Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

How lower power consumption is transforming wearables and enabling new and different use cases: a conversation with Ambiq

97 views

Published on

Wearable devices have become a daily part of consumer’s lives, but batteries never last long enough. The battery-life problem continues to degrade the user experience due to the frequent need to recharge or replace batteries. However, recent advancements in low power technology​ are changing the game. In this webinar, we've teamed up with the experts from Ambiq Micro, leading provider of low power semiconductors, to share what's possible with lower power consumption and how energy efficient architectures enable new and different use cases across a variety of industries.

Published in: Technology
  • Be the first to comment

  • Be the first to like this

How lower power consumption is transforming wearables and enabling new and different use cases: a conversation with Ambiq

  1. 1. How Lower Power Consumption is Transforming Wearables and Enabling New Use Cases Scott Hanson CTO and Co-Founder Ambiq Micro
  2. 2. The World’s Most Energy-Efficient ICs 2 Ambiq Micro offers the world’s most energy- efficient MCUs and SoCs for wearables…
  3. 3. The World’s Most Energy-Efficient ICs 3 …and the broader IoT
  4. 4. The Battery-Powered Internet of Things 4 The future of IoT is exciting – but not if we need to replace billions of batteries per day/month/year [Image source: Cisco]
  5. 5. Battery Strain in Wearables • Segment display • Basic watch functions • Multi-year battery life 5 • High res color display • Basic watch functions • Motion/activity tracking • Heart rate monitoring • GPS tracking • Altimeter • Thermometer • Bluetooth radio • Days/weeks battery life
  6. 6. Battery Strain in Wearables • Segment display • Basic watch functions • Multi-year battery life 6 • High res color display • Basic watch functions • Motion/activity tracking • Heart rate monitoring • GPS tracking • Altimeter • Thermometer • Bluetooth radio • Days/weeks battery life Smaller, thinner industrial designs New functions (cellular radios, new biosensors, more sensor analysis, etc.) New use cases (e.g., smart clothing with 1+ year life)
  7. 7. No Easy Solution with Batteries 7 Batteries just aren’t improving fast enough!
  8. 8. The Other Side of the Energy Equation 8 Component Power Consumption Display 10-100mW MCU/CPU 0.1-10mW Radio 10-30mW Heart Rate Monitor 0.1-1mW GPS Receiver 10-100mW Power Management 10-20% of system power MCU Radio Display Sensor PMIC 370mWh battery 24h/day * 7 days = 2.2mW average power for 1 week life 675mWh battery 24h/day * 14 days = 77µW average power for 1 year life 24h/day * 365 days
  9. 9. Addressing the IC Energy Problem (1/3) 9 Moore’s Law has been one of the strongest drivers of energy efficiency gains [Dreslinski et al., Proc. of the IEEE, 2010]
  10. 10. Addressing the IC Energy Problem (2/3) 10 • Growing “menu” of transistor options provide flexibility • Increasingly, small transistors are also low leakage • New device architectures offer promise
  11. 11. Addressing the IC Energy Problem (3/3) 11 0 Volts 1.2 Volts 0 Volts 0.3 Volts Energy ~ (Voltage)2 Conventional Circuit Design Sub-threshold Circuit Design Sub-threshold design was first conceived >30 years ago, but Ambiq was the first to build a comprehensive platform
  12. 12. We’ve Come a Long Way 12 Texas Instruments MSP430 Ambiq Micro Apollo2 Improvement Launch Date 2003 2016 -- Processor Core 16b Proprietary Core 32b ARM Core -- Max Speed 8 MHz 48 MHz 6X Flash 60 kB 1 MB 17X RAM 2 kB 256 kB 128X Power Consumption Per Clock Cycle 440 µW/MHz 33 µW/MHz 13X Power Consumption Per Unit of Work 400 µW/Coremark 14 µW/Coremark 29X Today’s best in class processor is 29X more energy efficient than the best in class processor in 2003
  13. 13. Addressing the Radio Energy Problem • New protocols have emerged • IC technology and architectures have improved • Application level techniques have proliferated (e.g., duty cycling) 13 Average power of ~10µW for idle BLE connection at 1s interval
  14. 14. Addressing the Display Energy Problem • New display technologies have emerged • Gesture/motion detection have also helped • New technologies will continue to emerge 14 Electrophoretic Displays (Today) MicroLED Displays (Emerging) [Courtesy Plessey][Courtesy eInk]
  15. 15. New Use Case: Battery-Free Wearables • When energy gets low enough energy harvesting is possible • Sources of energy can include temp differential, solar, RF, etc. • Becomes a complex system challenge 15 [Source: datasheet for Marlow EHA-PA1AN1-R03]
  16. 16. New Use Case: The Embedded Coach • Wearables collect – and discard – huge amounts of data • It’s impractical to store data, so we must analyze on the fly • We can then use this data to deliver coaching • All of this is being enabled with improving CPU efficiency 16 Sensor Type Data Collected Per Min Possible Uses of Data Accelerometer, Gyro ~70 kB Activity recognition, gesture recognition, posture analysis, skills assessment, etc. Heart Rate Monitor ~40 kB Real-time HR zones, HR recovery/response, VO2 and VO2 max, cardiac efficiency, pulse pressure, etc. Microphone ~1.9 MB Keyword detection, context detection, respiration analysis, etc.
  17. 17. New Use Case: Intelligent Wearables • Neural networks emerging as an incredible tool • Applications include activity recognition, skill assessment, gait analysis, emotion recognition, etc. 17 Hannink et al, “Sensor-based Gait Parameter Extraction with Deep Convolutional Neural Networks”
  18. 18. The Next Energy Problem to Solve • Basic computation is simple – multiply accumulate (MAC) • Number of MACs can be enormous and continuous • This is a new energy problem to solve 18 Application Estimated MMAC/s Estimated Model Size Source Keyword Detection 33 140 kB Arik et al, “CNNs for Small-Footprint Keyword Spotting” Gait Analysis 74 48 MB Hannink et al, “Sensor- based Gait Parameter Extraction with Deep CNNs” Abnormal Heart Sound Detection 4600 12 MB Rubin et al, “Recognizing Abnormal Heart Sounds Using Deep Learning”
  19. 19. No Easy Answers for Low Power 19 We’ve come a long way to enabling ultra-low power wearables – and there is still a long way to go

×