Overview of wearable device sensors2017 rev9
- 1. < Slide 1 > Overview of Wearable Device Sensors Walt Maclay President, Voler Systems Product Development
- 2. < Slide 2 > Agenda • Common physiological measurements • Battery limitations • Saving power
- 3. < Slide 3 > Innovation examples Azumio FitBit Activity Monitor Elder monitor Go Key Bluetooth hearing aid Game Golf
- 4. < Slide 4 > Common physiological measurements
- 5. < Slide 5 > Motion • The most studied and used parameter • Step counts • Gait analysis (illness) • Types of motion (walking, standing, sitting) • Dead reckoning (9-axis motion) • Works on wrist, ankle, torso, etc. • Different algorithms at different locations • Even used to count steps on dogs
- 6. < Slide 6 > Body temperature • Few good locations to measure core temperature • Axilla (under arm) or forehead are best locations • Not convenient for a wearable device • Extremeties (eg wrist) have variable temperature • Algorithms can partially adjust over time • Good contact is important – heat flow causes errors
- 7. < Slide 7 > Blood oxygen • Oxygen saturation in blood • Measured by pulse oximeter (infra-red) technology • Measure loss through body of 2 IR wavelengths • Separates changes in blood from other changes • Measure heart rate at the same time • Transmissive or reflective measurement • Reflective for wrist
- 8. < Slide 8 > Heart rate • Measured by • ECG electrodes – two are sufficient • Pulse oximeter sensing – reflected • Transmitted works on finger and ear • Pressure sensing of the pulse in the wrist • Wrist measurement works well
- 9. < Slide 9 > Respiration rate • Number of breaths per minute • Few good locations to measure • Movement of chest • Chest strap • Not convenient for a wearable device except shirt • Thoracic Impedance eliminates chest strap • Does not work on wrist • May work with opposite hand touching an electrode
- 10. < Slide 10 > Blood pressure • Measure of systolic and diastolic pressure • Accurate measurement requires pressure cuff that is compressed and released • Does not work on wrist • Pulse Transit Time – measure at wrist or elsewhere • Currently not accurate enough
- 11. < Slide 11 > EKG or ECG • Electrocardiogram • Measure of electrical activity of the heart • Measurement points have to be rather far apart • At least one and a half inches – larger devices needed • 2 leads for basic measurement • 12 leads for standard EKG • Does not work on the wrist
- 12. < Slide 12 > EMG / EEG • Electromyogram / electroencephalogram • Measure of electrical activity in muscles or brain • EMG electrodes must be placed over the muscle • Requires accurate placement (millimeters) • Can measure the wrong muscle • EEG must use electrodes on the head • Large number for standard EEG • EEG Does not work on the wrist
- 13. < Slide 13 > Blood sugar (glucose) • Measure of glucose level in blood sample • Widely used • Becoming a wearable • Closed loop system replaces the pancreas • Measure then control glucose with a pump • Normally use blood from finger tip • less accurate elsewhere • Patch with microneedles – requires recalibration • Not accurate on wrist
- 14. < Slide 14 > Agenda • Common physiological measurements • Battery limitations • Saving power
- 15. < Slide 15 > Battery Limitations • Slow pace of improvement If improved like semiconductors: Size of a pin head, could power your car, cost 1 cent • Must always work around limitations Long time between charging vs small size Battery life per charge
- 16. < Slide 16 > When will battery technology improve? • Chemical energy storage is approaching the limit of its efficiency • Nuclear energy is out of the question • A lot of research being done on higher density and better safety • Perhaps 2 times higher density in a few years • Will safety suffer?
- 17. < Slide 17 > Energy Density
- 18. < Slide 18 > Energy Density and Safety • As energy density has increased, safety has become more of a problem • Safety circuits are required on Lithium batteries • Poorly designed batteries can catch fire even with safety circuit • Shipping of Lithium batteries is restricted and regulated • Cells without safety circuit cannot ship by air
- 19. < Slide 19 > Agenda • Common physiological measurements • Battery limitations • Saving power
- 20. < Slide 20 > 6 areas that impact power • Sensors • Wireless • Displays • Microprocessors • Software
- 21. < Slide 21 > How much power do sensors use?
- 22. < Slide 22 > Agenda Sensors • Wireless • Displays • Microprocessors • Software
- 23. < Slide 23 > Three ways to get data into the cloud 1. Smart device directly to cloud 2. Sensor to gateway to cloud 3. Sensor to cell phone to cloud
- 24. < Slide 24 > Power– How much? How far? 10 bytes/sec 1 Kbytes/sec 1 Mbytes/sec 1 m lowest power 100 m 1 km highest power All units in mW data rate distance
- 25. < Slide 25 > Power– How much? How far? 10 bytes/sec 1K bytes/sec 1 Mbytes/sec 1 m BLE/Zigbee 0.15 LoRa 0.5 Bluetooth 25 WiFi 50 BLE/Zigbee 7.5 Bluetooth 50 WiFi 75 WiFi 300 100 m LoRa 0.5 WiFi 100 3G Cellular 100 LTE Cellular 100 WiFi 100 3G Cellular 120 LTE Cellular 120 WiFi 400 LTE Cellular 500 1 km LoRa 1 3G Cellular 120 LTE Cellular 120 3G Cellular 150 LTE Cellular 150 LTE Cellular 700 All units in mW
- 26. < Slide 26 > Agenda Sensors Wireless • Displays • Microprocessors • Software
- 27. < Slide 27 > Display Technologies Cost Energy Usage LCD (grayscale) Digital paper uW WattsmW LED Color LCD backlit OLED
- 28. < Slide 28 > Emerging Technology: Digital Paper (eInk) • Nearly zero power when not changing But: • Not available in color (this is changing) • Slow – can’t display video • eInk kept prices high until they lost a patent fight in 2015 • Market may expand now
- 29. < Slide 29 > Agenda Sensors Wireless Displays • Microprocessors • Software
- 30. < Slide 30 > Microprocessor Power • Low data rate sensor data collection: 1 to 10 mW • Audio Compression: 10 to 100 mW • Video Compression: 100 to 1000 mW • Multi-processor running several Windows tasks: 5 to 50 Watts
- 31. < Slide 31 > Agenda Sensors Wireless Displays Microprocessors • Software
- 32. < Slide 32 > Common causes of power consumption issues • Inefficient use of the cellular & WiFi network • Sending small data packets • Not putting the processor to sleep • Keeping the display backlight on too long • Sampling data too often • Using high power sensors when lower power sensors are available • Inefficient (frequent) messages from an app
- 33. < Slide 33 > SUMMARY: Total Power of the System • Sensor + Processor + Display + Wireless • Low: 0.01 mW • 3 axis accelerometer, processor asleep, no display, Bluetooth LE sends one sample every hour • Runs years on a coin cell • Medium: 1 mW • GPS every minute, processor making decisions, LCD display, no backlight, WiFi transmits once a minute • Runs 2 months on one AA Alkaline battery • High: 1000 mW • Cell phone, many sensors, high power processor, color LCD display with backlight, always connected to WiFi and cellular • Runs a few hours
- 34. < Slide 34 > Latency for the Same Examples • Low Power, 1 Hour latency • Bluetooth LE sends one sample every hour • Medium Power, 1 Minute Latency • WiFi transmits once a minute • High Power, Latency of milliseconds • Always connected to WiFi and cellular
- 35. < Slide 35 > SUMMARY Common physiological measurements Battery limitations Saving power
- 36. < Slide 36 > Walt Maclay, Voler Systems Walt@volersystems.com Quality Electronic Design & Software Sensor Interfaces Wireless Motion Control Medical Devices Slides are available at volersystems.com/news/ultra-low-power
- 37. < Slide 37 >
- 38. < Slide 38 > Energy Density 300 3000 30000 Gasoline Lithium Lithium Ion (rechargeable) Alkaline MJ/g MJ/g
- 39. < Slide 39 > Laptops are no longer the definition of low power Watts MicroWattsMilliWatts
- 40. < Slide 40 > Energy Harvesting • Gather energy from environment • Motion, temperature difference, radio frequencies • Smaller battery • Usually need some storage • Major limitation – only microwatts of power • Few devices can operate on so little power • Photovoltaic cells can provide more power • Large size • Small strip (as in a calculator) generates microwatts
- 41. < Slide 41 > Wireless • Trade off between • Low transmission power • High data rates • Long transmission distances • Different standards optimized for different trade-offs
- 42. < Slide 42 > Power– How much? How far? 10 bytes/sec 1K bytes/sec 1 Mbytes/sec 1 m BLE/Zigbee 0.15 BLE/Zigbee 7.5 100 m 1 km All units in mW
- 43. < Slide 43 > Power– How much? How far? 10 bytes/sec 1K bytes/sec 1 Mbytes/sec 1 m BLE/Zigbee 0.15 Bluetooth 25 BLE/Zigbee 7.5 Bluetooth 50 100 m 1 km All units in mW
- 44. < Slide 44 > Power– How much? How far? 10 bytes/sec 1K bytes/sec 1 Mbytes/sec 1 m BLE/Zigbee 0.15 LoRa 0.5 Bluetooth 25 BLE/Zigbee 7.5 Bluetooth 50 100 m LoRa 0.5 1 km LoRa 1 All units in mW
- 45. < Slide 45 > Power– How much? How far? 10 bytes/sec 1K bytes/sec 1 Mbytes/sec 1 m BLE/Zigbee 0.15 LoRa 0.5 Bluetooth 25 WiFi 50 BLE/Zigbee 7.5 Bluetooth 50 WiFi 75 WiFi 300 100 m LoRa 0.5 WiFi 100 WiFi 100 WiFi 400 1 km LoRa 1 All units in mW
- 46. < Slide 46 > Power– How much? How far? 10 bytes/sec 1K bytes/sec 1 Mbytes/sec 1 m BLE/Zigbee 0.15 LoRa 0.5 Bluetooth 25 WiFi 50 BLE/Zigbee 7.5 Bluetooth 50 WiFi 75 WiFi 300 100 m LoRa 0.5 WiFi 100 3G Cellular 100 LTE Cellular 100 WiFi 100 3G Cellular 120 LTE Cellular 120 WiFi 400 LTE Cellular 500 1 km LoRa 1 3G Cellular 120 LTE Cellular 120 3G Cellular 150 LTE Cellular 150 LTE Cellular 700 All units in mW
- 47. < Slide 47 > Wireless Power Saving Tips • Select the right wireless technology • Distance • Data rate • Power • Need for receiving device • Sleep whenever possible • Continuous data transmission like audio and video is not ultra low power • Video is hundreds of milliwatts (WiFi) • Audio is milliwatts (Bluetooth) • Send burst of data then sleep • Send as little data as possible
- 48. < Slide 48 > Overhead vs Data
- 49. < Slide 49 > Agenda Wireless • LEDs • Displays • Sensors • Microprocessors • Software
- 50. < Slide 50 > LEDs & lumen-per-watt efficacy Color Wavelength range (nm) Typical efficiency coefficient Typical efficacy (lm/W) Red 620 < λ < 645 0.39 72 Red-orange 610 < λ < 620 0.29 98 Green 520 < λ < 550 0.15 93 Cyan 490 < λ < 520 0.26 75 Blue 460 < λ < 490 0.35 37 LED indicator uses 10 to 50 milliwatts LED illumination much more Bottom line: LEDs are not ultra low power Source:https://en.wikipedia.org/wiki/Light-emitting_diode
- 51. < Slide 51 > LED Power Saving Tips • Turn them on only when being viewed • Blinking them can dramatically reduce power • Turning on an LED for 50 mS every second is quite visible • 10 mW becomes 0.5 mW
- 52. < Slide 52 > Displays • Gray scale LCD displays are lowest power • Backlights are very power hungry • Color LCD requires backlighting
- 53. < Slide 53 > Display Power Saving Tips • Avoid back lighting or turn it on only when needed • Consider gray-scale LCD or digital paper displays • Don’t change the image frequently • Digital paper especially • Smaller displays use less power • Use sound or a single LED for user interface • Send data to a phone for display
- 54. < Slide 54 > How much power do sensors use? • Camera chip – 300mW • Illumination for camera at night – 200 mW • GPS (Position) – 20 mW • Load cell (Weight) – 10 mW • Pulse Oximeter (Blood Oxygen) – 10 mW • EKG/Heart Rate – 1 mW • 9-axis Motion Sensor – 0.5 mW • Microphone – 0.1 to 10 mW • Light Intensity – 0.1 to 5 mW • 3-axis Accelerometer – 0.01 to 0.1 mW
- 55. < Slide 55 > Sensor Power Saving Tips • Turn off sensors to save power • If not sampling frequently • Audio and Video often require continuous sampling • Use lower power sensors • Capacitance load cell instead of resistance type • ie: Motion sensing chip instead of GPS • Lower power, but less accurate • Use camera or GPS in cell phone instead
- 56. < Slide 56 > Microprocessors • Many microprocessors have ultra low power • Few milliWatts • Depends on processing power • Sleep and draw even less • microWatts • Interrupt line is often used to wake them up with an event from a sensor
- 57. < Slide 57 > Sleeping Adds Delay • Reduced power but with a trade-off • Short delay with interrupt • Sensor must be on to generate interrupt • Long delay with polling – wake up to see if anything needs to be done • Latency determined by time between polling
- 58. < Slide 58 > Microprocessor Power Saving Tips • Select the right processor for the task • Minimize power hungry processing • Sleep whenever possible • Compressing takes processing power but compressed data usually saves more on wireless transmission power
- 59. < Slide 59 > Software has a huge impact on power
- 60. < Slide 60 > Software Power Saving Tips • Power down subsystems when not being used • Dim the display when no input from user • Use wireless connections efficiently • Transmit bursts of data • Do not sleep too long (cellular & WiFi need to re- establish connection) • Off-load energy intensive processing to mobile application or cloud • Use a motion sensor instead of GPS
- 61. < Slide 61 > GPS power saving tip: Use motion sensor chip • Baseband Technologies has firmware solution that provides location as good as GPS alone • Achieved by mixed use • Motion sensor chip 75% of the time • GPS 25% of the time • < 2 milliseconds to calculate position • Cuts power 75%
- 62. < Slide 62 > Software Testing • Testing can be tricky because of the many different states • Some of the states only happen briefly, such as the high power states • Low power states require careful testing – can’t use software to query the status of a microprocessor that is asleep • Result of incomplete testing – more power use than there should be
- 63. < Slide 63 > Turning off wireless causes latency • Latency = Delay from request to response • Time between wireless listening = maximum latency • Turning on receiver frequently or continuously reduces latency, increases power • Listening takes less power than transmitting
- 64. < Slide 64 > Short On-Time – Less Efficient Turn on time Transmit data Turn off time Overhead Data (payload) CRC
Editor's Notes
- #2: Talk focused on IoT and Wearables
- #3: So we must improve the electronics and software – what I will talk about.
- #15: So we must improve the electronics and software – what I will talk about.
- #18: Lithium have higher energy density than Alkaline batteries by 4X (2X for rechargeable Lithium Ion) Gasoline is 10 X Lithium, about the highestchemical energy density.
- #19: Samsung battery problems
- #20: So we must improve the electronics and software – what I will talk about.
- #23: So we must improve the electronics and software – what I will talk about.
- #26: Not mesh What connects to cell phones/PCs/hot spots?
- #28: LED display – different from LED LCD – no backlight
- #33: Software determines when the power is on or off and how much power is dissipated. Slower sampling – lower sensor power, lower wireless power, less microprocessor power.
- #34: Cell phone uses advanced software to control power AED monitor w cell modem – 2 AA Alkaline last 2 years. Sends data once/day or less.
- #36: So we must improve the electronics and software – what I will talk about.
- #39: Lithium have higher energy density than Alkaline batteries by 4X (2X for rechargeable Lithium Ion) Gasoline is 100 X Lithium, about the highest chemical energy density.
- #41: So we must improve the electronics and software – what I will talk about.
- #43: BLE = Bluetooth LE Zigbee 30 meters at higher power Low data rates, powered off part time
- #44: Low data rates, powered off part time
- #45: LoRa – new standard, between BLE & Bluetooth – but allows devices to transmit at much larger distances Low data rates, powered off part time
- #46: WiFi supports higher data rates including video
- #47: Cellular has many new modes that can help to reduce power (like narrow band LTE for IoT).
- #48: Receiving device - cell phones/PCs/hot spots to connect to How much data is really needed? Ex: tracking device only sends data when moving.
- #49: Each technology has different optimum data packet size.
- #51: Eye most sensitive to green.
- #52: Being viewed – ex. LED watch button, motion detector Turning off – careful UI and device design
- #53: Backlights consumer most power, little in display itself.
- #54: Turning off – careful UI and device design
- #55: 30,000 to 1 ratio of power! Pick right type sensor
- #59: High power processor consumes more power when not active. Avoid greatest possible compression
- #61: Reestablishing connection is a fraction of a second extra time – Keep Alive
- #64: Less important if no commands are sent to device – data is old
- #65: Trade-off when turning on a very short time. CRC = error checking/correction