A massive shift is happening in technology: everything is getting connected to the Internet. The convenience and power of the Internet makes these products much more valuable, so companies are rushing to get their Internet of Things (IoT) (Figure 1) products to market. But success in this rush requires expertise in sensors, wireless communication, and data collection technology.
Figure 1. Internet of Things
Some common sensors measure temperature, pressure, vibration, fluid flow, and acceleration. Additionally, cameras and microphones are very common. Now, newer phones have gyroscopes and magnetometers. There is now single-chip solutions available that contain three accelerators, three gyroscopes, and three magnetometers. We will see continued improvement in accuracy and sensitivity of current sensors, as well as the ability to detect other forms of energy and substances.
As sensors proliferate, designers increasingly choose to bring them together through wireless networks. But different sensors have different needs, and different wireless standards offer different features. Which wireless standard makes sense for your application?
Multiple standards exist for wireless, such as GPRS, GSM, CDMA, Zigbee, Bluetooth, WiFi, global positioning system (GPS), near field communications (NFC), and radio-frequency identification (RFID). Each wireless standard has its advantages and disadvantages, so depending on your application, different technologies are recommended to ensure you get the one you need. Which wireless standard makes sense for your application? See Table 1.
Table 1. Guide to Wireless Technology Standards
|Range||Short range up to 3 meters unless the power is increased||Short range up to 3 to 10 meters||Medium range up to 100 meters. Longer range requires more power||Medium range up to 100 meters|
|Operating Frequency||Low and high-speed data up to 24 Mbps in version 3. Most devices today have a maximum of 3 Mbps||Low and high-speed data up to 480 Mbps, but many wireless USB chips cannot handle this rate||Low-speed data up to 250 Kbps||High-speed data up to 600 Mbps|
|Implementation Cost||Low cost||Low cost||Low cost of parts, but implementation can be expensive||Low cost with new WiFi modules|
|Power Consumption||Low power – it can sleep between transmissions, but the wake-up time can be 3 seconds, compared to a few milliseconds for Zigbee||Low power – it can sleep between transmissions||Low power – it can sleep between transmissions||Generally not suited to battery applications, but there are several standards, some include low power|
|Networking Topology/Configuration||Point-to-point||Point-to-point||Self-organizing multinode mesh network||Point-to-multipoint|
|Typical Applications||– Connecting cell phone to headsets or computers to peripheral devices
|Primarily used in PC peripherals. Good for transmitting high-speed sensor data, such as vibration||Widely used in industrial and commercial application, such as lighting control, process control, and measurement of data that is on portable devices||Excels at carrying Ethernet signals wirelessly. It is also used to connect consumer electronics devices, such as hand held game consoles and cameras to PCs.|
|Strengths||Designed to replace wires in portable applications. Good for audio and supports video||– Software compatible with USB
– Easy to convert serial or USB to wireless USB
|– Excellent for very low data rate battery powered applications
– Ability to conserve power by sleeping yet has high power for long distance transmission- Can extend transmission distance by passing data from node to node
|– Designed for stationary base station with multiple portable devices-Ethernet data rates|
|Weaknesses||– Not good at sleeping between transmissions
– Limited range
|Limited range. Even shorter range at high speed (3 meters at 480 Mbps)||Complicated software to implement – some vendors have Zigbee-like solutions to overcome this||Not good for battery operation|
When you integrate and coordinate a group of sensors through a wireless network, you have begun to create a data acquisition subsystem. Data acquisition is the sampling of continuous real-world information to generate data that can be manipulated by software. Acquired data can be displayed, analyzed, and stored on a device. A PC, phone, or other wearable device can be used to provide data acquisition of real-world information from sensors. The components of data acquisition systems include appropriate sensors, filters, signal conditioning, data acquisition devices, and application software. Ultimately data analysis can only be as good as the input data, so acquisition is responsible for providing high quality data. And careful design is required to get meaningful data.
Let’s review some of the current challenges and constraints. What is limiting adoption of sensors? With many IoT devices, batteries are probably the major limitation. Batteries will improve slowly, but will continue to be a significant constraint as we try to make them last longer or to make them smaller. This slow progress is an issue because battery power and capacity are the key limits on wireless transmission range. For example, Bluetooth and WiFi do not go very far. Bluetooth is relatively low power. Cellular transmits a few miles, but uses a lot of power.
One workaround is to transmit in bursts, turning on that really powerful long-range transmitter for a fraction of a second, to send data and then turning it off again. However, this can create a problem for the receiver, which only listens for information when turned on: if one device only sends now and then and the other only listens now and then, they may never hear each other. Protocols like WiFi and cellular require regular or almost continuous connection, making it difficult to save battery power.
These constraints lead to some interesting design trade-offs. It is easy and inexpensive today to establish the location of a device using a GPS locator chip. But even GPS, which is only a receiver, is moderately power-hungry. Accelerometers, in contrast, are very low power. Often you can use an accelerometer instead of a GPS module to reduce power. The data is different, but with clever software an accelerometer is often able to give good position information.
In addition to power constraints, there are other characteristics of sensors the IoT designer has to keep in mind. Sensors themselves have limitations in accuracy and linearity. Extremely low-cost sensors are really cool, but are less accurate than the expensive ones, mainly in linearity, offset, and drift. These errors can be greatly reduced by calibration. Factory calibration can be expensive, and it does not overcome problems with drift easily. Calibration during use requires knowing when calibration makes sense.
For example, the offset errors in an inexpensive accelerometer can be removed by storing the readings when the accelerometer is stationary. Unfortunately, the accelerometer itself cannot distinguish stationary from a slow constant acceleration. This could lead to an incorrect calibration. Software can get around many of these limitations, but you need good electronics to collect the data.
The security challenges posed by IoT-connected devices may be particularly difficult because people are usually not willing to pay or plan for rare events, even if the events are known to have high impact. But in fact, we may have to develop new risk management strategies for extreme but rare events, such as major storms (Hurricane Sandy), large increases in electrical interference caused by solar flares, or just hacking of a website.
For example, our electrical and electronic infrastructure has never been tested by the extreme solar flare activity that happened more than 100 years ago. Today it would be devastating, because we are so dependent on electricity. With the IoT, that dependence is growing rapidly. We may have to consider the impact of loss of the power grid on IoT-connected systems.
But we do not have to look at once-in-a-century events. People are constantly being surprised by their own lack of preparation for hacking attacks. Tens of millions of personal data records are stolen every year as a result of such attacks. When even more data is available electronically—not just about your identity, but about your home security system, and your personal whereabouts–new threats are likely. A thief will be able to remotely monitor data about your home to know where you are, thus knowing when to break in. Security is more important than ever on the IoT. But can the sensors share the responsibility?
Another limitation of sensors is that each has their own data rate, precision, and signal processing needs. How can these be handled in a consistent way on the IoT? For example, temperature is normally sampled once a second or slower. Accelerometers, depending on the application, may need to be sampled at thousands of samples per second. If a device is sensing both, the packet content varies over time. This is not hard to handle, but there is no universal standard, making it harder to take data from many sources without understanding anything about the source.
Should the data be sent as raw binary, minimizing the processing needs at the sensor, or should it be converted to engineering units? Binary is meaningless unless you know how to convert the data into engineering units, making it a poor universal standard. If data is converted into engineering units, which units are standard? Temperature could be in degrees C or degrees F. Pressure has so many units it can drive people crazy: Pascals, bar, psi, inches of water, mm of mercury, and many more. So the units either need to be standardized or specified in the message.
In addition, the signal processing for some sensors is quite involved. Converting thermocouples readings into degrees C requires cold junction compensation, amplification, and linearization. The linearization is done in software to get the best accuracy. This requires either large tables or high-order polynomial calculations, which can be a significant burden on tiny low-power devices with limited memory. The burden will become less significant as digital processing advances over the next few years.
I hope that a data interchange standard is adopted that has the flexibility to handle home, industrial, medical, and other applications worldwide without being burdensome for the sensors nodes. You can see why this is not trivial.
Will we overcome the challenges to have an IoT future? The IoT is made possible because of improvements in sensors, which drives sensor innovation further. We will see major changes. I hope we can avoid most of the negative ones.
Mr. Walt Maclay, President and founder of Voler Systems, is committed to delivering quality electronic products on time and on budget. Voler Systems provides integrated design, development, and risk assessment of new devices for medical, industrial, aerospace, and instrumentation applications. Walt Maclay is recognized as a domain expert in Silicon Valley technical consulting associations. He is an instructor for the Product Realization NPI Program. He has also been past President of the Professional and Technical Consultants Association (PATCA). He has applied his outstanding leadership to many multidisciplinary teams that have delivered quality electronic devices. Mr. Maclay holds a BSEE degree in Electrical Engineering from Syracuse University. Voler clients, from early stage startups to large established corporations, are delighted with our quality and service. Past clients include Applied Biosystems, Applied Materials, BAE Systems, Boeing, Intel, JDS Uniphase, Lockheed Martin, Maxon Lift, Merck, NASA Ames, Northrop Grumman, Orbital Sciences, Pioneer Speakers, Puget Sound Naval Shipyard, Radiant Medical, Rain Bird, San Francisco Muni, Sandia Labs, Siemens, Spectra Physics, Stanford University, Teikoku Pharma, Thoratec, Tyco Valves, US Bureau of Reclamation, and the University of California at Davis.