polymorphic radios a new design paradigm for ultra low
play

Polymorphic Radios: A new design paradigm for ultra-low power - PowerPoint PPT Presentation

Polymorphic Radios: A new design paradigm for ultra-low power communication Mohammad Rostami , Jeremy Gummeson, Ali Kiaghadi, Deepak Ganesan, University of Massachusetts Amherst Why do we need a new low-power radio? Evolving communication needs


  1. Polymorphic Radios: A new design paradigm for ultra-low power communication Mohammad Rostami , Jeremy Gummeson, Ali Kiaghadi, Deepak Ganesan, University of Massachusetts Amherst

  2. Why do we need a new low-power radio?

  3. Evolving communication needs Gateway Cloud Connectivity Streaming (circa 1995) (circa 2015)

  4. Evolving communication needs Edge Cloud Cloud Connectivity Streaming (circa 1995) (circa 2015)

  5. What about radio power consumption? cloud offload Challenge: Low-power radios optimized for sporadic rather than streaming communication.

  6. What about radio power consumption? cloud offload Goal: Design a low-power streaming radio that provides low- latency connectivity and is reliable under dynamics.

  7. How can we optimize a streaming radio? -62 RSS (dBm) 30dB gap -92 RX sensitivity

  8. How do radios leverage the gap? 0 dBm -30 dBm - 62 RSS (dBm) Transmit softly (aka power control) 30dB gap - 92 RX sensitivity active sleep Transmit quickly (aka duty-cycling)

  9. How do radios leverage the gap? 0 dBm Oscillator low efficiency at LNA/PA low output power -30 dBm Mixer Transmit softly (aka power control) Transmit quickly (aka duty-cycling)

  10. How do radios leverage the gap? 0 dBm State-of-art low-power active radio (Nordic nRF5): - 16mW @ 0dBm -30 dBm - 8mW @ -40dBm Transmit softly (aka power control) Transmit quickly (aka duty-cycling)

  11. How do radios leverage the gap? 0 dBm State-of-art BLE (Nordic nRF5): - 16mW @ 0dBm - 8mW @ -40dBm -30 dBm Transmit softly (aka power control) Transmit quickly (aka duty-cycling)

  12. How do radios leverage the gap? 0 dBm State-of-art BLE (Nordic nRF5): - 16mW @ 0dBm - 8mW @ -40dBm -30 dBm Transmit softly (aka power control) Transmit quickly (aka duty-cycling) - Faster ⇒ higher on-off overhead - Shorter ⇒ less channel visibility

  13. Can we use passive radios? Active Radios Passive Radios Backscatter Oscillator TX LNA/PA Envelope Mixer Detector RX

  14. Can we use passive radios? Active Radios Passive Radios power efficiency Backscatter Oscillator TX LNA/PA reliability Envelope Mixer Detector RX

  15. How about passive radios? Active Radios Passive Radios power efficiency Reflected Signal Carrier Wave Carrier Wave reliability

  16. How about passive radios? Active Radios Passive Radios power efficiency Active Rx Passive Rx RX LNA reliability Sensitivity = -50dBm Sensitivity = -92dBm

  17. Key Challenge Active RF Passive RF reliable but inefficient efficient but unreliable

  18. Key Challenge Active RF Passive RF reliable but inefficient efficient but unreliable

  19. Polymorphic Radios Passive RF Active RF R Polymorphic radios : Combine active and Latency e l i a b passive building blocks to design low- i l i t y power streaming radios. Power

  20. Two modes of operation active-assisted passive Passive RF Active RF

  21. Mode 1: Active-assisted Backscatter Active Radio Received Signal Strength Passive Radio Receive Sensitivity @ 100kbps

  22. Mode 1: Active-assisted Backscatter Active Radio Received Signal Strength Passive Radio

  23. Two modes of operation active-assisted passive Passive RF Active RF passive-assisted active

  24. Mode 2: Backscatter-assisted Active Active Radio RSS @ 100kbps @ 100kbps RSS Passive Radio @ 3kbps Rx sensitivity depends on energy-per-bit

  25. Mode 2: Backscatter-assisted Active Active Radio RSS @ 100kbps Near-Zero Power RSS Passive Radio Channel Measurement @ 3kbps

  26. Polymorphic radio in a nutshell When passive works well, use active sparingly for reliability Passive RF Active RF When passive works poorly, use to monitor channel and optimize active duty-cycling.

  27. Roadmap: Network Stack Streaming Video/Audio Application Radio Selection/Switching MAC Polymorphic Radio HW PHY

  28. Hardware Overview Channel Meas. Shift Reg. Backscatter Switch Tx Baseband PA ~ Splitter LNA Rx Baseband Fast Envelope Detector Slow Envelope Detector

  29. Hardware Overview Channel Meas. Shift Reg. Backscatter Switch Active Radio Tx Baseband PA ~ Splitter LNA Rx Baseband Fast Envelope Detector Slow Envelope Detector

  30. Hardware Overview Channel Meas. Shift Reg. Backscatter Switch Tx Baseband Passive - Backscatter PA ~ Splitter LNA Rx Baseband Fast Envelope Detector Slow Envelope Detector

  31. Hardware Overview Channel Meas. Shift Reg. Backscatter Switch Tx Baseband PA ~ Splitter LNA Passive - Envelope Detector Rx Baseband Fast Envelope Detector Slow Envelope Detector

  32. Hardware Benchmarks Channel Meas. Shift Reg. Backscatter Switch ASK Mod RF Osc. Tx Baseband PA Ant. Splitter ~ ASK Demod Splitter LNA Rx Baseband Fast Envelope Detector Slow Envelope Detector Envelope detectors Mode Switching - Latency 30µs Active Mode 5.2mW @ 1.1dBm, 900MHz 10µW (measurement) Backscatter Mode 50µW (data)

  33. MAC - Decision Engine RSS measured in active mode TX mode Decision TX bitrate Engine RX mode Imputed RSS RX bitrate RSS measured in passive mode P ( RSS > RSS t in next k slots | RSS t , . . . , RSS t − 10 )

  34. MAC Evaluation - Datasets Wrist IMU Streaming IMU data @ 100 samples/sec from a Smartwatch Lapel Audio Streaming audio @ 4kHz sampling rate from a Lapel accessory (dialog) Eyeglass camera Streaming video @ 30fps from low power camera on an eyeglass

  35. Energy-efficiency vs. Reliability 100 T 1 T 2 T 3 Backscatter Energy efficiency (bits/nJ) 10 T 1 1 T 3 T 2 T 1 Duty-cycled Active T 2 T 3 0.1 0 10 20 30 40 50 Packet loss rate (%)

  36. Energy-efficiency vs. Reliability 100 T 1 T 2 T 3 Energy efficiency (bits/nJ) 10 Polymorphic ( 5x better T 1 1 T 3 efficiency than active ) T 2 T 1 T 2 T 3 0.1 0 10 20 30 40 50 Packet loss rate (%)

  37. Application: Audio Streaming Passive Q" Bit"to"IQ"mapping" 90 0 " +" RF"spli5er/combiner" I" RF"out" Data"bits" On/Off" ~" Tx"mode" …" Active Goal : Demonstrate low-power yet high quality audio streaming using a polymorphic radio

  38. Application: Audio Streaming Passive 100 Energy Efficiency (bits/nJ) 10 Polymorphic ( 6x better 1 efficiency than active ) Duty-cycled Active 0.1 Excellent Good Bad Poor Fair Mean Opinion Score (MOS)

  39. Application: Video Streaming Goal : Demonstrate tradeoff between sensing cost and communication cost using a polymorphic radio

  40. Application: Video Streaming + Sub-sample Gaze Goal : Demonstrate tradeoff between sensing cost and communication cost using a polymorphic radio

  41. Application: Video Streaming Dense sampling Passive Q" Bit"to"IQ"mapping" 90 0 " +" RF"spli5er/combiner" + I" RF"out" Data"bits" On/Off" ~" Tx"mode" …" Active Sparse sampling Passive radio has low cost, hence more energy is available for sampling, and vice-versa for active radio

  42. Application: Video Streaming Dense sampling Passive Q" Bit"to"IQ"mapping" 90 0 " +" RF"spli5er/combiner" + I" RF"out" Data"bits" On/Off" ~" Tx"mode" …" Active Sparse sampling Acgve Backscaher Polymorphic 0 5 10 15 20 25 30 Pupil Tracking Error (pixels)

  43. Conclusions R Latency e l i + a b i l i t y Active radio Backscatter radio Power Combining active and passive architectures allows us to design low-power streaming radios.

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend