine how a given business data they acquire. Tech Briefs posed questions to problem can be addressed and at this point, the best way to executives in the sensors and IoT component indus- understand it is hands-on. Start small and fail often. You may tries for their views on issues concerning the role of not know what business problems can be solved by the IoT the IoT in manufacturing. until you dig in and try. Our executive panel members are: Jen Gilburg, Senior Director of Strategy, Sensor Solutions, at TE Grau: Smart companies are experimenting with IoT initiatives Connectivity; Dr. Boris Golubovic, Vice President, and adopting a “fail forward fast” approach. Initiatives that Marketing & Strategy, at Littelfuse, Inc.; Alan Grau, enable organizations to learn what is practical and what will Vice President of IoT, Embedded Solutions, at produce business results are critical, enabling them to Sectigo; Benson Hougland, Vice President, Market- quickly adjust their initiatives and begin to see the business ing and Product Strategy, at Opto 22; and Robbie impacts. The companies that make large bets without vali- Paul, Director, IoT Business Development, at Digi- dating their assumptions will struggle to produce business Key Electronics. results. cono0430/Shutterstock.com

20 SMART SENSORS EBOOK SENSORS & THE IoT

Gilburg: IoT was so broad when it started and then we saw a more pragmatic approach. The companies that are use cases have a clear return “ At this point, the best way to understand IoT technology on investment (ROI). For example, any manufac- is hands-on. Start small and fail often. You may not know turer knows the cost of taking a machine offline. what business problems can be solved by the IoT until Being able to do predictive maintenance has a clear ROI. If you can show that you will recoup an you dig in and try.” investment and be more efficient with your Benson Hougland processes and achieve better growth, it tran- scends the business size.

Paul: There is a wide variety of opportunities in the IoT space. There Gilburg: For 5G, I think use cases that will leverage it immediately is some low-hanging fruit that can be easily capitalized for immediate are automotive and then we will start seeing it evolve for asset productivity or performance gains. The initial solutions are in home tracking, smart city, and some predictive maintenance. The chal- automation because is it such a large market opportunity and the lenges for smart cities is that the decision-making is fragmented large companies like Amazon, Apple, and Google can invest heavily. and large cities and municipalities tend to have other priorities for The real gains in performance and productivity are going to come spending. I think state/federal governments and large vested cor- with the integration of AI with IoT. porations will issue grants to get the momentum going but this will be a challenge. Golubovic: Consumers are often seen as early adopters of IoT; how- ever, companies also experiment on a smaller scale before rolling out Tech Briefs: When it comes to hardware and software for IoT, how IoT features to revenue-generating infrastructure. Deploying IoT is can companies determine what they need? often the next phase of what many have already been doing in au- tomation and controlling equipment within their localized facilities or Paul: The good news is that there is no shortage of options. We are campus. We should not forget that before IoT there was M2M com- also seeing improved bundling and integration of hardware and soft- munication and control on a localized scale. ware, leading to better solutions. Both hardware and software com- panies are realizing IoT is all about the ecosystem and they need to Tech Briefs: How will 5G expand the proliferation of IoT? work together to provide the best IoT solutions.

Paul: The low-power, wide-area (LPWA) technologies like LTE-M and Gilburg: For small companies building out a full end-to-end system, NB-IoT are game-changers. They provide a cost-performance ratio if they want to do condition monitoring, they buy company x’s con- that’s compelling and one that will open a host of opportunities. In dition monitoring system and it comes with the analytics platform, the U.S., reallocation of 2G and 3G spectrum the gateway, and the sensors. More realisti- will enable more efficient use of bandwidth. cally, there are system integrators pulling to- Even more IoT applications will be enabled as gether solutions and selling that platform. costs continue to come down. “ 5G will enable new use They have shifted the way they do business cases that previously were because of IoT. They work with different cloud Golubovic: 5G and provisions such as NB-IoT not cost-effective with providers and gateway vendors like Intel. They and LTE-M are targeted towards equipment look at the solution and work with the sensor connectivity — particularly mobile equip- current communication vendors to choose the sensors. They pull it to- ment connectivity. As such, these enable technologies. Many IoT gether for a customer and then sell it to five easy, location-independent deployment of other customers. applications require the simple, low-speed (e.g. alarm systems and access control), and data-rich applications higher data rates enabled Grau: In some cases, this will happen through (e.g. remote real-time image and video pro- by 5G. trial and error. I am a big believer in pilot cessing for process control or pipeline in- ” Alan Grau projects to learn what works. In other in- spections). stances including logistics and transporta- tion, business requirements and customer Hougland: Newer, faster, and more ro bust needs will drive innovation and implementa- communication technologies will only help the proliferation of indus- tion. For example, one large trash collection company is implement- trial IoT (IIoT) projects and 5G promises to address many current is- ing an IoT service to tell them when dumpsters are full to optimize sues related to latency, bandwidth, and reach. Not all IoT projects will drivers’ routes. The cost savings of not emptying partially full dump- require or benefit from 5G, though. It’s just another tool/technology sters is significant. that’s part of the larger IoT project picture. Hougland: Modularity, expansion, and scalability are critical elements Grau: 5G will enable new use cases that previously were not cost-effec- in considering a given hardware or software solution. Because IoT is tive with current communication technologies. Many IoT applications re- still new and companies can’t yet know all the ways its data can help quire the higher data rates enabled by 5G. Numerous network operators them, being able to try pilot projects and then expand proofs of con- and service providers are already envisioning or trialing implementations cept to scale is critical. Also important is the flexibility of a given hard- in smart cities, transportation, and a host of other vertical markets. ware solution to work with many different software components,

SMART SENSORS EBOOK 21 SENSORS & THE IoT

data being captured, revealing new, unexpected patterns and information. “ Both hardware and software companies are realizing IoT is all about the ecosystem and they need to work together to Golubovic: For many applications, IoT still may need to demonstrate value be cause its collected provide the best IoT solutions. ” data may be “hiding” the true value. The true in- Robbie Paul novation is often in transforming large amounts of data into actionable information. For example, an IoT system for remote monitoring of power lines may now be able to collect data, perform whether hardware-vendor supplied, off-the-shelf, or custom-devel- real-time analysis, deduce a power line breakage occurred, transmit oped. a control signal, and shut off the power, all before the broken con- ductor touches the ground, possibly preventing a wildfire. Golubovic: To help meet customer expectations, hardware in mobile or remote locations must be designed to account for the variety of Tech Briefs: Are companies still hesitant to implement an IoT net- adverse operating conditions. For outdoor applications, one can draw work because of security risks? on reliability standards developed for telecom and utility grid sys- tems. In mobile applications, the automotive industry has many years Paul: Security continues to be one of the last elements of IoT design of experience and established engineering guidelines that can be ap- and development to be worked on. More companies are taking a ho- plicable to IoT solutions. listic approach and including security in their initial plan. There has been a growing awareness that security is not something that can be Tech Briefs: The IoT enables the capture of different kinds and higher a consideration after deployment. Also, the number of security solu- volumes of data. Are companies prepared to effectively use all of tions is rapidly growing. There are more off-the-shelf security solution this data? pieces, both hardware and software, that are readily available in the market. Hougland: IoT is much more than just about the data. It’s an all-en- compassing approach that should align technologies like artificial Grau: What is unsettling to me is that many companies are moving intelligence, machine learning, operational equipment effectiveness, forward without addressing security concerns — add ing large num- and more with well-defined business objectives. The common bers of devices to their operations but failing to adequately secure denom inator in any IoT solution is the data itself but it’s just a com- those devices. This leaves data, devices, and network operations at ponent. You may collect all kinds of data but use only a subset of risk. The good news is that chip vendors have begun to address se- that data for a particular project or business objective. Once you’ve curity at the hardware level. garnered a successful implementation on one project, you should be able to use the Hougland: Anytime operational systems are data collection mechanisms to scale into connected to a network for the purpose of ad- I wish security was a other IoT projects. “ dressing IoT objectives, security is a concern Gilburg: When we talk about sensors and I bigger issue. Cybersecu- and should be considered fully before any im- hear the word “analytics,” I wonder where rity is my background and plementation is considered. Traditional au- they are getting their data and how good it I know how things can go tomation systems were separated from is. You have this garbage-in, garbage-out computer systems, often on proprietary net- problem if you’re not going with a well-cal- horribly wrong with IoT. works. Opening these systems in order to share ibrated, well-designed sensor. Sensors lose The good news is that 99% their data is valuable but requires thorough their calibration over time — especially analysis, so being hesitant is a virtue. of security breaches are cheap ones — so you need to make sure that the data you’re getting to analyze is ac- preventable. The bad news Gilburg: I wish it was a bigger issue. Cyber - curate and secure and not compromised. is that most people aren’t security is my background and I know how things can go horribly wrong with IoT. Not Paul: Customers are finding out quickly that doing the ounce of enough people are asking about security and they have too much data; most companies prevention. unfortunately, it is still very reactionary. only use a fraction of what they generate. ” Jen Gilburg There are lightweight things that can be One of the major trends in IoT is intelligence done. Are they being done? Probably not al- at the edge. Sensors and other end nodes are ways. Back when a system wasn’t connected, filtering and analyzing the data before send- there was no security risk. Now the systems ing it to the cloud. This results in more meaningful data for storage are connected. The good news is that anytime there is a breach, and further analysis. they are usually preventable with a little bit of common sense and a little bit of technology. The bad news is that most people aren’t Grau: In many cases, no, manufacturers are not yet ready to consume doing the ounce of prevention. and understand all the data produced by the IoT. New AI systems and analytics engines are being developed that will find actionable busi- Golubovic: Robust data and IT security practices are already critical ness information in this data. Companies will find ways to utilize the in today’s enterprises. If in place and maintained, these make for a

22 SMART SENSORS EBOOK SENSORS & THE IoT

ter at looking at problems deeper and more holistically. The big data Deploying IoT is often the next phase of what generated by IoT nodes are meeting up with the application of AI “ tools and this will be a strong trend in years to come. The real value many have already been doing in automation will be unlocked using AI to make sense and find trends that were and controlling equipment within their not even imagined.

localized facility. ” Gilburg: The evolution of IoT started with a lot of hype but it will be Boris Golubovic quite a bit slower than intended because we are pumping the brakes and thinking about security. It is challenging because IoT is an ecosys- tem — it takes a village — and there are a lot of different players trying to figure out which use cases will evolve and what data is needed. good foundation but each company must assess IoT-specific risks in Hougland: The IoT is just getting underway. We can’t see yet where their environment on a case-by-case basis. it will go but we know it will significantly change the way business is done — how processes are automated and products manufactured, Tech Briefs: What is the future of IoT? how remote sites are monitored and controlled, and how companies interact with one another, their suppliers, and their customers. Golubovic: Difficult to say. There are so many possibilities; for exam- ple, the use of NB-IoT and LTE-M-based IoT products using the mo- Grau: Increased connectivity is not new with the IoT but it is acceler- bile telecom infrastructure enables many remote monitoring and ating and reaching to smaller, lower-cost devices. The hope is that control solutions — virtually anyplace one can use a mobile phone. companies developing these devices will learn from the painful les- There are truly many possibilities for invention and innovation using sons of their predecessors and not ignore security. We will continue IoT for consumers and enterprises. to see headlines of data leaks, malware botnets, and IoT security breaches but with each publicized attack, the market will learn and Paul: Hardware and connectivity will continue to get cheaper, en- increasingly become aware of the risk/reward of not investing in se- abling more applications and solutions. Companies are getting bet- curing their IoT devices, data, or network operations.

FOR MORE INFORMATION:

Digi-Key Sectigo Littelfuse TE Connectivity Opto 22

SMART SENSORS EBOOK 23