Kim Fowler Sensor Survey: Part 1 The Current State of Sensors and Sensor Networks During the summer of 2008, I created a sensor survey and sent it to the IEEE I&M Society, the Sensors and Transducers Journal, and Sensorsmag.com. Conservatively, the appeal to respond to the survey went out to more than 50,000 people. My goal was to better understand how sensors are used and how they might be used within 5 years time. 468 people took the survey; of those 21 had not had any meaningful involvement with sensors within the past 5 years and this group did not answer any questions regarding sensors. The remaining 447 people, who were involved in developing or using sensors, took the survey; of that group, 168 responded to more questions about the sensors that they currently use. SurveyMonkey.com hosted the survey. The survey did not collect any personal information about respondents to maintain their anonymity. I did not restrict the survey to one response per computer, in the event that several people used the same computer to respond to the survey. I also did not force answers to each question to avoid frustrating respondents and possibly causing them to exit the survey early; I would rather have some answers than none. I refined the survey through four successive pilot surveys sent out to very small groups of trusted colleagues between June and August 2008. Each time I sent the pilot survey to different colleagues to try to reduce the bias. The final survey was completed and became active for inputs in late August when the appeals were sent to fill in the survey. Part 1 of this column describes the current state of sensors and sensor networks according to the responses in the survey. The second part of this column in the April issue will show where the survey respondents think sensors and sensor networks are going within the next 5 years. Who is Doing What? Four of the first six questions are next. The other two questions were directional and did not give substantive results to report here. All of the questions presented in these survey results are renumbered from the original survey to present the results more clearly. Fig. 1. The vast majority of respondents were almost evenly split between universities and industry. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) Fig. 2. The majority of respondents worked in either research or design and development. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) February 2009 IEEE Instrumentation & Measurement Magazine 39 1094-6969/09/$25.00 2009IEEE
continued 1. Who has been your primary employer within the past 5 years? (Fig. 1) 2. What has been your primary type of work within the past 5 years? (Fig. 2) 3. What has been your predominate involvement with sensors over the past 5 years? (Fig. 3) 4. In what fields of application have you encountered or used sensors? Check all that apply. (Fig. 4) The vast majority of people answering the survey were in either academia or industry (Fig. 1). Figure 2 shows data that breaks down the type of work they do in a little more detail, while Figure 3 shows data that indicates the numbers of sensor systems with which respondents work. Finally, Figure 4 shows data that provides the field of application. Current State of Sensors and Sensor Networks Sensors and sensor networks have a wide variety of parameters and characteristics. The charts that follow reflect responses to queries about sensor parameters and characteristics. They indicate a spread of values and mild peaks within these parameters. Each respondent was asked to focus on a single, primary sensor that they had familiarity. In many questions the survey asked the respondent to answer each of the questions three times: Fig. 3. The majority of respondents developed either custom sensors or custom systems with sensors. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) Fig. 4. Other than basic research, respondents fields of applications spread all over the spectrum of possibilities. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) 40 IEEE Instrumentation & Measurement Magazine February 2009
Fig. 5. Most applications have sampling rates of 10 ksps or lower. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) Fig. 7. Sensor bandwidth ranges between 1 byte per second and 1 million bytes per second in most applications. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) Fig. 6. Most applications fit a range of accuracies between 8 and 16 bits. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) the first time for the minimum value experienced by their sensor application, the second time for the average value, and the third time for the maximum value experienced by their particular sensor application. Sensor Parameters and Characteristics The next four questions, each with three part answers for minimum, average, and maximum ranges of applications, were: 5. What are the sampling speeds for this particular type of sensor? Even if the sensor operates asynchronously, give your best estimate of speed in samples per second. (Fig. 5) 6. What is the range of accuracy for this particular type of sensor? (Fig. 6) Respondents were asked to reply in effective-number-ofbits (ENOB). The first available answer was for pulse or switch Fig. 8. Sensor power consumption varies widely from less than 1 μw to greater than 1 W. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) closures effectively 1 bit. The remaining available answers were: 2 bits (14 db), 4 bits (26 db), 6 bits (38 db), 8 bits (50 db), 10 bits (62 db), 12 bits (74 db), 14 bits (86 db), 16 bits (98 db), 18 bits (110 db), 20 bits (122 db), 22 bits (134 db), 24 bits (146 db). 7. What is the bandwidth of this particular sensor? Even if the data are not continuous, give your best estimate of bandwidth. (Fig. 7) 8. What is the power consumption of this particular sensor? Even if the level of power consumption is not fixed, give your best estimates of the average levels of power consumption. (Fig. 8) February 2009 IEEE Instrumentation & Measurement Magazine 41
continued 9. For this particular sensor, what is the range of distances from the transducer to the data collection module or computer? (Fig. 9) Fig. 9. Distances from the sensors transducers to the final data collector vary widely from less than 20 cm to greater than 50 m. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) Figure 5 is the first chart; it shows data that includes a wide spread in sampling speeds of sensors. Sampling speeds range from less than once a second to greater than 10GSamples/sec. Interestingly, most respondents indicated that they use sensors with minimum rates below 1KSamples/sec and maximum rates not much above 100KSamples/sec. Figure 6 shows data for the spread in sensor accuracy in terms of ENOB. The minimum seems to be lumped between 8 and 12 bits (not 10 bits) and the maximum between 12 and 16 bits (not 14 bits). There is no way to tell if this is because those resolutions are sufficient or if analog-to-digital converters (ADCs) with only those resolutions are available. The results in Figure 7 for choice of sensor bandwidth range follow the pattern in Figure 5 for sampling speed. Minimum bandwidth clumps around the range between 1 byte/sec and 10Kbytes/second, and the maximum bandwidth between 100 bytes/second and 1Mbyte/second. Figure 8 shows data that indicates a broad range of power consumption; between 10μW/sensor and 1 W/sensor. Figure 9 shows data that indicates that the range of distances between sensors and data collection is very broad and evenly distributed except for the minimum and maximum range. Clearly a number of applications have sensor connections that are less than 20 cm in length, as well as a number of applications with connections much longer than 50 m. Sensor networks The next three questions were: 10. For your chosen sensor and field of application, what format is the data transmission from the sensor to the data collection module or computer? (Fig. 10) Fig. 10. Most sensor networks have a wired form, though wireless networks are gaining in popularity. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) Fig. 11. Most applications seem to have 3 or fewer different types of sensors; the maximum per system is somewhat evenly spaced from 1 to beyond 10 different types per system. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) 42 IEEE Instrumentation & Measurement Magazine February 2009
Fig. 12. Average applications seem to have 10 or fewer total sensors per system. Maximum applications range evenly from 3 to more than 100 sensors per system. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) 11. How many different TYPES of sensors are typically incorporated into a system? (Fig. 11) This question required three different answers: minimum, average, and maximum. 12. How many TOTAL sensors are typically incorporated into a system? (Fig. 12) This question required three different answers: minimum, average, and maximum. The data in Figure 10 strongly indicates that wired sensor networks are the most prevalent, whether digital or analog. Digital wireless networks appear to be up and coming. Figure 11 shows data that indicates that the different types of sensors within any one network are limited to five or fewer types. At least this is true for most respondents in minimum or average configurations of sensor networks. Only in the more extreme cases can sensor networks have significant numbers (> 6) of different types of sensors. Figure 12 shows data that indicates that currently most networks have 50 or fewer total sensors per system at least for minimum or average configurations of sensor networks. Only in the more extreme cases can sensor networks have total numbers of sensors greater than 50. The final two questions were: 13. Are any security measures used in the data transmission from the sensor to the data collection module or computer? (Fig. 13) 14. What is the environment that this particular sensor must endure in this particular field of application? (Fig. 14) This question required three different answers: minimum, average, and maximum. Office: - Temperature ranges from 0 to 50 C - Mild vibration and shock (drop to concrete floor from 1 m) - Humidity ranges from 10% to 80% - Pressure from sea level to 5000 m - EMI is nominal, power drops and brown outs, surges, spikes, ESD Medical: - Temperature ranges from 0 to 50 C - Mild vibration and shock (drop to concrete floor from 1m) - Humidity ranges from 0% to 100% condensing, sometimes immersion - Corrosion in implanted devices - Pressure from sea level to 5000m - EMI can be moderate, power drop outs, surges, spikes, ESD, RFI, reversed battery leads Industrial: - Temperature ranges from -40 to 85 C - Potentially severe vibration and shock - Humidity ranges from 0% to 100% condensing, sometimes immersion - Corrosion in caustic liquids and gases - Potentially volatile gases or liquids - Pressure from multiple atmospheres to near vacuum - EMI can be severe, power drop outs, surges, spikes, ESD, RFI Military: - Temperature ranges from -40 to 125 C - Potentially severe vibration and shock - Humidity ranges from 0% to 100% condensing, sometimes immersion - Corrosion in caustic liquids and gases - Pressure from multiple atmospheres to near vacuum - EMI can be severe, power drop outs, surges, spikes, ESD, RFI, Extreme (any or all of the following): - Temperature ranges much greater than from -40 to 125 C - Severe vibration and shock - Humidity ranges from 0% to immersion - Corrosion in caustic liquids and gases - Volatile gases or liquids - Pressure from many atmospheres to vacuum - EMI is severe, power drop outs, surges, spikes, ESD, RFI - Ionizing radiation, high energy particles and rays Figure 13 shows data that indicates that currently most sensor networks are not secure in data transmission between sensor and collector. A small percentage restricts access or have hardened enclosures but very few used encrypted data. Many survey respondents seemed to operate in the industrial realm, see Figure 14, in the typical or predominate environment. This requires more expensive components and structures to endure the environment. February 2009 IEEE Instrumentation & Measurement Magazine 43
continued Fig. 13. Most sensor applications have no security, or very little, to protect the data. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) Discussion I wish I could say that these results were from a scientific and statistically-valid survey. I can t. While 447 responded to the first part about the type of sensor work they did, only 168 people in a wide variety of fields gave responses to the more detailed questions about their current applications. I had no real control over ensuring the consistency of the responses to the questions with respect to application context; respondents perspectives could vary widely as to what they meant by important or minimum or maximum. A better method, though far more expensive in time and effort, would be to interview people directly. Summary I don t think that there were any real surprises in the current state of sensors and sensor networks. Sensors are used in a vast array of applications. Most typical applications seem to use 5 or fewer different sensor types and a total of less than 50 sensors per system. Sampling speed has a very wide range while resolution has some local peaks at 8, 10, 12, and 16 bits. Most Fig. 14. Many respondents had a harsh industrial environment for their sensors. ( 2008 by Kim Fowler. Used with permission. All rights reserved.) systems use a wired network or a combination of wired and wireless network. The one thing that did stand out is very few sensors systems have security measures for the data transmission between the sensor and the collector. The next column will give predictions from survey respondents as to what will happen to sensors and sensor networks over the next five years. Kim Fowler (kimf@ieee.org) lectures and develops concepts for new products. His web site, www.cool-stream.com, provides notes and advice on designing embedded, real-time products. He is researching the market and seriously considering starting a new sensors business. He is Executive Vice President of the I&M society for 2008 and 2009. 44 IEEE Instrumentation & Measurement Magazine February 2009