Yann Le Guillou - Cognitive Radio opportunities and RF receiver Challenges

Transcript

1. 1 Cognitive Radio opportunities and RF Receiver Challenges. Yann Le Guillou Séminaire SCEE, Supélec, 19 Mars 2014
2. Overview 2 Evolution Trends of Spectrum Usage TV White Spaces oInterest oUsage example oHardware Radio Challenges Cognitive Radio and ADCs challenges and directions Receiver challenges Some Solution Directions oRF filtering Interference oInterference Cancellation oDigital assistance of dirty RF Conclusion Séminaire SCEE, Supélec, 19 Mars 2014
3. 2008 Trend: Increasing growth of mobile traffic More than 4 billion mobile phone users today and growing at a startling rate “7 trillion wireless device serving 7 billion people in 2017” – Wireless World Research Forum Mobile Internet traffic handled by mobile operators grow from 7 billion megabytes worlwilde [in 2008] into 63 billion megabytes in 2013 (CAGR 54%)” - Informa 2008 1924 Gap Between Wireless Traffic Supply and Demand Widens 3 Séminaire SCEE, Supélec, 19 Mars 2014
4. Spectrum Scarce? Spectrum not well managed in time domain and in geographical area [Swisscom] TV Bands Séminaire SCEE, Supélec, 19 Mars 2014 4
5. Unlicensed secondary re-use of TV broadcast spectrum [Swisscom] Séminaire SCEE, Supélec, 19 Mars 2014 5
6. Overview Séminaire SCEE, Supélec, 19 Mars 2014 6 Evolution Trends of Spectrum Usage TV White Spaces oInterest oUsage example oHardware Radio Challenges Cognitive Radio and ADCs challenges and directions Receiver challenges Some Solution Directions oRF filtering Interference oInterference Cancellation oDigital assistance of dirty RF Conclusion
7. Why the interest in TV White Space? Switch-off of Analog TV oThe release underutilized and valuable spectrum in UHF bands that can be used for next generation wireless network and low-cost communications UHF Signal properties oMore spectrum available! oGood propagation ability (long range for low power) oGood building penetration (e.g. public safety applications) Séminaire SCEE, Supélec, 19 Mars 2014 7
8. Frequency: 470-698MHz Channels: 6-8MHz Incumbent protection (WSDB) Power Levels and control of secondary users 8 © 2013 General Agreement at high level Séminaire SCEE, Supélec, 19 Mars 2014 … but Roadblocks are still there harmonization across regulatory regimes are difficult Industry is still waiting…
9. Application Example – Rural Broadband 9 © 2013 Outstanding propagation characteristic in the VHF/UHF reduce build-out costs and open era to Rural Broadband & Smart Agriculture Séminaire SCEE, Supélec, 19 Mars 2014
10. Example - Offload 3G/4G Traffic “Super-WiFi” in TV White Spaces: similar coverage as mobile broadband with just a 20% deployment density and 10-50 times faster than 3G [Maziar Nekovee, BT] 5.8GHz 2.4GHz ~0.8GHz 1 km² area in Central London, 5 000 homes 100 mW (WiFi), 20 mW (TVWS) Feasibility study Central London Séminaire SCEE, Supélec, 19 Mars 2014 10
11. PHY Receiver Challenge o Wide dynamic range (especially in ADC) o Highly sensitive o Interference Management o Spectrum Sensing over wide frequency range PHY Transmitter Challenge (=SDR + extreme frequency agility) o Spectral Emission o Low Power 11 TVWS Radio Hardware Challenges © 2013 Both PHY o Frequency Agility: Operation Frequency range from ~ 10’ MHz to several GHz o Bandwidth Agility o Flexibility/upgradability: Standard agnostic transceiver able to Rx/Tx current and future standards o Low cost o Low Power Séminaire SCEE, Supélec, 19 Mars 2014
12. Overview 12 Evolution Trends of Spectrum Usage TV White Spaces oInterest oUsage example oHardware Radio Challenges Cognitive Radio and ADCs challenges and directions Receiver challenges Some Solution Directions oRF filtering Interference oInterference Cancellation oDigital assistance of dirty RF Conclusion Séminaire SCEE, Supélec, 19 Mars 2014
13. ADC Bandwidth versus SNDR Far from feasible … [Murmann- ADC Performance Survey 1997-2014] http://www.stanford.edu/~murmann/adcsurvey.html Cognitive Radio (6GHz and 100dB) TV WS (900MHz and 100dB) Séminaire SCEE, Supélec, 19 Mars 2014 13 Jitter=1psRMS Jitter=100fsRMS Jitter=10fsRMS Jitter=1fsRMS
14. Energy/fs versus SNDR Today, FOM=1pJ/conv-step for high speed high SNDR 10fJ/conv-step 100fJ/conv-step 1pJ/conv-step Séminaire SCEE, Supélec, 19 Mars 2014 14
15. Required ADC Power (w/o clocking system) Suppose: o 1pJ/conversion o Signal -100dBm … 0dBm ↔17 bits Cognitive Radio 50MHz-6GHZ↔ fsample ~12GHz P=Econv .fsample .2#bits=10-12.12.109.217≈1.5kW TVWS Radio 700MHz-900MHZ↔ fsample ~1.8GHz P=Econv .fsample .2#bits=10-12.1,8.109.217≈235W RF-ADC most flexible « software-defined » receiver far from being Green today ….but also ~225x(1.8GS/s for 8MHz channel) to 1000x(10GS/s for ~10MHz channel) overkill Séminaire SCEE, Supélec, 19 Mars 2014 15
16. Required clock power Suppose o100dB SNDR ↔ Clock Jitter, σt ~1fsRMS (see slide 16) oState of the art FOMPLL ~-230dB where Then PPLL ~ 10 kW !! Séminaire SCEE, Supélec, 19 Mars 2014 16 Not achievable today neither tolerable for low power. Today 20mW dissipation is tolerable, resulting in ~0,7psRMS clock jitter
17. Down-conversion / filtering required This radio is still very flexible with software f-shift (Mixer) ADC Band Filter LNA Anti-aliasing Filter DFE Down-conversion Filtering IMEC-Renesas- 11b- 410MSPs – TI SAR ADC - ~1psRMS jitter Séminaire SCEE, Supélec, 19 Mars 2014 17
18. TI-SAR ADC 2x time-interleaved pipeline SAR ADC in 28nm CMOS 11-bit, 0-410MSPs, 2.1mW 6.5fJ/conv-step Power consumption scale linearly with sampling SNDR=55dB@205MHz On-chip calibration engine for gain and offset errors Time Skew issue [Verbruggen, VLSI, 2013] Séminaire SCEE, Supélec, 19 Mars 2014 18
19. To push TI-ADC SNDR above 11-bits need to calibrate frequency dependant spurs due to time-skew and bandwidth mismatches. Digital calibration is required to get the right accuracy Pushing TI- ADC SFDR Frequency dependant analytic expression SFDR= ~60dB ~120dB [Paquelet, Kamdem De Teyou, Le Guillou, IEEE NEWCAS, 2013] Séminaire SCEE, Supélec, 19 Mars 2014 19 Estimation accuracy needed to reach the desired SFDR of 90 dB in 99.9% of cases with a 4TI-ADCs, an input signal x(t) =1.5 sin(2πfot), fo = 146.29MHz and fs = 320MHz. Gain Offset Time-Skew Bandwidth Accuracy [ppm] 20 120 7 540
20. Overview 20 Evolution Trends of Spectrum Usage TV White Spaces oInterest oUsage example oHardware Radio Challenges Cognitive Radio and ADCs challenges and directions Receiver challenges Some Solution Directions oRF filtering Interference oInterference Cancellation oDigital assistance of dirty RF Conclusion Séminaire SCEE, Supélec, 19 Mars 2014
21. A - Interference from CR to Primary networks ( CR Transmitter) B - Interference from Primary networks to CR networks (CR Receiver) C - Interference from CR to CR ( CR Transmitter and Receiver) Interference Management Challenge TVWS Radio are envisioned to be capable of sensing and reasoning about the operating radio environments and thereby autonomously adjusting their transceiver parameters to exploit the underutilized radio resources in a dynamic fashion Cognitive Radio operates in interference-intensive environments. Minimum interference is therefore essential to the coexistence of Primary and TVWS Radio (Secondary System) 21 Séminaire SCEE, Supélec, 19 Mars 2014
22. On top classical Rx noise due to Receiver RF Front End Impairment (Phase noise, linearity, Noise etc.) Tx noise due to Tx leakage IntPrimary due to Primary Network Interferers CRNetworks due to all CR networks 22 Additional RF impairment for TVWS device S I G N A L Thermal Noise Phase Noise Image Harmonics Isolation Xmod IP2 IP3 SNR R x N O I S E Tx leakage CR networks IntPrimary ADC LNA LO 90° ADC DSP PA LO 90° ADC ADC Duplexer Séminaire SCEE, Supélec, 19 Mars 2014
23. Linearity requirements challenges Assumption: 2 tones at Pin [dBm] IM level equal to Noisefloor N=-100dBm Challenge: high linearity broadband receivers Required IIP3 [dBm] Required IIP2 [dBm] Séminaire SCEE, Supélec, 19 Mars 2014 23
24. Overview 24 Evolution Trends of Spectrum Usage TV White Spaces oInterest oUsage example oHardware Radio Challenges Cognitive Radio and ADCs challenges and directions Receiver challenges Some Solution Directions oRF filtering Interference oInterference Cancellation oDigital assistance of dirty RF Conclusion Séminaire SCEE, Supélec, 19 Mars 2014
25. Tunable RF filtering challenges Frequency selection by coil L & capacitor C Tuned mechanically by variable plate-capacitor C Center frequency: On-chip CMOS filters: o limited Q inductors =fc/bandwidth<15 (lossy, bulky, not flexible) o Limited fc tuning-range ~30% o Active inductors (noisy, non-linear) Alternative Idea o Do not use inductor o Switch & capacitors are very linear o Switch & capacitors scale with CMOS Séminaire SCEE, Supélec, 19 Mars 2014 25
26. N-path filter concept N signal paths switched series R-C ϕ1 VBB1 ≠0 VBB1 =0 [Franks, ISSCC1960] Ghaffari, RFIC2010/JSSC2011] [Mirzaei,VLSI10/TCAS2011] [Darvishi, ISSCC2013] VRF , fRF = fLO VRF , fRF = 1.5fLO In-band (fRF ≈fLO ): High ZRF Out of band: low ZRF (Short circuit) RC>>TON TLO TON =TLO /N ϕ1 ϕ2 ϕN VRF C R ϕ1 VBB1 C ϕ2 VBB2 C ϕN VBBN Séminaire SCEE, Supélec, 19 Mars 2014 26
27. N-path filter properties CMOS: Switches & capacitors TLO TON =TLO /N ϕ1 ϕ2 ϕN fLO 0 Frequency filtering @ RF Frequency filtering @ baseband High-Q RF filter Linear Widely tunable C ϕ1 VBB1 C ϕ2 VBB2 C ϕN VBBN Resistive (R) Séminaire SCEE, Supélec, 19 Mars 2014 27
28. [Darvishi, van der Zee, Nauta, ISSCC 2013] 65nm CMOS Séminaire SCEE, Supélec, 19 Mars 2014 Filtering properties: flexibly tunable 6th order 8-path 28
29. Frequency Hopping according to the free IM spot…. Take a look in the past … Interference Cancellation Techniques … 25-Year Old Hollywood Star invents frequency hopping 10th June 1940 Hedy Lamarr 29 Séminaire SCEE, Supélec, 19 Mars 2014
30. Exploit the spectrum Sensing and use the Free IM Spots Select the IM3-free spots Séminaire SCEE, Supélec, 19 Mars 2014 30
31. IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 44, NO. 6, JUNE 2009 …and in the future Bio Inspired RF Silicon Cochlea 100MHz-10GHz 31 © 2013 Séminaire SCEE, Supélec, 19 Mars 2014
32. • Today’s mobile device (Tablet PC etc.) have more processing capabilities than the original supercomputers ! • Capitalize on it to bring the required flexibility and assist dirty RF © 2013 Capitalize on trend in the embedded digital core processor Séminaire SCEE, Supélec, 19 Mars 2014 32
33. 33 © 2013 Some digitally assist analog and RF impairment directions … • Linearity (IP2, IP3) enhancement techniques • Noise reduction by cross-correlation techniques (as in Spectrum Analyser) • I and Q magnitude and phase imbalance • TI-ADC impairments • Clock spur cancellation … Séminaire SCEE, Supélec, 19 Mars 2014 33
34. Example - clock spur cancellation Clock spur reduced by 40dB Séminaire SCEE, Supélec, 19 Mars 2014 34
35. Overview 35 Evolution Trends of Spectrum Usage TV White Spaces oInterest oUsage example oHardware Radio Challenges Cognitive Radio and ADCs challenges and directions Receiver challenges Some Solution Directions oRF filtering Interference oInterference Cancellation oDigital assistance of dirty RF Conclusion Séminaire SCEE, Supélec, 19 Mars 2014
36. Conclusion Cognitive Radio receiver challenges are interdisciplinary in nature (RF, analog, digital signal processing, …) Capitalize on CMOS technology and DSP but think out of the box o Make optimum use of Switches and Cap in CMOS technologies to built RF functions o Make optimum use of technology speed and high speed clock possibility o Digitally assist dirty RF & Analog to boost the performances 36 Séminaire SCEE, Supélec, 19 Mars 2014